Nvidia's Forecast Assumes No Data Center Chip Sales to China. The Sept. 24 U.S.-China Summit Could Change That.

Nvidia's $108.0 billion fiscal Q3 revenue forecast assumes zero China data-center compute revenue, implying roughly 89% year-over-year growth from $57.0 billion despite export restrictions. Fiscal Q2 revenue rose 106% to $96.2 billion, including $89.0 billion in data-center sales, while China-headquartered customers accounted for $7.9 billion, or 8%, mostly from non-restricted products. A U.S.-China breakthrough on AI-chip controls could provide upside—prior H20 demand represented about $7 billion of potential sales—but Nvidia's outlook and investment case do not depend on reopening the Chinese AI-compute market.

Analysis

The market should treat China optionality as a low-probability, high-convexity upside rather than part of the base earnings case. A meaningful policy change would matter less through the first quarter of shipments than through a re-rating of NVDA's terminal growth narrative: renewed access would reduce the risk that China becomes a permanently captive market for Huawei Ascend, Cambricon, and domestic accelerator ecosystems. Conversely, a narrow license expansion is unlikely to be earnings-material if Beijing maintains procurement restrictions or if compliance costs and tariffs keep Nvidia's delivered price uncompetitive.

The more important second-order issue is competitive lockout duration. Each additional quarter of restricted access gives Chinese hyperscalers incentive to rewrite software stacks, qualify domestic hardware, and invest in inference-specific ASICs; that raises the switching cost even if restrictions later ease. This is most damaging to Nvidia's long-run China TAM, but it is supportive near term for non-China customers because constrained China supply reduces the risk of incremental high-end GPU demand competing for global allocation and potentially preserves pricing discipline.

For the next several days, summit headlines create asymmetric gap risk but little basis for underwriting a sustained revenue revision until licenses, end-customer approvals, and Chinese procurement rules are disclosed. Over 1-3 months, the relevant catalyst is whether management converts permitted shipments into reported revenue without incremental inventory reserves or gross-margin dilution. Over 6-18 months, the thesis is falsified if China revenue remains immaterial while domestic alternatives gain measurable hyperscaler deployment, indicating the market has structurally shifted rather than merely been deferred.

Contrarian view: the bullish framing may understate that reopening China can be economically inferior to restoring the historical market. Export-compliant products could carry lower ASPs, higher support costs, tariff leakage, and greater policy-reversal risk; investors should not capitalize every dollar of China revenue at the same multiple as unrestricted global data-center revenue. NVDA remains fundamentally supported by non-China demand, but China-specific upside should be valued as an option, not added mechanically to forward estimates.

Key Decisions for Investors

• Maintain a core NVDA long only if sized for event volatility; add on a post-summit pullback rather than pre-positioning for a diplomatic headline. A durable bullish trigger is disclosed broad-based shipment authorization plus evidence of Chinese customer acceptance; a negative trigger is another inventory charge or gross-margin pressure tied to China-compliant inventory.
• For a tactical 1-3 month expression, prefer a defined-risk NVDA call spread funded with an out-of-the-money put spread rather than outright shares if implied volatility remains below event-risk pricing. The thesis requires policy language that changes commercial eligibility, not merely AI-cooperation rhetoric; close if no concrete export-control development emerges within two weeks.
• Monitor a relative-value basket long NVDA versus a basket of China AI-accelerator beneficiaries/proxies, including Cambricon where accessible and relevant domestic semiconductor ETFs, only after verified relaxation. The trade works if Nvidia regains commercially viable access; avoid initiating before confirmation because Beijing procurement policy can offset Washington approval.
• Do not infer a read-through to NFLX or TME from this development. Any China technology-policy easing may marginally improve broader risk sentiment, but neither has a direct enough accelerator supply-chain exposure to justify a standalone position.