When the U.S. government announced at the end of 2024 that it would permit Nvidia H200 chips to be sold to China, it appeared to be a turning point in a years-long technological blockade. But between the permission and an actual deal, a gap emerged — and it has a specific address: Beijing.

Permission Exists, but Deliveries Don't

Lenovo publicly confirmed to Reuters that it is "one of several companies permitted to sell H200 in China under Nvidia's export license." The list of approved buyers also included Alibaba, Tencent, ByteDance, JD.com, and Foxconn — roughly a dozen firms in total. Each of them can purchase up to 75,000 chips.

However, according to Reuters citing sources, no actual deliveries have been made. Chinese companies withdrew after direct signals from their own government.

The "American Transit" Scheme and 25% to the U.S. Treasury

The deal contained conditions that the Chinese side viewed with suspicion. According to Reuters sources, Trump negotiated a scheme whereby 25% of the revenue from chip sales goes to the U.S. Treasury. To legally formalize this levy, the chips must physically pass through U.S. territory before being shipped to China — American law does not permit direct export tariffs. Beijing, according to the sources, sees this as a risk of outside interference or hidden vulnerabilities in the equipment.

"Their shift to Huawei underscores Nvidia's precarious position in China"

Reuters

Nvidia CEO Jensen Huang publicly warned that due to export restrictions, the company's share of the AI accelerator market in China has effectively fallen to zero. Before restrictions began, Nvidia controlled roughly 95% of China's advanced chip market, and China accounted for 13% of the company's global revenue.

Huawei Steps Into the Vacuum

While the deal stalls, Huawei is expanding its position. As Financial Times reports, Beijing instructed Chinese technology companies to limit Nvidia chip use to overseas operations while supporting domestic production. By forecasts, Huawei could become the leader of China's AI chip market by 2026 — a market valued at $67 billion by 2030.

This is not merely competitive struggle. The Council on Foreign Relations (CFR) notes a strategic paradox: the multi-year export restriction policy that the U.S. built over nearly a decade was de facto cancelled by the same administration that established it. And China's forced transition to its own solutions — Huawei Ascend and others — only accelerated precisely because the restrictions existed.

• H200 — Nvidia's second most powerful chip; the newest B200 is under even stricter restrictions
• 75,000 chips — the limit per approved buyer
• 25% of revenue — the U.S. share under Trump's scheme terms
• $50 billion — Huang's valuation of China's AI market in 2025

Notably, Nvidia halted H200 production in early 2025 — the company, according to Asia Times, decided to focus on products with clearer market prospects. A license to sell a chip that is no longer being manufactured looks like a symbolic concession without practical substance.

The key question here is not whether Nvidia will gain access to the Chinese market — but whether Washington can offer terms under which Beijing will agree to remain dependent on American chips, rather than making a final bet on its own technological stack. If the answer is "no" — the H200 deal will enter textbooks as the moment when the U.S. itself accelerated what it tried to prevent.