US Tightens Norms On AI Chips Export: All You Need To Know

The Joe Biden-led administration on Monday unveiled a new set of regulations to further limit access to the United States' Artificial Intelligence (AI) chips and technology to other nations, in a final bid to keep the most advanced models within the borders and among close allies. 

The new regulations, which were unveiled just a week before Mr Biden is scheduled to leave office, seek to cap the number of AI chips that can be exported to most countries, including China, while allowing unlimited access to US' closest allies such as Britain, Canada, Germany, Japan, South Korea and Taiwan. 

Besides China, Russia, Iran and North Korea have been blocked from buying US' most advanced chips and will continue to remain subject to the new restrictions.

Since 2022, the Biden administration has imposed sweeping restrictions on China's access to AI and semiconductor technologies, updating the controls annually to tighten restrictions and capture countries at risk of diverting the technology to China.

When will the new norms kick in?

The new norms are expected to take effect 120 days from publication, giving President-elect Donald Trump's incoming administration time to weigh in.

It was not immediately clear how the Trump administration will enforce the new rules, but the two administrations have shared similar views on the growing competitive threat over the use of AI technology from countries like China. 

"The US leads AI now, both AI development and AI chip design, and it's critical that we keep it that way," US Commerce Secretary Gina Raimondo said.

Which AI chips have been restricted?

Graphics processing units (GPUs), which are used to power 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 such as those made by US-based industry leader Nvidia to process different pieces of data simultaneously has made them valuable for training and running AI models.

For instance, OpenAI's ChatGPT is trained and improved on tens of 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 itself and the time the developer wants to spend training it.

Why has the US tightened curbs on the export of GPUs?

It has tightened curbs to mainly control global access to AI. 

The limits on GPUs for most countries under the fresh norms are set by compute power, to account for 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 caps on compute power are restricted to a total of 790 million TPP through 2027.

The cap translates into the equivalent of nearly 50,000 H100 Nvidia GPUs, according to Divyansh Kaushik, an AI expert at Beacon Global Strategies, a Washington-based advisory firm. "This is an enormous amount of power - enough to fuel cutting-edge research, run entire AI companies or support the most demanding AI applications on the planet," he was quoted as saying by Reuters.

Those could include running a global-scale chatbot service or managing advanced real-time systems like fraud detection or personalised recommendations for massive companies like Amazon or Netflix, Kaushik added.

Who has been exempted from the caps?

Companies like Amazon Web Services or Microsoft's Azure cloud unit that meet the requirements for special authorisations - also known as "Universal Verified End User" status - are exempt from the caps.

National authorisations also are available to companies headquartered in any destination that is not a "country of concern". Those with national Verified End User status have caps of roughly 320,000 advanced GPUs over the next two years.

Are there any exemptions to licensing?

Yes. If a buyer orders small quantities of GPUs - the equivalent of up to some 1,700 H100 chips - they will not count toward the caps, and only require government notification, not a license.

Most chip orders fall below the limit, especially those placed by universities, medical institutions, and research organisations, according to US authorities. This exception is designed to accelerate low-risk shipments of US chips globally.

There also are exceptions for GPUs for gaming.

Which regions can get unlimited access to AI chips from the US?

Eighteen destinations are exempt from country caps on advanced GPUs, according to a senior administration official.

Those are Australia, Belgium, Britain, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, South Korea, Spain, Sweden and Taiwan.