US clears H200 sales to 10 Chinese firms but no deliveries land

TL;DR:

• The US has cleared around 10 Chinese firms — including Alibaba, Tencent, ByteDance, and JD.com — to buy Nvidia’s second-most-powerful AI chip, the H200, with Lenovo and Foxconn among the approved distributors, but no deliveries have yet been made.
• Each approved buyer can purchase up to 75,000 chips under US licensing terms. Nvidia CEO Jensen Huang has joined a White House delegation to Beijing this week to push the deals across the line.
• Trump negotiated an unusual revenue-share arrangement requiring 25% of chip-sale proceeds to flow to the US, with chips physically routing through US territory en route to China — a structure designed around the legal prohibition on direct export fees.

The shape of the stalled deal is the story. Before US export curbs tightened, Nvidia commanded roughly 95% of China’s advanced-chip market and China accounted for 13% of Nvidia revenue. Huang has previously estimated the Chinese AI market alone would be worth $50bn this year. Today, Huang says Nvidia’s share of AI accelerators in China has effectively fallen to zero — Chinese AI firms like DeepSeek now publicly tout reliance on domestic chips, including those developed by Huawei.

Why no deliveries

Despite approval, Chinese firms have pulled back after guidance from Beijing, sources told Reuters. The Chinese State Council issued two supply-chain security regulations recently, triggering a government-wide effort to identify and eliminate foreign-tech dependencies in critical infrastructure. Commerce Secretary Howard Lutnick told a Senate hearing last month that Beijing has not let domestic buyers take the chips because it is trying to keep investment focused on China’s own chip industry. The 25%-to-Treasury revenue share has also raised unease in Beijing over potential tampering or hidden vulnerabilities.

UK angle: compute access and sovereign-AI strategy

For UK sovereign-AI strategy, the H200 episode is a reminder that compute access is geopolitical infrastructure. AI Growth Zones, the Compute Strategy, and the next round of AI Opportunities Action Plan funding all rest on assumptions about chip availability that the US-China dynamic can change without warning. The Council on Foreign Relations’ Chris McGuire framed the hardliner view that “any deal that allows Nvidia to sell more chips to China means fewer Nvidia chips for U.S. firms, and a smaller U.S. lead in AI over China” — a logic that, applied to UK compute allocation, could equally argue for prioritising UK access ahead of broader market.

Looking forward

Two questions over the next quarter. First, whether any of the 10 approved Chinese firms actually take delivery, or whether Beijing’s domestic-chip pivot makes the H200 clearance a moot diplomatic gesture. Second, whether the UK government formally raises with Washington how the H200 episode affects UK compute roadmaps — given that Nvidia is also the principal supplier for UK AI Growth Zone procurement, supply-allocation politics in the US-China lane bleed directly into UK timelines.