Back to News
Market Impact: 0.35

Donald Trump government wants to take 25% 'cut' on the sale of Nvidia chips that China is not sure it wants to buy

NVDAAMD
Artificial IntelligenceTechnology & InnovationSanctions & Export ControlsTrade Policy & Supply ChainRegulation & LegislationAntitrust & Competition
Donald Trump government wants to take 25% 'cut' on the sale of Nvidia chips that China is not sure it wants to buy

The U.S. will allow exports of Nvidia’s H200 AI chips to China under strict conditions including a reported 25% government take on sales, caps limiting shipments to China to 50% of U.S. customer volumes, U.S. third‑party testing, and certification of sufficient U.S. supply; the Department of Commerce filing also applies similar rules to AMD’s MI325X. Nvidia welcomed the decision and says H200 production has restarted amid strong Chinese demand, but uncertainty over Chinese regulatory approval and the export restrictions limit near‑term revenue upside and leave the market impact moderate.

Analysis

Market structure: Nvidia (NVDA) is the direct beneficiary — US-approved H200 exports unlock incremental China revenue but with a de facto 25% margin haircut and a 50% shipment cap versus U.S. customer volumes, which limits upside and compresses gross margins on China sales by a material slice. Beneficiaries also include US/Taiwan fabs and semiconductor-equipment names (AMAT, LRCX, TSM) from restarted production; losers are firms reliant on undifferentiated GPU pricing power (AMD) and Chinese AI‑chip challengers as demand may concentrate on Nvidia’s proven stack over 6–24 months. Risk assessment: Key tail risks include Beijing refusing to permit imports (30–40% subjective near-term probability), US tightening of caps or adding licensing frictions, and third‑party testing bottlenecks creating supply delays; any of these could remove China demand within 30–90 days and knock ~5–15% off NVDA near-term revenue guidance assumptions. Time horizons: expect market reaction in days (news-driven), positionable moves over 1–3 months if purchase orders appear, and structural share shifts over 12–36 months as China builds domestic capacity. Hidden dependencies include certification/test lab capacity (single-point choke), contract-law risk for export licensing, and potential Chinese non-tariff retaliation targeted at US cloud/software customers. Trade implications: Favor tactical NVDA exposure to capture China demand while hedging regulatory outcomes — prefer limited-cost bullish option structures and small outright longs sized to conviction (2–4% position) with defined stops; overweight semiconductor capital equipment for 3–12 months (AMAT, LRCX, TSM ~1–2% each). Consider a relative-value pair (long NVDA, short AMD) to express Nvidia’s lead in high-end AI chips while neutralizing sector beta; size modest (1–2% net). Use 3–6 month protective puts or buy‑call spreads to cap downside from a sudden policy reversal. Contrarian angles: Consensus may underweight the probability China declines purchases — if Beijing allows buys, NVDA upside is underappreciated because current caps still permit meaningful revenue at scale (orders could plateau then re-accelerate), making short-term option skew rich; conversely, the market may be underpricing the negative earnings margin impact of the 25% government cut and compliance costs. Historical parallel: past US export relaxations (e.g., telecom equipment) show initial demand spikes followed by durable localization; monitor early order flow (first 60 days) as the decisive signal for sequenced investment scaling.