AMD wins major Alibaba MI308 chip order to challenge Nvidia H200

Nvidia CEO Jensen Huang secured US President Donald Trump’s approval to allow the H200 accelerator to return to the China market, with shipments to Chinese customers reportedly fast-tracked for around February 2026. Reinstating H200 exports would reopen a large AI compute market for Nvidia and could materially boost revenue and China-facing product shipments, while also signaling a calibrated US stance on high-end AI export controls that affects global supply chains and competitors.

Analysis

Market structure: Approval to ship H200s back into China is a material re-opening of a vast addressable market — expect direct beneficiaries to be NVDA, Chinese cloud providers (BIDU, BABA) and AI data-center integrators that will capture incremental GPU capacity. Pricing power for NVDA's data-center stack should remain intact near-term because H200 supply will be constrained; estimate a 5–15% incremental uplift to NVDA datacenter revenue over 12–18 months if shipments commence Feb 2026 as reported. Cross-asset: risk-on flows should support equities and EM FX (CNY), compress credit spreads for high-quality tech issuers, and push implied vols on NVDA calls lower as headline risk recedes.

Risk assessment: Tail risks include a policy reversal or new export controls (high-impact, low-probability) that could wipe expected China revenue and cause sharp re-pricing; another tail risk is Chinese countermeasures limiting NVDA market share domestically. Short-term (days–weeks) expect headline-driven volatility; medium-term (months to early-2026) execution risk around shipping windows; long-term (years) the structural lead in accelerator IP persists but depends on TSMC/Samsung capacity, HBM supply, and US-China geopolitics. Hidden dependencies: Nvidia’s ability to supply is gated by foundry/packaging/HBM bottlenecks and customer certification cycles in China.

Trade implications: Primary trade is directional NVDA exposure while hedging geopolitical execution risk. Use staggered entries: accumulate now into small core (3–5% portfolio) and add on confirmed channel/ship notices; employ LEAP call buys to capture convexity and short-dated call sales to finance theta if IV compresses. Pair trades: long NVDA vs short AMD (AMD) or INTC (INTC) to isolate GPU share gains; rotate 1–2% into China cloud names on confirmed order flows.

Contrarian angles: Consensus may underprice the chance NVDA offers price concessions or limited functionality SKUs to re-enter China, which would boost volumes but compress ASPs and margins — downside not fully discounted. Equally, the market may underweight the potential for this re-entry to accelerate China’s investment in domestic accelerators, eroding long-run TAM; historically similar relaxations were followed by policy swings, so prepare for a 20–40% volatility regime if diplomacy turns sour.

Key Decisions for Investors

• Establish a phased long NVDA (NVDA) position equal to 3–5% of portfolio: buy 50% immediately and tranche remaining 50% on any NVDA pullback >8% or upon formal US export-license confirmation; set a hard stop at -18% from average entry or if US policy publicly reverses before Feb 2026.
• Buy NVDA Jan 2027 LEAP calls ~15–25% OTM (target delta ≈0.30) sized to 0.5% of portfolio notional to capture upside into shipments (Feb 2026) while limiting capital at risk; finance by selling 1–3m covered calls 20–30% OTM after rallies >25%.
• Execute a relative-value pair: go long NVDA and short AMD (AMD) on equal dollar basis for a 0.5–1% portfolio position to isolate Nvidia-specific China re-entry upside; rebalance if spread closes >15% or NVDA guidance misses by >5% revenue consensus.
• Allocate a tactical 1–2% into Chinese cloud beneficiaries (BIDU, BABA) on confirmed procurement announcements or if each name falls >15% from current levels, and exit/trim if shipments to China are delayed beyond Apr 2026 or if Chinese regulatory action targets cloud AI deployments.