Trump Weighing Advanced Nvidia Chip Sales to China, Lutnick Says

President Trump is weighing whether to permit Nvidia to export advanced AI chips (Bloomberg has reported the H200 model) to China, and Commerce Secretary Howard Lutnick said the president will make the final decision after hearing advice from multiple advisers. The deliberation signals potential changes to U.S. export controls that could materially affect Nvidia's addressable market and China’s access to high-end AI hardware, while also carrying geopolitical and regulatory implications that could move semiconductor supply chains and investor positioning depending on the outcome.

Analysis

Market structure: Approval would concentrate incremental demand on Nvidia (NVDA) and its foundry partners, boosting NVDA’s pricing power and potentially lifting near-term revenue by a low-teens percentage vs a denial scenario; competitors (AMD) and domestic Chinese fabs would face slower share gains. Limited TSMC wafer capacity means any China allocations are supply-constrained, so unit volumes likely rise modestly while ASPs remain intact, supporting margin expansion rather than a pure volume surge. Cross-asset flows: a positive decision favors risk assets (equities), tightens corporate credit spreads by 10–30bps in tech-heavy IG, and could firm semiconductor-related commodities (copper, palladium) modestly; FX moves likely small but USD could strengthen on risk-on.

Risk assessment: Tail events include a revocation/rollback of licenses or new blanket bans (low probability, high impact) that could cut 10–20% of NVDA addressable market within 12 months and force inventory markdowns. Near term (days–weeks) expect binary volatility spikes around announcements; medium term (3–12 months) depends on contract fulfillment and TSMC capacity; long term (1–3 years) hinge on China’s ability to indigenize GPUs and US political cycles. Hidden dependencies: license language (software/toolchain vs hardware), export logistics, payment/contract terms and Chinese cloud procurement cycles.

Trade implications: Favor asymmetric overweight to NVDA via limited-cost directional option structures timed to the decision window, and a relative-value long NVDA / short AMD pair to express market-share capture; size initial exposure small (1–3% portfolio) and scale on license clarity. Use 30–90 day event-driven strangles when the announcement date is firm, but cap premium at 3% notional; avoid outright levered longs until license text is public and TSMC capacity confirmations arrive. Rotate modest capital from China cloud capex beneficiaries into US AI hardware names if approval seems likely within 30 days.

Contrarian angles: Consensus treats the outcome as binary and underweights logistics and software-stack frictions — even with a license, meaningful shipments could lag 2–4 quarters, creating a buy-the-news opportunity later. Markets may also underprice a denial if investors focus only on short-term demand loss and ignore NVDA’s ability to reallocate supply to other markets; historical parallels (Huawei-era controls) show durable vendor pricing power can persist despite geopolitical shocks. Unintended consequence: a narrow approval could invite follow-on restrictions, making staged, scale-in positions preferable to full conviction buys.

Key Decisions for Investors

• Establish a 2% portfolio long in NVDA using 12-month LEAP call spreads (buy Jan-2026 ~40% delta call, sell 20–30% OTM call) to capture upside while capping cost; if a favorable license is announced within 30 calendar days, add another 1% via a 3–6 month call spread.
• Initiate a 6-month pair trade: long NVDA equal-dollar / short AMD (AMD) equal-dollar, size combined = 1.5% portfolio; target 20% relative outperformance, stop-loss at 10% relative move against the pair.
• When an official decision date is scheduled within 7 trading days, buy a 30–45 day NVDA strangle (10% OTM calls and puts) sized so premium ≤3% of notional; close within 2 trading days post-announcement or if implied vol rises >40% from entry.
• Reduce China cloud exposure by rotating 2–4% portfolio weight from BABA/TCEHY into US AI infra suppliers (NVDA, TSM) if internal tracking shows license language granting H200-class exports; reverse within 10 trading days if license is denied or materially narrowed.