The United States strengthens its grip on AI chip flows around the world

The U.S. government announced Monday that it will further restrict exports of chips and artificial intelligence technology, dividing the world to keep advanced computing power in the United States and among its allies, while finding more ways to block access from China.

The new regulations will limit the number of AI chips that can be exported to most countries and allow unrestricted access to US AI technology for America’s closest allies, while maintaining a block on exports to China, Russia, Iran and North Korea.

The lengthy new rules unveiled in the final days of outgoing President Joe Biden’s administration go beyond China and aim to help the United States maintain its dominant status in AI by controlling it in the whole world.

“The United States is now leading in AI, both in AI development and in chip design, and it is critical that it stays that way,” said U.S. Commerce Secretary Gina Raimondo .

The regulations end a four-year effort by the Biden administration to hinder China’s access to advanced chips that could improve its military capabilities and seek to maintain U.S. leadership in AI by filling gaps and adding new guardrails to control the flow of chips and global AI development. .

Although it is unclear how the new administration of President-elect Donald Trump will enforce the new rules, the two administrations share similar views on the competitive threat posed by China. The regulations are expected to take effect 120 days after publication, giving the Trump administration time to decide.

New limits will be placed on advanced graphics processing units (GPUs), which are used to power the data centers needed to train AI models. Most are made by Nvidia, based in Santa Clara, California, while Advanced Micro Devices also sells AI chips.

Major cloud service providers, such as Microsoft, Google And Amazonwill be able to apply for global permits to build data centers, an important part of new rules that will exempt their projects from national quotas on AI chips.

To earn a seal of approval, licensed businesses must meet strict conditions and restrictions, including safety requirements,

Until now, the Biden administration had imposed drastic restrictions on China’s access to advanced chips and the equipment to produce them, updating controls each year to tighten restrictions and catch countries risking hijacking the technology to China.

Because the rules are changing the landscape of AI chips and data centers around the world, powerful industry voices criticized the plan even before it was released.

Nvidia on Monday called the rule a “drastic overstep” and said the White House would crack down on “technology already available in consumer gaming PCs and consumer hardware.” Data Center Provider Oracle argued earlier this month that the rules would hand “most of the global AI and GPU market to our Chinese competitors.”

The rules impose global licensing requirements on certain chips, with exceptions, and also establish controls for so-called “model weights” of the most advanced “closed-weight” AI models. Model weights help drive machine learning decision making and are typically the most valuable elements of an AI model.

The settlement divides the world into three levels. About 18 countries, including Japan, Britain, South Korea and the Netherlands, will essentially be exempt from the rules. Some 120 other countries, including Singapore, Israel, Saudi Arabia and the United Arab Emirates, will face national caps. And countries subject to an arms embargo, such as Russia, China and Iran, will absolutely not be able to receive this technology.

Additionally, U.S.-based vendors eligible to receive global authorizations, such as AWS and Microsoft, will only be allowed to deploy 50% of their total AI computing power outside the United States, no more than 25% outside tier 1 countries, and no more. more than 7% in a single country not belonging to level 1.

AI has the potential to improve access to healthcare, education and food, among other benefits, but can also contribute to the development of biological and other weapons, support cyberattacks and contribute to surveillance and other human rights violations.

“The United States must prepare for a rapid increase in AI capabilities in the coming years, which could have a transformative impact on the economy and our national security,” the US security adviser said National, Jake Sullivan.