Washington moved on Monday to close the biggest remaining gap in its AI chip controls: remote access. The Commerce Department’s Bureau of Industry and Security issued an interim final rule requiring a license before advanced AI chips or servers can be sent to data centers anywhere in the world if those facilities might provide computing power to companies in China or other embargoed countries.

The rule also adds licensing requirements for advanced chip shipments to Singapore, Malaysia, Indonesia, Thailand, and Brazil, five countries the department says carry a higher-than-normal risk of chips being diverted. Data centers already serving covered customers have 90 days to get a license once the rule takes effect.

The cutoff is drawn in computing power, not geography. It covers data centers or networks able to supply more than 12 million total processing performance, an export-control metric known as TPP. TechCrunch estimated that is roughly the maximum output of about 750 Nvidia H100 or H200 chips, enough horsepower to train a serious AI model.

The timing isn’t a coincidence.

Four days earlier, Nvidia co-founder Jensen Huang resigned as chief executive after revelations that he had a relationship with a woman US intelligence officials suspected of being a Chinese intelligence operative. NPR has not independently verified those claims. Alice Huang, no relation to Jensen Huang, is a former executive at Megaspeed International, a Singapore-based data center firm with facilities in Malaysia. Megaspeed denies sending chips to China.

“The United States is the global leader in advanced computing and super intelligence technology and is committed to promoting the global adoption of American technology for legitimate commercial purposes while ensuring that sensitive technologies are not exploited in ways that threaten U.S. national security,” the Bureau of Industry and Security said in a statement on Monday.

What the new US AI chip export controls actually do

Until now, the controls mostly governed physical shipments. Since October 2022, US companies have typically needed a license to sell advanced computing chips above set thresholds to China, including Hong Kong and Macau. But a Chinese company didn’t need to buy a chip at all. It could rent time on one in a data center abroad. Monday’s rule treats that remote access the same way as a physical sale.

The remote-access provisions target companies based in the D:5 group of embargoed countries, China, Russia, Iran, North Korea, and 18 others, and cover indirect access through go-betweens or brokers, so routing a rental through a middleman doesn’t dodge the rule. End users that are, or are owned by, US-headquartered companies are exempt. Fortune reported that Microsoft, Oracle, and Amazon are expected to qualify, though the remote-access limits would still apply to them.

Commercial recommendation systems, the kind used by e-commerce sites and social media feeds, are explicitly excluded.

Five countries now facing license checks

Singapore, Malaysia, Indonesia, Thailand, and Brazil are the five destinations singled out for new chip-shipment licenses, chosen for diversion risk: the concern that chips sent there could end up somewhere they shouldn’t. The Megaspeed case put that risk in sharp focus. The firm is based in Singapore, runs facilities in Malaysia, and stands accused of funneling Nvidia chips to China. Megaspeed denies it.

The Nvidia scandal behind the rule

Huang’s resignation on Thursday, November 12, ended a 33-year run at the company he co-founded and built into the world’s most valuable chipmaker. He had personally lobbied President Trump to loosen restrictions on Nvidia chip sales to China. The New York Times reported Trump was first briefed by intelligence officials about the relationship in October 2025.

Both Huang and Commerce Secretary Howard Lutnick are expected to testify before Congress. Rep. Yvette Clarke told NPR the episode needs a full airing under oath: “We need to get the facts investigated, bring in individuals to give us as much information as possible under oath.”

Why this matters

This rule marks a shift in how Washington polices AI. The old question was where a chip physically goes. The new question is who gets to use the computing power, wherever the hardware sits. That turns on knowing your customer, not just your shipping address.

It also formally buries the Biden administration’s AI Diffusion Rule, the January 2025 framework that would have set country-level limits on advanced chip exports. The Trump administration said in May it would not enforce it, calling it ill-conceived and counterproductive. Monday’s rule rescinds it outright.

The rule takes effect once it is published in the Federal Register, according to Fortune. Nvidia reports quarterly earnings on Tuesday, November 17, and the first questions on that call will almost certainly be about how much of its business now runs through a license queue.

Frequently asked questions

What are the new US AI chip export controls?

The Commerce Department’s interim final rule requires a license before advanced AI chips or servers can go to data centers that may provide computing power to companies in China or other embargoed countries, including through remote access.

Which countries need licenses for AI chip shipments now?

Advanced chip shipments to Singapore, Malaysia, Indonesia, Thailand, and Brazil now require a license. Shipments meant for US-based companies stay exempt, with exporters filing monthly tracking reports.

What is the 12 million TPP threshold?

The rule covers data centers or networks able to supply more than 12 million total processing performance, an export-control measure of computing power. TechCrunch estimated that is roughly the output of about 750 Nvidia H100 or H200 chips.

Why did the US tighten the rules now?

The move follows the Nvidia scandal: CEO Jensen Huang resigned on November 12 after revelations about a relationship with a woman US intelligence officials suspected of being a Chinese intelligence operative.

Sources: TechCrunch, Fortune, NPR.