Nvidia Earns 69% of Its Revenue From the U.S. Here's How Tariffs Could Affect That in 2026.

Nvidia remains a dominant AI-accelerator supplier with ~69% of revenue from U.S. customers and a roughly 90% market share in AI chips; fiscal 2026 Q3 data center sales rose 66% YoY to $51 billion, diluted EPS was $1.30 (up 67%), and gross margin is near 74%. The company took a $4.5 billion Q1 charge for H20 chips it could not sell to China, but the U.S. has now permitted H200 sales to China provided Nvidia remits 25% of those sales to the U.S. government; China currently accounts for about 5% of chip sales (vs. 25% in 2022). While the allowance to sell H200s could modestly increase China revenue in 2026, the 25% revenue cut and Chinese policy to develop domestic chips leave the scale and margin impact uncertain, with core U.S. data-center demand expected to remain robust.

Analysis

Market structure: Nvidia (NVDA) remains the primary beneficiary — ~90% AI accelerator share, 69% revenue from U.S. buyers, and 74% gross margin give it persistent pricing power even if China sales tick up from ~5% to mid‑teens. Winners include U.S. cloud/data‑center operators (META, GOOGL) that will keep buying; losers are marginal Chinese AI hardware entrants whose addressable market shrinks once H200 is available but taxed. The U.S. government is an unconventional beneficiary via a 25% revenue carve‑out on H200 China sales.

Competitive dynamics & supply/demand: Tight foundry capacity at TSMC (TSM) and ramp plans for H200 suggest supply will remain a constraint through 2026, supporting ASPs and durable margin tailwinds for NVDA; however, the 25% revenue transfer is an effective margin tax on China sales and caps profitability of incremental China growth. If NVDA scales China sales meaningfully (>8–10% of revenue), the net contribution margin will be materially lower than domestic sales and should be treated as lower‑quality growth.

Risk assessment: Tail risks include sudden reimposition of full export bans, a Taiwan supply disruption, or a rapid drop in data‑center spending (demand shock >20%) — any of which could cut NVDA revenue by double digits in quarters. Immediate effects (days) will be policy‑driven headline moves; short term (weeks–months) inventory and guidance resets; long term (2026–2030) depends on capex trajectory ($3–4T data‑center spend to 2030).

Trade implications & cross‑asset: Expect NVDA equity outperformance, tighter credit spreads for large cloud buyers, higher call IVs and richer option skew on NVDA, modest upward pressure on copper/energy (data‑center build), and episodic CNY demand swings. Catalysts to watch: US final rule text on the 25% carve‑out, NVDA quarterly guidance, TSMC capacity commentary, and Chinese subsidy announcements for domestic chips.

Key Decisions for Investors

• Establish a 2–3% long NVDA equity position (buy into weakness or size in tranches over 4 weeks). Add another 1% if NVDA trades down >10% intraday or if guidance implies China revenue <8%. Target +30% in 6–12 months; hard stop at -15% from entry.
• Initiate a 1–2% long TSM position as a foundry play (12–24 month horizon) but hedge tail risk by buying 3‑month TSM ATM puts sized to 50% of the equity notional if Taiwan geopolitical risk spikes; trim if TSM guidance shows capital allocation >20% toward mature nodes.
• Open a relative‑value pair: long NVDA (1% notional) / short SMIC (0981.HK) (0.5–1% notional) to express widening gap between world‑class fabless demand and lagging Chinese foundry capability; review after each quarterly report and adjust if SMIC gross margin narrows <10%.
• Options tactic: Buy 6‑month NVDA calls (target delta ~0.30) sized to 50% of a full equity position, finance by selling 1‑month OTM calls (calendar/diagonal) to collect premium. Alternatively, sell cash‑secured NVDA puts 15% OTM if willing to acquire at a discount; reprice if IV spikes >40% above 90‑day average.