The US gives a formal green light to sales of Nvidia's H200 chips to China, which cannot receive more than 50% of the total amount of chips sold to US customers
Context & Ripple Effects
The decision implements the earlier policy direction to permit H200 shipments to approved Chinese buyers, following the administration’s conditional opening for H200 exports. It turns a broad political commitment into an enforceable allocation rule.
The move also sits alongside a same-day case-by-case licensing framework for H200 and AMD MI325X exports, making access conditional rather than broadly restored.
First-order effects
- Nvidia can pursue H200 sales to China under a formal US authorization, but Chinese deliveries are capped at no more than half of volumes sold to US customers.
- Chinese buyers gain a defined route to obtain H200 capacity, while Nvidia must allocate supply in a way that preserves the required US-sales base.
Second-order effects
- The cap makes US demand part of Nvidia’s China export capacity, strengthening the commercial importance of domestic allocations when supply is constrained.
- AMD’s MI325X is subject to the same case-by-case licensing approach, so China-facing AI-chip competition remains shaped by export approval as well as product demand.
Third-order effects
- AI-chip trade is moving toward conditional market access: governments can permit sales while retaining leverage over destinations, buyers, volumes, and supply allocation.
- If this model persists, chipmakers’ regional sales strategies will increasingly be designed around compliance constraints rather than a single global supply pool.
The trend: This is one data point in the shift from blanket AI-chip restrictions toward tightly managed, conditional access to advanced compute.