The Joe Biden-led administration on Monday unveiled a new set of rules to limit the United States’ access to artificial intelligence (AI) chips and technology to other countries, in order to keep the most advanced models within the borders and beyond their borders. Can be kept in the middle. Close associates.
The new rules, which were unveiled just a week before Mr Biden left office, seek to limit the number of AI chips that can be exported to most countries, including China, while extending unlimited access to America’s closest allies such as Britain. Allow access. , Canada, Germany, Japan, South Korea and Taiwan.
In addition to China, Russia, Iran and North Korea have been barred from buying America’s most advanced chips and remain subject to the new sanctions.
Starting in 2022, the Biden administration has imposed sweeping restrictions on China’s access to AI and semiconductor technologies, tightening restrictions and updating controls annually to catch countries at risk of diverting technology to China.
When will the new norms come into effect?
The new criteria are expected to take effect within 120 days from publication, giving the incoming administration of President-elect Donald Trump time to consider.
It was not immediately clear how the Trump administration would implement the new rules, but both administrations have shared similar views on the growing competitive threat over the use of AI technology from countries like China.
US Commerce Secretary Gina Raimondo said, “The United States is now the leader in AI, both in AI development and AI chip design, and it is vital that we keep it that way.”
Which AI chips are banned?
Graphics processing units (GPUs), which are used to power the data centers needed to train AI models. Most are made by Santa Clara, California-based Nvidia, while Advanced Micro Devices also sells AI chips.
Although known largely for their role in gaming, the ability of GPUs made by US-based industry leader Nvidia to process different pieces of data simultaneously makes them valuable for training and running AI models. Is.
For example, OpenAI’s ChatGPT has been trained and improved on thousands of GPUs.
The number of GPUs needed for an AI model depends on how advanced the GPU is, how much data is being used to train the model, the size of the model, and how much time the developer wants to spend training it. .
Why has the US tightened restrictions on GPU exports?
It has tightened restrictions primarily to control global access to AI.
Under the new standards GPU limits for most countries are set by compute power, taking into account differences in individual chips.
Total processing performance (TPP) is a metric used to measure the computational power of a chip. Under the regulation, countries with limits on compute power are limited to a total of 790 million TPP by 2027.
According to Divyansh Kaushik, an AI expert at Beacon Global Strategies, a Washington-based consulting firm, this limit is equivalent to about 50,000 H100 Nvidia GPUs. “That’s a massive amount of power – enough to fuel cutting-edge research, run entire AI companies or support the most demanding AI applications on the planet,” Reuters quoted him as saying.
These could include running a global-scale chatbot service or managing advanced real-time systems like fraud detection or personalized recommendations for large companies like Amazon or Netflix, Kaushik said.
Who is exempt from the cap?
Companies like Amazon Web Services or Microsoft’s Azure cloud unit that meet the requirements for special authorizations — also known as “universal verified end user” status — are exempt from the limit.
National authorizations are also available to companies whose headquarters are in a destination that is not a “country of concern”. Those with National Verified End User status have a limit of approximately 320,000 advanced GPUs over the next two years.
Are there any exemptions in licensing?
Yes. If a buyer orders small quantities of GPUs – the equivalent of about 1,700 H100 chips – they will not count against the limit, and will only require government notification, not a license.
According to US officials, most chip orders fall below the limit, especially orders placed by universities, medical institutions and research organizations. This exception is designed to speed up the shipment of low-risk US chips globally.
There are also exceptions for GPUs for gaming.
Which regions can get unlimited access to AI chips from the US?
Eighteen destinations have been exempted from the country’s limit on advanced GPUs, according to a senior administration official.
They are Australia, Belgium, Britain, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, South Korea, Spain, Sweden and Taiwan.
