Jensen Huang says he’s found a ‘brand new’ $200B market for Nvidia

Nvidia said Vera CPUs have already generated $20 billion in standalone sales this year and framed the product as opening a new $200 billion TAM. The company also reported record quarterly revenue of $81.6 billion and guided to $91 billion next quarter, reinforcing strong momentum in AI infrastructure. The article is upbeat on Nvidia’s growth outlook, though it notes rising competitive pressure from cloud providers and CPU rivals.

Analysis

NVDA is trying to extend the AI stack from accelerator supplier to platform landlord. If Vera is accepted as the default control plane for agent workloads, the margin pool expands beyond GPUs into a much stickier, higher-attach CPU layer that can be bundled, standardized, and refreshed on a cadence that looks more like enterprise infrastructure than a one-off chip sale. That matters because the economic moat shifts from raw performance to systems-level integration, where hyperscalers often pay up for fewer vendors, tighter power/performance optimization, and simpler procurement.

The second-order effect is pressure on CPU incumbents and on cloud vendors building in-house silicon. If agentic workloads proliferate, the winner is not necessarily the best general-purpose CPU, but the vendor that can co-optimize inference scheduling, networking, memory, and software tooling. That structurally favors NVDA’s full-stack approach and makes it harder for single-product challengers to compete on TCO alone, especially if customers value time-to-deployment over headline chip cost. The most exposed names are those whose AI narrative depends on owning the CPU layer without owning the surrounding software ecosystem.

The market is likely underestimating how long adoption friction can persist. The base case is not a clean displacement of x86 over the next few quarters, but a multi-year procurement cycle where design wins accumulate first at hyperscalers and then percolate to enterprise and robotics use cases. The key risk to the upside thesis is that cloud customers use the announcement to renegotiate pricing and diversify suppliers, compressing NVDA’s share of wallet even if unit volumes grow.

Contrarian take: the ‘billions of agents’ framing may be directionally right but premature in timing. Agent workloads will initially be uneven, bursty, and heavily managed, which means CPU demand may grow slower than the narrative suggests, while custom ASICs from hyperscalers capture enough early share to cap NVDA’s TAM capture. In that scenario, the stock can remain structurally strong, but the incremental upside from this specific CPU story may be less explosive than the market is likely to price in.

Key Decisions for Investors

• Stay long NVDA on pullbacks over the next 1-3 months, but express it via call spreads rather than outright shares; the setup favors continued multiple support, while capped upside from custom-chip competition makes defined-risk convexity preferable.
• Pair trade: long NVDA / short INTC over 3-6 months. The thesis is that AI CPU attach goes to the vendor with the strongest software and systems integration, not the legacy general-purpose CPU incumbent.
• Tactically underweight AMD versus NVDA for the next 2 quarters. AMD can participate in AI demand, but this CPU narrative increases the risk that the market re-rates platform breadth over point-product competitiveness.
• Buy AMZN and META selectively on weakness as a hedge against the ‘Nvidia wins everything’ narrative. If they succeed in internal CPU deployment, it validates the agentic workload thesis while preserving optionality on in-house silicon economics.
• Watch for a multi-week slowdown in hyperscaler capex commentary; if procurement budgets shift toward diversification or internal silicon, trim NVDA exposure into strength because the TAM story is more vulnerable to share leakage than to outright demand failure.