Washington closes the cloud loophole: licences now required for AI chips rented abroad to China-linked firms
The Commerce Department's new rule targets data centers anywhere in the world that rent more than 12 million TPP of computing power to firms headquartered in China and other restricted countries.
📊 Scorecard: our Bearish call on $NVDA is scored from $156.14, the last close before it ran. Its first result comes after the next close.

The News
The U.S. Commerce Department's Bureau of Industry and Security issued an interim final rule on Monday, November 16, titled "Addressing Remote Access to Advanced Semiconductors" (RIN 0694-AK50). It closes the route by which Chinese firms rented advanced American computing power from overseas data centers instead of buying restricted chips — the workaround known as the cloud loophole.
Under the rule, a licence is now required to supply advanced chips or servers to any data center, anywhere in the world, that provides more than 12,000,000 aggregate total processing performance (TPP) of compute — roughly the power of 750 Nvidia H100 or H200 chips — to companies headquartered in Macau or a Country Group D:5 destination, a list that includes China, Russia, Iran and North Korea. Separately, advanced-chip exports to Indonesia, Malaysia, Singapore, Thailand and Brazil now need a licence, with a reporting-based exclusion for U.S.-headquartered recipients. Commercial recommender systems are exempt.
The rule takes effect when it is published in the Federal Register; no publication date has been set yet, so no effective date exists on the calendar. November 19 is the date of the Trump-Xi meeting in Shenzhen. Existing operators get a 90-day window to obtain licences, and there is a 60-day comment period. The rule also formally rescinds the Biden-era AI Diffusion Rule, which the Bureau says was never enforced. It lands days before President Trump meets President Xi in Shenzhen on Thursday, November 19, at the APEC summit, and days after Reuters reported Chinese AI firms were rushing to reserve overseas Nvidia-powered capacity ahead of tighter controls.
Washington just put a licence on the cloud itself — and the market is pricing in demand that may never come back.
Why It Matters
The rule hits the chip industry's fastest-growing way of selling computing power: renting it, not selling the chips. Chinese AI firms that could not buy Nvidia's most advanced chips could still rent them by the hour from data centers in Singapore, Malaysia and elsewhere. That workaround kept Nvidia's chips in demand even under export controls. Closing it takes a real chunk of that demand off the table, and investors dislike demand that disappears by government order.
The reach is unusually wide. The licence requirement applies to any data center anywhere in the world, not just in the listed countries. A server farm in Southeast Asia serving a Chinese customer now needs a U.S. licence, so chipmakers and customers must track who ultimately uses their computing power. That adds paperwork, delay and legal risk to deals that used to close in days.
The timing adds political risk. The rule lands days before the Trump–Xi meeting in Shenzhen, so the market will read any retaliation — or any softening — as a signal about where U.S.–China tech policy goes next. The 90-day window gives operators time, but it also gives the market 90 days to wonder how many licences get granted and how many get denied.
Analyst's Take
Our call is Bearish. This is a structural demand cut, not a one-day headline: the world's largest AI market just lost its easiest legal route to advanced American computing power, and the licence process is a choke point the U.S. controls.
The near-term question is how the licence process actually works — approval rates, processing times, and whether the 90-day window becomes a grace period or a cliff. If licences flow freely, the damage is modest. If they do not, overseas data center operators will cut orders for Nvidia, AMD and TSMC chips, and the chip rout gets a second leg.
For readers, the practical point is that chip demand now has a political ceiling as well as a commercial one. Watch for the Federal Register publication — no date has been set yet — and signals from the Trump–Xi meeting on November 19. Reader poll: Bullish or bearish on $NVDA after the new rule? Vote: Bullish, Bearish, Mixed.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is TPP?
TPP stands for total processing performance, a measure of how much computing power a chip or a data center can deliver. The rule sets the licence trigger at 12 million TPP, roughly the power of 750 top-end Nvidia chips.
Why does this hurt Nvidia?
Chinese firms that could not buy Nvidia's most advanced chips were renting them by the hour from overseas data centers instead. Closing that route removes a source of demand for Nvidia's chips, and for the TSMC-made chips inside them.
What was the AI Diffusion Rule?
A Biden-era framework for controlling AI chip exports that the new rule formally rescinds. The Bureau of Industry and Security says the old rule was never enforced.
👁
Source: SOFX