NVIDIA Gives Up On China

Nvidia has halted production of H200 chips destined for the Chinese market amid persistent regulatory barriers from both Washington and Beijing, according to the FT, abandoning expected Chinese demand that was rumored to reach into the hundreds of thousands of units. CEO Jensen Huang flagged no China revenue in mid‑2025 guidance, yet the company still projects robust growth—cited here as roughly 70% quarterly top‑line growth—and maintains a market capitalization above $4 trillion, underscoring that lost China sales may be manageable against Nvidia’s broader AI leadership and outlook.

Analysis

Market structure: Nvidia’s decision to stop H200 shipments to China benefits US cloud providers (AMZN, MSFT, GOOGL) and Nvidia’s pricing power because constrained Chinese demand lowers geopolitical execution risk while leaving Western demand intact; expect NVDA to retain >50% gross margin leadership and sustain ~50–70% YoY revenue growth next 2–4 quarters if guidance holds. Chinese foundries (SMIC) and system integrators lose near-term addressable GPU demand—this shifts semiconductor CAPEX from China to Taiwan/US suppliers (TSM, LRCX, AMAT) and compresses China hardware multiples by 10–30% relative to U.S. peers.

Risk assessment: Tail risks include accelerated export controls, Chinese tech self-sufficiency (3–5 year horizon), or a Taiwan supply disruption that would cut Nvidia wafer supply (TSMC dependency); each could move NVDA ±30–60% in stressed scenarios. Short-term (days-weeks) watch for NVDA guidance updates and Treasury/BIS announcements; medium-term (3–12 months) monitor Chinese IP/system releases (DeepSeek) and foundry capacity reports as catalysts.

Trade implications: Favor convex long exposure to NVDA via defined-risk structures (3–6 month call spreads 10–20% OTM) sized 1–3% portfolio; hedge tail risk with 6–12 month 15–25% OTM puts (0.5–1% notional). Pair trade: long US cloud/AI infra (MSFT, AMZN, LRCX/AMAT) and short Chinese semiconductor plays (SMIC) for 3–12 months to capture demand reallocation and multiple compression.

Contrarian angles: The market underestimates that cutting China exposure may raise NVDA’s near-term free cash flow by concentrating sales in higher-margin enterprise/cloud channels, so upside is underpriced; conversely, consensus underestimates China’s ability to close the algorithmic gap in 36–60 months, which could cap hardware TAM by 10–30% long-term. Watch for unintended consequence: software/model efficiency gains reducing GPU demand growth — if telemetry shows per-inference FLOP decline >15% YoY, reprice GPU-capex plays.

Key Decisions for Investors

• Establish a 2% portfolio long in NVDA via a 3–6 month call spread 10–20% OTM (defined-risk) sized to ~2% notional; hold through next two earnings, trim on a 30%+ price move or after two consecutive beats.
• Purchase 6–12 month NVDA puts 15–25% OTM sized to 0.5–1% notional as insurance against sudden regulatory/geopolitical shocks; roll or unwind on clarified export-control language within 30–60 days.
• Implement a pair trade: long MSFT (1.5%) and AMZN (1.5%) vs short SMIC (1%); target 6–12 month horizon—exit if SMIC reports >20% QoQ revenue growth or if US export controls materially change.
• Reduce direct China semiconductor hardware exposure (SMIC/Hua Hong/etc.) by ~50% within 30 days and redeploy proceeds into US AI infra suppliers (LRCX, AMAT, TSM) sized 0.5–1% each to capture capacity-driven wins over 6–18 months; monitor TSMC capacity and ASML shipment notices as stop/scale signals.