Not the GPUs. This Overlooked Piece of Nvidia's Business Could Matter Most for the Stock's Next Leg Up.

Nvidia reported recent-quarter revenue of $96 billion, up 100% year over year, and profit of $59 billion, while its shares are on track for roughly a 20% annual gain. The company’s Vera Rubin platform introduces its first stand-alone CPU, targeting a $200 billion market opportunity and projected to generate $20 billion in CPU sales this year. The article views Nvidia’s CPU push as a potential new growth driver as agentic AI increases demand for compute infrastructure beyond GPUs.

Analysis

The relevant issue is not whether Nvidia can sell a CPU, but whether it can attach one to accelerated-computing racks without sacrificing system economics. A credible standalone CPU increases Nvidia's share of AI infrastructure spend and strengthens its ability to bundle compute, networking and software; that raises switching costs and could shift a larger portion of hyperscaler capex from component purchasing to Nvidia-designed systems. The first-order pressure falls on AMD's EPYC growth narrative and Intel's data-center recovery, especially where customers value validated full-stack deployments over CPU price/performance alone.

The market should discount management's CPU revenue target until customer mix, standalone versus bundled sales, and gross-margin contribution are disclosed. Much of the apparent opportunity may be internal bill-of-materials capture rather than incremental silicon TAM, while hyperscalers retain meaningful bargaining power and are advancing Arm-based in-house designs. Over the next 1-3 months, channel commentary around rack configurations and cloud-instance launches matters more than promotional TAM claims; over 6-18 months, sustained CPU attach rates would justify a further Nvidia platform premium and challenge AMD/Intel consensus server-share assumptions.

Contrarian view: Nvidia's greatest CPU success may be strategically negative for near-term gross margin if it uses aggressive bundle pricing to lock in rack-level standards. Intel could also be less exposed than implied if enterprise refresh demand, sovereign deployments, or x86 software inertia remain decisive outside frontier AI clusters. The thesis is falsified if Nvidia reports meaningful CPU sales but no uplift in system ASP, networking attach, or data-center gross margin, or if AMD/Intel maintain server-unit share and guide data-center revenue above expectations.

Key Decisions for Investors

• Maintain NVDA overweight, but add only on evidence of system-level monetization: initiate/add after the next earnings release if management quantifies CPU attach and data-center gross margin is stable to up. Target a 6-12 month hold; reduce if gross margin declines more than 150bp without a corresponding acceleration in data-center revenue.
• Express relative exposure via long NVDA / short AMD in equal-dollar notional for 3-6 months, sized modestly. The trade captures risk to AMD server-CPU estimates if Nvidia wins integrated AI-rack designs; stop out if AMD raises data-center guidance or demonstrates sustained EPYC share gains in hyperscaler deployments.
• Avoid a directional INTC short solely on this development. Use INTC as a watch item around its next data-center guide: a revenue-guide cut or weak Xeon mix would validate competitive pressure, while evidence of enterprise/sovereign demand resilience would make Intel a potential crowded-negative reversal candidate.
• Set an alert for independent verification of Nvidia CPU demand: named cloud instances, disclosed external CPU revenue, and rack BOM/ASP data. Absent those disclosures, treat the CPU narrative as valuation support rather than a new earnings driver and do not pay incremental option premium for it.