Test report DSG-6302 · Rev C · tested November 17, 2026

Supply Chain & PolicyDevice under test

US issues interim final rule blocking Chinese remote access to AI servers

BIS has issued an interim final rule restricting Chinese firms' remote access to AI servers and cloud infrastructure, effective in 90 days, closing rental workarounds to chip export controls.

Read
2 min
Words
463
Node
5nm
Operator
Amara Osei

Spec summary

  1. BIS released an interim final rule restricting remote access to US-based AI technology, effective in 90 days.
  2. Nvidia Blackwell Ultra B300 GPUs were found in a China Telecom data center despite export controls.
  3. Nvidia required data center operators to source 51% of business from non-Chinese companies earlier this month.
  4. Licenses will be required for advanced chip exports to firms headquartered in Indonesia, Malaysia, Singapore, Thailand or Brazil.
  5. Higher-risk countries named in the rule include China, Russia and Iran.
New US rules aim to restrict Chinese firms’ remote access to AI servers and infrastructure
Fig. ANew US rules aim to restrict Chinese firms’ remote access to AI servers and infrastructure — blickpixel / pixabay

The US Bureau of Industry and Security (BIS) has published an interim final rule that restricts Chinese firms' remote access to AI servers and cloud infrastructure, with the new requirements taking effect in 90 days. The rule, developed with the Department of Treasury and the National Security Council, targets cloud-rental workarounds to existing hardware export controls.

The regulation backs up existing limits on exporting advanced AI hardware built by companies like Nvidia to non-whitelisted firms based in countries such as China. Its purpose, according to an official statement, is clear: "The rule is designed to ensure that existing restrictions on access to sensitive American technology cannot be circumvented through remote computing arrangements" using data centers in third-party countries.

Why did BIS act now?

Evidence has mounted that physical export restrictions are leaking. Nvidia may export its high-performance AI chips only to whitelisted companies, but reports of black-market GPU sales persist, and authorities recently cracked down on overseas AI GPU rentals from China.

The most concrete signal: Nvidia Blackwell Ultra B300 GPUs turned up in a data center used by China Telecom. Physical controls alone, in other words, did not stop the hardware from reaching a Chinese operator.

Nvidia has already moved partway on its own. Earlier this month the company told data center operators they must limit remote access to their services — no more than 49% of their business can come from Chinese companies, meaning 51% must come from non-Chinese customers. How effective that measure has been remains unclear, since powerful AI capabilities can be reached remotely rather than in person.

The takeaway: vendors' self-policing was not enough, and regulators have now stepped in with binding rules.

What does the rule change?

The IFR imposes three main requirements:

  • Overseas data centers must do more to limit access to their services wherever US-based technology is in use.
  • Restrictions apply to data centers supplying remote access to companies headquartered in higher-risk countries — China, Russia and Iran are named.
  • A license will be required to export advanced chips to companies headquartered in Indonesia, Malaysia, Singapore, Thailand or Brazil.

The requirements take effect in 90 days from publication.

The rule closes a gap the original export controls left open: computing power could be rented even where chips could not be bought.

What is the context at Nvidia?

The rule lands amid ongoing uncertainty over Nvidia, the US AI hardware leader at the center of the controls. CEO Jensen Huang resigned after confirming an affair, and now faces a potential quick-fire comeback — despite reports that the woman involved in the affair is an alleged Chinese spy.

Leadership turbulence at the company whose products the rule is written around adds another layer of instability to US AI export policy.

via tomshardware.com (Source)

Filed under

  • export-controls
  • ai-chips
  • cloud-infrastructure
  • nvidia
  • china
Share this article:

More from Amara Osei

Amara Osei

Show full bio

Staff writer covering money, funding and market shifts around hardware layer of artificial intelligence at Die Signal.

698 articles

Same lot · LOT-362D

« Previous articleNext article »