US prepares AI chip export controls targeting remote access to computing power

Executive Briefing
“The US administration is preparing to expand AI chip export controls to address access from...”
Key Takeaways
- 1.The supplied report says the administration is preparing controls aimed at companies that remotely access AI infrastructure in other countries to get around existing export restrictions.
- 2.Under the interim rule described in the report, companies would need licenses to export, reexport or transfer AI chips and servers to data centers that may provide computing power to businesses in China or other restricted countries. The targeted group includes China, Russia, Iran and North Korea.
- 3.The reported threshold covers data centers or networks able to supply more than 12 million TPP, a US export-control measure of computing power. The report estimates that this is roughly the maximum computing power of about 750 Nvidia H100 or H200 chips.
- 4.The reported rule also covers indirect access through intermediaries or brokers and requires exporters and data-center operators to monitor customers’ ownership and activities.
- 5.The report says new license requirements also cover certain advanced-chip transactions involving Malaysia, Singapore, Indonesia, Thailand and Brazil. It attributes the selection to the Commerce Department’s assessment of elevated diversion risk and says the requirement does not apply to chips destined for US-based companies.
The US administration is preparing to expand AI chip export controls to address access from China and other restricted countries to computing power hosted abroad. The supplied report says some shipments serving data centers would require licenses.
Key Developments and Technical Context
The supplied report says the administration is preparing controls aimed at companies that remotely access AI infrastructure in other countries to get around existing export restrictions.
Under the interim rule described in the report, companies would need licenses to export, reexport or transfer AI chips and servers to data centers that may provide computing power to businesses in China or other restricted countries. The targeted group includes China, Russia, Iran and North Korea.
The reported threshold covers data centers or networks able to supply more than 12 million TPP, a US export-control measure of computing power. The report estimates that this is roughly the maximum computing power of about 750 Nvidia H100 or H200 chips.
The reported rule also covers indirect access through intermediaries or brokers and requires exporters and data-center operators to monitor customers’ ownership and activities.
The report says new license requirements also cover certain advanced-chip transactions involving Malaysia, Singapore, Indonesia, Thailand and Brazil. It attributes the selection to the Commerce Department’s assessment of elevated diversion risk and says the requirement does not apply to chips destined for US-based companies.
Strategic Analysis and Industry Implications
If implemented as described, the controls would extend scrutiny beyond where chips are shipped to who uses the computing power they provide. Customer and ownership checks could therefore become a significant part of data-center compliance.
A license requirement does not mean every covered transaction is prohibited. Its practical effect will depend on how regulators assess applications.
Outlook and Key Factors to Watch
The supplied material does not establish the rule’s effective date or the criteria for granting licenses.
It remains unclear how many existing data centers and transactions exceed the TPP threshold or would require licenses.







