Nvidia says it will see $20 billion in CPU sales this year

Nvidia said it is targeting the $200 billion CPU market and now sees nearly $20 billion in standalone CPU revenue this year, expanding beyond its core GPU dominance. Management said every major hyperscaler and system maker is partnering to deploy its CPUs, and it is preparing CPU-only Grace servers now with Vera-only servers planned for 2027. The comments reinforce Nvidia's AI infrastructure leadership and point to a broader product opportunity across agentic AI and robotic physical AI.

Analysis

The market is likely underestimating how quickly the AI stack is shifting from a GPU-only narrative to a full-system architecture race. If Nvidia can credibly own both the compute engine and the orchestration layer, it raises switching costs for hyperscalers and compresses the strategic runway for standalone server CPU vendors, even if near-term demand is broadening rather than cannibalizing. The second-order effect is that CPU scarcity can become a gating item for AI deployment velocity, which means the bottleneck may move from accelerators to host infrastructure faster than consensus expects.

Intel’s rally may be more about narrative repricing than durable fundamentals. The key question is not whether demand exists, but whether supply, platform readiness, and software enablement allow it to convert into margins before competitors match the roadmap; that can take multiple quarters, not weeks. If execution slips, the stock is vulnerable because expectations have already moved ahead of the operating data, especially if enterprise PC/server digestion or capex discipline offsets AI enthusiasm.

AMD sits in a more nuanced spot: it benefits from the same AI-server expansion, but it is less positioned to capture the platform lock-in effect that Nvidia is building by pairing CPUs and GPUs. That makes AMD more exposed to pricing pressure if customers use it as a secondary source while reserving strategic platform share for Nvidia. The contrarian view is that this is not a zero-sum CPU winner-take-all market yet; the real trade may be in relative exposure to ecosystem control, not raw unit growth.

The biggest risk to the bullish CPU thesis is timing. Over the next 1-2 quarters, the market may overpay for headlines while actual revenue mix changes lag behind design wins and capacity ramps. If AI inference economics improve faster than expected, however, CPU pull-through could remain stronger than investors model for the next 12-18 months, keeping the theme alive well beyond the current earnings cycle.

Key Decisions for Investors

• Long NVDA / short INTC for 3-6 months: best expression of platform control versus narrative-driven rerating; expect upside asymmetry if Nvidia converts CPU design wins into bundled share, but stop if Intel shows two consecutive quarters of meaningful supply relief plus margin expansion.
• Buy NVDA pullbacks on any post-earnings compression over the next 2-4 weeks: use weakness to own the company with the highest leverage to both GPU and CPU attach; risk/reward remains favorable as long as hyperscaler capex stays intact.
• Fade the recent INTC rally with a defined-risk put spread 2-4 months out: the stock has likely discounted too much near-term success, and execution risk around supply, platform qualification, and mix recovery is still high.
• Relative long NVDA / short AMD into the next earnings window: AMD benefits from AI demand, but Nvidia’s bundled CPU-GPU strategy should capture a greater share of the strategic moat and wallet share if the agentic AI thesis proves durable.
• Consider selling out-of-the-money covered calls on INTC if already long for tactical exposure: the move has been sentiment-led, and upside may be capped until actual shipment data validates the growth story.