Nvidia wins US license to export limited H200 chips to China

Nvidia has received a US license to export a limited number of its less-advanced H200 AI chips to China, according to the company.

The license allows shipments of a limited number of these chips, subject to US inspection and a 25% duty.

Nvidia said it has not yet generated revenue from these exports and is uncertain if Chinese authorities will permit the imports.

The company did not include China data-center revenue in its first-quarter sales forecast.

The license marks a step in Nvidia’s efforts to re-enter the Chinese market, which has been restricted by US-China tensions.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Licensing Nvidia’s H200 chip reads as a tactical shift meant to slow China’s push for tech self-reliance, not a policy U-turn.
US officials made the call after reviewing progress in Huawei’s homegrown AI chips, which are getting more competitive 1.
The aim is to keep Chinese AI work tied to Nvidia’s CUDA software platform for developing and running AI models, which helps preserve the “dominance of the American tech stack” 1.
Allowing exports of a powerful, older chip seeks to reduce incentives for China to build a separate AI ecosystem that cannot work with US tools 1.

These export rules matter beyond Nvidia’s quarterly results because they speed up a break in the global AI market.
China has answered with caution, telling its largest tech companies to avoid H200 purchases “unless necessary” 2.
This struggle is pushing the industry toward two tech paths. One leans on Nvidia’s CUDA platform, while the other grows around Chinese options such as Huawei’s MindSpore, an AI software framework 3.
The momentum could create a “5G-style” decision for companies and governments, steering them toward either the US or Chinese technology stack and creating long-term interoperability problems 3.