Balancing national security and scientific research, US chip restrictions are marking a historical turning point that could reshape the dynamics of artificial intelligence development around the world.
The US government has announced a new set of regulations aimed at controlling the flow of American chips and related technologies abroad. This measure regulates the sale of advanced semiconductors essential for powering the most sophisticated artificial intelligence applications, amid growing concerns that these technologies could reach entities that might misuse them. In this article, we cover the key details of these rules and their implications for the chip industry and countries seeking advanced AI solutions.
What types of chips are covered by the restrictions?
The chips in question are known as graphics processing units (GPUs), which were originally used for gaming and advanced image rendering. However, their ability to process massive amounts of data in parallel has made them the backbone for training and running artificial intelligence models. An example is Nvidia, a pioneer in this field. The higher the performance of these chips, the greater their potential use in running large-scale artificial intelligence systems, such as generative models and advanced conversational assistants.
Read also: Digital arms race: Will artificial intelligence ignite a new cold war?
Details of the new restrictions
The restrictions focus primarily on what is termed “total computing power,” imposing a specific limit on units capable of performing 790 million operations per second through 2027. This cap is roughly equivalent to 50,000 Nvidia H100 chips, which possess extraordinary computing capabilities. These measures aim to ensure that an excessive share of chips is not concentrated within specific countries or entities, making the development of ultra-advanced AI systems restricted and tightly monitored. Additionally, licensed companies recognized as “validated end-users” can exceed these limits, provided they obtain prior authorization from US authorities.
Exceptions and special licenses
There are exceptions for universities, medical institutions, and research organizations purchasing limited quantities of chips—roughly equivalent to 1,700 H100 chips—requiring only notification rather than a license for these transactions. Home gaming chips are also exempt. Furthermore, a list of 18 international destinations considered close allies or partners of the United States, such as Japan, South Korea, and the United Kingdom, will not be subject to any purchasing cap. Meanwhile, other countries are required to comply with agreed-upon limits or undergo national licensing procedures.
Weights and their significance
The decision also covers what are known as “weights” in artificial intelligence models, which are the numerical parameters extracted after training models on massive data. This step aims to prevent the leakage of critical algorithms that could be exploited to develop advanced systems for illicit purposes. US law requires safeguarding these weights and prohibits their transfer to foreign entities except after an intensive security review.
Read also: Major breakthrough in artificial intelligence development is about to change the world
Expected impact on industry and research
Chip manufacturers are expected to face challenges in complying with US chip restrictions, as they will be required to closely track and document export movements. Additionally, some countries reliant on US imports may experience tighter access to the most powerful chips, prompting them to attempt developing local alternatives. At the same time, academic and research bodies will not be significantly harmed, thanks to exemptions dedicated to supporting medical and university research.
The United States aspires to maintain strict control over the export of the most advanced artificial intelligence technologies by regulating the movement of processing chips and advanced software systems. Although these restrictions are intended to protect innovation and national security, they raise questions about their reflection on the global market and the balance between safeguarding security interests and encouraging technological progress.
