Nvidia says its forecast for $200 billion CPU market includes China

Nvidia said its new Vera CPU line opens access to a $200 billion market, including China, reinforcing long-term AI demand despite U.S.-China tensions. Huang also said H200 chips have been licensed for shipment to China, though no deliveries have occurred and Chinese approval is still pending. The article highlights continued supply-chain demand in Taiwan and ongoing export-control and smuggling risks around advanced AI chips.

Analysis

The incremental signal is not just that Nvidia can still monetize China; it’s that the company is broadening the AI stack from GPU-led capex to a more durable systems spend cycle anchored by CPUs, networking, and platform integration. That matters because it lowers the cyclicality of the AI buildout: if agentic workloads require more inference and orchestration, then the spending pool shifts from a one-time model-training binge to a recurring fleet-refresh model. In that regime, NVDA’s moat becomes less about raw accelerator scarcity and more about the full rack-level architecture, which should support valuation even if headline GPU unit growth normalizes.

AMD is the cleaner near-term beneficiary on relative basis because any acceleration in Taiwan assembly and advanced packaging capacity tends to lift the whole non-TSMC AI supply chain, but AMD’s upside depends on proving it can convert design wins into actual volume shipments faster than NVDA can use its ecosystem advantage. TSM is the hidden leverage point: the combination of higher-end CPU/GPU mix and a busier second half for Taiwan vendors likely improves utilization and advanced-node loading, but it also increases concentration risk if export-control headlines trigger customer prebuild pauses. The market is still underestimating how much of the AI trade is now a manufacturing throughput story, not just a chip demand story.

The biggest risk is policy, not demand. A single delayed license, enforcement action, or additional smuggling investigation can temporarily freeze China-linked revenue assumptions across the group, and that risk likely plays out over days-to-weeks rather than quarters. Over months, however, the more important catalyst is whether agentic AI actually proves inference-heavy enough to require a second wave of capex; if yes, the current AI trade has room to extend, and if no, multiple compression will hit the high-expectation names first.

Consensus is probably too focused on the China access headline and not focused enough on the structural expansion of the TAM across CPUs and AI infrastructure. The market may be treating China as a binary upside option, when in reality the larger prize is that NVDA can use policy friction to justify a broader platform narrative and keep customers engaged on non-China geographies. That makes the trade less about reopening one market and more about sustaining the capital intensity of the entire AI ecosystem.

Key Decisions for Investors

• Long NVDA / short SOXX on a 1-3 month horizon: own the platform leader while fading the basket if policy headlines reintroduce dispersion; target 8-12% relative outperformance, stop if broad AI capex momentum rolls over.
• Add to AMD on weakness into Taiwan-capex headlines for a 2-4 month trade: the stock should benefit if the market re-rates second-source and non-NVDA AI exposure; risk/reward is favorable if it can reclaim narrative premium versus semiconductor peers.
• Initiate a modest long TSM position or call spread into the next 6-10 weeks: higher AI supply-chain utilization should support earnings revisions, with limited downside unless export-control enforcement expands materially.
• Use NVDA downside puts as a hedge rather than a directional short ahead of any China policy or enforcement event: 30-45 day puts can monetize headline risk while preserving upside from continued AI demand.
• Pair long TSM / short a China-exposed semiconductor proxy if available: the cleaner exposure is to Taiwan manufacturing capacity rather than direct China licensing risk, with the pair benefiting if AI buildout continues but China approvals stay messy.