Reuters reports that several Chinese artificial-intelligence companies are negotiating to reserve computing time on foreign data centers that use high-end Nvidia chips. These powerful machines are essential for training and running large language models and other AI services.
What is happening and why it matters
The United States has restricted shipments of the most advanced Nvidia chips to China. Despite those constraints, Chinese developers can still access the technology remotely: they do not need to import the chips physically, but can instead use machines housed in overseas data centers via the internet and pay rental fees for the computation they consume.
A proposed U.S. law could narrow that access. The U.S. House of Representatives has passed a measure intended to limit remote access to foreign machines using restricted chips, but the bill has not yet been taken up by the Senate. If enacted, the proposal would curtail a route that many Chinese firms currently rely on.
Fears of further tightening after the Nvidia CEO controversy
According to Reuters sources, Chinese companies are rushing to secure access because they fear additional U.S. restrictions in the wake of controversy surrounding Nvidia CEO Jensen Huang. Huang has acknowledged infidelity in his marriage, and the New York Times reported that the woman involved — formerly an executive at Singapore‑based Megaspeed — was allegedly linked to Chinese intelligence.
Megaspeed has previously been accused of helping Chinese clients obtain Nvidia chips that are subject to export controls. These reports, together with potential legislative changes, have increased uncertainty in China’s AI sector and prompted firms to lock in necessary compute capacity in advance.
Potential consequences
If the Senate approves the House‑passed measure, limiting remote access to overseas data centers that run restricted GPUs would significantly complicate Chinese companies’ ability to use the latest hardware. That could slow development and reduce competitiveness for Chinese AI efforts, while firms try to preserve the compute resources they need to operate.
