Sora 2 and ChatGPT are consuming so much power that OpenAI just did another 10 gigawatt deal

OpenAI has partnered with Broadcom to design and develop custom AI chips and systems, a deal expected to consume 10 gigawatts of power and driving Broadcom shares up 12%. This strategic collaboration aims to secure critical computing resources for OpenAI's rapidly expanding user base, including 800 million weekly ChatGPT users, and allows OpenAI to embed its model learnings directly into hardware for enhanced capabilities. The initiative underscores the escalating power demands of the AI boom, with data centers projected to consume up to 12% of total US electricity by 2028, highlighting significant infrastructure and environmental considerations.

Analysis

OpenAI's strategic partnership with Broadcom to design and develop 10 gigawatts of custom AI chips and systems represents a significant development for both companies. Broadcom shares surged 12% following the announcement, confirming earlier speculation about a new $10 billion customer. This collaboration secures a substantial revenue stream for Broadcom and critical computing resources for OpenAI's rapidly expanding user base, which includes 800 million weekly ChatGPT users.

For OpenAI, this initiative allows for greater control over its hardware infrastructure, enabling the company to embed its frontier model learnings directly into custom accelerators. This bespoke hardware development is expected to unlock new levels of capability and intelligence for AI services like ChatGPT and the fast-growing Sora video generation app. The move also diversifies OpenAI's chip supply chain, reducing reliance on existing partners like Nvidia and AMD for these specialized needs.

The deal underscores the escalating power demands of the AI boom, with the 10-gigawatt requirement equating to the electricity consumption of a large city or 8 million US households. This massive energy need contributes to projections that data centers will consume 6.7% to 12% of total US electricity by 2028, up from 4.4% in 2023. Such figures highlight growing environmental and infrastructure concerns related to AI's rapid expansion, with deployment expected to commence in the second half of 2026.

Key Decisions for Investors

• Investors should closely monitor Broadcom's execution and integration of this substantial custom AI chip contract, as it significantly bolsters their long-term revenue and strategic positioning
• Consider the potential shift in OpenAI's hardware procurement strategy and its implications for existing chip suppliers like Nvidia and AMD, particularly regarding future custom chip opportunities
• Evaluate the increasing energy demands of AI infrastructure as a critical factor influencing utility providers, data center operators, and ESG-focused investments, given the projected surge in electricity consumption