Trump snubs Jensen Huang for his trip to China

Nvidia’s China opportunity remains blocked despite a U.S. licensing scheme that would allow limited H200 sales in exchange for a 25% government cut. Chinese authorities have blocked H200 imports, and Reuters reports Jensen Huang was not invited on Trump’s China trip, underscoring continued policy and market-access uncertainty. Nvidia said it is still not certain it can sell any H200s in China, even as the market is estimated at $54 billion.

Analysis

The market should read this less as a headline about diplomacy and more as a reminder that China monetization is becoming a policy variable, not a demand variable. NVDA’s core risk is no longer just export approval; it is approval without effective channel access, which turns a licensing win into stranded inventory and working-capital drag. That creates a second-order headwind for TSM as well: if demand visibility weakens, advanced-node ordering can shift from “build ahead” to “wait for clarity,” compressing utilization expectations over the next 1-2 quarters.

The key competitive dynamic is that any delay in high-end US accelerator shipments to China accelerates local substitution, but in a nonlinear way. Once Chinese buyers commit to domestic clusters and software stacks, reversion to imported hardware becomes harder even if restrictions later ease, because switching costs move into inference tooling, model deployment, and procurement standards. That is modestly negative for AMD too, but more importantly it broadens the window for BABA and other domestic AI/cloud platforms to capture incremental capex and ecosystem share, even if near-term absolute performance lags NVDA.

The contrarian setup is that the near-term reaction may be overdone if investors are already discounting zero H200 sales in China; in that case the real issue becomes mix, not volume. If NVDA can redirect supply to non-China hyperscaler demand or pull forward Vera Rubin demand, the earnings hit may be deferred rather than destroyed. The catalyst sequence to watch is the May 20 print and any commentary on backlog, lead times, and whether China uncertainty is creating enough supply flexibility to offset lost revenue within the next 1-2 quarters.

Key Decisions for Investors

• Short NVDA into the May 20 earnings window or buy short-dated put spreads; the setup is asymmetric if management confirms China remains effectively closed, with 10-15% downside possible on guidance compression and inventory risk.
• Pair trade: long BABA / short NVDA for a 1-3 month horizon; if domestic China AI procurement accelerates, BABA should benefit from capex substitution while NVDA faces policy-constrained monetization. Keep tight risk controls because any export-policy thaw would unwind the spread quickly.
• Reduce exposure to TSM on any strength over the next 2-4 weeks; the risk/reward is poor if AI supply chain orders become less certain, though the downside should be limited unless broader semi capex rolls over.
• Avoid chasing AMD on a China-license headline until there is evidence of actual shipments; this is a classic ‘approval without revenue’ setup, and the first clean tradeable confirmation would be channel checks rather than the policy headline itself.
• If already long NVDA structurally, hedge with a short-dated collar into earnings: sell upside calls to finance downside protection. This preserves exposure if China issues are contained while protecting against a 1-day gap lower on guidance.