The Bureau of Industry and Security said in an announcement of the new remote access rule that it is targeting access to advanced American computing technology through data centers outside the United States. The agency said the interim rule is intended to align safeguards for remote use with controls that already apply when technology is physically exported.

The change reflects a basic shift in how artificial intelligence infrastructure is being bought and used. A restricted company does not necessarily need to import a shipment of advanced chips if it can rent enough computing power from a facility in another country. Under the new framework, Washington would treat that remote access as an export-control issue.

A new test based on computing capacity

The rule establishes licensing requirements for exports, reexports, and transfers of advanced semiconductors and related equipment to overseas data centers. The requirement applies when a facility can provide more than 12 million aggregate total processing performance, or TPP, to entities headquartered in Macau or countries in the government’s D:5 group, according to the interim final rule published by the Commerce Department.

That threshold is designed to capture facilities capable of supporting serious AI development rather than ordinary cloud workloads. TechCrunch reported that the capacity is roughly comparable to the computing power of about 750 Nvidia H100 or H200 chips. The publication also reported that the rule is aimed at preventing Chinese companies and other restricted organizations from securing advanced compute indirectly through foreign data centers.

The practical effect is to make cloud capacity part of the compliance chain. Chip exporters and data-center operators would need to investigate who owns a facility, who its customers are, and whether brokers or intermediaries are helping restricted entities gain access. The rule therefore reaches beyond the question of where a processor is installed. It asks who can actually use it.

For operators, that may require more than checking a customer’s billing address. Ownership structures, corporate affiliates, reseller arrangements, and patterns of activity could all become relevant. A data center that appears to serve a local company might still create risk if that company is acting for a prohibited customer or allowing access to a restricted organization.

More scrutiny for several countries

The rule also adds licensing requirements for advanced semiconductor shipments to Malaysia, Singapore, Indonesia, Thailand, and Brazil. The Commerce Department identifies those destinations as presenting elevated diversion risks, according to the Federal Register document.

The controls would not apply to chips destined for U.S.-based companies, the rule says. That distinction leaves American firms operating abroad with a different compliance position from local entities, while still placing pressure on suppliers to understand the final destination and intended user of their products.

Existing data-center operators affected by the rule would have 90 days to apply for licenses after the interim rule takes effect. That period gives companies time to determine whether their facilities cross the capacity threshold, review customer relationships, and seek authorization before continuing affected operations.

The government is also rescinding the Biden-era AI Diffusion Rule issued in January 2025 and replacing it with what it describes as a narrower structure focused on particular chips, data centers, and compute levels. Commercial recommendation systems used in areas such as e-commerce and social media are excluded, according to the rule.

Pressure from China’s overseas demand

The policy arrives as Chinese AI companies are reportedly competing for access to overseas computing capacity. TechCrunch reported that companies are rushing to secure foreign compute as U.S. restrictions make direct access to the newest American processors more difficult.

That demand has turned cloud infrastructure into a geopolitical asset. A data center can function as a physical warehouse for chips, but it can also act as a remote gateway to those chips. Engineers may be able to train or operate advanced models without the processors ever entering the country where the company is based.

The issue has also drawn attention to Megaspeed, an Nvidia customer that U.S. officials reportedly suspect of acquiring large quantities of Nvidia processors for Chinese businesses, according to TechCrunch. The report did not establish that Megaspeed violated the rules, but the case illustrates the problem regulators are trying to address: restrictions can be weakened when transactions pass through layers of customers and intermediaries.

Sen. Maria Cantwell and other Democratic senators have criticized Commerce for allegedly allowing advanced AI chips to reach China through overseas data centers and companies such as Megaspeed. In a letter to Commerce Secretary Howard Lutnick, they urged the department to act quickly to close what they described as export-control loopholes. No opposing response was included in the materials available for this report.

Compliance becomes part of infrastructure

The immediate burden will fall on companies that sell chips, build data centers, and provide cloud access. They may need systems capable of linking hardware inventories with customer identities and beneficial ownership records. They may also have to monitor how computing capacity is used after a contract is signed.

For AI companies, the broader change is strategic. Access to computing will no longer depend only on price, location, and technical performance. It may also depend on whether a provider can demonstrate that the customer behind a workload is permitted to use the hardware.

That could raise costs and slow the expansion of overseas AI infrastructure. It could also encourage companies to build clearer ownership structures and more transparent customer relationships. Either way, the rule signals that the next contest over advanced AI will not be fought only over who manufactures the chips. It will also be fought over who controls the rooms where those chips operate, and who is allowed to reach them from afar.

#U.S. Commerce Department#Bureau of Industry and Security#China#Nvidia#Megaspeed#Maria Cantwell#Howard Lutnick

Daniel Reyes is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.

The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.

How a story is made

  • Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
  • Writing. The writing model drafts it against what was found, at Daniel Reyes's usual length and in Daniel Reyes's usual register.
  • The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
  • Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
  • Fact check. A last pass hunts for claims the article makes and its sources do not.
  • A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.

If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.

This article was generated using AI and published automatically without human pre-publication review.

How this article was made

The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.