US Shuts the Cloud Loophole Letting China Rent Top AI Chips Abroad
Direct answer: The United States moved on Monday, November 16, 2026, to close a gap that let Chinese companies rent America’s most powerful AI chips from data centers in other countries. The Commerce Department’s Bureau of Industry and Security issued an interim final rule that treats rented computing power much like a physical chip shipment, and it now requires a licence in key cases.
What happened
- The Bureau of Industry and Security, known as BIS, issued an interim final rule, reference RIN 0694-AK50, titled Addressing Remote Access to Advanced Semiconductors. Reuters reported that the administration announced the measure on Monday, November 16, 2026.
- The rule covers data centers anywhere in the world that give more than 12 million aggregate TPP of computing power to companies headquartered in China, Macau or another Country Group D:5 country. TPP is the US measure of chip power, and the level works out to roughly 750 Nvidia H100 chips or about 330 Blackwell B200 chips.
- Exporters will also need a licence to send advanced AI chips and servers to Indonesia, Malaysia, Singapore, Thailand and Brazil, which the rule describes as carrying a higher than normal risk of diversion. CBS News and Reuters reported this part of the rule.
- Data centers already selling that much computing power to firms in the restricted group have 90 days to get a licence or face further export restrictions.
- The rule formally rescinds the Biden administration’s January 2025 AI Diffusion Rule, which sorted the world into tiers for chip access.
- BIS said the rule was written at the direction of the Treasury Department and the National Security Council.
- Companies headquartered in the United States are exempt from the new licensing requirement if they meet certain reporting conditions, and computing power used only for commercial recommender systems is carved out. You can read a detailed account of the rule, which cites the BIS text directly, here.
Why it matters for beginners
For four years Washington has banned direct sales of its best AI chips to China, but Chinese labs could rent the same chips from cloud companies in third countries. This rule closes that rental route, so the ban now covers rented access as well as bought chips.
It lands days before Nvidia reports earnings on Tuesday, November 17, its first results since chief executive Jensen Huang resigned on November 12. Treasury Secretary Scott Bessent has already told Nvidia to hire an outside chief executive and answer chip questions in 10 days, as covered in our earlier post. Nvidia shares were down about 2 percent in Monday trading.
The rule also arrives days before President Donald Trump meets Chinese President Xi Jinping in Shenzhen on November 19.
For beginners, the key point is simple. Chips are the engine of AI, and governments now treat access to computing power as a national security question, not just a trade question. When the rules change, the shares of chip makers and cloud companies can move fast, which is one reason AI stocks have been so jumpy in recent weeks.
What to watch next
- The rule still needs formal publication in the Federal Register, with public comments due 60 days later.
- Watch whether data center operators in the five named countries apply for licences or move customers elsewhere.
- Nvidia’s earnings on Tuesday will show whether the AI spending boom is still strong.
- The Trump-Xi meeting on November 19 may bring a reaction from Beijing.
FAQ
What was the cloud loophole?
Chinese firms could not buy top US AI chips directly, but they could rent computing time on those chips from data centers in other countries. The new rule requires a licence for that rented access above a set level.
What is TPP?
TPP stands for total processing performance, the US measure of how much computing power chips provide. The new threshold is 12 million aggregate TPP.
Does the rule ban all chip sales to the five countries?
No. It adds a licence requirement for the most advanced chips and servers going to Indonesia, Malaysia, Singapore, Thailand and Brazil. Ordinary business use is not banned.
Why is this happening now?
The rule follows fresh scrutiny of how restricted chips reached Chinese buyers after Nvidia chief executive Jensen Huang resigned on November 12. Reuters reported last week that Chinese AI firms were racing to reserve Nvidia-powered capacity from overseas providers before the rule landed.
