Restrictions Tighten on Incoming Microchip Shipments from the U.S.
The Biden Administration has imposed new restrictions on the export of U.S.-designed computer chips, aiming to prevent advanced AI technologies and semiconductors from reaching strategic adversaries, particularly China. The new measures establish a tiered, risk-based export control framework that depends on the destination country's relationship with the U.S. and the sensitivity of the technology involved.
Under this framework, the U.S. and its key allies or partners (Tier 1) face minimal restrictions and can access advanced computing chips without a license if security protocols are met. Most other countries (Tier 2) require specific licensing pathways or participation in validated end-user programs to receive controlled items. Countries under U.S. arms embargoes and high-risk areas like Macau continue to face strict import prohibitions on advanced AI semiconductors and model weights.
Recently, there has been a shift towards more targeted export controls, aiming to prevent diversion of advanced chips, especially through certain Southeast Asian countries like Malaysia and Thailand. These countries have become focal points due to their growing chip shipments and data center infrastructure, raising concerns that chips could be transferred onward to China.
The new rules cap exports of GPUs at 790 million TPP until 2027, roughly equivalent to 50,000 Nvidia H100 chips. Entities with "Verified End-User" status are limited to around 320,000 advanced GPUs over the next two years.
Tech giants Nvidia and Oracle have criticized the new measures, arguing that they amount to overregulation rather than safeguarding the interests of the U.S., its partners, and allies. Nvidia's Vice President of Government Affairs, Ned Finkle, stated that the new rules threaten to disrupt innovation and economic growth worldwide and undermine America's leadership. Ken Glueck, Executive Vice President at Oracle, echoed similar sentiments.
The new restrictions also impose a complete ban on receiving the technology for Russia, China, and Iran. Tier 2 includes 120 nations, including Singapore, Israel, Saudi Arabia, and the UAE, with limits on the amount of computing power they can purchase and the option to apply for extra capacity via license requests. A 120-day comment period for the new rules has been opened. Specific licensing requirements exempt some companies like Amazon Web Services or Microsoft from these limitations.
The rules introduce mandatory licenses for certain chips, and Tier 1 includes allied states such as Australia, Japan, South Korea, Taiwan, and 13 other countries, facing no new restrictions. The AI startup xAI launched a training cluster called Colossus using 100,000 Nvidia H100 GPUs, demonstrating the potential impact of these restrictions on the tech industry.
In conclusion, the Biden Administration's restrictions on exporting U.S.-designed computer chips represent a strategic effort to control the global diffusion of high-performance AI chips while allowing trusted allies access and imposing layered controls to mitigate security risks from less trusted countries. The new rules aim to prevent the transfer of advanced AI technologies to strategic adversaries, particularly China, while maintaining supply and cooperation with allies and tech partners in Tier 1 countries.
Data-and-cloud-computing technologies are a focal point of the new export control framework implemented by the Biden Administration, aimed at preventing advanced AI technologies and semiconductors from reaching strategic adversaries such as China. The technology sector, including tech giants Nvidia and Oracle, have expressed concerns that the new measures amount to overregulation and could disrupt innovation and economic growth worldwide.
