Incident sheet · /security

Filed
Read
3 min · 699 words
Severity
Routine

U.S. Tightens AI Chip Export Controls on Overseas Data Centers

New U.S. licenses cover AI chips and servers feeding overseas data centers above 12 million TPP, closing a remote-compute route into China and other embargoed countries.

US tightens AI chip export controls to deter Chinese companies from accessing compute remotely
FeedUS tightens AI chip export controls to deter Chinese companies from accessing compute remotely — frabre / pixabay

Event log

  1. Coverage threshold: data centers or networks above 12 million TPP, roughly 750 Nvidia H100 or H200 chips.
  2. Remote-access reach: the 22-country D:5 group comprises China, Russia, Iran, North Korea and 18 others.
  3. Compliance deadline: existing covered data center operators must obtain licenses within 90 days.
  4. Diversion-risk markets: Malaysia, Singapore, Indonesia, Thailand and Brazil face new advanced-semiconductor licensing, except chips destined for U.S.-based companies.
  5. Policy reversal: the rule rescinds the Biden administration's AI Diffusion Rule issued in January 2025.

By Olivia Hart

A single overseas data center with the maximum computing power of about 750 Nvidia H100 or H200 chips will trigger U.S. export licensing if it may provide compute to a company in China or another embargoed country.

The Trump administration is preparing to roll out the interim rule on Monday. It targets a remote-access loophole by requiring licenses for exports, reexports and transfers of AI chips and servers to covered data centers anywhere in the world.

Existing U.S. rules already require chip companies to obtain licenses before shipping AI chips to companies based in, or owned by a business in, the “D:5” embargo group. The 22-country group includes China, Russia, Iran, North Korea and 18 others.

“This new control is a logical extension of existing controls,” the interim rule reads. “These restrictions will undermine the ability of companies headquartered in countries of concern from using U.S. advanced Super Intelligence (SI) semiconductors to train advanced SI models.”

The 12 million TPP trigger

The rule covers any data center or network able to supply more than 12 million total processing performance (TPP), the export-control metric for computing power. Twelve million TPP roughly matches the maximum computing power of about 750 Nvidia H100 or H200 chips.

The main controls are:

  • Exporters need licenses for covered AI chips and servers whenever the destination data center may serve a company in China or another restricted country.
  • Indirect access through brokers and other go-betweens falls within the rule.
  • Exporters and data center operators must monitor customers’ ownership structures and activities to ensure that people from the D:5 group do not run those customers.
  • Existing data center operators with covered services must obtain licenses within 90 days after the rule takes effect.

New diversion-risk licensing

The Commerce Department also added licensing requirements for exports, reexports and transfers of advanced semiconductors to Malaysia, Singapore, Indonesia, Thailand and Brazil.

It included those five countries because they present a “higher-than-normal risk of diversion,” although the new requirement does not apply to chips “destined” for companies based in the United States.

The rule does not apply to commercial recommendation systems, including those used in e-commerce and social media.

What does this change for vendors and data center operators?

For exporters, the change makes customer ownership and intended use part of the licensing decision. For data center operators, the rule creates a 90-day licensing deadline for existing covered services and a continuing obligation to monitor ownership and activity.

For data center customers, the threshold turns remote capacity into a controlled relationship: once a facility can provide more than 12 million TPP, access tied to a D:5 customer can bring the service into the licensing regime.

The policy shift came days after Nvidia co-founder Jensen Huang resigned amid a scandal involving an alleged Chinese spy linked to one of the chipmaker’s customers. The U.S. government reportedly wanted that customer, Megaspeed, added to a special entity list. Officials believed it had bought thousands of Nvidia chips to provide compute to Chinese businesses and work around existing export restrictions.

Reuters reported last week that Chinese AI companies were racing to secure computing capacity overseas as they anticipated tighter U.S. controls.

The interim rule also rescinds the Biden administration’s AI Diffusion Rule from January 2025, which imposed country-level limits on exports of advanced AI chips and restricted transfers of certain powerful AI models. In May, the Trump administration said it would discard that framework because it placed unnecessary restrictions on U.S. companies and the country’s allies.

“By making these changes, this IFR prevents the use of certain SI items in data centers and at a computing power level that can be used for training or running of advanced SI models,” the rule reads. It adds that such compute could support military and intelligence activities that threaten U.S. national security or foreign policy interests.

For AI vendors and operators, the practical line is now clear: the 12 million TPP threshold, not the physical border of a chip shipment, determines when overseas compute becomes a U.S. export-control issue.

via techcrunch.com (Source)

Filed under

  • ai-chip-export-controls
  • overseas-data-centers
  • semiconductor-regulation
  • export-licensing
Share this article:

More from Olivia Hart

Olivia Hart

Show full bio

News editor covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.

207 articles

The newsroom digest

One email a day with the stories that matter. No noise.

By subscribing you agree to the privacy policy.

Cross-references

‹ Previous articleNext article ›