Science & Tech · AI · 2 hrs ago
How China accesses Nvidia AI chips despite US export controls
US export rules block Nvidia’s most advanced AI chips from being shipped to China over security concerns.
Nvidia’s graphics processing units can handle the large amounts of data needed to train and run artificial intelligence.
Chinese AI companies have released powerful models, and analysts say some are still using advanced Nvidia chips despite the restrictions.
One route is renting computing power from data centres in other countries, where the chips are installed and can be accessed online.
This can let a Chinese customer use a chip abroad without it being shipped into China, though the article says export rules restrict where chips are shipped, not who uses them.
US officials have also accused Chinese AI labs of using high-end Nvidia chips, while a technology outlet reported that DeepSeek used chips brought into China after being dismantled overseas; Nvidia said it had seen no evidence of that.
China’s access matters because advanced chips can make AI development faster, even as Chinese chipmakers offer alternatives that may take more time and engineering to use.
US export controls bar Nvidia’s top-end Blackwell and Rubin AI chips from China, but Chinese firms may access Nvidia computing power through overseas data centres or smuggling.
Chinese companies can rent computing power at data centres in third countries, including Malaysia, while their customers log in from China.
The United States eased controls on Nvidia’s less powerful H200 chip in 2026, but conditions have made shipments difficult to approve.
US officials have also said Chinese AI labs accessed advanced Nvidia chips, while reports have alleged that chips were smuggled into China.
Chinese-made chips, including those from Huawei, are improving, though training AI models on them can take more time and engineering.
- Who
- Chinese AI companies, Nvidia and US officials are involved.
- What
- Chinese firms are obtaining or accessing Nvidia AI computing power despite US export controls, including through overseas data centres and possibly smuggling.
- When
- The articles were published on November 10, 2026; they also describe developments in 2025, 2026 and 2027.
- Where
- China, with computing power potentially accessed from third-country data centres, including in Malaysia and Thailand.
- Why
- Nvidia chips help train and run AI models; US restrictions cite security concerns and the military applications of advanced AI.
US government
Jensen Huang
Export controls
US government
The US bars exports of Nvidia’s top-end Blackwell and Rubin chips to China, citing security concerns and advanced AI’s military applications.
Jensen Huang
Huang has lobbied the US government to allow Nvidia chip sales to China and warned that restrictions are fuelling the growth of domestic alternatives.
ChatGPT’s launch showcased AI’s potential and helped spark an investment boom.
The Information reported that DeepSeek developed a new model using thousands of chips dismantled overseas and brought into China.
Washington eased controls on Nvidia’s H200 chip, while taking a 25% cut of sales.
Huawei said its next-generation AI chip would be ready in the first quarter of 2027, earlier than planned.
Huawei’s next-generation AI chip is expected to be ready.
- Restricted chips
- Nvidia Blackwell and Rubin GPUs
- H200 sales cut
- 25% for the United States
- Overseas data-centre example
- Malaysia’s Johor region
- Access claim
- White House adviser Michael Kratsios said Moonshot AI accessed high-end Nvidia chips in Thailand.
- Chinese chipmaker
- Huawei
- Huawei chip target
- First quarter of 2027
2 sources
What people are saying
Chinese vendors are still using cutting-edge Nvidia chips to train their models
US export controls restrict where a chip is shipped, not who uses it
Nvidia chips still make top-level AI work faster and easier, but they are no longer the only path
seems far-fetched