Goldman Sachs reiterates Nvidia stock Buy rating on AI growth outlook

Goldman Sachs reiterated a Buy rating and $300 price target on Nvidia, citing its AI market outlook, supply dynamics and an estimated $3 trillion-$4 trillion AI total addressable market by 2030. Nvidia reaffirmed expectations for roughly 70% year-over-year calendar 2027 growth despite constraints in advanced packaging, wafers, memory, power and datacenter sites; its revenue rose 83% over the past 12 months and its PEG ratio was cited at 0.22. Additional bullish analyst actions from Piper Sandler, BMO and Cantor Fitzgerald, alongside Rackspace's Blackwell infrastructure partnership, reinforce strong AI-compute demand and Nvidia's ecosystem position.

Analysis

The actionable signal is not another sell-side Buy, but the shift in bottleneck economics: constrained packaging, HBM memory, power delivery and datacenter commissioning can cap system shipments even if accelerator demand remains intact. That favors suppliers with capacity-linked pricing and less valuation embedded AI demand—particularly MU (HBM), AVGO (custom silicon/networking) and VRT/ETN (power and thermal infrastructure)—over pure server integrators such as SMCI and RXT, whose revenue can grow rapidly while working-capital needs and gross-margin competition intensify.

NVDA’s financial commitments to secure ecosystem capacity are a second-order risk worth monitoring. They can defend near-term deployment velocity, but convert part of the AI cycle from a customer-capex story into an exposure to datacenter developer financing, power availability and end-customer utilization. A higher-rate backdrop raises the hurdle for marginal GPU-cloud projects; the first evidence of stress would be rising lease incentives, weaker GPU-rental pricing, or a material increase in NVDA-related guarantees/receivables rather than a decline in chip orders.

Consensus is likely underweighting the distinction between supply-constrained revenue and durable incremental demand. Over the next 1-3 months, favorable estimates can support NVDA, but the stock’s downside is asymmetric if growth remains high yet merely decelerates faster than implied expectations. For the 6-18 month horizon, AMD is the cleaner share-gain optionality if customers prioritize second-source supply and lower total system cost; INTC remains a higher-risk turnaround rather than a direct beneficiary until external-foundry execution and product competitiveness are independently demonstrated.

Do not treat generic fair-value and PEG references in a promotional, AI-generated article as valuation evidence. The thesis is falsified by sustained improvement in advanced-packaging/HBM lead times without corresponding unit growth, NVDA datacenter revenue growth falling below management’s implied trajectory, or material gross-margin pressure from product mix and ecosystem financing.

Key Decisions for Investors

• Maintain an existing NVDA core only; do not add solely on the rating reiteration. Over the next earnings cycle, use a data-center revenue-growth or gross-margin guide-down as a risk trigger; a growth deceleration without an offsetting backlog/visibility extension warrants reducing exposure.
• Express the infrastructure bottleneck theme over 3-9 months via long MU and/or VRT versus short SMCI in equal dollar risk. MU/VRT monetize scarce components and power buildout, while SMCI is more exposed to server pricing, inventory and working-capital pressure. Exit if HBM availability normalizes materially or SMCI demonstrates sustained gross-margin expansion.
• For a higher-beta 6-18 month diversification trade, accumulate AMD on broad semiconductor or rate-driven pullbacks rather than chasing momentum; pair against a partial NVDA trim rather than a naked NVDA short. The payoff requires meaningful accelerator share gains or inference adoption, and is invalidated if AMD fails to show sequential data-center GPU revenue traction.
• Put RXT on a watchlist rather than initiating. Its regulated-cloud positioning can create AI-services upside, but the required diligence is contract backlog, GPU utilization, debt/refinancing costs and gross-margin conversion; absent those data, it is a financing-sensitive infrastructure story, not a clean AI beta.