11 Words From Nvidia CFO Colette Kress That Should Have AMD and Intel Investors Worried

Nvidia says its Vera CPU opens a new $200 billion TAM, with CFO Colette Kress citing nearly $20 billion in CPU revenue visibility this year. The CPU expands Nvidia beyond GPUs into a full-stack AI infrastructure offering, potentially increasing customer switching costs and share of AI data center capex. The article argues the market has not yet priced in this new growth vector, implying upside to NVDA valuation.

Analysis

The market is still treating NVDA as a single-product GPU story, but the strategic significance is that it is moving up the stack from component supplier to architecture owner. That matters because once the CPU becomes part of the reference design, Nvidia can monetize adjacency: networking, software, orchestration, and later service attach. The second-order effect is not just incremental revenue; it is better pricing power and stickier multi-year procurement cycles, which should reduce earnings volatility versus a pure accelerator cycle.

The real competitive pressure shifts to INTC and AMD, but the damage is not linear. If Nvidia wins design-ins for AI control-plane workloads, incumbents lose the most attractive enterprise entry point into AI data centers: the low-risk CPU layer that historically served as the wedge. That also raises the bar for custom silicon efforts, because hyperscalers may accept ASICs for a narrow inference function but still prefer Nvidia’s integrated stack for the broader agentic workflow where software compatibility and deployment speed matter more than raw silicon cost.

The stock setup looks underappreciated in the medium term. Near-term the market may still cap multiple expansion because CPU ramp is usually modeled as commoditized, but that underestimates how much of AI capex is shifting from training-only to persistent inference + orchestration spend. The catalyst is not a single quarter; it is a 6-18 month design-win cycle that can gradually lift consensus revenue and gross margin durability if CPU attach rates rise alongside Blackwell/Rubin deployments.

Contrarian risk: the bull case fails if agentic AI adoption remains mostly pilot-stage and never scales into production budgets, or if hyperscalers standardize on in-house control-plane silicon to protect platform economics. In that scenario, Vera becomes a niche product with limited TAM conversion and investors are left overpaying for optionality. The key tell over the next 2-3 quarters is whether management commentary starts quantifying CPU attach per system and whether that pull-through shows up in networking and software mix, not just headline chip revenue.

Key Decisions for Investors

• Go long NVDA on a 6-12 month horizon; use any post-earnings multiple compression to add. Risk/reward is attractive if the market starts capitalizing CPU attach as a second growth leg rather than a one-off product launch.
• Pair trade: long NVDA / short INTC, 3-6 months. The thesis is that Nvidia is taking the highest-value AI-adjacent CPU slots while Intel remains exposed to commoditized enterprise refresh cycles. Tight stop if Intel demonstrates material AI socket share gains.
• Pair trade: long NVDA / short AMD, 3-6 months. AMD remains vulnerable if hyperscalers prefer a single-vendor AI stack for deployment speed and software coherence; upside for the short if AI infrastructure buyers prioritize integration over component price.
• Buy NVDA call spreads 6-9 months out to express upside from valuation expansion without paying full premium for a still-uncertain CPU adoption curve. Best entry on any selloff tied to GPU-cycle or ASIC concerns.
• Avoid chasing momentum in AI suppliers until the next 1-2 quarters clarify whether CPU guidance converts into actual mix shift; if adoption is slower than expected, the overhang is a model miss rather than a thesis break.