What is U.S.’s new rule for exporting AI chips? | Explained

New U.S. regulations on AI chip exports aim to enhance national security and control technology diffusion globally.

The story so far: With an objective to advance increased control over circulation of Artificial Intelligence (AI) chips and technology, the U.S. Bureau of Industry and Security (BIS) introduced a tiered framework for licensing and exports. It held the regulations were in tandem with “national security and foreign policy interests”. Further, it would help “cultivate a secure and trusted technology ecosystem for the responsible use and diffusion of AI”.

Broadly, the regulation concerns advanced computing chips and certain closed AI model weights. AI models are software programs that comprise a series of mathematical operations. When data is entered into the program, it executes those operations on the entered data (provided by the user) to produce outputs which could be information, analysis or media. It is the design of these operations and their arrangement, known as the model architecture, that determines the nature and quality of the output.

The new regulations update existing rules for export, re-export and transfers (in-country) by segregating countries into three tiers — each harbouring different rules for licensing and authorisation. The first of these tiers entail no restrictions for the export, re-export or transfer of advanced computing chips to 18 U.S. allies and partners, including Australia, Belgium, Canada, South Korea, U.K. and Japan, among others. The second tier introduces caps on volume and exemptions based on specifications alongside mandatory authorisation and licensing. It holds that transactions that could contribute to the development of advanced AI models would require a Validated End User (VEU) authorisation. Those that do not contribute to the development of advanced AI models, such as chips with a collective computation power of roughly 1,700 advanced GPUs, would not need an authorisation. China and India together have been categorised under this sub-head. The third tier includes arms-embargoed countries, such as North Korea, Iraq, Iran and Russia, among others. These countries will have no access to the technology.

One of the main objectives has been to ensure that the technology (or equipment) do not reach the ‘countries of concern’ or U.S. adversaries. Furthermore, as detailed in the Federal Register, it strives to ensure that model weights are stored outside the U.S. “only under stringent security conditions” and that the large clusters of advanced integrated circuits (ICs) necessary to train those models are “built in destinations that pose comparatively low risks of diversion or misuse”. The BIS determined that adversaries could potentially use the advanced AI systems to improve speed and accuracy of their military decision making, planning and logistics. BIS also observed that access to systems could potentially lower the barrier for non-experts to develop weapons of mass destruction, support offensive cyber operations and assist in human rights violations (such as through mass surveillance). Separately, it is essential to note that the BIS placed concerns about Chinese companies utilising “foreign subsidiaries in a range of uncontrolled destinations to buy ICs”.

The primary concern relates to the threat to U.S.’ global competitiveness in the realm. Ned Finkle, Vice President of Govt Affairs at NVIDIA — among the largest chip makers globally, wrote in a blogpost that it would undermine the innovation that has kept the U.S. ahead. Mr. Finkle argued that the rules would do nothing to enhance U.S. security. “The new rules would control technology worldwide, including technology that is already widely available in mainstream gaming PCs and consumer hardware.”

Ken Glueck, Executive Vice President at Oracle, wrote in December 2024, that the rules assumed there were no other non-U.S. suppliers to procure GPU technology from. Explaining the aspect about competitiveness, the Oracle executive stated that by adding “more chips to the problem, you can keep playing the game”. “If your alternate supplier has less performance, you can achieve parity by just adding more GPUs for the task. Enter Huawei and Tencent. Do it a cheaper price. Enter the CCP. And deploy it globally, enter Alibaba,” he explained. The rules have been introduced less than a week before President-elect Donald Trump’s return to the White House. The tech industry have pinned their hopes on the incoming administration to withdraw the framework.

According to Pranay Kotasthane, Chairperson at High Tech Geopolitics Programme at Takshashila Institution, big Indian data centres wanting to deploy advanced AI chips might need to apply for the VEU authorisation to speed things up. He observed that Indian companies acquiring the VEU authorisation can utilise the exported items for civilian and military purposes except for nuclear-end uses. “All in all, it does not seem to be a big deal. The bigger story is that India is not in the trusted allies and partners category, probably because of the leakages of chips to Russia,” he said.