Q4 2025 NVIDIA Corp Earnings Call

February 26, 2025

Operator: NVIDIA's fourth quarter earnings call. All lines have been placed on mute to prevent any background noise. After the speaker's remarks, there will be a question and answer session. If you would like to ask a question during this time, simply press star followed by the number one on your telephone keypad. If you would like to withdraw your question, again, press star one. Thank you. Stewart Stecker, you may begin your conference.

All lines have been placed on mute to prevent any background noise.

Speaker Change: After the Speakers' remarks, there will be a question and answer session. If you would like to ask a question. During this time simply press star followed by the number one on your telephone keypad and if you'd like to withdraw your question again Press Star one. Thank you Stuart Becker you may begin your conference.

Speaker Change: Thank you good afternoon, everyone and welcome to <unk> conference call for the fourth quarter of fiscal 2025.

Stewart Stecker: Thank you. Good afternoon, everyone, and welcome to NVIDIA's conference call for the fourth quarter of fiscal 2025. With me today from NVIDIA are Jensen Huang, President and Chief Executive Officer, and Colette Kress, Executive Vice President and Chief Financial Officer. I'd like to remind you that our call is being webcast live on NVIDIA's investor relations website. The webcast will be available for replay until the conference call to discuss our financial results for the first quarter of fiscal 2026. The content of today's call is NVIDIA's property. It can't be reproduced or transcribed without prior written consent. During this call, we may make forward-looking statements based on current expectations. These are subject to a number of significant risks and uncertainties, and our actual results may differ materially.

Colette Kress: Our Q4 data center compute revenue jumped 18% sequentially and over 2x year-on-year. Customers are racing to scale infrastructure to train the next generation of cutting-edge models and unlock the next level of AI capabilities. With Blackwell, it will be common for these clusters to start with 100,000 GPUs or more. Shipments have already started for multiple infrastructures of this size. Post-training and model customization are fueling demand for NVIDIA infrastructure and software as developers and enterprisers leverage techniques such as fine-tuning, reinforcement learning, and distillation to tailor models for domain-specific use cases. Hugging Face alone hosts over 90,000 derivatives, graded from the Lama Foundation model.

Colette Kress: Our Q4 data center compute revenue jumped 18% sequentially and over 2X year-on-year. Customers are racing to scale infrastructure to train the next generation of cutting-edge models and unlock the next level of AI capabilities. With Blackwell, it will be common for these clusters to start with 100,000 GPUs or more. Shipments have already started for multiple infrastructures of this size. Post-training and model customization are fueling demand for NVIDIA infrastructure and software as developers and enterprises leverage techniques such as fine-tuning, reinforcement learning, and distillation to tailor models for domain-specific use cases. Hugging Face alone hosts over 90,000 derivatives created from the Llama Foundation model. The scale of post-training and model customization is massive and can collectively demand orders of magnitude more compute than pre-training. Our inference demand is accelerating, driven by test time scaling and new reasoning models like OpenAI's O-3, DeepSeek R-1, and Grok-3.

Good afternoon. My name is Krista and I will be your conference operator today at this time I would like to welcome.

Speaker Change: With me today from Nvidia are Jensen, Huang President and Chief Executive Officer, and Colette, Kress Executive Vice President and Chief Financial Officer.

All lines have been placed on mute to prevent any background noise. After the speakers' remarks, there will be a question and answer session. If you would like to ask a question. During this time simply press star followed by the number one on your telephone keypad and if you'd like to withdraw your question again press Star one thank you.

Speaker Change: To remind you that our call is being webcast live on <unk> Investor Relations website.

Speaker Change: Webcast will be available for replay until the conference call to discuss our financial results for the first quarter of fiscal 2026.

Becker: The word Becker you may begin your conference.

Speaker Change: The content of today's call is <unk> property, it can't be reproduced or transcribed without prior written consent.

Speaker Change: Thank you good afternoon, everyone and welcome to <unk> conference call for the fourth quarter of fiscal 2025.

Speaker Change: During this call we may make forward looking statements based on current expectations. These are subject to a number of significant risks and uncertainties and our actual results may differ materially for a discussion of factors that could affect our future financial results and business. Please refer to the disclosure in today's earnings release, our <unk>.

Becker: With me today from Nvidia are Jensen Huang President.

Becker: Incident, and Chief Financial Officer.

Stewart Stecker: For discussion of factors that could affect our future financial results and business, please refer to the disclosure in today's earnings release, our most recent Forms 10-K and 10-Q, and the reports that we may file on Form 8-K with the Securities and Exchange Commission. All our statements are made as of today, February 26, 2025, based on information currently available to us. Except as required by law, we assume no obligation to update any such statements. During this call, we will discuss non-GAAP financial measures. You can find a reconciliation of these non-GAAP financial measures to GAAP financial measures in our CFO commentary, which is posted on our website. With that, let me turn the call over to Colette.

Becker: To remind you that our call is being webcast live on <unk> Investor Relations website.

Colette Kress: The scale of post-training and model customization is massive and can collectively demand orders of magnitude more compute than pre-training.

Becker: Webcast will be available for replay until the conference call to discuss our financial results for the first quarter of fiscal 2026.

Speaker Change: Most recent forms 10-K, and 10-Q and the reports that we may file on form 8-K, with the Securities and Exchange Commission. All our statements are made as of today February 26, 2025 based on information currently available to us.

Today's call is continuous property it can't be reproduced or transcribed without prior written consent.

Colette Kress: or Inference Demand is accelerating. driven by test time scaling and new reasoning models like OpenAIS-03, DeepSeq R1, and Grok3. Long-thinking reasoning AI can require 100x more compute per task compared to one-shot inferences. Blackwell was architected for reasoning, AI, and Blackwell supercharges reasoning AI models with up to 25x higher token throughput and 20x lower cost versus Hopper 100. It is revolutionary. Transformer Engine is built for LLM and a mixture of experts and And its NVLink domain delivers 14x the throughput of DCIe Gen5, ensuring the response time, throughput, and cost efficiency needed to tackle the growing complexity of infants at scale.

Becker: During this call we may make forward looking statements based on current expectations. These are subject to a number of significant risks and uncertainties and our actual results may differ.

Speaker Change: It's required by law, we assume no obligation to update any such statements.

Speaker Change: During this call we will discuss non-GAAP financial measures can find a reconciliation of these non-GAAP financial measures to GAAP financial measures in our CFO commentary, which is posted on our website with that let me turn the call over to Colette.

Colette Kress: Long-thinking reasoning AI can require 100X more compute per task compared to one-shot inferences. Blackwell was architected for reasoning AI inference. Blackwell supercharges reasoning AI models with up to 25X higher token throughput and 20X lower cost versus Hopper 100. It is revolutionary. Transformer Engine is built for LLM and mixture of experts inference, and its NVLink domain delivers 14X the throughput of PCIe Gen 5, ensuring the response time, throughput, and cost efficiency needed to tackle the growing complexity of inference at scale. Companies across industries are tapping into NVIDIA's full stack inference platform to boost performance and slash costs. NAP tripled inference throughput and cut costs by 66% using NVIDIA TensorRT for its screenshot feature. Perplexity sees 435 million monthly queries and reduced its inference costs 3X with NVIDIA Triton Inference Server and TensorRT LLM.

Becker: Results and business. Please refer to the disclosure in today's earnings release, our most recent forms 10-K, and 10-Q and the reports that we may file on form 8-K, with the Securities and Exchange Commission.

Colette Kress: Thanks Stuart.

Colette Kress: Q4 was another record quarter revenue of 39, 3 billion was up 12% sequentially and up 78% year on year.

Colette Kress: Thanks, Stewart. Q4 was another record quarter. Revenue of $39.3 billion was up 12% sequentially and up 78% year-on-year, and above our outlook of $37.5 billion. For fiscal 2025, revenue was $130.5 billion, up 114% from the prior year. Let's start with data center. Data center revenue for fiscal 2025 was $115.2 billion, more than doubling from the prior year. In the fourth quarter, data center revenue of $35.6 billion was a record, up 16% sequentially and 93% year-on-year as the Blackwell ramp commenced and Hopper 200 continued sequential growth. In Q4, Blackwell sales exceeded our expectations. We delivered $11 billion of Blackwell revenue to meet strong demand. This is the fastest product ramp in our company's history, unprecedented in its speed and scale. Blackwell production is in full gear across multiple configurations, and we are increasing supply quickly to meet expanding customer adoption.

Becker: All our statements are made as of today February 26, 2025 based on information currently available to us except as required by law, we assume no obligation to update any such statements.

Colette Kress: Above our outlook of $37 5 billion.

Colette Kress: For fiscal 2025 revenue was 135 billion up 114% prior year.

Becker: During this call we will discuss non-GAAP financial measures.

Becker: A reconciliation of these non-GAAP financial measures.

Colette Kress: Let's start with data center.

Colette Kress: Data center revenue for fiscal 2025, with $115 2 billion more than doubling from the prior year.

Becker: Posted on our website with that let me turn the call over to Colette.

Colette Kress: Companies across industries are tapping into NVIDIA's full-stack inference platform to boost performance and slash cost. now tripled inference throughput and cut costs by 66% using NVIDIA TensorRT for its screenshot feature. Complexity sees 435 million monthly queries and reduced its inference costs 3x with NVIDIA Triton Inference Server and Tensor RTLLM. Microsoft Bing achieved a 5x speed up and major TCO savings for visible search across billions of images with NVIDIA, TensorRT, and Acceleration Library. Blackwell has great demand for NVIDIA. Many of the early GB200 deployments are earmarked for inference, a first for a new architecture. Blackwell addresses the entire AI market from pre-training, post-training, to inference across cloud, to on-premise, to enterprise.

Colette: Thanks Stuart.

Colette Kress: In the fourth quarter data center revenue of $35 6 billion was a record up 16% sequentially and 93% year on year.

Colette: <unk> was another record quarter revenue of $39 3 billion that was up 12% sequentially and up 78% year on year and above our outlook of $37 5 billion.

Colette Kress: <unk> come back.

Colette Kress: In Harper 200 continued sequential growth.

Colette: For fiscal 2025 revenue was $135.

Colette Kress: In Q4, Blackwell sales exceeded our expectations, we delivered 11 billion Blackwell revenue to meet strong demand.

Colette: 114% from the prior year.

Colette: Let's start with data center.

Colette Kress: This is the fastest product ramp in our company's history unprecedented speed and scale.

Colette: Data center revenue for business.

Colette Kress: Microsoft Bing achieved a 5X speedup and major TCO savings for visual search across billions of images with NVIDIA TensorRT and acceleration libraries. Blackwell has great demand for inference. Many of the early GB200 deployments are earmarked for inference, a first for a new architecture. Blackwell addresses the entire AI market from pre-training, post-training to inference across clouds to on-premise to enterprise. CUDA's programmable architecture accelerates every AI model and over 4,400 applications, ensuring large infrastructure investments against obsolescence in rapidly evolving markets. Our performance and pace of innovation is unmatched. We're driven to a 200X reduction in inference costs in just the last two years. We deliver the lowest TCO and the highest ROI. Full stack optimizations for NVIDIA and our large ecosystem, including 5.9 million developers, continuously improve our customers' economics. In Q4, large CSPs represented about half of our data center revenue.

Colette: More than doubling from the prior year.

Colette Kress: Black oil production is in full gear across multiple configurations, and we are increasing supply quickly expanding customer adoption.

Colette: In the fourth quarter data center revenue of $35 6 billion was a record up 16% sequentially and 93% year on year.

Colette Kress: Our Q4 data center compute revenue jumped 18% sequentially and over two X year on year.

Colette: That's the black bar graph come back.

Colette: And Hopper 200 continued sequential growth.

Colette Kress: Tumors are racing to scale infrastructure to train the next generation of cutting edge model and unlock the next level of AI capabilities.

Colette: In Q4, Blackwell sales exceeded our expectations, we delivered 11 billion Blackwell revenue to meet strong demand. This.

Colette Kress: With Blackwell it will be common for these clusters to start with 100000 Gpus.

Colette Kress: CUDA's programmable architecture accelerates every AI model and over 4,400 applications, ensuring large infrastructure investments against obsolescence in rapidly evolving markets.

Colette: This is the fastest product ramp in our company's history unprecedented.

Colette Kress: More.

Colette Kress: Shipments have already started for multiple infrastructures of this size.

Colette: It is in full gear across multiple configurations, and we are increasing supply quickly expanding customer adoption.

Colette Kress: Post training and model customization are fueling demand for Nvidia infrastructure and software as developers and enterprises leverage techniques, such as fine tuning reinforcement learning and distillation to tailor models for domain specific use cases.

Colette Kress: Our performance and pace of innovation is unmatched. were driven to a 200x reduction in inference cost in just the last two years. We deliver the lowest GCO and the highest ROI. and full-stack optimizations for NVIDIA and our large ecosystem, including 5.9 million developers, continuously improve our customers' economics.

Colette: Our Q4 data center compute revenue jumped 18% sequentially and over two X year on year.

Colette: Customers are racing to scale infrastructure to train the next generation of cutting edge model and unlock the next level of AI capabilities.

Colette Kress: Again faced alone hosts over 90000 derivatives traded from Obama Foundation model.

Colette: With Blackwell it will be common for these clusters to start with 100000 Gpus or more chip.

Colette Kress: The scale of post training and model customization is massive and collectively demand orders of magnitude more compute than pre training.

Colette: Shipments have already started for multiple infrastructures of this size.

Colette Kress: Thank you for... Large CSPs represented about half of our data center revenue. and these sales increased nearly 2x year on year. Large CSPs were some of the first to stand up Blackwell with Azure, GCP, AWS and OCI bringing GB200 systems to cloud regions around the world to meet surging customer demand for AI. Regional cloud hosting NVIDIA GPUs increased as a percentage of data center revenue, reflecting continued AI factory build-outs globally and rapidly rising demand for AI reasoning models and agencies.

Colette: Post training and Marvel customization are fueling demand for it and video infrastructure and software as developers and enterprises leverage techniques, such as fine tuning reinforcement learning and distillation to tailor models for domain specific use cases.

Colette Kress: R and friends demand is accelerating.

Colette Kress: These sales increased nearly 2X year-on-year. Large CSPs were some of the first to stand up Blackwell, with Azure, GCP, AWS, and OCI bringing GB200 systems to cloud regions around the world to meet surging customer demand for AI. Regional clouds hosting NVIDIA GPUs increased as a percentage of data center revenue, reflecting continued AI factory buildouts globally and rapidly rising demand for AI reasoning models and agents. CoreWeave launched a 100,000 GB200 cluster-based instance with NVLink Switch and Quantum II InfiniBand. Consumer Internet revenue grew 3X year-on-year, driven by an expanding set of generative AI and deep learning use cases. These include recommender systems, vision language understanding, synthetic data generation search, and agentic AI. For example, xAI is adopting the GB200 to train and inference its next generation of Grok AI models.

Colette Kress: Given by test time, scaling and new reasoning models like opening is owned Green deep CCAR, one and <unk> three.

Colette Kress: Long thinking reasoning AI can require 100 X more compute her test compared to one shot influences.

Colette Kress: Long-thinking reasoning AI can require 100X more compute per task compared to one-shot inferences. Blackwell was architected for reasoning AI inference. Blackwell supercharges reasoning AI models with up to 25X higher token throughput and 20X lower cost versus Hopper 100. It is revolutionary. Transformer Engine is built for LLM and mixture of experts inference, and its NVLink domain delivers 14X the throughput of PCIe Gen 5, ensuring the response time, throughput, and cost efficiency needed to tackle the growing complexity of inference at scale. Companies across industries are tapping into NVIDIA's full-stack inference platform to boost performance and slash costs. NAP tripled inference throughput and cut costs by 66% using NVIDIA TensorRT for its screenshot feature. Perplexity sees 435 million monthly queries and reduced its inference costs 3X with NVIDIA Triton Inference Server and TensorRT LLM.

Colette: I mean face although hosts over 90000 derivatives traders from Obama Foundation model.

Colette Kress: <unk> was architected for reasoning AI in France.

Colette: The scale of post training and Marvell customization is massive.

Colette Kress: Blackwell supercharged his reasoning AI models with up to 25 X.

Collectively demand orders of magnitude more compute than pre training.

Colette Kress: Higher token throughput and 20 ex lower cost versus hunker 100.

Colette: Our N friends demand is accelerating.

Colette Kress: CoreWeave launched a 100,000 GV200 cluster-based instance with NVLink Switch and Quantum2 InfiniBand. Consumer Internet Revenue Group 3x year-on-year, driven by an expanding set of generative AI and deep learning use cases. These include Recommender Systems, Vision Language Understanding, Synthetic Data Generation Search, and Agentic AI. For example, XAI is adopting the GB200 to train and inference its next generation of grog AI models. Meta's cutting-edge Andromeda advertising engine runs on NVIDIA's Grace Hopper Superchip, serving vast quantities of ads across Instagram, Facebook applications. Andromeda harnesses Grace Hopper's fast interconnect and large memory to boost inference throughput by 3X, enhance ad personalization, and deliver meaningful jumps in monetization and ROI.

Colette Kress: It is revolutionary transformer engine is built for L O them and make sure of experts in France.

Colette: Even by test time scaling our new reasoning models like opening it is old.

Colette: Screen deep seek our one and <unk> three.

Colette Kress: And it hadn't been link domain delivered 14 X the throughput of Pcie Gen. Five ensuring the response time throughput and cost efficiency needed to tackle the growing complexity of infants at scale.

Colette: Long thinking reasoning AI can require 100 X more compute her tests compared to one shot thanks Francis.

Speaker Change: Blackwell was architected for reasoning AI.

Colette: Yeah.

Colette Kress: Companies across industries are tapping into invidious, all stock and <unk> platform to boost performance and slash costs.

Colette: Blackwell supercharged his reasoning AI models with up to 25 X.

Colette: Your token throughput and once he acts lower cost versus hunker one contract.

Colette Kress: Now tripled inference throughput and cut costs by 66% using Nvidia tensor RT for its screenshots feature.

Colette Kress: Meta's cutting-edge Andromeda advertising engine runs on NVIDIA's Grace Hopper superchip, serving vast quantities of ads across Instagram and Facebook applications. The Andromeda harnesses Grace Hopper's fast interconnect and large memory to boost inference throughput by 3X, enhance ad personalization, and deliver meaningful jumps in monetization and ROI. Enterprise revenue increased nearly 2X year on the accelerating demand for model fine-tuning, RAG, and agentic AI workflows, and GPU-accelerated data processing. We introduced NVIDIA Llama Neumatron model family NIMs to help developers create and deploy AI agents across a range of applications, including customer support, fraud detection, and product supply chain and inventory management. Leading AI agent platform providers, including SAP and ServiceNow, are among the first to use new models.

Colette: It is revolutionary transformer engine is built for L O them and mixture of experts in France.

Colette Kress: Perplexity fees 435 million monthly queries and reduced its in fringe costs three apps within video Triton and print server and tensor RT go alone.

Colette: And it hadn't been linked demand delivers 14 X the throughput of D. C. I E. Gen fives, ensuring the response time throughput and cost efficiency needed to tackle the growing complexity of infants at scale.

Colette Kress: Microsoft being achieved a five X speed up a major tcl savings for visual search across billions of images with Nvidia tensor RT and acceleration libraries.

Colette Kress: Microsoft Bing achieved a 5X speedup and major TCO savings for visual search across billions of images with NVIDIA TensorRT and acceleration libraries. Blackwell has great demand for inference. Many of the early GB200 deployments are earmarked for inference, a first for a new architecture. Blackwell addresses the entire AI market from pre-training, post-training to inference across clouds to on-premise to enterprise. CUDA's programmable architecture accelerates every AI model and over 4,400 applications, ensuring large infrastructure investments against obsolescence in rapidly evolving markets. Our performance and pace of innovation is unmatched. We're driven to a 200X reduction in inference costs in just the last two years. We deliver the lowest TCO and the highest ROI. Full-stack optimizations for NVIDIA and our large ecosystem, including 5.9 million developers, continuously improve our customers' economics.

Colette: Companies across industries are tapping into in videos, all stock inference platform to boost performance and slash costs.

Colette Kress: Enterprise revenue increased nearly 2x year-on-accelerating demand for model fine-tuning, RAG, and agentic AI workflows, and GPU-accelerated data processing.

Colette Kress: So that's all have great demand for entrants many of the early GBP 200 deployments are earmarked for infants.

Colette: Now tripled inference throughput and cut costs by 66% using Nvidia tensor RT screenshots feature.

Colette Kress: First for a new architecture.

Colette Kress: We introduced NVIDIA LLAMA Numitron Model Family NIMS to help developers create and deploy AI agents across a range of applications. including customer support, fraud detection, and product supply chain and inventory management. Leading AI agent platform providers including SAP and ServiceNow are among the first to use new models. Healthcare leaders IQVIA, Illumina, and Mayo Clinic are well as AHRQ Institute are using NVIDIA AI to speed drug discovery, enhance genomic research, and pioneer advanced healthcare services with generative and agentic AI.

Speaker Change: Perplexing, six 435 million monthly queries and reduced its in fringe costs three apps within video Triton and print server and tensor RT all along.

Blackwell addresses the entire AI market from pre training post training to entrance across cloud to on premise to enterprise.

Colette Kress: Who knows programmable architecture accelerates every AI model and over 4400 applications insuring large infrastructure investments against obsolescence and rapidly evolving market.

Microsoft: Microsoft being achieved a five X speed up a major T C O savings for visual search across billions of images with Nvidia tensor RT and acceleration libraries.

Colette Kress: Healthcare leaders IQVIA, Illuminon, and Mayo Clinic, as well as the AHRQ Institute, are using NVIDIA AI to speed drug discovery, enhance genomic research, and pioneer advanced healthcare services with generative and agentic AI. As AI expands beyond the digital world, NVIDIA infrastructure and software platforms are increasingly being adopted to power robotics and physical AI development. One of the early and largest robotics applications and autonomous vehicles where virtually every AV company is developing on NVIDIA in the data center, the car, or both. NVIDIA's automotive vertical revenue is expected to grow to approximately $5 billion this fiscal year. At CES, Hyundai Motor Group announced it is adopting NVIDIA technologies to accelerate AV and robotics development and smart factory initiatives. Vision transformers, self-supervised learning, multimodal sensor fusion, and high-fidelity simulation are driving breakthroughs in AV development and will require 10X more compute.

Microsoft: So that's all have great demand for entrants many of the early G. P 200 deployments are earmarked for inference.

Colette Kress: Our performance and pace of innovation is unmatched.

Colette Kress: We are driven to a 200 X reduction in insurance costs in just the last two years, we've delivered the lowest to Seattle and the highest or what.

Microsoft: First for a new architecture.

Microsoft: Blackwell addresses the entire AI market I'm pretty training post training to entrance across cloud to on premise to enterprise.

Colette Kress: And full stock optimization spin Nvidia and our large ecosystem, including 5.9 million developers continuously improve our customers' economics.

Colette Kress: As AI expands beyond the digital world, NVIDIA infrastructure and software platforms are increasingly being adopted to power robotics and physical AI development. One of the early and largest robotics applications and autonomous vehicles where virtually every AV company is developing on NVIDIA in the data center, the car, or both. NVIDIA's automotive vertical revenue is expected to grow to approximately $5 billion this fiscal year. At CES, Hyundai Motor Group announced it is adopting NVIDIA technologies to accelerate AV and robotics development and smart factory initiatives Vision Transformers, Self-Supervised Learning, Multimodal Sensor Fusion, and High-Fidelity Simulation are driving breakthroughs in AV development and will require 10x more compute.

Microsoft: Who knows programmable architecture accelerates every AI model and over 4400 applications insuring large infrastructure investments against obsolescence and rapidly evolving market.

Colette Kress: In Q4.

Colette Kress: Large CSP represented about half of our data center revenue.

Colette Kress: In Q4, large CSPs represented about half of our data center revenue, and these sales increased nearly 2X year-on-year. Large CSPs were some of the first to stand up Blackwell, with Azure, GCP, AWS, and OCI bringing GB200 systems to cloud regions around the world to meet surging customer demand for AI. Regional clouds hosting NVIDIA GPUs increased as a percentage of data center revenue, reflecting continued AI factory buildouts globally and rapidly rising demand for AI reasoning models and agents. CoreWeave launched a 100,000 GB200 cluster-based instance with NVLink Switch and Quantum II InfiniBand. Consumer Internet revenue grew 3X year-on-year, driven by an expanding set of generative AI and deep learning use cases. These include recommender systems, vision language understanding, synthetic data generation search, and agentic AI. For example, xAI is adopting the GB200 to train and inference its next generation of Grok AI models.

Colette Kress: And these sales increased nearly two <unk> year on year large csp's, where some of the first to standup Blackwell with Azure TCP, AWS and OCI, bringing GBP 200 systems to cloud regions around the world to meet surging customer demand for AI.

Speaker Change: Our performance and pace of innovation is unmatched.

Speaker Change: We are driven to a 200 X reduction in insurance costs and just the last two years, we've delivered the lowest to Seattle and the highest or what.

Speaker Change: And full stock optimization spin Nvidia and our large ecosystem, including 5.9 million developers continuously improve our customers' economics.

Colette Kress: Retail and cloud hosting and video Gpus increased as a percentage of data center revenue, reflecting continued AI factory build out globally and rapidly rising demand for AI reasoning models and agents.

Speaker Change: In Q4.

Speaker Change: Large CSP represented about half of our data center revenue.

We've launched a 100000 G V 200 cluster based instance, with NV link switch and quantum two infiniband.

Speaker Change: A nice sales increase nearly two X year on year large csp's, where some of the first to standup Blackwell with Azure TCP, AWS and OCI, bringing GBP 200 systems to cloud regions around the world to meet surging customer demand for AI.

Colette Kress: At CEX, we announced the NVIDIA Cosmo World Foundation model platform. Just as language foundation models have revolutionized language AI, Cosmos is a physical AI to revolutionize robotics. Leading robotics and automotive companies, including ride-sharing giant Uber, are among the first to adopt the platform. From a geographic perspective, sequential growth in our data center revenue was strongest in the U.S., driven by the initial ramp of black boxes.

Colette Kress: At CES, we announced the NVIDIA Cosmos World Foundation Model Platform. Just as language foundation models have revolutionized language AI, Cosmos is a physical AI to revolutionize robotics. Leading robotics and automotive companies, including ride-sharing giant Uber, are among the first to adopt the platform. From a geographic perspective, sequential growth in our data center revenue was strongest in the U.S., driven by the initial ramp of Blackwell. Countries across the globe are building their AI ecosystems, and demand for compute infrastructure is surging. France's €200 billion AI investment and the EU's €200 billion Invest AI initiative offer a glimpse into the buildout to set redefined global AI infrastructure in the coming years. Now, as a percentage of total data center revenue, data center sales in China remained well below levels seen at the onset of export controls.

Colette Kress: Consumer Internet revenue grew three X year on year, driven by an expanding set of generative AI and deep learning use cases. These include recommend or systems.

Speaker Change: Retail cloud hosting and video Gpus increased as a percentage of data center revenue, reflecting continued AI factory build out globally and rapidly rising demand for AI recently models and agents.

Colette Kress: <unk> language understanding synthetic data generation search and Agentic AI.

Colette Kress: For example, X AI is adopting the GBP 202 train and influence its next generation across AI models.

Speaker Change: We've launched a 100000 G V 200 cluster based instance, with NV link switch and quantum two infiniband.

Colette Kress: Meadows cutting age Andromeda advertising engine runs on and videos Grace Hopper Super chip, serving vast quantities of ads across Instagram Facebook applications with Ramadan harnesses Grace Hopper's fast interconnect and large memory to boost inference throughput.

Colette Kress: Meta's cutting-edge Andromeda advertising engine runs on NVIDIA's Grace Hopper superchip, serving vast quantities of ads across Instagram and Facebook applications. The Andromeda harnesses Grace Hopper's fast interconnect and large memory to boost inference throughput by 3X, enhance ad personalization, and deliver meaningful jumps in monetization and ROI. Enterprise revenue increased nearly 2X year on the accelerating demand for model fine-tuning, RAG, and agentic AI workflows, and GPU-accelerated data processing. We introduced NVIDIA Llama NeMo model family NIMs to help developers create and deploy AI agents across a range of applications, including customer support, fraud detection, and product supply chain and inventory management. Leading AI agent platform providers, including SAP and ServiceNow, are among the first to use new models.

Colette Kress: Countries across the globe are building their AI ecosystems and demand for compute infrastructure is surging. France's 200 billion euro AI investment, and the EU's 200 billion euro Invest AI initiative offer a glimpse into the build-out to set redefined global AI infrastructure in the coming year.

Speaker Change: Consumer Internet revenue grew three X year on year, driven by an expanding set of generative AI and deep learning use cases. These include recommend or systems.

Colette Kress: Three acts enhance AD personalization and deliver meaningful jumps in monetization and ROI.

Speaker Change: Language understanding synthetic data generation search and <unk> AI.

Speaker Change: For example, X AI is adopting the GBP 202 train and influence its next generation AI models.

Colette Kress: Now, as a percentage of total data center revenue, data center sales in China remained well below levels seen on the onset of export controls. Absent any change in regulations, we believe that China shipments will remain roughly at the current percentage. The market in China for data center solutions remains very competitive. We will continue to comply with export controls while serving our customers.

Colette Kress: Enterprise revenue increased nearly two X year on accelerating demand model fine tuning rack and <unk>, AI workflows and GPU accelerated data processing.

Speaker Change: Meadows cutting age Andromeda.

Colette Kress: Absent any change in regulations, we believe that China's shipments will remain roughly at the current percentage. The market in China for data center solutions remains very competitive. We will continue to comply with export controls while serving our customers. Networking revenue declined 3% sequentially. Our networking attached to GPU compute systems is robust at over 75%. We are transitioning from small NVLink 8 with InfiniBand to large NVLink 72 with Spectrum-X. Spectrum-X and NVLink Switch revenue increased and represent a major new growth sector. We expect networking to return to growth in Q1. AI requires a new class of networking. NVIDIA offers NVLink Switch systems for scale-up compute. For scale-out, we offer Quantum InfiniBand for HPC supercomputers and Spectrum-X for Ethernet environments. Spectrum-X enhances the Ethernet for AI computing and has been a huge success.

Speaker Change: Advertising engine runs on and videos Grace Hopper Super chip, serving mass quantities of ads across Instagram Facebook applications with Ramadan.

Colette Kress: We introduced Nvidia Lama <unk> model family Nims to help developers create and deploy AI agents across a range of applications, including customer support fraud detection and product supply chain and inventory management.

Speaker Change: Harnesses grasshoppers fast interconnect and large memory to boost and friends throughput by three acts enhance AD personalization and deliver meaningful jumps to monetization and are a lot.

Colette Kress: Networking revenue declined 3% sequentially. Our networking attached to GPU compute systems is robust at over 75%. We are transitioning from small NVLink 8 with InfiniBand to large NVLink 72 with Spectrum X. Spectrum X and NVLink Switch revenue increased and represents a major new growth.

Colette Kress: Leading AI agent platform providers, including SAP TM service now are among the first to use new models.

Speaker Change: Enterprise revenue increased nearly two X year on its accelerating into bound model fine tuning rack and Egencia AI workflows and GPU accelerated data processing.

Colette Kress: Health care leaders IQ via alumina and Mayo clinic are well as arc Institute are using Nvidia AI to speed drug discovery enhanced genomic research and pioneer advanced health care services with generative and age into chaos.

Colette Kress: Healthcare leaders IQVIA, Illumina, and Mayo Clinic, as well as the AHRQ Institute, are using NVIDIA AI to speed drug discovery, enhance genomic research, and pioneer advanced healthcare services with generative and agentic AI. As AI expands beyond the digital world, NVIDIA infrastructure and software platforms are increasingly being adopted to power robotics and physical AI development. One of the early and largest robotics applications is autonomous vehicles, where virtually every AV company is developing on NVIDIA in the data center, the car, or both. NVIDIA's automotive vertical revenue is expected to grow to approximately $5 billion this fiscal year. At CES, Hyundai Motor Group announced it is adopting NVIDIA technologies to accelerate AV and robotics development and smart factory initiatives. Vision transformers, self-supervised learning, multimodal sensor fusion, and high-fidelity simulation are driving breakthroughs in AV development and will require 10X more compute.

Speaker Change: We introduced Nvidia Lama numerous Ron model family Nims to help developers create and deploy AI agents across a range of applications, including customer support fraud detection and <unk>.

Colette Kress: We expect networking to return to growth in Q1. AI requires a new class of networking. NVIDIA offers NVLink switch systems for scale-up compute. For scale-out, we offer Quantum InfiniBand for HPC supercomputers and Spectrum X for ethernet environments. Spectrum X enhances the ethernet for AI computing and has been a huge success. Microsoft Azure, OCI, CoreWeave, and others are building large AI factories with Spectrum X. The first Stargate data centers will use Spectrum X. Yesterday, Cisco announced integrating Spectrum X into their networking portfolio to help enterprises build AI in With its large enterprise footprint and global reach, Cisco will bring NVIDIA Ethernet to every industry.

Colette Kress: It's AI expands beyond the digital world Nvidia infrastructure and software platforms are increasingly being adopted to power robotics and physical AI development.

Speaker Change: That supply chain and inventory management.

Speaker Change: Leading AI agent platform providers, including our SAP TM service now are among the first to use new models.

Colette Kress: One of the early and largest robotics applications and autonomous vehicles, where virtually every a the company is developing an nvidia in the datacenter in the car or bowls.

Speaker Change: Health care leaders IQ via alumina and Mayo clinic are well as Ark Institute are using Nvidia AI to speed drug discovery enhanced genomic research and pioneer advanced health care services with generative and agents of chaos.

Colette Kress: Microsoft Azure, Oracle Cloud Infrastructure, CoreWeave, and others are building large AI factories with Spectrum-X. The first Stargate data centers will use Spectrum-X. Yesterday, Cisco announced integrating Spectrum-X into their networking portfolio to help enterprises build AI infrastructure. With its large enterprise footprint and global reach, Cisco will bring NVIDIA Ethernet to every industry. Now, moving to gaming and AI PCs. Gaming revenue of $2.5 billion decreased 22% sequentially and 11% year-on-year. Full-year revenue of $11.4 billion increased 9% year-on-year. Demand remained strong throughout the holiday. However, Q4 shipments were impacted by supply constraints. We expect strong sequential growth in Q1 as supply increases. The new GeForce RTX 50 Series desktop and laptop GPUs are here. Built for gamers, creators, and developers, they fuse AI and graphics, redefining visual computing. Powered by the Blackwell architecture, fifth-generation Tensor cores and fourth-generation RT cores, and featuring up to 3,400 AI TOPS.

And videos automotive vertical revenue is expected to grow to approximately 5 billion this fiscal year.

Colette Kress: I see us at CES, Hyundai Motor Group announced it is adopting and video technologies to accelerate Avi and robotics development and smart factory initiatives.

Speaker Change: It's AI expands beyond the digital world and video infrastructure and software platforms are increasingly being adopted to power robotics and physical AI development.

Colette Kress: Vision Transformers self supervised learning multimodal sensor fusion and high fidelity simulation are driving breakthroughs in AB development and will require 10 X more compute.

Speaker Change: One of the early and largest robotics applications and autonomous vehicles, where virtually every a the company is developing an nvidia in the datacenter in the car or bowls.

Colette Kress: Now moving to gaming and ARPC. Gaming revenue of $2.5 billion decreased 22% sequentially and 11% year-on-year. Full-year revenue of $11.4 billion increased 9% year-on-year, and demand remained strong throughout the holiday.

Colette Kress: At CES, we announced the Nvidia Cosmo.

Speaker Change: And videos automotive vertical revenue is expected to grow to approximately 5 billion this fiscal year.

Colette Kress: World Foundation model platform.

Colette Kress: As language Foundation models have revolutionized language AI Cosmos is a physical AI to revolutionize robotics mini robotics, and automotive companies, including Ridesharing giant Uber are among the first to adopt the platform.

Speaker Change: I see us at CES Hyundai Motor group.

Speaker Change: Now it is adopting and video technologies to accelerate Avi and robotics development and smart factory initiatives.

Colette Kress: However, Q4 shipments were impacted by supply cuts. We expect strong sequential growth in Q1 as supply increases.

Speaker Change: Vision Transformers self supervised learning multimodal sensor fusion and high fidelity simulation are driving breakthroughs in development and will require 10 X more compute.

Colette Kress: From a geographic perspective sequential growth in our data center revenue was strongest in the U S driven by the initial ramp up like Wow.

Colette Kress: The new GeForce RTX 50 Series Desktop and Laptop GPUs are here. Built for gamers, creators, and developers, they fuse AI and graphics, redefining visual computing. powered by the Blackwell architecture, fifth-generation Tensor cores and fourth-generation RT cores and featuring up to 3,400 AI tops. These GPUs deliver a 2X performance leap and new AI-driven rendering, including neural shaders, digital human technologies, geometry, and lighting. The new ZLSS4 boosts frame rates up to 8x with AI-driven frame generation, turning one rendered frame into three. It also features the industry's first real-time application of transformer models, packing 2x more parameters and 4x to compute for unprecedented visual fidelity.

Colette Kress: Countries across the globe are building their AI ecosystem of demand for compute infrastructure is surging.

Speaker Change: Let's see yes, we announced the Nvidia Cosmo World Foundation model platform.

Colette Kress: France is 200 billion Euro AI investment and the EUR 200 billion Euro invest AI initiatives offer a glimpse into the build out.

Speaker Change: Just as language Foundation models have revolutionized language AI Cosmos is a physical AI to revolutionize robotics mini robotics, and automotive companies, including Ridesharing giant Uber are among the first to adopt the platform.

Colette Kress: These GPUs deliver a 2X performance leap and new AI-driven rendering, including neural shaders, digital human technologies, geometry, and lighting. The new DLSS 4 boosts frame rates up to 8X with AI-driven frame generation, turning one rendered frame into three. It also features the industry's first real-time application of transformer models, packing 2X more parameters and 4X the compute for unprecedented visual fidelity. We also announced a wave of GeForce Blackwell laptop GPUs with new NVIDIA Max-Q technology that extends battery life by up to an incredible 40%. These laptops will be available starting in March from the world's top manufacturers. Moving to our professional visualization business, revenue of $511 million was up 5% sequentially and 10% year-on-year. Full-year revenue of $1.9 billion increased 21% year-on-year. Key industry verticals driving demand include automotive and healthcare. NVIDIA technologies and generative AI are reshaping design, engineering, and simulation workloads.

Colette Kress: <unk> redefined global AI infrastructure in the coming years.

Colette Kress: Now as a percentage of total data center revenue data center sales in China remained well below levels seen on the onset of export controls absent any change in regulations, we believe that China's shipments will remain roughly at the current percentage.

Speaker Change: From a geographic perspective sequential growth in our data center revenue was strongest in the U S driven by the initial ramp up like Wow.

Speaker Change: Countries across the globe are building their AI ecosystem of demand for compute infrastructure is surging.

Colette Kress: Absent any change in regulations, we believe that China's shipments will remain roughly at the current %. The market in China for data center solutions remains very competitive. We will continue to comply with export controls while serving our customers. Networking revenue declined 3% sequentially. Our networking attached to GPU compute systems is robust at over 75%. We are transitioning from small NVLink 8 with InfiniBand to large NVLink 72 with Spectrum-X. Spectrum-X and NVLink Switch revenue increased and represents a major new growth sector. We expect networking to return to growth in Q1. AI requires a new class of networking. NVIDIA offers NVLink Switch systems for scale-up compute. For scale-out, we offer Quantum InfiniBand for HPC supercomputers and Spectrum-X for Ethernet environments. Spectrum-X enhances the Ethernet for AI computing and has been a huge success.

Colette Kress: The market in China for datacenter solutions remains very competitive we will continue to comply with export controls while serving our customers.

Speaker Change: France is 200 billion Euro AI investment and easy to use 200 billion euro invest AI initiatives offer a glimpse into the build out.

Colette Kress: We also announced a wave of GeForce Blackwell laptop GPUs with new NVIDIA Max-Q technology that extends battery life by up to an incredible 40%.

Networking revenue declined 3% sequentially.

Speaker Change: It redefined global AI infrastructure in the coming years.

Our networking attached to GPU compute system is robust at over 75%. We are transitioning from small and be link eight with infiniband to large and we link 72 with spectrum acts.

Speaker Change: Now as a percentage of total data center revenue.

Colette Kress: These laptops will be available starting in March from the world's top manufacturers.

Speaker Change: Sales in China remained well below levels seen on the onset of export controls.

Colette Kress: Moving to our professional visualization. Revenue of $511 million was up 5% sequentially and 10% year-on-year. Full year revenue of $1.9 billion, increased 21% year-on-year. key industry verticals driving demand include automotive and health NVIDIA technologies and generative AI are reshaping design, engineering, and simulation workloads. Increasingly, these technologies are being leveraged in leading software platforms, from ANSYS, Cadence, and Siemens, fueling demand for NVIDIA RTX workloads.

Speaker Change: Any change in regulations, we believe that China's shipments will remain roughly at the current percentage of the market in China for datacenter solutions remains very competitive we will continue to comply with export controls while serving our customers.

Colette Kress: I drove X and building switch revenue increased and represents a major new growth sector.

Colette Kress: We expect NEK with networking, who returned to growth in Q1.

Colette Kress: AI requires a new class of networking and video offers and be like switch systems for scale up compute for scale out we offer quantum infiniband for H B C supercomputers.

Speaker Change: Networking revenue declined 3% sequentially, our networking attached to GPU compute system is robust at over 75%.

Colette Kress: Increasingly, these technologies are being leveraged in leading software platforms from Ansys, Cadence, and Siemens, fueling demand for NVIDIA RTX workstations. Now, moving to automotive, revenue was a record $570 million, up 27% sequentially and up 103% year-on-year. Full-year revenue of $1.7 billion increased 55% year-on-year. Strong growth was driven by the continued ramp in autonomous vehicles, including cars and robotaxis. At CES, we announced Toyota, the world's largest automaker, will build its next-generation vehicles on NVIDIA O-RAN, running the safety-certified NVIDIA Drive OS. We announced Aurora and Continental will deploy driverless trucks at scale powered by NVIDIA Drive 4. Finally, our end-to-end autonomous vehicle platform, NVIDIA Drive Hyperion, has passed industry safety assessments by Husu and Hufreiland, two of the industry's foremost authorities for automotive-grade safety and cybersecurity. NVIDIA is the first AV platform to receive a comprehensive set of third-party assessments.

Colette Kress: Spectrum X for Ethernet environments.

Speaker Change: We are transitioning from small and be link eight with infiniband to large and we link 72 the spectrum acts.

Colette Kress: Dramatics enhances the Ethernet for AI computing and has been a huge success, Microsoft Azure OCI or we and others are building barge AI factories with spectrum ex the.

Colette Kress: Now moving to Autumn Modo. Revenue was a record $570 million, up 27% sequentially, and up 103% year-on-year. Full-year revenue of $1.7 billion increased 55% year-on-year. Strong growth was driven by the continued ramp in autonomous vehicles, including cars and robotics. At CES, we announced Toyota, the world's largest automaker, will build its next-generation vehicles on NVIDIA OREN, running the safety-certified NVIDIA DRIVE OS. We announced Aurora and Continental will deploy driverless trucks at scale powered by NVIDIA DRIVE 4. Finally, our end-to-end autonomous vehicle platform, NVIDIA DRIVE Hyperion, has passed industry safety assessments by Hsu Hsu and Hsu Ryland, two of the industry's foremost authorities for automotive-grade safety and cybersecurity.

Speaker Change: Spectrum ex an enviable switch revenue increased and represents a major new growth sector.

Colette Kress: The first Star Gate data centers will use spectrum Act yes.

Speaker Change: We expect network networking to return to growth in Q1.

Colette Kress: Yesterday, Cisco announced integrating spectrum X into their networking portfolio to help enterprises build AI infrastructure.

Speaker Change: Hey requires a new class of networking and video offers that would be like switch system.

Colette Kress: With its large enterprise footprint and global reach Cisco will bring Nvidia Ethernet to every industry.

Speaker Change: So bad for H, B C supercomputers and spectrum X for Ethernet environments.

Speaker Change: Now moving to gaming and AI P. CS.

Speaker Change: Dramatics enhances the Ethernet for AI computing and it has been a huge success, Microsoft Azure OCI or we and others are building barge factories with spectrum ex.

Speaker Change: <unk> revenue of $2 5 billion decreased 22% sequentially and 11% year on year.

Speaker Change: Full year revenue of $11 4 billion increased 9% year on year and.

Speaker Change: The first Star Gate data centers will use spectrum Act.

Speaker Change: And demand remains strong throughout the holiday. However, Q4 shipments were impacted by supply constraints, we expect strong sequential growth in Q1 as supply increases.

Speaker Change: Yesterday, Cisco announced integrated spectrum X into their networking portfolio to help enterprises build AI infrastructure.

Speaker Change: With its large enterprise footprint and global reach Cisco will bring Nvidia Ethernet to every industry.

Speaker Change: The New G Force architect 50 series desktop and laptop Gpus are here.

Colette Kress: NVIDIA is the first AV platform to receive a comprehensive set of third-party assistants.

Speaker Change: For gamers creators and developers they use AI and graphics redefining visual computing.

Speaker Change: Now moving to the gaming and AI P series.

Speaker Change: Gaming revenue of $2 5 billion decreased 22% sequentially and 11% year on year.

Colette Kress: Okay, moving to the rest of the PML. GAAP gross margins was 73% and non-GAAP gross margins was 73.5% down sequentially as expected with our first deliveries of the Blackwell RPC. As discussed last quarter, Blackwell is a customizable AI infrastructure with several different types of NVIDIA-built chips, multiple networking options, and for air and liquid-cooled data center. We exceeded our expectations in Q4 in ramping Blackwell, increasing system availability, providing several configurations to our customers. As Blackwell ramps, we expect gross margins to be in the low 50s. in the low 70s. We, initially, we are focused on expediting the manufacturing of Blackwell systems to meet strong customer demand as they race to build out Blackwell infrastructure.

Colette Kress: OK, moving to the rest of the P&L. GAAP gross margins were 73% and non-GAAP gross margins were 73.5%, down sequentially as expected, with our first deliveries of the Blackwell architecture. As discussed last quarter, Blackwell is a customizable AI infrastructure with several different types of NVIDIA-built chips, multiple networking options, and for air and liquid-cooled data center. We exceeded our expectations in Q4 in ramping Blackwell, increasing system availability, and providing several configurations to our customers. As Blackwell ramps, we expect gross margins to be in the low 70s. Initially, we are focused on expediting the manufacturing of Blackwell systems to meet strong customer demand as they race to build out Blackwell infrastructure. When fully ramped, we have many opportunities to improve the cost and gross margin will improve and return to the mid-70s late this fiscal year.

Speaker Change: Powered by the Blackwell architecture.

Speaker Change: Generation tensor cores and fourth generation, our T cars and featuring up to 3400 AI was hot.

Colette Kress: Powered by the Blackwell architecture, fifth-generation Tensor cores and fourth-generation RT cores, and featuring up to 3,400 AI TOPS, these GPUs deliver a 2X performance leap and new AI-driven rendering, including neural shaders, digital human technologies, geometry, and lighting. The new DLSS 4 boosts frame rates up to 8X with AI-driven frame generation, turning one rendered frame into three. It also features the industry's first real-time application of transformer models, packing 2X more parameters and 4X the compute for unprecedented visual fidelity. We also announced a wave of GeForce Blackwell laptop GPUs with new NVIDIA Max-Q technology that extends battery life by up to an incredible 40%. These laptops will be available starting in March from the world's top manufacturers. Moving to our professional visualization business, revenue of $511 million was up 5% sequentially and 10% year-on-year. Full-year revenue of $1.9 billion increased 21% year-on-year.

Full year revenue of $11 4 billion increased 9% year on year.

Speaker Change: And demand remains strong throughout the holiday. However, Q4 shipments were impacted by supply constraints, we expect strong sequential growth in Q1 as supply increases.

Speaker Change: These gpus deliver a two X performance leap and new AI, driven rendering including neuro shatters gender with human technologies geometry and lighting.

Speaker Change: The news the L. S us for boost frame rate up two eight acts with AI driven frame generation, turning one rendered frame into three.

Speaker Change: The New G Force Archie at 50 series desktop and laptop Gpus are here.

Speaker Change: For gamers creators and developers they use AI and graphics redefining visual computing.

Speaker Change: It also features the industry's first real time application of transformer models talking to X more parameters and Forex the compute for unprecedented visual fidelity.

Speaker Change: Powered by the Blackwell architecture.

Speaker Change: Generation tensor cores and fourth generation, our T cars and featuring of Qs 3400, AI was hot.

Speaker Change: We also announced a wave of G Force Blackwell laptop Gpus with new Nvidia next to technology that extends battery life by up to an incredible 40%. These laptops will be available starting in March from the worlds top manufacturers.

Speaker Change: These gpus deliver a two X performance leap and new AI, driven rendering including neural shatters gender with human technologies geometry and lighting.

Colette Kress: When fully ramped, we have many opportunities to improve the cost and gross margin, will improve, and return to the mid-70s late this fiscal year. Sequentially, GAAP operating expenses were up 9% and non-GAAP operating expenses were 11%, reflecting higher engineering development costs and higher compute and infrastructure costs for new product introductions.

Speaker Change: The news the L. S. S. Four boost frame rate up two eight acts with AI driven frame generation.

Speaker Change: Moving to our professional visualization business revenue.

Speaker Change: Revenue of $511 million was up 5% to question sequentially and 10% year on year.

Speaker Change: One rendered frame into three.

Colette Kress: Sequentially, GAAP operating expenses were up 9% and non-GAAP operating expenses were 11%, reflecting higher engineering development costs and higher compute and infrastructure costs for new product introductions. In Q4, we returned $8.1 billion to shareholders in the form of share repurchases and cash dividends. Let me turn to the outlook in the first quarter. Total revenue is expected to be $43 billion, plus or minus 2%. Continuing with its strong demand, we expect a significant ramp of Blackwell in Q1. We expect sequential growth in both data center and gaming. Within data center, we expect sequential growth from both compute and networking. GAAP and non-GAAP gross margins are expected to be 70.6% and 71% respectively, plus or minus 50 basis points. GAAP and non-GAAP operating expenses are expected to be approximately $5.2 billion and $3.6 billion, respectively.

Speaker Change: It also features the industry's first real time application of transformer models talking to X more parameters and Forex the compute for unprecedented visual.

Speaker Change: Full year revenue of $1 9 billion increased 21% year on year.

Speaker Change: Key industry verticals driving demand include automotive and health care.

Speaker Change: Okay.

Colette Kress: Key industry verticals driving demand include automotive and healthcare. NVIDIA technologies and generative AI are reshaping design, engineering, and simulation workloads. Increasingly, these technologies are being leveraged in leading software platforms from Ansys, Cadence, and Siemens, fueling demand for NVIDIA RTX workstations. Now, moving to automotive, revenue was a record $570 million, up 27% sequentially and up 103% year-on-year. Full-year revenue of $1.7 billion increased 55% year-on-year. Strong growth was driven by the continued ramp in autonomous vehicles, including cars and robotaxis. At CES, we announced Toyota, the world's largest automaker, will build its next-generation vehicles on NVIDIA O-RAN, running the safety-certified NVIDIA Drive OS. We announced Aurora and Continental will deploy driverless trucks at scale powered by NVIDIA Drive 4. Finally, our end-to-end autonomous vehicle platform, NVIDIA Drive Hyperion, has passed industry safety assessments by Husu and Hufreiland, two of the industry's foremost authorities for automotive-grade safety and cybersecurity.

Colette Kress: In Q4, we returned $8.1 billion to shareholders in the form of share repurchases and cash dividends.

Speaker Change: And video technologies in general today are reshaping design engineering and simulation work loans increasingly these technologies are being leveraged and leading software platforms from answers cadence and Siemens fueling demand for Nvidia Our T X workstations.

Colette Kress: Let me turn to the Outlook in the first order. Total revenue is expected to be $43 billion, plus or minus 2%. Continuing with its strong demand, we expect a significant ramp of Blackwell in Q1. We expect sequential growth in both data center and gaming. Within Datacenter, we expect sequential growth from both compute and networking.

Speaker Change: Moving to our professional visualization business.

Speaker Change: Now moving to automotive.

Speaker Change: Revenue of 511 million was up 5% to question sequentially and 10% year on year.

Revenue was a record $570 million up 27% sequentially and up 103% year on year.

Speaker Change: Full year revenue of $1 9 billion increased 21% year on year.

Speaker Change: Full year revenue of $1 7 billion increased 55% year on year strong growth was driven by the continued ramp in autonomous vehicles, including cars and Robo taxis.

Speaker Change: Industry verticals driving demand include automotive and health care.

Colette Kress: GAAP and non-GAAP gross margins are expected to be 70.6% and 71% respectively, plus or minus 50 days. GAAP and non-GAAP operating expenses are expected to be approximately $5.2 billion and $3.6 billion, respectively. We expect full fiscal year 26 operating expenses to grow to be in the mid-30s. GAAP and non-GAAP other incoming expenses are expected to be an income of approximately $400 million. excluding gains and losses from non-marketable and publicly held equity securities.

Speaker Change: And video technologies in general today are reshaping design engineering and simulation workloads.

Speaker Change: At CES, we announced Toyota the World largest automaker will build its next generation vehicles on Nvidia Oren running the safety certified Nvidia drive relapse, we announced Aurora in Continental will deploy driverless trucks at scale powered by Nvidia drive store.

Speaker Change: Presumably these technologies are being leveraged in leading software platforms from answers cadence at Siemens fueling demand for video of our T X workstations.

Colette Kress: We expect full-year fiscal year 2026 operating expenses to grow to be in the mid-30s. GAAP and non-GAAP other income expenses are expected to be an income of approximately $400 million, excluding gains and losses from non-marketable and publicly held equity securities. GAAP and non-GAAP tax rates are expected to be 17%, plus or minus 1%, excluding any discrete items. Further financial details are included in the CFO commentary and other information available on our IR website, including a new financial information AI agent. In closing, let me highlight upcoming events for the financial community. We will be at the TD Cowan Healthcare Conference in Boston on March 3 and at the Morgan Stanley Technology, Media, and Telecom Conference in San Francisco on March 5. Please join us for our annual GTC Conference starting Monday, March 17, in San Jose, California.

Speaker Change: Now moving to automotive.

Speaker Change: Revenue was a record $517 million up 27% sequentially and up 103% year on year.

Speaker Change: Finally, our end to end economists vehicle platform Nvidia drive high period has passed industry safety assessments like.

Speaker Change: Full year revenue of $1 7 billion increased 55% year on year strong growth was driven by the continued ramp in autonomous vehicles, including cars and Robo taxis.

Speaker Change: Soon and to Roland.

Speaker Change: Two of the industry's foremost authorities for automotive grade safety and cyber security.

Colette Kress: GAAP and non-GAAP tax rates are expected to be 17%, plus or minus 1%, excluding any discrete items. Further financial details are included in the CFO commentary and other information available on our IR website, including a new financial information AI page For more information visit www.fema.gov In closing let me highlight upcoming events for the financial community. We will be at the TD Cowan Healthcare Conference in Boston on March 3rd and at the Morgan Stanley Technology, Media and Telecom Conference in San Francisco on March 5th. Please join us for our annual GTC conference starting Monday, March 17th, in San Jose, California.

Speaker Change: At CES, we announced Toyota the worlds largest automaker will build its next generation vehicles on Nvidia Oren running the safety certified Nvidia drive for Abbas Wheelhouse Aurora in Continental will deploy driverless trucks at scale powered by Nvidia drive store.

Speaker Change: Nvidia is the first AZ platform to receive a comprehensive set of third party assessments.

Colette Kress: NVIDIA is the first AV platform to receive a comprehensive set of third-party assessments. OK, moving to the rest of the P&L. GAAP gross margins were 73% and non-GAAP gross margins were 73.5%, down sequentially as expected, with our first deliveries of the Blackwell architecture. As discussed last quarter, Blackwell is a customizable AI infrastructure with several different types of NVIDIA-built chips, multiple networking options, and for air and liquid-cooled data center. We exceeded our expectations in Q4 in ramping Blackwell, increasing system availability, and providing several configurations to our customers. As Blackwell ramps, we expect gross margins to be in the low 70s. Initially, we are focused on expediting the manufacturing of Blackwell systems to meet strong customer demand as they race to build out Blackwell infrastructure.

Speaker Change: Okay moving to the rest of the P&L GAAP.

Speaker Change: GAAP gross margin was 73% and non-GAAP gross margin was 73, 5% down sequentially as expected with our first deliveries of the Blackwell architecture.

Speaker Change: Finally, our end to end the autonomous vehicle platform Nvidia drive high period has passed industry safety assessments.

Speaker Change: As discussed last quarter Blackwell, it's a customizable AI infrastructure with several different types of Nvidia build ships multiple networking options and for air and liquid cooled due to summer we exceeded our expectations in Q4 and ramping Blackwell increasing system availability, providing several.

It was soon and to Roland.

Speaker Change: Two of the industry's foremost authorities for automotive grade safety and cyber security.

Speaker Change: Nvidia is the first AZ platform to receive a comprehensive set of third party assessments.

Colette Kress: Jensen will deliver a news-packed keynote on March 18, and we will host a Q&A session for our financial analysts the next day, March 19. We look forward to seeing you at these events. Our earnings call to discuss the results for our first quarter of fiscal 2026 is scheduled for May 28, 2025. We are going to open up the call operator to questions. If you could start that, that would be great.

Colette Kress: Jensen will deliver a news-packed keynote on March 18th, and we will host a Q&A session for our financial analysts the next day, March 19th. We look forward to seeing you at these events.

Speaker Change: Configurations to our customers.

Speaker Change: Okay moving to the rest of the P&L GAAP.

Speaker Change: As Blackwell ramps, we expect gross margins to be in the.

Speaker Change: GAAP gross margins was 73% and non-GAAP gross margin was 73, 5% down sequentially as expected with our first deliveries of the Blackwell architecture.

Speaker Change: In the low seventies.

Colette Kress: Our earnings call to discuss the results for our first quarter of fiscal 2026 is scheduled for May 28, 2025.

Speaker Change: We initially we are focused on expediting the manufacturing of Blackwell systems to meet strong customer demand as they race to build out Blackwell infrastructure.

Speaker Change: As discussed last quarter Blackwell, it's a customizable AI infrastructure were several different types of Nvidia build ships multiple networking options and for air and liquid cooled due to summer we exceeded our expectations in Q4 and ramping Blackwell increasing system availability, providing several.

Operator: We are going to open up the call operator to questions. If you could start that, that would be great. Thank you.

Speaker Change: When fully ramped we have many opportunities to improve the cost and gross margin will improve and return to the mid seventies late this fiscal year.

Colette Kress: When fully ramped, we have many opportunities to improve the cost and gross margin will improve and return to the mid-70% late this fiscal year. Sequentially, GAAP operating expenses were up 9% and non-GAAP operating expenses were 11%, reflecting higher engineering development costs and higher compute and infrastructure costs for new product introductions. In Q4, we returned $8.1 billion to shareholders in the form of share repurchases and cash dividends. Let me turn to the outlook in the first quarter. Total revenue is expected to be $43 billion, plus or minus 2%. Continuing with its strong demand, we expect a significant ramp of Blackwell in Q1. We expect sequential growth in both data center and gaming. Within data center, we expect sequential growth from both compute and networking. GAAP and non-GAAP gross margins are expected to be 70.6% and 71%, respectively, plus or minus 50 basis points.

Operator: Thank you. At this time, I would like to remind everyone, in order to ask a question, please press star then the number one on your telephone keypad. I also ask that you please limit yourself to one question. For any additional questions, please re-queue. Your first question comes from C.J. Muse with Cantor Fitzgerald. Please go ahead.

Operator: At this time, I would like to remind everyone, in order to ask a question, please press star, then the number one on your telephone keypad. I also ask that you please limit yourself to one question. For any additional questions, please recue.

Speaker Change: Sequentially GAAP operating expenses were up 9% and non-GAAP operating expenses were 11%, reflecting higher engineering development costs, and higher compute and infrastructure cost for new product introductions in.

Speaker Change: Configurations to our customers.

Speaker Change: Blackwell ramps, we expect gross margins to be in the.

CJ Mews: And your first question comes from CJ Mews with Cancer Fitzgerald. Please go ahead. Yeah, good afternoon. Thank you for taking the question. I guess, for me, Johnson, as Tefcon Compute and Reinforcement Learning shows such promise, we're clearly seeing increasing blurring of the lines between training and inference. What does this mean for the potential future of potentially inference dedicated clusters? And how do you think about the overall impact to NVIDIA and your customers? Thank you. Yeah, I appreciate that, CJ.

Speaker Change: In the low seventies.

Speaker Change: We initially we are focused on expediting the manufacturing of Blackwell systems to meet strong customer demand as they race to build out Blackwell infrastructure.

Speaker Change: In Q4, we returned $8 1 billion to shareholders in the form of share repurchases and cash dividends.

[Analyst]: Good afternoon. Thank you for taking the question. I guess for me, Jensen, as test time compute and reinforcement learning show such promise, we're clearly seeing an increasing blurring in the lines between training and inference. What does this mean for the potential future of potentially inference-dedicated clusters? How do you think about the overall impact to NVIDIA and your customers? Thank you.

Speaker Change: Let me turn to the outlook in the first quarter.

Speaker Change: When fully ramped we have many opportunities to improve the cost and gross margin will improve and return to the mid seventies late this fiscal year.

Speaker Change: Total revenue is expected to be 43 billion plus or minus 2%.

Speaker Change: Continuing with its strong demand, we expect a significant ramp of Blackwell in Q1.

Jensen Huang: I appreciate that, C.J. There are now multiple scaling laws. There's the pre-training scaling law, and that's going to continue to scale because we have multimodality. We have data that came from reasoning that are now used to do pre-training. The second is post-training scaling law using reinforcement learning human feedback, reinforcement learning AI feedback, reinforcement learning verifiable rewards. The amount of computation you use for post-training is actually higher than pre-training. It's kind of sensible in the sense that you could, while you're using reinforcement learning, generate an enormous amount of synthetic data or synthetically generated tokens. AI models are basically generating tokens to train AI models. That's post-training. The third part, this is the part that you mentioned, is test time compute or reasoning, long thinking, inference scaling. They're all basically the same ideas. There you have a chain of thought. You have search.

Speaker Change: We expect sequential growth in both data center and gaming.

Jensen Huang: There are now multiple scaling There's the pre-training scale and log. and that's going to continue to scale because We have multimodality. We have data that came from reasoning that are now used to do pre-training. and then the second is post-training scaling. using reinforcement learning human feedback, reinforcement learning AI feedback, reinforcement learning verifiable rewards. Amount of computation you use for post-training actually higher than pre-training and it's kind of sensible in the sense that you could while you're using reinforcement learning generate an enormous amount of synthetic data or synthetically generated token. AI models are basically generating tokens to train AI models.

Speaker Change: Within data center, we expect sequential growth from both compute and networking.

Speaker Change: GAAP and non-GAAP gross margins are expected to be 76% and 71%, respectively, plus or minus 50 basis points.

Speaker Change: In Q4, we returned $8 1 billion to shareholders in the form of share repurchases and cash dividends.

Speaker Change: Let me turn to the outlook in the first quarter.

Speaker Change: GAAP and non-GAAP operating expenses are expected to be approximately $5 2 billion and $3 6 billion respectively.

Colette Kress: GAAP and non-GAAP operating expenses are expected to be approximately $5.2 billion and $3.6 billion, respectively. We expect full-year fiscal year 2026 operating expenses to grow to be in the mid-30%. GAAP and non-GAAP other income expenses are expected to be an income of approximately $400 million, excluding gains and losses from non-marketable and publicly held equity securities. GAAP and non-GAAP tax rates are expected to be 17%, plus or minus 1%, excluding any discrete items. Further financial details are included in the CFO commentary and other information available on our IR website, including a new financial information AI agent. In closing, let me highlight upcoming events for the financial community. We will be at the TD Cowan Healthcare Conference in Boston on March 3 and at the Morgan Stanley Technology, Media, and Telecom Conference in San Francisco on March 5.

Speaker Change: Total revenue is expected to be 43 billion plus or minus 2%.

Speaker Change: We expect full year fiscal year 'twenty six operating expenses to grow to be in the mid thirties.

Speaker Change: Continuing with its strong demand, we expect a significant ramp of Blackwell in Q1, we expect.

Speaker Change: Sequential growth in both data center and gaming.

Speaker Change: GAAP and non-GAAP other income and expenses are expected to be an income of approximately 400 million exclude.

Speaker Change: Within data center, we expect sequential growth from both compute and networking.

Speaker Change: Excluding gains and losses from non marketable and publicly held equity securities.

Speaker Change: GAAP and non-GAAP gross margins are expected to be 76% and 71%, respectively, plus or minus 50 basis points.

Speaker Change: GAAP and non-GAAP tax rates are expected to be 17% plus or minus 1%, excluding any discrete items.

Speaker Change: GAAP and non-GAAP operating expenses are expected to be approximately $5 2 billion and $3 6 billion respectively.

Jensen Huang: And that's post-training. And the third part, this is the part that you mentioned, is test time compute, or reasoning, long thinking, inference scaling. They're all basically the same idea. and there's your chain of thought. Search. The amount of tokens generated the amount of inference compute needed. There's already a hundred times more than the one-shot examples. and the one-shot capabilities of... large language models in the beginning. And that's just the beginning. This is just the beginning. The idea that the next generation could have thousands of times. and even, hopefully, extremely thoughtful and simulation-based and search-based. models that could be hundreds of thousands, millions of times more compute than today is in our future.

Speaker Change: Further financial details are included in the CFO commentary and other information available on our IR website.

Speaker Change: We expect full year fiscal year 'twenty six operating expenses to grow to be in the mid thirties.

Speaker Change: Putting a new financial information AI.

Speaker Change: Right.

Speaker Change: In closing, let me highlight upcoming events for the financial community, we will be at the TD Cowen Health Care Conference in Boston on March 3rd and at the Morgan Stanley Technology Media and Telecom conference in San Francisco on March 1st please.

Speaker Change: GAAP and non-GAAP other income and expenses are expected to be an income of approximately 400 million, excluding gains and losses from non marketable and publicly held equity securities.

Jensen Huang: The amount of tokens generated, the amount of inference compute needed is already 100 times more than the one-shot examples and the one-shot capabilities of large language models in the beginning. That's just the beginning. This is just the beginning. The idea that the next generation could have thousands of times and even hopefully extremely thoughtful and simulation-based and search-based models that could be hundreds of thousands, millions of times more compute than today is in our future. The question is, how do you design such an architecture? Some of the models are autoregressive. Some of the models are diffusion-based. Sometimes you want your data center to have disaggregated inference, sometimes it's compacted. It's hard to figure out what is the best configuration of a data center, which is the reason why NVIDIA's architecture is so popular. We run every model. We are great at training.

Speaker Change: Please join us for our annual GTC Conference, starting Monday March 17th in San Jose, California, Jensen will deliver a news packed keynote on March 15th and we will host a Q&A session for financial analysts. The next day March 19, we look forward to seeing you at these events.

Speaker Change: And non-GAAP tax rates are expected to be 17% plus or minus 1%, excluding any discrete items.

Colette Kress: Please join us for our annual GTC Conference starting Monday, March 17, in San Jose, California. Jensen will deliver a news-packed keynote on March 18, and we will host a Q&A session for our financial analysts the next day, March 19. We look forward to seeing you at these events. Our earnings call to discuss the results for our first quarter of fiscal 2026 is scheduled for May 28, 2025. We are going to open up the call operator to questions. If you could start that, that would be great.

Speaker Change: Further financial details are included in the CFO commentary and other information available on our IR website, including a new financial information AI.

Speaker Change: Our earnings call to discuss the results for our first quarter of fiscal 2026 is scheduled for May 28 2025.

Speaker Change: Jet.

Speaker Change: In closing, let me highlight upcoming events for the financial community, we'll be at the TD Cowen Health Care Conference in Boston on March 3rd and at the Morgan Stanley Technology Media and Telecom conference in San Francisco on March 1st.

Speaker Change: We are going to open up the call operator to questions. If you could start that that would be great.

Jensen Huang: And so the question is how do you design such an architecture? Some of the models are auto-regressive, some of the models are diffusion-based. Some of the times you want your data center to have disaggregated inference, sometimes it's compacted. And so it's hard to figure out what is the best configuration of a data center, which is the reason why NVIDIA's architecture is so popular. We run every model. We are great at training. The vast majority of our compute today is actually inference. and Glidewell. takes all of that to a new level. We designed Blackwell with the idea of reasoning models in mind.

Speaker Change: Thank you at this time I would like to remind everyone in order to ask a question. Please press Star then the number one on your telephone keypad.

Speaker Change: I also ask that you please limit yourself to one question for any additional questions. Please re queue.

Speaker Change: And your first question comes from C. J Muse with Cantor Fitzgerald. Please go ahead.

Speaker Change: Our earnings call to discuss the results for our first quarter of fiscal 2026 is scheduled for May 28, two.

C.J. Muse: Yeah. Good afternoon. Thank you for taking the question I guess.

[Analyst]: Yeah, good afternoon. Thank you for taking the question. I guess for me, Jensen, as test time compute and reinforcement learning show such promise, we're clearly seeing an increasing blurring in the lines between training and inference. What does this mean for the potential future of potentially inference-dedicated clusters? How do you think about the overall impact to NVIDIA and your customers? Thank you.

For me Josef computing reinforcement learning show such promise, we're clearly seeing increasing blurring of the lines between training and inference. What does this mean for the potential future potentially influence dedicated clusters and how do you think about the overall impact to Nvidia and your customers. Thank you.

Speaker Change: 2020 fives.

Speaker Change: We are going to open up the call operator to questions. If you could star stuff that would be great.

Jensen Huang: The vast majority of our compute today is actually inference. Blackwell takes all of that to a new level. We designed Blackwell with the idea of reasoning models in mind. When you look at training, it's many times more performant. What's really amazing is for long thinking, test time scaling, reasoning AI models were tens of times faster, 25 times higher throughput. Blackwell is going to be incredible across the board. When you have a data center that allows you to configure and use your data center based on are you doing more pre-training now, post-training now, or scaling out your inference, our architecture is fungible and easy to use in all of those different ways. We're seeing, in fact, much, much more concentration of a unified architecture than ever before.

Speaker Change: Thank you at this time I would like to remind everyone in order to ask a question. Please press Star then the number one on your telephone keypad I also ask that you. Please limit yourself to one question for any additional questions. Please re queue.

Jensen Huang: And when you look at training as many times more performant. What's really amazing is, for long thinking... Test Time Scaling reasoning AI models were tens of times faster, 25 times higher throughput. So Blackwell is going to be incredible across the board.

C.J. Muse: I appreciate that.

C.J. Muse: There are now multiple scaling loss.

C.J. Muse: There's the pre training scaling law.

Operator: And your first question comes from C. J Muse with Cantor Fitzgerald. Please go ahead.

C.J. Muse: And that's going to continue to scale because.

C.J. Muse: We have multi modality, we have data that came from.

Speaker Change: Yes. Good afternoon. Thank you for taking the question I guess.

Speaker Change: For me Josef as Jeff kind of computing reinforcement learning show such promise, we're clearly seeing increasing blurring of the lines between training and inference.

C.J. Muse: Reasoning that are now used to do pre training.

Jensen Huang: And when you have a data center that allows you to configure and use your data center based on, are you doing more pre-training now, post-training now, or scaling out your inference, our architecture is fungible and easy to use in all of those different ways. We're seeing, in fact, much, much more concentration of a unified architecture than ever before.

C.J. Muse: And then the second is post training scaling loan.

Meaning for the potential future potentially influence dedicated clusters and how do you think about the overall impact to Nvidia and your customers. Thank you.

C.J. Muse: Usually reinforcement learning human feedback reinforcement learning AI feedback reinforcement learning verifiable rewards.

Speaker Change: Yes, I appreciate that C J.

C.J. Muse: Hum.

C.J. Muse: The amount of computation you used for post training.

Speaker Change: There are now multiple scaling walls.

Speaker Change: There's the pre training scaling law.

C.J. Muse: Is actually higher than pre training and it's kind of sensible in the sense that you could.

Speaker Change: And that's going to continue to scale because.

Speaker Change: We have multi modality.

C.J. Muse: While you're using reinforcement learning generate an enormous amount of synthetic data, we're synthetically generated tokens.

Joe Moore: Your next question comes from the line of Joe Moore with JPMorgan. Please go ahead.

Operator: Your next question comes from the line of Joe Moore with JPMorgan. Please go ahead.

Speaker Change: That came from.

Speaker Change: Reasoning.

Speaker Change: Are now used to do pre training.

C.J. Muse: My models are basically generating tokens to training models and.

Jensen Huang: Morgan Stanley, actually, thank you. I wonder if you could talk about GB200 at CES you sort of talked about the complexity of the rack level systems and the challenges you have and then as you said in the prepared remarks we've seen a lot of general availability you know where are you in terms of that ramp are there still bottlenecks to consider at a systems level above and beyond the chip level and just you know have you maintained your enthusiasm for the NGL72 platforms you Well, I'm more enthusiastic today than I was at... And the reason for that is because we've shipped a lot more.

[Analyst]: Morgan Stanley, thank you. I wonder if you could talk about GB200. At CES, you talked about the complexity of the rack-level systems and the challenges you have. As you said in the prepared remarks, we've seen a lot of general availability. Where are you in terms of that ramp? Are there still bottlenecks to consider at a systems level above and beyond the chip level? Have you maintained your enthusiasm for the NVL 72 platforms?

Speaker Change: And then the second.

C.J. Muse: That's post treat.

C.J. Muse: And the third the third part is the part that you mentioned is tough time compute or reasonably long thinking do French scaling.

Speaker Change: Whose training scaling law.

Speaker Change: Using reinforcement learning humans feedback reinforcement learning AI feedback reinforcement learning verifiable rewards.

C.J. Muse: Basically the same ideas there as you have chain of thought.

Speaker Change: The amount of computation you used for post training.

Search.

C.J. Muse: The amount of.

Speaker Change: Is actually higher than pre training and it's kind of sensible in the sense that you could while.

C.J. Muse: Tokens generated the amount of influence compute need it.

C.J. Muse: It's already a 100 times more than the one shot examples.

Jensen Huang: I'm more enthusiastic today than I was at CES. The reason for that is because we've shipped a lot more to the CES. We have some 350 plants manufacturing the 1.5 million components that go into each one of the Blackwell racks, Grace Blackwell racks. Yes, it's extremely complicated. We successfully and incredibly ramped up Grace Blackwell, delivering some $11 billion in revenues last quarter. We're going to have to continue to scale as demand is quite high and customers are anxious and impatient to get their Blackwell systems. You've probably seen on the web a fair number of celebrations about Grace Blackwell systems coming online. We have them, of course. We have a fairly large installation of Grace Blackwells for our own engineering and our own design teams and software teams. CoreWeave has now been quite public about the successful bring-up of theirs. Microsoft has. Of course, OpenAI has.

Speaker Change: While you're using reinforcement learning generate an enormous amount of synthetic data, we're synthetically generated tokens.

C.J. Muse: One shot capabilities of.

C.J. Muse: <unk> language models from the beginning and that's just the beginning this is just the beginning the idea that the.

Jensen Huang: We have some 350 plans. Manufacturing The one and a half million components that go into each one of the Blackwell racks, Grace Blackwell racks. Yes, it's extremely complicated, and we successfully and incredibly ramped up Grace Blackwell. delivering some $11 billion in revenues last quarter. We're going to have to continue to scale as demand is quite high and customers are anxious and impatient to get their Blackwell system. You've probably seen on the web a fair number of celebrations about Grace Blackwell systems coming online and we have them of course. We have a fairly large installation of Grace Blackwells for our own engineering and our own design teams and software teams. CoreWeave has now been quite public about the successful bring up of theirs.

Speaker Change: AI models are basically generating tokens to training models.

Speaker Change: That's post street.

C.J. Muse: The next generation could have.

Speaker Change: And then the third part is the part that you mentioned is his time compute or reasonably long. Thank you.

C.J. Muse: <unk> times.

C.J. Muse: Even hopefully.

C.J. Muse: Extremely thoughtful and simulation based in search space.

Speaker Change: French scaling.

Speaker Change: Basically the same ideas and there is you have chain of thought.

C.J. Muse: Models that could be hundreds of thousands millions of times more compute than today.

Speaker Change: Search.

Speaker Change: The amount of.

Speaker Change: Tokens generate at the amount of inference compute need it.

C.J. Muse: As is in our future.

C.J. Muse: And so.

Speaker Change: It's already a 100 times more than the one shot examples.

C.J. Muse: The question is how do you design such a such an architecture some of it some of the models are auto regressive. Some of the models are diffusion based.

Speaker Change: The one shot capabilities of.

Speaker Change: Larsen language models from the beginning and that's just the beginning this is just the beginning.

C.J. Muse: Some of it some of them at the time do you want your.

Speaker Change: Due to the.

Speaker Change: The next generation could have.

C.J. Muse: Data center to have disaggregated influenced sometimes it's.

Speaker Change: <unk> times.

Speaker Change: Even hopefully.

C.J. Muse: As compacted and so it's.

Speaker Change: Extremely thoughtful and simulation based in search space.

C.J. Muse: It's hard to it's hard to figure out.

C.J. Muse: What is the best configuration of a data center, which is the reason why videos architecture. So popular.

Speaker Change: Models that could be hundreds of thousands millions of times more compute than today.

C.J. Muse: We run a remodel.

Jensen Huang: Microsoft has, of course OpenAI has. and you're starting to see many, many come online. and so I think the answer to your question is... Nothing is easy about what we're doing, but we're doing great, and all of our partners are doing great.

C.J. Muse: We are great training.

Speaker Change: As is in our future.

Jensen Huang: You're starting to see many, many come online. I think the answer to your question is nothing is easy about what we're doing, but we're doing great. All of our partners are doing great.

C.J. Muse: The vast majority of our compute today is actually inference.

Speaker Change: And so.

Speaker Change: The question is how do you design such a such an architecture some of it some of the models are auto regressive. Some of the models are diffusion based.

C.J. Muse: And Blackwell.

C.J. Muse: It takes all of that to a new level redesign blend well with the idea of reasoning models in mind.

C.J. Muse: And when you look at training is many times.

Speaker Change: Some of it some of them at the time do you want your.

C.J. Muse: More performance.

Speaker Change: Data center to have disaggregated infringe sometimes it's.

C.J. Muse: What's really amazing is for long thinking.

Vivek Arra: Your next question comes from the line of Vivek Arra with Bank of America Securities. Please go ahead. Thank you for taking my question.

Operator: Your next question comes from the line of Vivek Arya with Bank of America Securities. Please go ahead.

Speaker Change: As compacted and so it's.

C.J. Muse: Test time scaling.

Speaker Change: It's hard to it's hard to figure out.

C.J. Muse: Reasoning AI models.

What is the best configuration of a data center, which is the reason why videos architecture. So popular.

[Analyst]: Thank you for taking my question. Colette, if you wouldn't mind confirming if Q1 is the bottom for gross margins. Jensen, my question is for you. What is on your dashboard to give you the confidence that this strong demand can sustain into next year? Has DeepSeek and whatever innovations they came up with, has that changed that view in any way? Thank you.

Where tens of times faster 25 times higher throughput.

Colette Kress: Kulak, if you wouldn't mind confirming if Q1 is the bottom for gross margins. And then, Jensen, my question is for you. What is on your dashboard to give you the confidence that this strong demand can sustain into next year? And has DeepSeek and whatever innovation they came up with, has that changed that view in any way?

Speaker Change: We run a remodel.

C.J. Muse: And so Blackwell is going to be incredible across the board and when you have a data center that allows you to.

Speaker Change: We are great training.

Speaker Change: The vast majority of our compute today is actually inference.

Configure and use your data center base.

Speaker Change: And Blackwell.

Speaker Change: Based on are you doing more pre trading now post training now or stealing out your inference.

Speaker Change: It takes all of that to a new level, we designed blend well with the idea of reasoning models in mind and when you look at training is many times more.

Speaker Change: Architecture is fungible and easy to use in all of those different ways.

Colette Kress: Let me first take that first part of the question regarding the gross margin. During our Blackwell ramp, our gross margins will be in the low 70%. At this point, we are focusing on expediting our manufacturing, expediting our manufacturing to make sure that we can provide to customers as soon as possible. Our Blackwell has fully ramped. Once it does, I'm sorry, once our Blackwell fully ramps, we can improve our cost and our gross margin. We expect to probably be in the mid-70% later this year. Walking through what you heard Jensen speak about, the systems and their complexity, they are customizable in some cases. They've got multiple networking options. They have liquid-cooled and water-cooled. We know there is an opportunity for us to improve these gross margins going forward.

Colette Kress: Let me first take the first part of the question there regarding the gross margin. During our Blackwell ramp, our gross margins will be in the low 70s. At this point, we are focusing on expediting our manufacturing. We are expediting our manufacturing to make sure that we can provide to customers as soon as possible. Our Blackwell has fully ramped, and once it does, I'm sorry, once our Blackwell fully ramps, we can improve our cost and our growth margin. So we expect to probably be in the mid-70s later this year. You know, walking through what you heard Johnson speak about the systems and their complexity.

Speaker Change: More performance.

Speaker Change: But what's really amazing is for long thinking.

Speaker Change: And so we're.

Speaker Change: We're seeing in fact, much much more concentration of a unified architecture than ever before.

Speaker Change: Test time scaling.

Speaker Change: Reasoning AI models.

Speaker Change: Where hundreds of times faster 25 times higher throughput.

Speaker Change: Your next question comes from the line of Joe Moore with J P. Morgan. Please go ahead.

Speaker Change: And so Blackwell is going to be incredible across the board and when you have a data center that allows you to.

Operator: Your next question comes from the line of Joe Moore with JPMorgan. Please go ahead.

Speaker Change: Morgan Stanley actually thank you.

Speaker Change: I Wonder if you could.

[Analyst]: Morgan Stanley, actually, thank you. I wonder if you could talk about GB200. At CES, you sort of talked about the complexity of the rack-level systems and the challenges you have. As you said in the prepared remarks, we've seen a lot of general availability. Where are you in terms of that ramp? Are there still bottlenecks to consider at a systems level above and beyond the chip level? Have you maintained your enthusiasm for the NVL72 platforms?

Speaker Change: Configure and use your data center.

Speaker Change: Talk about GBP 200 at CES, you sort of talked about the complexity of the rack level systems and the challenges you have and then as you said in the prepared remarks, we've seen a lot of general availability.

Speaker Change: Based on are you doing more pre trading now post training now or stealing out your inference of ours.

Speaker Change: The texture is fungible and easy to use in all of those different ways.

Speaker Change: Where are you in terms of that ramp are there still bottlenecks to consider at a systems level above and beyond the chip level and just.

Colette Kress: They are customizable in some cases, they've got multiple networking options, they have liquid cool and water cool. So we know there is an opportunity for us to improve these gross margins going forward.

Speaker Change: So.

Speaker Change: We're seeing in fact, much much more concentration of a unified architecture than ever before.

Speaker Change: Have you maintained your enthusiasm for the NPL 72 platforms.

Speaker Change: Well I'm more enthusiastic today than I was at CES.

Speaker Change: Your next question comes from the line of Joe Moore with J P. Morgan. Please go ahead.

Jensen Huang: But right now, we are gonna focus on getting the manufacturing complete and to our customers as soon as possible. We know several things. We have a fairly good line of sight of the Amount of capital investment that Data centers are building, building out towards We know that Going forward, the vast majority of software is going to be based on machine learning. So accelerated computing and generative AI, reasoning AI, are going to be the type of architecture you want in your data set. The number of startups are still quite vibrant, and each one of them need a fair amount of computing infrastructure.

Colette Kress: Right now, we are going to focus on getting the manufacturing complete and to our customers as soon as possible.

Speaker Change: And the reason for that is because we shipped a lot more success.

Speaker Change: We have we have some 350.

Speaker Change: Morgan Stanley actually thank you.

Jensen Huang: We know several things, Vivek. We have a fairly good line of sight of the amount of capital investment that data centers are building out towards. We know that going forward, the vast majority of software is going to be based on machine learning. Accelerated computing and generative AI, reasoning AI, are going to be the type of architecture you want in your data center. We have, of course, forecasts and plans from our top partners. We also know that there are many innovative, really exciting startups that are still coming online as new opportunities for developing the next breakthroughs in AI, whether it's agentic AIs, reasoning AIs, or physical AIs. The number of startups is still quite vibrant. Each one of them needs a fair amount of computing infrastructure.

Speaker Change: <unk>.

Speaker Change: I Wonder if you could.

Speaker Change: Manufacturing.

Speaker Change: The one and a half million components that go into each one of the Platteville racks.

Speaker Change: <unk>, yes, it's extremely complicated and we successfully.

Speaker Change: Where are you in terms of that ramp are there still bottlenecks to consider at a systems level above and beyond the chip level and just.

Speaker Change: And.

Speaker Change: Incredibly ramped up.

Speaker Change: Grace Blackwell.

Speaker Change: Delivering some $11 billion in revenues last quarter.

Speaker Change: Have you maintained your enthusiasm for the NGL 72 platforms.

Speaker Change: We're going to have to continue to scale as demand is quite high and customers are anxious and impatient to get their Blackwell systems.

Speaker Change: Well I'm more enthusiastic today than I was at CES.

Speaker Change: And the reason for that is because we shipped a lot more success.

Speaker Change: You've probably seen on on on the web.

Speaker Change: We have we have some 350.

Speaker Change: A fair number of celebrations about about Grace Blackwell systems coming online.

Speaker Change: Lance.

Speaker Change: Manufacturing.

Speaker Change: We have them of course, we have a fairly large installation of <unk> both for our own engineering.

Speaker Change: The one and a half million components that go into each one of the Platteville racks.

Speaker Change: <unk>, yes, it's extremely complicated.

Speaker Change: Our own design teams and software teams.

Speaker Change: Core <unk> costs and so on.

Speaker Change: We successfully.

Speaker Change: And.

Speaker Change: <unk> been quite public about the successful bring up Theres a Microsoft has of course open AI has and youre starting to see many many come online and.

Speaker Change: Incredibly ramped up.

Speaker Change: Chris Blackwell.

Speaker Change: Delivering some $11 billion in revenues last quarter.

Speaker Change: We're going to have to continue to scale as demand is quite high and customers are anxious and impatient to get their Blackwell systems.

Jensen Huang: I think whether it's the near-term signals or the mid-term signals, near-term signals, of course, are POs and forecasts and things like that. Mid-term signals would be the level of infrastructure and CapEx scale-out compared to previous years. The long-term signals have to do with the fact that we know fundamentally software has changed from hand coding that runs on CPUs to machine learning and AI-based software that runs on GPUs and accelerated computing systems. We have a fairly good sense that this is the future of software. Maybe as you roll it out, another way to think about that is we've really only tapped consumer AI and search and some amount of consumer generative AI, advertising, recommenders, kind of the early days of software.

Jensen Huang: So I think whether it's the near-term signals or the mid-term signals, near-term signals, of course, are POs and forecasts and things like that. Mid-term signals would be the level of infrastructure and CapEx scale-out compared to previous years. And then the long-term signals has to do with the fact that we know fundamentally software has changed. from hand coding that runs on CPUs to machine learning and AI-based software that runs on GPUs and accelerated computing. Search, and some amount of consumer generative AI. advertising, recommenders, kind of the early days of software. The next wave is coming, agentic AI for enterprise, physical AI for robotics.

Speaker Change: So so.

Speaker Change: I think the answer to your question is.

Speaker Change: You've probably seen on on on the web.

Speaker Change: Nothing is easy about what we're doing but we're doing great.

Speaker Change: A fair number of celebrations about about Grace Blackwell systems coming online.

Speaker Change: All of our partners are doing great.

Speaker Change: We have them of course, we have a fairly large installation of <unk> both for our own engineering.

Speaker Change: Your next question comes from the line of Vivek.

Operator: Your next question comes from the line of Vivek Arya with Bank of America Securities. Please go ahead.

Vic: Vic Iraq with Banc of America Securities. Please go ahead.

Speaker Change: Our own design teams and software teams.

Speaker Change: Core <unk> as well.

Vivek: Thank you for taking my question.

Speaker Change:

Vivek: If you wouldn't mind confirming if Q1 is the bottom for gross margin and then my question is for you.

Speaker Change: Quite public about the successful bring up theirs.

Speaker Change: Microsoft has of course open AI has and Youre starting to see many many come online.

What is on your dashboard to give you the confidence that the strong demand can sustain into next year and has deep seek and whatever innovations. They came up with has that changed that view in any way. Thank.

Speaker Change: So.

Speaker Change: So I think the answer to your question is nothing is easy about what we're doing but we're doing great.

Vivek: Thank you.

Speaker Change: All of our partners are doing great.

Vivek: Well, let me first take the first part of the question there regarding the gross margin.

Colette Kress: Let me first take that first part of the question regarding the gross margin. During our Blackwell ramp, our gross margins will be in the low 70%. At this point, we are focusing on expediting our manufacturing, expediting our manufacturing to make sure that we can provide to customers as soon as possible. Our Blackwell has fully ramped. Once our Blackwell fully ramps, we can improve our cost and our gross margin. We expect to probably be in the mid-70% later this year. Walking through what you heard Jensen speak about, the systems and their complexity, they are customizable in some cases. They've got multiple networking options. They have liquid-cooled and water-cooled. We know there is an opportunity for us to improve these gross margins going forward. Right now, we are going to focus on getting the manufacturing complete and to our customers as soon as possible.

Speaker Change: Your next question comes from the line of Vivek <unk> with Bank of America Securities. Please go ahead.

Vivek: During our Blackwell route our gross margins will be in about seven days at this point, we are focusing on expediting our manufacturing expediting our manufacturing to make sure that we can provide our customers as soon as possible.

Vivek <unk>: Thank you for taking my question.

Speaker Change: You wouldn't mind confirming if Q1 is the bottom for gross margin and then my question is for you.

Jensen Huang: The next wave is coming, agentic AI for enterprise, physical AI for robotics, and sovereign AI as different regions build out their AI for their own ecosystems. Each one of these is barely off the ground. We can see them because obviously, we're in the center of much of this development. We can see great activity happening in all these different places. These will happen. Near-term, mid-term, long term.

Speaker Change: What is on your dashboard to give you the confidence that the strong demand can sustain into next year and has deep seek and whatever innovations. They came up with has that changed that view in any way.

Vivek: Once it's fully ramped and once it does I'm sorry, once our Blackwell fully rounds, we can improve our cost and our gross margin. So we expect to probably be in the mid seventies later this year.

Jensen Huang: and Sovereign AI as different regions build out. their AI for their own ecosystems. And so each one of these are barely off the ground and we can see them. We can see them because, you know, obviously we're in the center of much of this development and we can see great activity happening in all these different places and these will happen.

Speaker Change: Okay.

Speaker Change: Walking through what you've heard Johnson speak about the systems and their complexity. They are customizable in some cases, they've got multiple networking options. They have liquid cooling water cooled. So we know there is an opportunity for us to improve these gross margins going forward I'll provide now we are going to focus on.

Speaker Change: Well, let me first take the first part of the question there regarding the gross margin.

Speaker Change: During our Blackwell route our gross margins will be in about seven days at this point, we are focusing on expediting our manufacturing expediting our manufacturing to make sure that we can provide to customers as soon as possible our block wasn't fully ramped.

Jensen Huang: near-term, mid-term, long-term.

Harlan Sur: Your next question comes from the line of Harlan Sir with JPMorgan. Please go ahead. Yeah, good afternoon. Thanks for taking my question. Your next generation Blackwell Ultra is set to launch in the second half of this year in line with the team's annual product cadence. Jensen, can you help us understand the demand dynamics for Ultra, given that you'll still be ramping the current generation Blackwell solutions? How do your customers and the supply chain also manage the simultaneous ramps of these two products? Is the team still on track to execute Blackwell Ultra in the second half of this year?

Operator: Your next question comes from the line of Harlan Sur with JPMorgan. Please go ahead.

Speaker Change: On getting the manufacturing complete them to our customers as soon as possible.

[Analyst]: Yeah, good afternoon. Thanks for taking my question. Your next-generation Blackwell Ultra is set to launch in the second half of this year in line with the team's annual product cadence. Jensen, can you help us understand the demand dynamics for Ultra, given that you'll still be ramping the current-generation Blackwell solutions? How do your customers and the supply chain also manage the simultaneous ramps of these two products? Is the team still on track to execute Blackwell Ultra in the second half of this year?

Speaker Change: We know several things.

Speaker Change: And once it does I'm sorry, once our Blackwell fully brands, we can improve our cost and our gross margin. So we expect to probably be in the mid seventies later this year.

Speaker Change: We have a fairly good line of sight of the amount.

Jensen Huang: We know several things, Vivek. We have a fairly good line of sight of the amount of capital investment that data centers are building out towards. We know that going forward, the vast majority of software is going to be based on machine learning. Accelerated computing and generative AI, reasoning AI, are going to be the type of architecture you want in your data center. We have, of course, forecasts and plans from our top partners. We also know that there are many innovative, really exciting startups that are still coming online as new opportunities for developing the next breakthroughs in AI, whether it's agentic AI, reasoning AI, or physical AI. The number of startups is still quite vibrant, and each one of them needs a fair amount of computing infrastructure.

Speaker Change: Amounts of capital investment.

Speaker Change: But data centers or building building out towards.

Walking through what you've heard Johnson speak about the systems and their complexity. They are customizable in some cases, they've got multiple networking options. They have liquid cooling water cooled. So we know there is an opportunity for us to improve these gross margins going forward I'll provide now we are going to focus on.

Speaker Change: We know that.

Speaker Change: Going forward.

Speaker Change: The vast majority of software, it's going to be based on machine learning.

Speaker Change: And so accelerated computing in general today, I recently, AI or would you be the type of architecture you want in your data center.

Speaker Change: We have of course.

Jensen Huang: Yes, Blackwell Ultra is second half. As you know... The first black wall was, we had a hiccup that probably cost us a couple of... were fully recovered of course. The team did an amazing job recovering and all of our supply chain partners and just so many people helped us recover at the speed of light. And so now we've successfully ramped production of Blackwell. But that doesn't stop the next train. The next train is on an annual rhythm and Blackwell Ultra with new networking, new memories, and of course new processors and all of that is coming online.

Jensen Huang: Yes. Blackwell Ultra is second half. As you know, the first Blackwell was we had a hiccup that probably cost us a couple of months. We're fully recovered, of course. The team did an amazing job recovering, and all of our supply chain partners and just so many people helped us recover at the speed of light. Now we've successfully ramped production of Blackwell. That doesn't stop the next train. The next train is on an annual rhythm. Blackwell Ultra with new networking, new memories, and of course, new processors, all of that is coming online. We've been working with all of our partners and customers laying this out. They have all of the necessary information, and we'll work with everybody to do the proper transition. This time between Blackwell and Blackwell Ultra, the system architecture is exactly the same.

Speaker Change: Our forecast and plans from.

Speaker Change: On getting the manufacturing complete them to our customers as soon as possible.

Speaker Change: Our top partners and we also.

Speaker Change: Several things.

Speaker Change: No that there many innovative really exciting startups.

Speaker Change: We have a fairly good line of sight of the amount.

Speaker Change: We're still coming online.

Speaker Change: Amount of capital investment.

Speaker Change: New opportunities for developing the next big breakthroughs in AI, where there is a genetic reasoning.

Speaker Change: Our.

Speaker Change: Data centers or building building out towards.

Speaker Change: We know that.

Speaker Change: Going forward the vast majority.

Speaker Change: Yeah.

Speaker Change: Or physical.

Speaker Change: Majority of software is going to be based on machine learning.

Speaker Change: Number of startups or are still quite vibrant and each one of them need a fair amount of computing computing infrastructure.

Speaker Change: And so accelerated computing in general today.

Speaker Change: We're going to be the type of architecture, you want in your data center.

Speaker Change: So.

Speaker Change: I think the weather.

Speaker Change: We have of course.

Speaker Change: Whether it's the near term.

Our forecast and plans from our top partners.

Speaker Change: Signals.

Speaker Change: Or the mid term signals near term signals of course, our Pos and forecast and things like that midterm signals would be.

Speaker Change: And we also.

Jensen Huang: We've been working with all of our partners and customers. Laying this out, we have all of the necessary information and we'll work with everybody to do the proper transition. This time between Blackwell and Blackwell Ultra, the system architecture is exactly the same. It's a lot harder going from Hopper to Blackwell because we went from an NVLink 8 system to a NVLink 72 base. So the chassis, the architecture of the system, the hardware, the power delivery, all of that had to change. This was quite a challenging transition, but the next transition will slot right in. Brace Blackwell Ultra will slot right in.

Speaker Change: No that there are many innovative really exciting startups.

Speaker Change: The level of.

Speaker Change: We're still coming online.

Speaker Change: Our infrastructure and Capex scale out compared to previous years.

Speaker Change: New opportunities for developing the next break breakthroughs in AI, whether it's <unk> reasoning.

Speaker Change: And then the long term signals it has to do with the fact that we know fundamentally software has changed.

Speaker Change: Or physical.

Jensen Huang: It's a lot harder going from Hopper to Blackwell because we went from an NVLink 8 system to an NVLink 72-based system. The chassis, the architecture of the system, the hardware, the power delivery, all of that had to change. This was quite a challenging transition. The next transition will slot right in. Grace Blackwell Ultra will slot right in. We've also already revealed and been working very closely with all of our partners on the click after that. The click after that is called Vera Rubin, and all of our partners are getting up to speed on the transition of that and preparing for that transition. Again, we're going to provide a big, big, huge step up. Come to GTC, and I'll talk to you about Blackwell Ultra, Vera Rubin, and then show you what's the one click after that. Really, really exciting new product. Come to GTC, please.

Speaker Change: From hand coding that runs on Cpus to machine learning and AI based software that runs on Gpus and accelerated computing computing systems and so we have a fairly good sense, but this is the future of software and then maybe as you roll it out another way to think.

Speaker Change: A number of startups or are still quite vibrant and each one of them need a fair amount of computing computing infrastructure.

Speaker Change: So.

Speaker Change: I think the weather.

Speaker Change: Whether it's the near term.

Speaker Change: Signals.

Speaker Change: For the mid term signals near term signals of course, our Pos and forecast and things like that mid term signals would be.

Speaker Change: That is is we've really only tapped a consumer.

Speaker Change: The level of.

Speaker Change: Consumer.

Jensen Huang: We've also already revealed and been working very closely with all of our partners on the click after that. And the click after that is called Vera Rubin. And all of our partners are getting up to speed on the transition of that. And so preparing for that transition, and again, we're going to provide a big, big, huge step up.

Speaker Change: AI and <unk>.

Speaker Change: Infrastructure, and Capex scale out compared to previous years.

Speaker Change: Search and some amount of consumer generative at yacht.

Speaker Change: And then the long term signals it has to do with the fact that we know fundamentally software has changed.

Speaker Change: Advertising.

Speaker Change: Recommended.

Speaker Change: The early days of software.

Speaker Change: From Hana Cody that runs on Cpus to machine learning and AI based software that runs on Gpus and accelerated computing computing systems and so we have a fairly good sense, but this is the future of software and then maybe as you roll it out another way to think about.

Speaker Change: The next the next waves coming Agentic AI for enterprise.

Jensen Huang: The next wave is coming, agentic AI for enterprise, physical AI for robotics, and sovereign AI as different regions build out their AI for their own ecosystems. Each one of these is barely off the ground. We can see them because obviously, we're in the center of much of this development, and we can see great activity happening in all these different places. These will happen. Near-term, mid-term, long term.

Speaker Change: Physical AI for robotics.

Speaker Change: And Ah.

Jensen Huang: And so come to GTC, and I'll talk to you about Blackwell Ultra, Vera Rubin, and then show you what's the one click after that. Really, really exciting new products. So come to GTC.

Speaker Change: Software and AI is as different regions build out.

Speaker Change: There are there.

Speaker Change: For their own ecosystems, and so each one of these are barely off the ground and we can see that we can see them because obviously, we're in the center of much of this development and we can see great activity happening in all these different places and this will happen.

Speaker Change: That is.

Speaker Change: We've really only helped.

Speaker Change: Consumer.

Speaker Change: AI and <unk>.

Timothy Arcuri: Your next question comes from the line of Timothy Arcuri with UBS. Please go ahead. Thanks a lot. Jensen, we hear a lot about custom ASICs. Can you kind of speak to the balance between custom ASIC and merchant GPU? We hear about some of these heterogeneous superclusters to use both GPU and ASIC. Is that something customers are planning on building? Or will these infrastructures remain fairly distinct?

Operator: Your next question comes from the line of Timothy Arcuri with UBS. Please go ahead.

Speaker Change: Search and some amount of consumer generative AI.

Speaker Change: Near term midterm long term.

[Analyst]: Thanks a lot. Jensen, we hear a lot about custom ASICs. Can you kind of speak to the balance between custom ASIC and merchant GPU? We hear about some of these heterogeneous superclusters to use both GPU and ASIC. Is that something customers are planning on building, or will these infrastructures remain fairly distinct? Thanks.

Speaker Change: Advertising.

Speaker Change: Recommended.

Harlan Sur: Your next question comes from the line of Harlan sur with J P. Morgan. Please go ahead.

Speaker Change: The early days of software.

Operator: Your next question comes from the line of Harlan Sur with JPMorgan. Please go ahead.

The next the next waves coming agency AI for enterprise.

Harlan Sur: Yes. Good afternoon. Thanks for taking my question. Your next generation Blackwell Ultra set to launch in the second half of this year in line with the team's annual product cadence Jonathan can you help us understand the.

Speaker Change: Physical AI for robotics.

[Analyst]: Good afternoon. Thanks for taking my question. Your next-generation Blackwell Ultra is set to launch in the second half of this year in line with the team's annual product cadence. Jensen, can you help us understand the demand dynamics for Ultra, given that you'll still be ramping the current-generation Blackwell solutions? How do your customers and the supply chain also manage the simultaneous ramps of these two products? Is the team still on track to execute Blackwell Ultra in the second half of this year?

Speaker Change: And Ah.

Speaker Change: Software and AI is.

Speaker Change: As a different regions buildout.

Speaker Change: They're they're AI for their own ecosystems, and so each one of these are barely off the ground and we can see them. We can see them because obviously, we're in the center of much of this development and we can see great activity happening in all these different places.

Jensen Huang: Well, we build very different things than ASICs, in some ways, completely different in some areas we intersect. We're different in several ways. One, NVIDIA's architecture is general, you know, whether you've optimized for autoregressive models or diffusion-based models or vision-based models or multimodal models or text models, we're great at all of it. We're great at all of it because our software stack is so, our architecture is flexible, our software stack ecosystem is so rich that we're the initial target of, you know, most exciting innovations and algorithms. And so, by definition, we're much, much more general than narrow.

Jensen Huang: We build very different things than ASICs. In some ways, completely different. In some areas, we intersect. We're different in several ways. One, NVIDIA's architecture is general. You know, whether you've optimized for autoregressive models or diffusion-based models or vision-based models or multimodal models or text models, we're great at all of it. We're great at all of it because our software stack is so, our architecture is flexible. Our software stack ecosystem is so rich that we're the initial target of most exciting innovations and algorithms. By definition, we're much, much more general than narrow. We're also really good from the end-to-end, from data processing, the curation of the training data, to the training of the data, of course, to reinforcement learning used in post-training, all the way to inference with test time scaling. You know we're general. We're end-to-end. We're everywhere.

Harlan Sur: Demand dynamics for OXXO, given that youll still be ramping.

Speaker Change: The current generation Blackwell solutions, how do your customers in the supply chain and also manage the simultaneous ramps of these two products and is the team still on track to execute blackboard ultra in the second half of this year.

Speaker Change: This will happen near term midterm long term.

Harlan Sur: Yes.

Your next question comes from the line of Harlan sur with J P. Morgan. Please go ahead.

Harlan Sur: Blackwell Ultra is second half.

Jensen Huang: Yes. Blackwell Ultra is second half. As you know, the first Blackwell was we had a hiccup that probably cost us a couple of months. We're fully recovered, of course. The team did an amazing job recovering, and all of our supply chain partners and just so many people helped us recover at the speed of light. Now we've successfully ramped production of Blackwell, but that doesn't stop the next train. The next train is on an annual rhythm. Blackwell Ultra with new networking, new memories, and of course, new processors, and all of that is coming online. We've been working with all of our partners and customers laying this out. They have all of the necessary information, and we'll work with everybody to do the proper transition. This time between Blackwell and Blackwell Ultra, the system architecture is exactly the same.

Harlan Sur:

Harlan Sur: As you know.

First Blackwell was.

Harlan Sur: We had a hiccup.

Harlan Sur: Probably cost us a couple of months.

Harlan Sur: Were fully recovered of course.

Harlan Sur: The team did an amazing job recovery and all of our supply chain partners.

Harlan Sur: Demand dynamics for ultra given that youll still be ramping.

Jensen Huang: We're also really good from the end-to-end, from data processing, the curation of the training data. to the training of the data, of course, to reinforcement learning used in post-training all the way to inference with test-time scaling. So we're general, we're end-to-end, and we're everywhere. And because we're not in just one cloud, in every cloud we could be on-prem, we could be in a robot. Our architecture is much more accessible and a great target. Initial target for anybody who's starting up a new company. And so we're everywhere. And then the third thing I would say is that our performance and our rhythm is so incredibly fast.

Harlan Sur: So many people helped us.

Harlan Sur: The current generation Blackwell solutions, how do your customers in the supply chain and also manage the simultaneous ramps of these two products and is the team still on track to execute block will occur in the second half of this year.

Harlan Sur: Recover at the speed of light and so now we have successfully ramped production of Blackhawk.

Speaker Change: Stop the next trains next trimas.

Speaker Change: On an annual rhythm and Blackwell ultra.

Speaker Change: With new networking, new new memories and of course, new new processors in them.

Harlan Sur: Yes.

Harlan Sur: Blackwell Ultra is second half.

Speaker Change: All of that is coming online.

Harlan Sur: As you know.

Speaker Change: We've.

Harlan Sur: The first black well was we had a hiccup.

Speaker Change: Ben Ben.

Speaker Change: Working with all of our partners and customers.

Jensen Huang: Because we're not in just one cloud, we're in every cloud. We could be on-prem. We could be in a robot. Our architecture is much more accessible and a great initial target for anybody who's starting up a new company. We're everywhere. The third thing I would say is that our performance and our rhythm is so incredibly fast. Remember that these data centers are always fixed in size. They're fixed in size or they're fixed in power. If our performance per watt is anywhere from 2X to 4X to 8X, which is not unusual, it translates directly to revenues. If you have a 100-megawatt data center, if the performance or the throughput in that 100-megawatt or that gigawatt data center is four times or eight times higher, your revenues for that gigawatt data center are eight times higher.

Harlan Sur: That probably cost us a couple of months will.

Speaker Change: Laying this out.

Harlan Sur: We will fully recovered of course.

Speaker Change: They have all of the necessary information and we'll work with everybody to do the proper transition.

Harlan Sur: The team did an amazing job recovery and all of our supply chain partners.

Speaker Change: At this time between Bleichwehl Blackwell Ultra system architecture is exactly the same it's a lot harder going from Harper to Blackwell, because we went from an enviable MP linked ate system to a <unk> 72 based system.

Harlan Sur: So many people helped us.

Harlan Sur: Recover at the speed of light and so now we have successfully ramped production of Black Hawk.

Harlan Sur: Stop the next trains next trimas.

Harlan Sur: On an annual rhythm and Blackwell ultra.

Speaker Change: The chassis the architecture of the system the hardware the power delivery all of that has to change.

Harlan Sur: New networking new.

Jensen Huang: Remember that these data centers are always fixed in size. They're fixed in size or they're fixed in power. And if our performance per watt. is anywhere from 2x to 4x to 8x, which is not unusual. It translates directly to revenues. And so if you have a 100 megawatt data center, if the performance or the throughput in that 100 megawatt or that gigawatt data center is four times or eight times higher, your revenues for that gigawatt data center is eight times higher. and the reason that it's so different than data centers of the past. because AI factories are directly monetizable through its tokens generated.

Harlan Sur: New memories and of course, new new processors in them.

Harlan Sur: All of that is coming online.

Speaker Change: This was quite a challenging transition, but the next transition will slot right in Bridgeport, Blackwell ultra low slot right here.

Harlan Sur: We've.

Harlan Sur: Ben Ben.

Harlan Sur: Working with all of you.

Harlan Sur: All of the necessary information.

Speaker Change: We've also.

Speaker Change: Already revealed in and it's been working very closely with all of our partners on the click after that and the click after that is called Vera Rubin.

Harlan Sur: We will work with everybody to do that.

Harlan Sur: The proper transition.

Harlan Sur: At this time between Blackwell Blackwell Ultra system architecture is exactly the same it's a lot harder going from Harper to Blackwell, because we went from an enviable and fueling eight system to a <unk> 72 based system.

Speaker Change: All of our partners are are getting up to speed on on on the transition of that and so preparing for that transition and again, we're going to provide a big big huge step up and so come the GCC and I'll talk to you about about Blackwell Ultra Vera Rubin.

Jensen Huang: The reason that is so different than data centers of the past is because AI factories are directly monetizable through its tokens generated. The token throughput of our architecture being so incredibly fast is just incredibly valuable to all of the companies that are building these things for revenue generation reasons and capturing the fast ROIs. I think the third reason is performance. The last thing that I would say is the software stack is incredibly hard. Building an ASIC is no different than what we do. We have to build a new architecture. The ecosystem that sits on top of our architecture is 10 times more complex today than it was two years ago. That is fairly obvious because the amount of software that the world is building on top of our architecture is growing exponentially. AI is advancing very quickly.

Harlan Sur: The chassis the architecture of the system the hardware the power delivery all of that has to change. This was a this was quite a challenging transition, but the next transition will slot right in various blood Blackwell ultra low slot Ryan.

Speaker Change: And then show you show you whats the one click after that.

Jensen Huang: And so the token throughput of our architecture being so incredibly fast is just incredibly valuable to all of the companies that are building these things for revenue generation reasons and capturing the fast ROI. And so I think the third reason is performance.

Speaker Change: Really really exciting new products that come to the street sweepers.

Harlan Sur: It's been working very closely with all of our partners on the click out.

Speaker Change: Your next question comes from the line of Timothy Arcuri with UBS. Please go ahead.

Operator: Your next question comes from the line of Timothy Arcuri with UBS. Please go ahead.

Timothy Arcuri: Thanks, a lot gentlemen, we hear a lot about custom asics.

Harlan Sur: And the click after that is called Vera Rubin.

Harlan Sur: And.

Timothy Arcuri: Can you kind of speak to the balance between custom ASIC and merchant GPU, we hear about some of these heterogeneous supercluster to use both GPU at Asa because that's something customers are planning on building or will these infrastructure it remains fairly distinct things.

Harlan Sur: All of our partners are getting up to speed on on the transition of that and so preparing for that transition and again, we're going to provide a big big huge step up and so come the GCC.

Jensen Huang: And then the last thing that I would say is... The software stack is incredibly hard. Building an ASIC is no different than what we do, we have to build a new architecture and the ecosystem that sits on top of our architecture 10 times more complex today than it was two years ago. and that's fairly obvious because the amount of software that the world is building on top of architecture is growing exponentially and AI is advancing very quickly. So bringing that whole ecosystem on top of multiple chips is hard.

Timothy Arcuri: Yeah.

Speaker Change: I'll, let him talk to you about about.

Timothy Arcuri: Yeah.

Timothy Arcuri: Well, we feel very different things the basics.

Speaker Change: Blackwell Ultra fair Ruben and then show you show you, what's the play capture that really really exciting new products that come to the street sweepers.

Timothy Arcuri: In some ways completely different in some areas we intercept.

Timothy Arcuri: We are different in several ways one.

Speaker Change: Three with UBS. Please go ahead.

Timothy Arcuri: And various architectures general.

Timothy Arcuri: Whether you're you've optimized for auto regressive models or diffusion based models or vision based models or more.

Jensen Huang: Bringing that whole ecosystem on top of multiple chips is hard. I would say that those four reasons. Finally, I will say this. Just because the chip is designed does not mean it gets deployed. You have seen this over and over again. There are a lot of chips that get spilled. When the time comes, a business decision has to be made. That business decision is about deploying a new engine, a new processor into a limited AI factory in size, in power, and in time. Our technology is not only more advanced, more performant, it has much, much better software capability. Very importantly, our ability to deploy is lightning fast. These things are not for the faint of heart, as everybody knows now. There are a lot of different reasons why we do well, why we win.

Speaker Change: Thanks, a lot gentlemen, we hear a lot about custom asics.

Speaker Change: Can you kind of speak to the balance between custom ASIC and merchant GPU, we hear about some of these heterogeneous supercluster to use both GPU in Asia, because that's something customers are planning on building or will these infrastructure remain fairly distinct things.

Jensen Huang: And so I would say that those four reasons, and then finally, I will say this. Just because a chip is designed doesn't mean it gets deployed. and you've seen this, you know, over and over again. There are a lot of chips that get spilled, but when the time comes, a business decision has to be made. and that business decision is about deploying a new. Engine, a new processor into a limited AI factory in size, in power, and in time. and our technology. is not only more advanced, more performant. has much much better software capability and very importantly our ability to deploy is lightning So these things are enough for the faint of heart, as everybody knows now.

Timothy Arcuri: Multimodal models or types of models worked great in all of it were greater than all of it because our software stack and so our architecture responsible our software stack is ecosystem is so rich.

Timothy Arcuri: What are the initial target of.

Speaker Change: Yeah.

Speaker Change: Most exciting innovations and algorithms.

Speaker Change: Okay.

Timothy Arcuri: So by definition, we're much much more general.

Speaker Change: In some ways completely different in some areas we intercept.

Speaker Change: Harrow where also.

Speaker Change: We are different in several ways one.

Speaker Change: Really good from the end to end from data processing deterioration as the training data too.

Speaker Change: And videos architectures general.

Speaker Change: Whether you're you've optimized for auto regressive models or diffusion based models are.

Speaker Change: Two.

Speaker Change: The training of the data of course to reinforcement learning.

Speaker Change: Houston Post training all the way to.

Speaker Change: Were created all of it.

Speaker Change: Grid on all of it because our software stack is so our architecture responsible our software stack is ecosystem.

Speaker Change: Difference with.

Speaker Change: Tough times scaling so.

Speaker Change: General where end to end.

Speaker Change: Most exciting innovations and algorithms.

Speaker Change: We're everywhere.

Speaker Change: Because we're not in just one cloud in every cloud we could be on Prem we could be.

Speaker Change: And so by definition, we're much much more general narrow we're also.

Jensen Huang: So there's a lot of different reasons why. why we do well.

Speaker Change: Robot architecture is much more accessible and a great target.

Speaker Change: Really good from the end to end from data processing deterioration as the training.

Ben Reitz: Your next question comes from the line of Ben Reitz with Malleus Research. Please go ahead. Yeah, hi, Ben Reitz is here. Hey, thanks a lot for the question. Hey, Jensen, it's a geography related question. You know, you did a great job explaining some of the demand underlying, you know, factors here on the strength, but US was up about 5 billion or so sequentially. And I think, you know, there is a concern about whether US can pick up the slack if there's regulations towards other geographies. And I was just wondering, as we go throughout the year, you know, if this kind of surge in the US continues and it's going to be, whether that's, that's okay.

Operator: Your next question comes from the line of Ben Reitzes with Malleus Research. Please go ahead.

Speaker Change: Initial target for anybody who is starting up.

Speaker Change: New company and so so.

[Analyst]: Yeah, hi, Ben Reitzes here. Hey, thanks a lot for the question. Hey, Jensen, it's a geography-related question. You did a great job explaining some of the demand underlying factors here on the strength. The U.S. was up about $5 billion or so sequentially. I think there is a concern about whether the U.S. can pick up the slack if there's regulations towards other geographies. I was just wondering, as we go throughout the year, if this kind of surge in the U.S. continues and it's going to be, whether that's OK. If that underlies your growth rate, how can you keep growing so fast with this makeshift towards the U.S.? Your guidance looks like China is probably up sequentially. I'm just wondering if you could go through that dynamic and maybe Colette can weigh in. Thanks a lot.

Speaker Change: Of course to reinforcement learning.

Speaker Change: We're everywhere and then the third thing I would say.

Speaker Change: Houston Post training all the way too.

Speaker Change: Our performance and our rhythm is so incredibly fast.

Speaker Change: The difference with with.

Speaker Change: Remember that these data centers are always fixed in size they.

Speaker Change: Tough times scaling.

Speaker Change: There are fixed in size or they're fixing power and.

Speaker Change: General where end to end.

Speaker Change: And if our performance per watt.

Speaker Change: Is anywhere from two onex to Forex to apex, which is not unusual.

Speaker Change: In every cloud we could be on Prem we could be.

Speaker Change: It translates directly to revenues.

Speaker Change: Robot our architecture is.

Speaker Change: And so if you have a 100 megawatt data center.

Speaker Change: Much more accessible and a great target.

Speaker Change: If the performance or the throughput and a 100 megawatts or the gigawatt datacenter is four times, our eight times higher your revenues for that gigawatt datacenter is eight times higher.

Jensen Huang: And if that, you know, underlies your growth rate, how can you keep growing so fast with this mix shift towards the US? Your guidance looks like China is probably up sequentially.

Speaker Change: Initial target for anybody who is starting up a new company and so so.

Speaker Change: Performance in our rhythm is so incredibly fast.

Speaker Change: And the reason the reason that is so different data.

Speaker Change: Remember that these data centers are always fixed in size they.

Jensen Huang: So just wondering if you could go through that dynamic and maybe Collette can weigh in. Thanks a lot. China is approximately the same percentage as Q4. and as previous quarter. it's it's about half of what it was before the export but it's approximately. With respect to geography...

Speaker Change: Data centers of the past is because.

Speaker Change: They are fixed in size or they're fixing power.

Speaker Change: Factories are directly monetize <unk> through its tokens generated.

Speaker Change: If our performance per watt.

Speaker Change: Is anywhere from.

Speaker Change: And so the token throughput of our architecture being so incredibly fast.

Jensen Huang: China is approximately the same % as Q4 and as previous quarters. It's about half of what it was before the export control, but it's approximately the same in %. With respect to geographies, the takeaway is that AI is software. It's modern software. It's incredibly modern software, but it's modern software. AI has gone mainstream. AI is used in delivery services everywhere, shopping services everywhere. If you were to buy a quarter of milk and it's delivered to you, AI was involved. Almost everything that a consumer service provides, AI is at the core of it. Every student will use AI as a tutor. Healthcare services use AI. Financial services use AI. No fintech company will not use AI. Every fintech company will. Climate tech companies use AI. Mineral discovery now uses AI. Every higher education, every university uses AI.

Speaker Change: Just incredibly valuable to all of the companies that are building these things for revenue generation reasons.

Speaker Change: It translates directly to revenues.

Speaker Change: And so if you have a 100 megawatt data center.

Speaker Change: The performance or the throughput and the 100 megawatt or that gigawatt datacenter is four times, our eight times higher your revenues for that.

Speaker Change: <unk> is a fast rois.

Speaker Change: So I think the the <unk>.

Speaker Change: Third reason is.

Speaker Change: Performance.

Speaker Change: And then and then the last thing that I would say is.

Jensen Huang: The takeaway is that AI is software. It's modern software. It's incredible modern software, but it's modern software. and AI has gone mainstream. AI is used in delivery services everywhere, shopping services everywhere. Steward by A quarter of milk is delivered to you, AI was involved. and so almost everything that a consumer service provides AIs at the core of. every every Student will use AI as a tutor. Healthcare services use AI. Financial services use AI. No fintech company will not use AI. Every fintech company will. Climate tech company use AI. Mineral Discovery now uses AI. The number of the number of every higher education, every university uses AI.

Speaker Change: The software stack is incredibly hard.

Speaker Change: This is the reason that is so different.

Speaker Change: Building in ASIC is no different than what we do to build a new architecture and the ecosystem that sits on top of our architecture.

Speaker Change: Data centers of the past is because.

Speaker Change: Factories are directly monetize <unk> through its tokens generated.

Speaker Change: Is 10 times more complex today than it was two years ago.

Speaker Change: And so the token throughput of our architecture being so incredibly fast.

Speaker Change: And that's fairly obvious because the amount of software that the world is building on top of architecture is growing exponentially and AI is advancing very quickly so bringing that whole ecosystem on top of multiple multiple chips is hard and so I would I would say that those four reasons and then finally I will say this.

Speaker Change: That are building these things for revenue generation reasons and capturing to fast rois.

Speaker Change: So I think the.

Speaker Change: The third reason is performance.

Speaker Change: And then and then.

Speaker Change: The last thing that I would say is the.

Speaker Change: The software.

Speaker Change: Just because the chip is designed doesn't mean it gets deployed.

Speaker Change: No different than what we do to build a new architecture.

Speaker Change: And you've seen this over and over again there are a lot of chips that gets built on but when the time comps a business decision has to be made and that business decision is about deploying a new.

Speaker Change: And the ecosystem that sits on top of our architecture.

Speaker Change: Yes.

Speaker Change: And that's fairly obvious because the amount of software that the world is building on top of architecture is growing exponentially.

Speaker Change: Engine that new processor into a limited AI factory inside an hour and in time and our our technology.

Jensen Huang: And so I think it is fairly safe to say that that AI has gone mainstream, that it's been integrated into every application. and our hope is that, of course.

Speaker Change: <unk> is advancing very quickly so bringing that whole ecosystem on top of multiple multiple chips is hard and so I would.

Jensen Huang: I think it is fairly safe to say that AI has gone mainstream and that it's being integrated into every application. Our hope is that, of course, the technology continues to advance safely and advance in a helpful way to society. With that, I do believe that we're at the beginning of this new transition. What I mean by that in the beginning is, remember, behind us has been decades of data centers and decades of computers that have been built. They've been built for a world of hand coding and general-purpose computing and CPUs and so on and so forth. Going forward, I think it's fairly safe to say that that world is going to be almost all software will be infused with AI. All software and all services will be based on, ultimately, based on machine learning.

Speaker Change: As you know not not only more at more advanced more performance.

Speaker Change: Just because the chip is designed doesn't mean it gets deployed.

Speaker Change: Much much better.

Speaker Change: Better software capability and very importantly, our ability to deploy is.

Jensen Huang: Technology continues to advance safely and advance in a helpful way to society and with that, you know, we're I do believe that we're at the beginning of this new transition and what I mean by that in the beginning is remember behind us has been Decades of data centers and decades of computers that have been built, and they've been built for a world of hand coding and general purpose computing and CPUs and so on and so forth. And going forward, I think it's fairly safe to say that the world is going to be, almost all software will be infused with AI.

Speaker Change: And you've seen this over and over again there are a lot of chips that gets built on but when the time comps a business decision has to be made and that business decision.

Speaker Change: Lightning fast and so these things are enough for the fate of heart as everybody knows now and so there's a lot of different reasons why.

Speaker Change: Why are why did we do well why we win.

Speaker Change: Processor into a limited AI factory inside an hour and in time and our our technology.

Speaker Change: Your next question comes from the line of Ben Reed with Melius Research. Please go ahead.

Operator: Your next question comes from the line of Ben Reitzes with Malleus Research. Please go ahead.

Ben Reitzes: Yeah, Hi, Ben Reitzes here, Hey, Thanks, a lot for the question, Hey, Jensen, it's a geography related question.

Speaker Change: As you know not not only more and more advanced more performance.

[Analyst]: Yeah, hi, Ben Reitzes here. Hey, thanks a lot for the question. Hey, Jensen, it's a geography-related question. You did a great job explaining some of the demand underlying factors here on the strength. The U.S. was up about $5 billion or so sequentially. I think there is a concern about whether the U.S. can pick up the slack if there's regulations towards other geographies. I was just wondering, as we go throughout the year, if this kind of surge in the U.S. continues and it's going to be, whether that's OK, and if that underlies your growth rate, how can you keep growing so fast with this makeshift towards the U.S.? Your guidance looks like China is probably up sequentially. I'm just wondering if you could go through that dynamic and maybe Colette can weigh in. Thanks a lot.

Speaker Change: You did a great job explaining some of the day.

Speaker Change: <unk> underlying factors here on the strength, but U S was up about $5 billion or so sequentially and and I think there is a concern about whether U S can pick up the slack if there's regulations sports other geographies.

Speaker Change: Our software capability and very importantly, our ability to deploy is.

Speaker Change: Lightning fast and so these things are enough for the fate of heart as everybody knows now and so there's a lot of different reasons why.

Jensen Huang: All software and all services will be based on, ultimately based on machine learning. The data flywheel is gonna be part of improving software and services. and that the future computers will be accelerated. The future computers will be based on AI. And we're really two years into that journey. and in modernizing computers that have taken decades to build out. So I'm fairly sure that we're in the beginning of this new era.

Speaker Change: Why are why did we do well.

Speaker Change: And I was just wondering as we go throughout the year. If this kind of surge in the U S continues in it and it's going to be.

Jensen Huang: The data flywheel is going to be part of improving software and services. The future computers will be accelerated. The future computers will be based on AI. We're really two years into that journey and in modernizing computers that have taken decades to build out. I'm fairly sure that we're in the beginning of this new era. Lastly, no technology has ever had the opportunity to address a larger part of the world's GDP than AI. No software tool ever has. This is now a software tool that can address a much larger part of the world's GDP more than any time in history. The way we think about growth and the way we think about whether something is big or small has to be in the context of that. When you take a step back and look at it from that perspective, we're really just in the beginnings.

Speaker Change: Question comes from the line of Ben Reed with Melius Research. Please go ahead.

Speaker Change: Whether that's that's okay, and if that you know.

Ben Reitzes: Yeah, Hi, Ben Reitzes here, Hey, Thanks, a lot for the question, Hey, Jensen, it's a geography related question.

Speaker Change: Underlies your growth rate, how can you keep growing so fast with this mix shift towards the U S.

Ben Reitzes: You did a great job explaining some of the.

Speaker Change: <unk> it looks like China.

Ben Reitzes: Demand underlying factors here and on the strength, but U S was up about $5 billion or so sequentially.

Speaker Change: It's probably up sequentially. So I'm just wondering if you could go through that dynamic and maybe collect can weigh in and thanks a lot.

Jensen Huang: And then lastly. No technology has ever had the opportunity to address a larger part of the world's GDP than AI. No, no software tool ever have. and so this is now a software tool that can address a much larger part of the world's GDP more than any time in history. And so the way we think about. Growth and the way we think about whether something is big or small has to be in the context.

Ben Reitzes: And I think there is a concern about whether U S can pick up the slack if there is regular.

Speaker Change: China is approximately the same percentage as Q4.

Speaker Change: And as in.

Ben Reitzes: As we go throughout the year, if this kind of surge in the U S.

Speaker Change: Previous quarters.

Speaker Change: It's it's about half of what it was before the export control.

Ben Reitzes: S continues in it and it's going to be.

Speaker Change: But it's approximately the same percentage.

Speaker Change: With respect to with respect to geographies.

Ben Reitzes: That's that's okay and if that.

Ben Reitzes: Underlies your growth rate, how can you keep growing so fast with this mix shift towards the U S.

Speaker Change: The takeaway is.

Aaron Rakers: and and when you when you take a step back and look at it from that perspective we're really just in the Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead. Aaron, your line is open.

Speaker Change: AI is software.

Speaker Change: It's modern software, it's incredible modern software, but its modern software and AI has gone to the street.

Ben Reitzes: Guidance looks like China.

Ben Reitzes: It was probably up sequentially. So I'm just wondering if.

Operator: Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead. Aaron, your line is open. Your next question comes from Mark Lipacis with Evercore ISI. Please go ahead.

Speaker Change: AI is Houston.

Ben Reitzes: China is approximately the same percentage as Q4.

Speaker Change: Delivery services everywhere shopping services everywhere.

Speaker Change: You know if you were to buy.

Ben Reitzes: And as in.

Ben Reitzes: Previous quarters.

Speaker Change: You know a quarter of milk is deliver to your AI was adult.

Ben Reitzes: It's it's about half of what it was before the exports.

Speaker Change: So almost everything that a consumer service provides.

Marc Lupicis: Your next question comes from Marc Lupicis with Evercore ISI. Please go ahead. Hi, that's Marc Lopatis. Thanks for taking the question. I had a clarification and a question. Colette, for the clarification, did you say that enterprise within the data center grew 2x year-on-year for the January quarter? And if so, would that make it faster going than the hyperscalers? And then, Jensen, for you, the question, hyperscalers are the biggest purchasers of your solutions, but they buy equipment for both internal and external workloads, external workloads being cloud services that enterprises use. So the question is, can you give us a sense of how that hyperscaler spend splits between that external workload and internal?

Speaker Change: At the core of it right.

Ben Reitzes: With respect to with respect to geographies.

Speaker Change: Every every.

Speaker Change: Students will use AI as a tutor.

[Analyst]: Hi, that's Mark Lipacis. Thanks for taking the question. I had a clarification and a question. Colette, for the clarification, did you say that enterprise within the data center grew 2X year on year for the January quarter? If so, does that make it faster growing than the hyperscalers? Jensen, for you, the question, hyperscalers are the biggest purchasers of your solutions, but they buy equipment for both internal and external workloads, external workloads being cloud services that enterprises use. The question is, can you give us a sense of how that hyperscaler spend splits between that external workload and internal? As these new AI workloads and applications come up, would you expect enterprises to become a larger part of that consumption mix? Does that impact how you develop your service, your ecosystem? Thank you.

Ben Reitzes: The takeaway is.

Ben Reitzes: AI is software.

Speaker Change: Our healthcare services use AI financial services use AI.

Ben Reitzes: It's modern software is incredible modern software, but its modern software and AI has gone.

Speaker Change: No Fintech company will not use AI every fintech company will.

Climate Tech company use AI, a mineral discovery now uses AI. The number the number of every all your education every University uses AI and so I think it's fairly safe to say that he.

Ben Reitzes: Services everywhere shopping services everywhere.

Ben Reitzes: You know if you were to buy.

Ben Reitzes: You know a quarter of milk is delivered to your AI was adult.

Ben Reitzes: And so almost everything that a consumer service provides.

Jensen Huang: I think it is fairly safe to say that AI has gone mainstream and that it's being integrated into every application. Our hope is that, of course, the technology continues to advance safely and advance in a helpful way to society. I do believe that we're at the beginning of this new transition. What I mean by that in the beginning is, remember, behind us has been decades of data centers and decades of computers that have been built. They've been built for a world of hand coding and general-purpose computing and CPUs and so on and so forth. Going forward, I think it's fairly safe to say that that world is going to be almost all software will be infused with AI. All software and all services will be based on, ultimately, based on machine learning. The data flywheel is going to be part of improving software and services.

Speaker Change: He has gone mainstream.

Speaker Change: <unk> integrated into every application.

Ben Reitzes: Every.

Speaker Change: Yeah.

Ben Reitzes: Students will use AI as a tutor.

Speaker Change: And our hope is that that of course the.

Ben Reitzes: Our healthcare services use AI financial services use AI.

Speaker Change: This technology continues to advance.

Colette Kress: And as these new AI workloads and applications come up, would you expect enterprises to become a larger part of that consumption mix and does that impact how you develop your service, your ecosystem? Thank you.

Speaker Change: Safely in advance.

Speaker Change: In a in a helpful way to to our society and with that we're in.

No Fintech company will not use AI every fintech company will.

Ben Reitzes:

Speaker Change: I do believe that we're at the beginning of this this new transition.

Ben Reitzes: AI the number the number of <unk>.

Speaker Change: And what I mean by that in the beginning is is it remember behind us.

Ben Reitzes: Every all your education every University uses AI and so I think it's fairly safe to say that.

Colette Kress: Sure, thanks for the question regarding our enterprise business. Yes, it grew to X and very similar to what we were seeing with our large CSPs. Keep in mind, these are both important areas to understand. Working with the CSPs can be working on large language models, can be working on inference on their own work, but keep in mind that is also where the enterprises are surfacing. Your enterprises are both with your CSPs as well as in terms of building on their own. They're both, correct, growing quite well. The CSPs are about half of our business. and, and, um...

Colette Kress: Sure. Thanks for the question regarding our enterprise business. Yes, it grew 2X and very similar to what we were seeing with our large CSPs. Keep in mind, these are both important areas to understand. Working with the CSPs can be working on large language models, can be working on inference on their own work. Keep in mind, that is also where the enterprises are surfacing. Your enterprises are both with your CSPs as well as in terms of building on their own. They're both growing quite well.

Speaker Change: It has been.

Speaker Change: Decades of data centers and decades of computers that had been built and they've been built for a world of hand coding and general purpose computing and.

Ben Reitzes: He has got midstream.

Ben Reitzes: It's being integrated into every application.

Ben Reitzes: And.

Yes.

Ben Reitzes: And our hope is that that of course.

Speaker Change: Cpus and so on so forth.

Ben Reitzes: Technology continues to advance.

Speaker Change: And going forward I think it's fairly safe to say that the world is going to be almost all software would be infused with AI all software and all services will be based on ultimately based on machine learning and the data flywheel, it's gonna be part of improving software and services.

Ben Reitzes: Safely in advance.

Ben Reitzes: And a helpful way to to our society and with that we're in.

Ben Reitzes: Sure.

Ben Reitzes: I do believe that we're at the beginning.

This new transition.

Ben Reitzes: And what I mean by that in the beginning is is remember behind us.

Speaker Change: And then the future computers will be accelerated future computers will be based on AI and we're really.

Jensen Huang: The future computers will be accelerated. The future computers will be based on AI. We're really two years into that journey and in modernizing computers that have taken decades to build out. I'm fairly sure that we're in the beginning of this new era. Lastly, no technology has ever had the opportunity to address a larger part of the world's GDP than AI. No software tool ever has. This is now a software tool that can address a much larger part of the world's GDP more than any time in history. The way we think about growth and the way we think about whether something is big or small has to be in the context of that. When you take a step back and look at it from that perspective, we're really just in the beginnings.

Jensen Huang: The CSPs are about half of our business. The CSPs have internal consumption and external consumption, as you say. We're, of course, used for internal consumption. We work very closely with all of them to optimize workloads that are internal to them because they have a large infrastructure of NVIDIA gear that they could take advantage of. The fact that we could be used for AI on the one hand, video processing on the other hand, data processing like Spark, we're fungible. The useful life of our infrastructure is much better. If the useful life is much longer, then the TCO is also lower. The second part is, how do we see the growth of enterprise or not CSPs, if you will, going forward? The answer is, I believe long term, it is by far larger.

Ben Reitzes: It has been.

Ben Reitzes: Decades.

Speaker Change: Two years into that journey.

Colette Kress: CSPs have internal consumption and external consumption, as you say. We're using, of course, used for internal consumption. We work very closely with. all of them to optimize workloads that are internal to them. because they have a large infrastructure of NVIDIA gear that they could take advantage of. And the fact that we could be used for AI on the one hand. Video Processing, on the other hand, data processing like Spark were fungible. So the useful life of our infrastructure is much better. If the useful life is much longer, then the TCO is also longer. and, and so The second part is...

Speaker Change: In in modernizing computers that have taken decades to build out.

Ben Reitzes: It's been built for a world of hand coding and general purpose computing.

Speaker Change: So I'm fairly sure that were in the beginning of this new era.

Ben Reitzes: Cpus and so on and so forth and.

Ben Reitzes: And going forward I think it's fairly safe to say that that world is going to be almost all software would be infused with AI.

And then lastly.

Hum.

Speaker Change: No no technology has ever had the opportunity to address a larger part of the world's GDP and AI.

Speaker Change:

Ben Reitzes: Totally based on machine learning and the data flywheel, it's going to be part of improving software and services.

Speaker Change: Oh, no software tool ever house.

Speaker Change: And so this is now a software tool that can address a much larger part of the world's GDP more than any time in history.

Ben Reitzes: And then the future computers will be accelerated the future computers will be based on AI and we're really.

Speaker Change: So the way we think about.

Ben Reitzes: Two years into that journey.

Speaker Change: Growth in the way, we think about whether something is big or small has to be in the context of that.

Ben Reitzes: In in modernizing computers that have taken decades to build out.

Speaker Change: And when you when you take a step back and look at it from that perspective, we're really just in the beginnings.

Ben Reitzes: So I'm fairly sure that were in the beginning of this new era.

Ben Reitzes: And then lastly.

Ben Reitzes:

Speaker Change: Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead.

Ben Reitzes: A larger part of the world's GDP is an AI.

Jensen Huang: How do we see the growth of enterprise, or not CSPs, if you will, going forward? And the answer is, I believe long-term, it is by far larger. And the reason for that is because if you look at the computer industry today, and what is not served by the computer industry is largely industrial. So let me give you an example. When we say enterprise, and let's say, let's use the car company as an example because they make both soft things and hard. And so in the case of a car company... employees would be what we call enterprise and agentic AI and software planning systems and tools and we have some really exciting things to share with you guys at GTC.

Ben Reitzes: Oh, no software tool ever house.

Speaker Change: Aaron Your line is open.

Ben Reitzes: And so this is now a software tool that can address a much larger part of the world's GDP more than any time in history.

Jensen Huang: The reason for that is because if you look at the computer industry today, and what is not served by the computer industry is largely industrial. Let me give you an example. When we say enterprise, and let's say let's use a car company as an example because they make both soft things and hard things. In the case of a car company, the employees would be what we call enterprise. Agentic AI and software planning systems and tools, and we have some really exciting things to share with you guys at GTC. Those agentic systems are for employees to make employees more productive, to design, to market, to plan, to operate their company. That's agentic AIs. On the other hand, the cars that they manufacture also need AI. They need an AI system that trains the cars, treats this entire giant fleet of cars.

Speaker Change: Your next question comes from Mark <unk> with Evercore ISI. Please go ahead.

Ben Reitzes: You about whether something is big or small has to be in the context of that.

Speaker Change: Hi.

Speaker Change: Martin deposits. Thanks for taking my question I had a clarification and a question Colette up for the clarification did you say the enterprise within the data center.

Ben Reitzes: And when you when you take a step back and look at it from that perspective, we're really just in the beginnings.

Speaker Change: Your next question comes from the line of Erin.

Speaker Change: <unk> grew two X year on year for the January quarter, and if so does that.

Speaker Change: Would that make it the faster growing than the hyper scaler and then Jason for you the question <unk>.

Speaker Change: Yeah.

Aaron: Aaron Your line is open.

Speaker Change: Scalar has had the biggest purchasers of your solutions, but they buy equipment for both internal and external workloads external workloads being cloud services that enterprises use. So the question is can you give us a sense of how that hyperscale expense splits between that external workload and internal and.

Jensen Huang: Those agentic systems are for employees to make employees more productive. to design, to market, to plan, to operate their company. On the other hand, Cars that they manufacture also need AI. They need an AI system that trains the cars, treats this entire giant fleet of cars. And today, there's a billion cars on the road. Someday, there'll be a billion cars on the road, and every single one of those cars will be robotic cars. And they'll all be collecting data, and we'll be improving them using an AI factory. Whereas they have a car factory today, in the future, they'll have a car factory and an AI.

Speaker Change: Your next question comes from Mark <unk> with Evercore ISI. Please go ahead.

Aaron: Hi.

Aaron: Martin deposits. Thanks for taking my question I had a clarification and a question Colette up for the clarification did you say the enterprise within the data center.

Speaker Change: And as.

Speaker Change: As these new AI.

Speaker Change: Workloads and applications come up would you would expect enterprises to become a larger part of that consumption.

Aaron: Glue two X year on year for the January quarter, and if so does that.

Jensen Huang: You know today there's a billion cars on the road. Someday, there'll be a billion cars on the road. Every single one of those cars will be robotic cars. They'll all be collecting data. We'll be improving them using an AI factory. Whereas they have a car factory today, in the future, they'll have a car factory and an AI factory. Inside the car itself is a robotic system. As you can see, there are three computers involved. There's the computer that helps the people. There's the computer that builds the AI for the machineries. It could be, of course, it could be a tractor. It could be a lawnmower. It could be a human or a robot that's being developed today. It could be a building. It could be a warehouse. These physical systems require a new type of AI we call physical AI.

Aaron: Yeah.

Speaker Change: And does that impact how you do.

Speaker Change: It all up your service your ecosystem. Thank you.

Aaron: For you the question <unk>.

Speaker Change: Hyperscale or some of the biggest purchasers of your solutions, but they buy equipment for both internal and external workloads external workloads being cloud services that enterprises use. So the question is can you give us a sense.

Speaker Change: Sure. Thanks for the question regarding our enterprise business, yes, it crude to act and very similar to what we were seeing with our large CSP is keep in mind. These are both important areas to understand working with the CSP can be working on large language models.

Colette Kress: Sure. Thanks for the question regarding our enterprise business. Yes, it grew 2X and very similar to what we were seeing with our large cloud service providers. Keep in mind, these are both important areas to understand. Working with the cloud service providers can be working on large language models, can be working on inference on their own work. Keep in mind, that is also where the enterprises are surfacing. Your enterprises are both with your cloud service providers as well as in terms of building on their own. They're both growing quite well.

Jensen Huang: and then inside the car itself is a robotic. So, as you can see, there are three computers involved. and there's the computer that helps the people, there's the computer that builds the AI for it. Machineries, it could be, of course, it could be a tractor, it could be a lawnmower, it could be a human or robot that's being developed today, it could be a building, it could be a warehouse. These physical systems require a new type of AI we call physical AI. We can't just understand the meaning of words and languages, but they have to understand the meaning of the world.

Speaker Change: In that external workload, an internal and an as is.

Speaker Change: And are you working on inference on their own work, but keep in mind that is also where the enterprises are surfacing Youre enterprises are both with your C. S piece as well as in terms of building on their own they are both correct growing quite quite well.

Speaker Change: These new AI workloads.

Speaker Change: Workloads and applications come up would you would expect enterprises to become a larger part of that can assumption on mix and does that impact how you.

Speaker Change: The csp's or about half of our business.

Speaker Change: Sure. Thanks for the question regarding our enterprise business, Yes. It grew two acts and very similar to what we were seeing with our large CSP is keep in mind. These are both important areas to understand.

Speaker Change: And.

Jensen Huang: They can't just understand the meaning of words and languages, but they have to understand the meaning of the world, friction and inertia and object permanence and cause and effect, and all of those types of things that are common sense to you and I. AIs have to go learn those physical effects. We call that physical AI. That whole part of using agentic AI to revolutionize the way we work inside companies is just starting. This is now the beginning of the agentic AI era. You hear a lot of people talking about it. We've got some really great things going on. There is the physical AI after that, and then there are robotic systems after that. These three computers are all brand new. My sense is that long term, this will be by far the larger of them all, which kind of makes sense.

Speaker Change: The CSP to internal consumption.

Speaker Change: And external consumption as you say.

Jensen Huang: friction and inertia object permanence and cause and effect and all of those type of things that are common sense to you and I But you know AI's have to go learn those physical effects we call that physical AI. That whole part of using agentic AI to revolutionize the way we work inside companies, that's just starting. This is now the beginning of the agentic AI era and you hear a lot of people talking about it and we've got some really great things going on. And then there's the physical AI after that and then there are robotic systems after that.

Speaker Change: And.

Speaker Change: We are using of course used for internal consumption.

Speaker Change: We work very closely with.

Speaker Change: This model can be working on inference on their own work, but keep in mind that is also where the enterprises are surfacing Youre enterprises are both with your <unk> as well as in terms of building on their own. They are both correct growing quite quite well.

Speaker Change: All of them to optimize.

Speaker Change: Workloads that are internal to them.

Speaker Change: Because they have a large infrastructure of <unk>.

Speaker Change: Video gear that they could take advantage of.

Speaker Change: And the fact that we.

Speaker Change: It could be used for AI on the one hand.

Speaker Change: Video processing on the other hand data processing like spark.

Speaker Change: Half of our business.

Speaker Change:

Speaker Change: And and.

Speaker Change: We are fungible and so so the.

Speaker Change: The CSP to internal consumption.

Speaker Change: External consumption as you say.

Speaker Change: The useful life of our of our infrastructure is much better.

Speaker Change: And.

Jensen Huang: And so these three computers are all brand new and my sense is that long-term this will be by far the larger of them all, which kind of makes sense. You know, the world's GDP is represented by either heavy industries or industrials and companies that are providing for those.

Speaker Change: We're using of course used for internal consumption.

Speaker Change: If the useful life is much longer than the Tcl was is also lower.

Speaker Change: We work very closely with.

Speaker Change: All of them to optimize.

Speaker Change: And and so.

Jensen Huang: The world's GDP is represented by either heavy industries or industrials and companies that are providing for those.

Speaker Change: Workloads that are internal to them.

Speaker Change:

Speaker Change: The second part.

Speaker Change: Because they have a large infrastructure.

Speaker Change: Yes.

Speaker Change: How do we see the growth of enterprise not Csp's. If you will are.

Speaker Change: Nvidia gear that they could take it.

Speaker Change: For AI on the one hand.

Speaker Change: Going forward and if the answer is I believe long term it is by far larger.

Aaron Rakers: Your next question comes from the line of Erin Rickers with Wells Fargo. Please go ahead. Yeah thanks for letting me back in.

Operator: Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead.

Speaker Change: Video processing on the other hand.

Speaker Change: And the reason for that is because if you look at the computer industry today.

Speaker Change: Data processing like spark.

[Analyst]: Yeah, thanks for letting me back in. Jensen, I'm curious, as we now approach the two-year anniversary of really the Hopper inflection that you saw in 2023 in Gen AI in general, and we think about the roadmap you have in front of us, how do you think about the infrastructure that's been deployed from a replacement cycle perspective and whether you know if it's GB200 or if it's the Rubin cycle where we start to see maybe some refresh opportunity? I'm just curious to how you look at that.

Speaker Change: And what is not served by the computer industry is largely industrial so let me give you an example.

Speaker Change:

Jensen Huang: Jensen, I'm curious as we now approach the two-year anniversary of really the hopper inflection that you saw in 2023 and in Gen AI in general, and we think about the roadmap you have in front of us, how do you think about the infrastructure that's been deployed from a replacement cycle perspective and whether you know if it's you know GB300 or if it's the Rubin cycle where we start to see maybe some refresh opportunity. I'm just curious to how you look at that. I appreciate it. First of all, people are still using Voltas, and Pascals, and Amperes.

Speaker Change: We are fungible.

Speaker Change: So the.

Speaker Change: The useful life of our of our infrastructure is much better.

Speaker Change: When we say enterprise.

Speaker Change: And let's say, let's use the car company as an example, because they make both soft and hard things.

Speaker Change: If the useful life is much longer than the <unk> is also lower.

Speaker Change: And.

Speaker Change: And so in the case of a car company.

Speaker Change: The second part.

Speaker Change: The employees will be what we call enterprise and a gentle AI.

Speaker Change: Is.

Speaker Change: How do we see the growth of enterprise not Csp's, if you will.

Speaker Change: <unk> software planning systems and tools and we have some really exciting things to share with you guys at GTC built a gentex systems are for employees to make employees more productive.

Speaker Change: Going forward and if the answer is I believe long term it is by far larger.

Jensen Huang: Yeah, I appreciate it. First of all, people are still using Voltas and Pascals and Amperes. The reason for that is because CUDA is so programmable, you could use it. One of the major use cases right now is data processing and data curation. You find a circumstance that an AI model is not very good at. You present that circumstance to a vision language model, let's say. Let's say it's a car. You present that circumstance to a vision language model. The vision language model actually looks at the circumstance and says, "This is what happened. And I wasn't very good at it." You then take that response, the prompt, and you go and prompt an AI model to go find in your whole lake of data other circumstances like that, whatever that circumstance was.

Speaker Change: And the reason for that is because.

Speaker Change: <unk> designed to market plan to operate their company.

Jensen Huang: And the reason for that is because there are always things that, because CUDA is so programmable, you could use it right, well, one of the major use cases right now is data processing and data curation. You find a circumstance that an AI model is not very good at. You present that circumstance to a vision language model, let's say. Let's say it's a car. You present that circumstance to a vision language model. The vision language model actually looks at the. So this is what happened and I wasn't very good at it. You then take that response, the prompt, and you go and prompt an AI model to go find in your holes lake of data.

Speaker Change: And what is not served by the computer industry is largely industrial so let me give you an example.

Speaker Change: <unk> on.

Speaker Change: On the other hand.

Speaker Change: When we say enterprise.

Speaker Change: The cars that the manufacturer also need AI.

Speaker Change: Let's say, let's use the car company as an example, because they make both soft and hard things.

Speaker Change: They need an AI system that trains the cars treats this entire giant fleet of cars.

Speaker Change: And.

Speaker Change: And so in the case of a car company.

Speaker Change: The employees will be what we call enterprise and <unk> AI.

Speaker Change: There's 1 billion cars on the road someday there'll be 1 billion cars on the road in every single one of those cars will be robotic cars and they'll all be collecting data and we'll be improving them using an AI factory, where the whereas they have a car factory today in the future that I'm a car factory.

Speaker Change: Software planning systems and tools and we have some really exciting things to share with you guys at GTC. Those agency systems are for employees to make employees more productive.

Speaker Change: Two designed to market plan to operate their company.

Speaker Change: And in that factory.

Speaker Change: And then inside of the car itself is a robotics system and.

Speaker Change: <unk> on.

Speaker Change: On the other hand.

Speaker Change: The cars that the manufacturer also need AI.

Speaker Change: So as you can see there are three computers involved.

Jensen Huang: and other circumstances like that. whatever that circumstance was. And then you use an AI to do domain randomization and generate a whole bunch of other examples. And then from that, you can go train the model. And so. You could use the Amperes to go and do data processing and data curation and machine learning-based search. and then you create the training data set which you then present to your Hopper systems for training. So each one of these architectures are completely, they're all CUDA-compatible, and so everything runs on everything. But if you have infrastructure in place, then you can put the less intensive workloads onto the install base of the platform.

Speaker Change: They need an AI system that trains the cars treats this entire giant fleet of cars.

And there's the computer that helps the people theres. The computers are built the AI for it.

Jensen Huang: You use an AI to do domain randomization and generate a whole bunch of other examples. From that, you can go train the model. You could use the Amperes to go and do data processing and data curation and machine learning-based search. You create the training data set, which you then present to your Hopper systems for training. Each one of these architectures are completely they're all CUDA compatible. Everything runs on everything. If you have infrastructure in place, then you can put the less intensive workloads onto the install base of the past. All of our CPUs are very well employed.

Speaker Change: Machineries that could be of course <unk>.

Speaker Change: There's 1 billion cars on the road.

Speaker Change: Tractor it could be a lawnmower it could be a humanoid robot that's being developed today it could be a building it could be a warehouse. These physical systems require a new type of AI, we call physical alley.

Speaker Change: Good day there'll be 1 billion cars on the road in every single one of those cars will be robotic cars and they'll all be collecting data and we will be improving.

Speaker Change: Can I just understand the meaning of words in languages, but they have to understand the meaning of the world.

Speaker Change: Car factory today in the future that I'm, a car factory and in AI factory.

Speaker Change: Friction and a nurse shut it.

Speaker Change: And then inside of the car itself is a robotic system.

Speaker Change: Object permanence and cause and effect in all of those type of things that are common sense to eni, but you know AI something called learn those physical effects. So we called out physically.

Speaker Change: So as you can see there are three computers involved.

Speaker Change: And there is the computer that helps the people there is the computer.

Speaker Change: That whole that whole part of us.

Speaker Change: If you're a tractor it could be a lawnmower it could be a humanoid robot that's being developed today it could be a building it could be a warehouse. These physical systems require a new type of AI, we call physical AI.

Speaker Change: Using <unk>.

Speaker Change: Agentic AI to revolutionize the way we work inside companies. That's just starting this is now at the beginning of the Agentic AI.

Jensen Huang: All of our GPUs are very well...

Speaker Change: Era, and you hear a lot of people talking about and we've got some really great things going on and then there's the physical AI after that and then the robotic systems after that and so these three computers are all brand new and my sense is that long term this will be by far the larger football, which kind of makes sense in.

Speaker Change: They cannot just understand that.

Atif Malik: We have time for one more question and that question comes from Atif Malik with Citi. Please go ahead. Hi, thank you for taking my question. I have a follow-up question, gross margins for Colette. Colette, I understand there are many moving parts, the Blackwell Yields and Relink 72 and Ethernet mix. And you kind of tiptoed the earlier question, if April quarter is the bottom. But second half would have to ramp like 200 basis points per quarter to get to the mid-70s range that you're giving for the end of the fiscal year. And we still don't know much about Terex's impact to broader semiconductors.

Operator: We have time for one more question. That question comes from Atif Malik with Citi. Please go ahead.

Speaker Change: Friction in inertia.

Speaker Change: Object permanent cause and effect in all of those type of things that are common sense to eni, but AI something called learn those physical effects. So we called out physically.

[Analyst]: Hi, thank you for taking my question. I have a follow-up question on gross margins for Colette. Colette, I understand there are many moving parts, the Blackwell yield, NVLink 72 and Ethernet mix. You kind of tiptoed the earlier question if April quarter is the bottom. The second half would have to ramp like 200 basis points per quarter to get to the mid-70% range that you're giving for the end of the fiscal year. We still don't know much about tariffs' impact to broader semiconductors. What kind of gives you the confidence in that trajectory in the back half of this year?

Speaker Change: The world the world of the world's GDP is representing represented by either heavy industries or industrials and companies that are providing for dose.

Jensen Huang: The world's GDP is represented by either heavy industries or industrials and companies that are providing for those.

Speaker Change: That hope that helps.

Speaker Change: AI to revolutionize the way we work inside companies. That's just starting this is now the beginning of the Agentic AI.

Speaker Change: Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead.

Operator: Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead.

Aaron Rakers: Yes, thanks for letting me back in.

Colette Kress: So what kind of gives you the confidence in that trajectory in the back half of this year? Yeah, thanks for the question. Our gross margins, they're quite complex in terms of the material and everything that we put together in a Blackwell system. Tremendous amount of opportunity to look at a lot of different pieces of that on how we can better improve our gross margins over time. Remember, we have many different configurations as well on Blackwell that will be able to help us do that. So together, working after we get some of these really strong ramping completed for our customers, we can begin a lot of that work.

Aaron Rakers: Just I'm curious as we now approach the two year anniversary of really the Hopper inflection that you saw in 2023 and Jen AI in general and we think about the roadmap you have in front of US how do you think about the infrastructure that's been deployed from a replacement cycle perspective and weather.

Colette Kress: Yeah, thanks for the question. Our gross margins, they're quite complex in terms of the material and everything that we put together in a Blackwell system. A tremendous amount of opportunity to look at a lot of different pieces of that on how we can better improve our gross margins over time. Remember, we have many different configurations as well on Blackwell that will be able to help us do that. Together, working after we get some of these really strong ramping completed for our customers, we can begin a lot of that work. If not, we're going to probably start as soon as possible if we can. If we can improve it in the short term, we will also do that. Tariffs, at this point, it's a little bit of an unknown. It's an unknown until we understand further what the U.S.

Speaker Change: World The world of the world's GDP is representing red presented by either heavy industries or industrials and companies that are providing for dose.

Aaron Rakers: If it's.

Aaron Rakers: GBP 300, or if it's the Ruben cycle, where we start to see maybe some refresh opportunity I'm just curious of how you look at that.

Speaker Change: Your next question comes from the line of Aaron Rakers with Wells Fargo. Please go ahead.

Aaron Rakers: I appreciate it first of all.

Aaron Rakers: People are still using voltage and Pascal <unk> and <unk> and the reason for that is because they're always.

Aaron Rakers: Yes, thanks for letting me back in.

Aaron Rakers: Things that because because cuda is so programmable you could use it right well what is it one of the major use cases right now.

Colette Kress: If not, we're gonna probably start as soon as possible if we can. And if we can improve it in the short term, we will also do that.

Aaron Rakers: Data processing and data curation.

Speaker Change: Hugh Hugh you point, a circumstance that an AI model is not very good at.

Colette Kress: Tariffs, at this point, it's a little bit of an unknown. It's an unknown until we understand further what the U.S. government's plan is, both its timing, its where, and how much. So at this time, we are awaiting. But again, we would, of course, always follow export controls and or tariffs in that manner.

Speaker Change: You present that circumstance to a vision language model, let's say, let's say, it's a car you present that that that circumstance to efficient language model Division language model actually looks at the circumstances.

Aaron Rakers: If it's.

Aaron Rakers: GBP 300, or if it's the Ruben cycle, where we start to see maybe some refresh opportunity I'm just curious of how you look at that.

Colette Kress: government's plan is, both its timing, its where, and how much. At this time, we are awaiting. Again, we would, of course, always follow export controls and/or tariffs in that manner.

Aaron Rakers: I appreciate it first of all.

Speaker Change: This isn't this is what happened and.

Aaron Rakers: People are still using voltage and Pascal <unk> and <unk> and the reason for that.

Speaker Change: I wasn't very good at it.

Speaker Change: You then take that response this prompt and you go and prompt an AI model to go find in your holes Lake of data.

Aaron Rakers: Because because kudos so programmable you could use it right what are the what are the major use cases right now.

Speaker Change: Other circumstances like that.

Operator: Ladies and gentlemen, that does conclude our question and answer session. I'm sorry. Thank you.

Operator: Ladies and gentlemen, that does conclude our question and answer session.

Aaron Rakers: Data processing and data curation.

Speaker Change: Whatever that circumstance was and then you're using AI to do domain randomization and generate a whole bunch of other examples and then from that you can go train the model.

[Analyst]: Thank you.

Operator: I'm sorry.

Speaker Change: Hugh Hugh you find a circumstance that an AI model is not very good at you.

[Analyst]: Thank you.

Jensen Huang: And we're going to open up to Jensen, and he has a couple of things. I just wanted to thank you. Thank you, Colette. The demand for Blackwell is extraordinary. AI is evolving beyond perception and generative AI into reasoning. With Reasoning AI, we're observing another scaling loss. inference time or test time scaling. more computation The more the model thinks Models like OpenAIs, ROT3, DeepSeq R1 are reasoning models that apply inference time Reasoning models can consume 100 times more compute. Future reasoning models can consume much more compute. DeepSake R1 has ignited global enthusiasm. It's an excellent innovation, but even more importantly, it has open-sourced a world-class reasoning AI model.

Colette Kress: We are going to open up to Jensen.

[Analyst]: I just wanted to thank you.

Speaker Change: You present that circumstance to a vision language model, let's say, let's say, it's a car.

Colette Kress: He has a couple of things.

[Analyst]: I just wanted to thank you. Thank you, Colette. The demand for Blackwell is extraordinary. AI is evolving beyond perception and generative AI into reasoning. With reasoning AI, we're observing another scaling law: inference time or test time scaling. The more computation, the more the model thinks, the smarter the answer. Models like OpenAI's ROT-3, DeepSeek R1 are reasoning models that apply inference time scaling. Reasoning models can consume 100 times more compute. Future reasoning models can consume much more compute. DeepSeek R1 has ignited global enthusiasm. It's an excellent innovation. Even more importantly, it has open-sourced a world-class reasoning AI model. Nearly every AI developer is applying R1 or chain-of-thought and reinforcement learning techniques like R1 to scale their model's performance. We now have three scaling laws, as I mentioned earlier, driving the demand for AI computing. The traditional scaling laws of AI remain intact.

Speaker Change: And so.

Speaker Change: Resent that that that circumstance to efficient language model Division language model actually looks at the circumstances.

Speaker Change: You could use that.

Speaker Change: The amperes.

Speaker Change: To go and do a data processing and data curation and machine learning based search.

Speaker Change: I'm very good at it.

Speaker Change: And then you create the training dataset, which you then present to your Hopper systems for training and so each one of these these are architectures are completely they're all cuda compatible and so everything once on everything.

Speaker Change: You then take that response.

Speaker Change: And you go and prompt an AI model to go find in your holes.

Speaker Change: Lake of data.

Speaker Change: Other circumstances like that.

Speaker Change: Whatever that circumstance was and then you're using AI to do domain randomization and generate a whole bunch of other examples and then from that you can go train the model.

Speaker Change: If you have infrastructure in place that you can put the less intensive workloads onto the installed base of the past.

Speaker Change: So.

Speaker Change: You could use that.

Speaker Change: All the more you have to use our very.

Speaker Change: The amperes.

Speaker Change: Very well employed.

Speaker Change: To go and do data processing and data curation and machine learning based search.

Speaker Change: We have time for one more question and that question comes from <unk> Malik with Citi. Please go ahead.

Speaker Change: And then you create the training dataset, which you then present to your Hopper systems for training and so each one of these these architectures are completely they're all cuda compatible and so everything once on everything but if you have infrastructure in place that you can.

Operator: We have time for one more question. That question comes from Atif Malik with Citi. Please go ahead.

Jensen Huang: Nearly every AI developer is applying R1. or chain of thought and reinforcement learning techniques like R1 to scale their model's performance. We now have three scaling laws, as I mentioned earlier, driving the demand for AI computing. The traditional scaling laws of AI remains intact. Foundation models are being enhanced with multi-modality and pre-training is still growing. But it's no longer enough. We have two additional scaling dimensions. Post-training scaling, where reinforcement learning, fine-tuning, model distillation require orders of magnitude more compute than pre-training alone. Improved time scaling and reason where a single query can demand 100 times more compute.

Speaker Change: Hi, Thank you for taking my question I have a follow up question gross margins for Colette Colette I understand there are many moving parts the backbone <unk> been willing sandy to an Ethernet mix and you're kind of Tiptoed that earlier question. The April quarter is the bottom.

Speaker Change: Second half would have to ask like 200 basis points per quarter to get to the mid seventies.

Speaker Change: The less intensive workloads onto the installed base of the past.

[Analyst]: Foundation models are being enhanced with multimodality. Pre-training is still growing. It's no longer enough. We have two additional scaling dimensions. Post-training scaling, where reinforcement learning, fine-tuning, model distillation require orders of magnitude more compute than pre-training alone. Inference time scaling and reasoning, where a single query can demand 100 times more compute. We designed Blackwell for this moment, a single platform that can easily transition from pre-training, post-training, and test time scaling. Blackwell's FP4 transformer engine and NVLink 72 scale-up fabric and new software technologies let Blackwell process reasoning AI models 25 times faster than Hopper. Blackwell in all of its configurations is in full production. Each Grace Blackwell NVLink 72 rack is an engineering marvel. One and a half million components produced across 350 manufacturing sites by nearly 100,000 factory operators. AI is advancing at light speed. We're at the beginning of reasoning AI and inference time scaling.

Speaker Change: Means that youre, giving for the end of the fiscal year and we still don't know much about tariff impact to broader semiconductor. So what kind of gives you the confidence.

Speaker Change: All the money you have to use our very well employed.

Speaker Change: In that.

Speaker Change: We have time for one more question and that question comes from <unk> Malik with Citi. Please go ahead.

Speaker Change: Could you actually in the back half of this year.

Speaker Change: Yeah. Thanks for thanks for the question or gross margins.

Speaker Change: Yeah.

Malik: Hi, Thank you for taking my question I have a follow up question gross margins for Colette Colette I understand there are many moving parts the backlog <unk> been willing sandy to an Ethernet mix and you're kind of Tiptoed that earlier question April quarter as the bottom.

Speaker Change: There are quite complex in terms of the material and everything that we've put together an ink Blackwell system.

Speaker Change: A tremendous amount of opportunity to look at a lot of different pieces of thought on how we can better improve our gross margins over time remember we have many different configurations as well Blackwell, although it will be able to help us do that.

Jensen Huang: We design Blackwell for this moment, a single platform that can easily transition. from pre-training, post-training, and test-time scaling. Blackwell's FP4 transformer engine. and NVLink 72 Scale-Up Fabric. and new software technologies let Blackwell process reasoning AI models 25 times faster than Hopper. Blackwell in all of this configuration. is in full production. Each Grace Blackwell NVLink 72 rack is an engineering marvel. One and a half million components produced across 350 manufacturing sites by nearly 100,000 factory operators. AI is advancing at light speed. We're at the beginning of reasoning AI and inference time scaling. But we're just at the start of the HMAI.

Malik: Back to the mid seventies, a range that you're giving for the end of the fiscal year and we still don't know much about tariff impact to broader semiconductor. So what kind of gives you the confidence.

Speaker Change: Together are working after we get some of these really strong ramping completed for our customers. We can begin a lot of that work if not we're going to probably start as soon as possible. If we can if we can improve it in the short term we've also stopped.

Malik: In that.

Malik: Could you please in the back half of this year.

Malik: Yeah. Thanks for thanks for the question our gross margins.

Malik: There are quite complex in terms of the material and everything that we've put together and then Blackwell system.

Speaker Change: Tariffs I'm at this point, it's a little bit of an unknown. That's an unknown until we understand further what U S government partners, both its timing its wear and and how much. So at this time, we are awaiting AR, but again, we would of course.

Malik: A tremendous amount of opportunity to look at a lot of different pieces of thought on how we can better.

Colette Kress: government's plan is, both its timing, its where, and how much. At this time, we are awaiting. We would, of course, always follow export controls and/or tariffs in that manner.

Malik: The durations as well on Blackwell are that we'll be able to help us do that.

Malik: So together are working.

Speaker Change: Always follow ex export controls and ore types in that manner.

Malik: After we get some of these really strong ramping completed for our customers. We can begin a lot of that work if not we're going to probably start as soon as possible. If we can if we can improve it in the short term we've also to dot.

[Analyst]: We're just at the start of the age of AI. Multimodal AIs and enterprise AI, sovereign AI, and physical AI are right around the corner. We will grow strongly in 2025. Going forward, data centers will dedicate most of CapEx to accelerated computing and AI. Data centers will increasingly become AI factories. Every company will have them, either rented or self-operated. I want to thank all of you for joining us today. Come join us at GTC in a couple of weeks. We're going to be talking about Blackwell Ultra, Rubin, and other new computing, networking, reasoning AI, physical AI products, and a whole bunch more. Thank you.

Jensen Huang: multi-modal Enterprise AI, Sovereign AI, and Physical AI are right around the corner. We will grow strongly in 2025. Going forward, data centers will dedicate most of CapEx to accelerated computing and AI. data centers will increasingly become AI factories. and every company will have either rented or self-operated.

Speaker Change: Ladies and gentlemen that does conclude our question and answer session.

Operator: Ladies and gentlemen, that does conclude our question and answer session.

Speaker Change: I'm sorry in Q.

Speaker Change: No no.

[Analyst]: Thank you.

Operator: I'm sorry.

Malik: Tariffs.

[Analyst]: Thank you. No, no.

Speaker Change: It's too Jensen.

Colette Kress: We are going to open up to Jensen.

Malik: At this point, it's a little bit of an unknown. It's an unknown until we understand further what U S government partners, both its timing its ware.

Ben Reitzes: A couple of things I just wanted to thank you. Thank you Colette.

[Analyst]: I just wanted to thank you.

Colette Kress: If he has a couple of things.

[Analyst]: I just wanted to thank you. Thank you, Colette. The demand for Blackwell is extraordinary. AI is evolving beyond perception and generative AI into reasoning. With reasoning AI, we're observing another scaling law: inference time or test time scaling. The more computation, the more the model thinks, the smarter the answer. Models like OpenAI's O-3, DeepSeek R1 are reasoning models that apply inference time scaling. Reasoning models can consume 100 times more compute. Future reasoning models can consume much more compute. DeepSeek R1 has ignited global enthusiasm. It's an excellent innovation. Even more importantly, it has open-sourced a world-class reasoning AI model. Nearly every AI developer is applying R1 or chain-of-thought and reinforcement learning techniques like R1 to scale their model's performance. We now have three scaling laws, as I mentioned earlier, driving the demand for AI computing. The traditional scaling laws of AI remain intact.

Speaker Change: Demand for Blackwell is extraordinary.

Speaker Change: AI is evolving beyond perception in general to the earlier two reasoning.

Speaker Change: With reasoning AI, we're observing another scaling law.

Malik: Are awaiting.

Malik: But again, we would of course always follow ex export controls and or tariffs in that manner.

Jensen Huang: I want to thank all of you for joining us today. Come join us at GTC in a couple of weeks. We're going to be talking about Blackwell Ultra, Rubin, and other new computing, networking, reasoning AI, physical AI products. and a whole bunch more. Thank you.

Speaker Change: Inference time or test times scaling.

Speaker Change: More computation.

Speaker Change: The more the model things smarter.

Speaker Change: The smarter the answer.

Speaker Change: Models like opening is brought three deep seek are one our reasoning models that apply inference times scaling.

Malik: Ladies and gentlemen that does conclude our question and answer session.

Speaker Change: I'm sorry.

Speaker Change: No no.

Speaker Change: Reasoning models can consume 100 times more compute fuse.

Speaker Change: It's to Janssen.

Operator: This concludes today's conference call. You may now disconnect.

Speaker Change: A couple of things I just wanted to thank you. Thank you Colette.

Speaker Change: Future reasoning models can consume much more compute.

Speaker Change: Deep CCAR one.

Speaker Change: Demand for Blackwell is extraordinary.

Speaker Change: Has ignited global enthusiasm.

Speaker Change: AI is evolving beyond perception agenda today, how you took reasoning.

Speaker Change: It's an excellent innovation, but even more importantly, and as open source a world class reasoning AI model near.

Speaker Change: With reasoning AI, we're observing another scaling law.

Speaker Change: Inference time or test time scaling.

Speaker Change: Nearly every AI developer is applying our one.

Speaker Change: The more computation.

Speaker Change: The more the model things.

Speaker Change: Or chain of thought and reinforcement learning techniques like our one to scale their models performance.

Speaker Change: <unk> the answer.

Speaker Change: Models like opening is brought three deep seek are one our reasoning models that apply inference types scaling.

Speaker Change: We now have three scaling laws as I mentioned earlier.

Speaker Change: The demand for AI computing, the traditional scaling laws of AI remains intact.

Speaker Change: The reasoning models can consume 100 times more compute future reasoning models can consume much more compute.

Speaker Change: Foundation models are in being enhanced with multi modality and pre training is still growing.

Speaker Change: Deep CCAR one.

[Analyst]: Foundation models are being enhanced with multimodality, and pre-training is still growing. It's no longer enough. We have two additional scaling dimensions: post-training scaling, where reinforcement learning, fine-tuning, model distillation require orders of magnitude more compute than pre-training alone; inference time scaling and reasoning, where a single query can demand 100 times more compute. We designed Blackwell for this moment, a single platform that can easily transition from pre-training, post-training, and test time scaling. Blackwell's FP4 transformer engine and NVLink 72 scale-up fabric and new software technologies let Blackwell process reasoning AI models 25 times faster than Hopper. Blackwell in all of its configurations is in full production. Each Grace Blackwell NVLink 72 rack is an engineering marvel. One and a half million components produced across 350 manufacturing sites by nearly 100,000 factory operators. AI is advancing at light speed.

Speaker Change: Has ignited global enthusiasm.

Speaker Change: But it's no longer enough.

Speaker Change: It's an excellent innovation, but even more importantly, and as open source a world class reasoning AI model near.

Speaker Change: Do you have two additional scaling dimensions.

Speaker Change: Post training scaling where reinforcement learning.

Speaker Change: Nearly every AI developer is applying our one.

Speaker Change: Tuning model of distillation require orders of magnitude more compute than pre training alone.

Speaker Change: Or chain of thought and reinforcement learning techniques like armor.

Speaker Change: Inference time, scaling and reasoning, where a single query and demand 100 times more compute.

Speaker Change: Well have three scaling laws as I mentioned earlier.

Speaker Change: Driving the demand for AI computing.

Speaker Change: We just like black well for this moment a single platform that can easily transition.

Speaker Change: Traditional scaling laws of AI remains intact found.

Speaker Change: Foundation models are in being enhanced with multi modality.

Speaker Change: From pre trading post training and test times scaling.

Speaker Change: And pre training is still growing.

Speaker Change: <unk> FP for transformer engine.

Speaker Change: But it's no longer enough when you have two additional scaling dimensions.

Speaker Change: NV link 72 scale up fabric.

Speaker Change: And new software technologies Blackwell process reasoning AI models, when <unk> five times faster than Hopper.

Speaker Change: Post training scaling where reinforcement learning.

Speaker Change: Fine tuning model of distillation require orders of magnitude more compute than pre training alone.

Speaker Change: Blackwell and all of its configurations.

Speaker Change: Is in full production.

Speaker Change: Inference.

Speaker Change: Each grace Blackwell envy linked 72 rack is an engineering Marvel.

Speaker Change: Man 100 times more compute.

Speaker Change: 5 million components produced across 350 manufacturing sites by nearly 100000 factory operators.

We just like black well for this moment a single platform that can easily transition.

Speaker Change: From pre trading post training and test times scaling.

Speaker Change: AI is advancing at light speed, where at the beginning of reasoning AI and inference time scaling.

Speaker Change: <unk> FP for transformer engine and NV link 72 Scaleup fabric.

[Analyst]: We're at the beginning of reasoning AI and inference time scaling. We're just at the start of the age of AI. Multimodal AIs and enterprise AI, sovereign AI, and physical AI are right around the corner. We will grow strongly in 2025. Going forward, data centers will dedicate most of CapEx to accelerated computing and AI. Data centers will increasingly become AI factories. Every company will have them, either rented or self-operated. I want to thank all of you for joining us today. Come join us at GTC in a couple of weeks. We're going to be talking about Blackwell Ultra, Rubin, and other new computing, networking, reasoning AI, physical AI products, and a whole bunch more. Thank you.

Speaker Change: But we're just at the start of the HMA on.

Speaker Change: And new software technologies Blackwell process reasoning AI models, when <unk> five times faster than Hopper.

Speaker Change: Multimodal Ai's.

Speaker Change: AI.

Speaker Change: <unk> AI and physical AI are right around the corner.

Speaker Change: Blackwell and all of its configurations.

Speaker Change: We will grow strongly in 2025.

Is in full production.

Speaker Change: Going forward data centers, we will dedicate most of Capex.

Speaker Change: Each Grace Blackwell Envy link 72 rack is an engineering Marvel.

$1 5 million components produced across 350 manufacturing sites by nearly 100000 factory operators.

Speaker Change: Accelerated computing and AI.

Speaker Change: Data centers will increasingly become AI factories.

Speaker Change: And every company will have them.

Speaker Change: AI is advancing at light speed, where at the beginning of reasoning AI and inference time scaling.

Speaker Change: Either rents in our self operated.

Speaker Change: I want to thank all of you for joining us today I'm joined us at GTC in a couple of weeks, we're going to be talking about black while ultra Rubin and other new computing networking reasoning AI physical AI products in them and a whole bunch more.

Speaker Change: But we're just at the start of the HMA odd.

Speaker Change: Multimodal.

Speaker Change: Enterprise AI sovereign AI and physical AI are right around the corner.

Speaker Change: We will grow strongly in 2025.

Speaker Change: Thank you.

Speaker Change: This concludes today's conference call you may now disconnect.

Speaker Change: Going forward data centers, we will dedicate most of capex to accelerated computing and AI.

Operator: This concludes today's conference call. You may now disconnect.

Speaker Change: Data centers will increasingly become AI factories.

Speaker Change: And every company will happen.

Speaker Change: Either rented or self operated.

Speaker Change: I want to thank all of you for joining us today I'm joined us at GTC in a couple of weeks.

Speaker Change: New computing networking reasoning AI physical AI products in them and a whole bunch more.

Speaker Change: Thank you.

Speaker Change: This concludes today's conference call you may now disconnect.

Speaker Change: Okay.

Hum.

Speaker Change: Yeah.

Speaker Change:

Q4 2025 NVIDIA Corp Earnings Call

Request a DemoDemo

NVDA

NVIDIA

Earnings