US Tightens AI Chip Export Controls to Block Remote Compute Access by Chinese Firms
The Trump administration is set to implement stringent new export controls on artificial intelligence chips and data centers located outside the Unite
S
··2 min readSciTech Editorial Team
Newsroom

The Trump administration is set to implement stringent new export controls on artificial intelligence chips and data centers located outside the United States. This regulatory move aims to prevent foreign entities, particularly in China, from bypassing existing restrictions by remotely accessing advanced AI infrastructure hosted in third-party countries. The interim rule, issued by the Commerce Department, mandates that companies obtain licenses to export, reexport, or transfer AI chips and servers to any global data center if there is a risk that such compute resources will be utilized by firms in restricted nations.
The new regulations specifically target the U.S. D:5 group of embargoed countries, which includes China, Russia, Iran, North Korea, and 18 other states. According to the interim rule, these restrictions are designed to undermine the capacity of companies headquartered in these 'countries of concern' to leverage U.S. advanced Super Intelligence (SI) semiconductors for training sophisticated AI models. The rule applies to any data center or network capable of supplying over 12 million total processing performance (TPP), a metric roughly equivalent to the maximum computing power of approximately 750 Nvidia H100 or H200 chips. Exporters and data center operators are now required to scrutinize their customers' ownership structures to ensure they are not controlled by entities from the D:5 group.
This escalation in export controls comes in the wake of a significant scandal involving Nvidia co-founder Jensen Huang, who recently resigned amid allegations linking a customer to a Chinese spy. Reports indicate that the U.S. government sought to place the customer, Megaspeed, on a special entity list due to suspicions that the company was purchasing thousands of Nvidia chips to provide compute power to Chinese businesses, thereby circumventing export controls. Reuters had previously reported that Chinese AI firms are aggressively securing overseas computing capacity in anticipation of such tighter U.S. restrictions.
Furthermore, the new rule introduces specific licensing requirements for the export of advanced semiconductors to Malaysia, Singapore, Indonesia, Thailand, and Brazil, citing a 'higher-than-normal risk of diversion' in these regions. However, these license requirements will not apply to chips destined for companies based in the United States. Existing data center operators falling under this new jurisdiction must secure licenses within 90 days of the rule's effective date. Notably, the regulations exclude commercial recommendation systems used in e-commerce and social media platforms from these strict controls.
In a significant policy shift, the administration has also rescinded the Biden-era AI Diffusion Rule from January 2025, which had imposed country-level limits on advanced AI chip exports. The Trump administration argued that the previous framework added unnecessary burdens on U.S. companies and allies. The new interim final rule asserts that preventing access to such high-level computing power is essential to stop the training of advanced SI models that could support military or intelligence activities threatening U.S. national security and foreign policy interests.



