Random Image Display on Page Reload

US to Introduce New Restrictions on China’s Access to Cutting-Edge Chips

Nov 27, 2024 5:50 PM

US to Introduce New Restrictions on China’s Access to Cutting-Edge Chips

The new limits, which are expected to be announced Monday, are intended to slow China’s ability to build large and powerful AI models.

People are seen inside a Huawei store at the Wangfujing shopping area in Beijing on November 26 2024 shortly after...

People are seen inside a Huawei store at the Wangfujing shopping area in Beijing on November 26, 2024, shortly after Huawei's new Mate 70 mobile phone was launched.Photograph: ADEK BERRY/Getty Images

The Biden administration is expected to announce a sweeping set of measures on Monday designed to further restrain China’s ability to develop advanced artificial intelligence, people familiar with the matter told WIRED. The controls could include sanctioning dozens of Chinese companies that produce equipment for making semiconductors, as well as placing restrictions on a handful of chip manufacturing plants, some of which have ties to the Chinese tech giant Huawei.

The US Department of Commerce has also discussed including controls on the sale of high-bandwidth memory, or HBM, an advanced kind of 3D-stacked computer memory component that is often used in high-performance GPUs and customized AI chips. Bloomberg previously reported that the Biden administration was considering clamping down on China’s access to HBM chips.

In total, the Biden administration could end up adding around 200 Chinese firms to an entity list maintained by the Bureau of Industry and Security—an agency within the Commerce department—which would require other companies to acquire special licenses to supply them with software or products from the United States. The US government has been discussing the new measures with its allies and representatives from the semiconductor industry for months, and the exact details of what will be announced on Monday were still in flux as of earlier this week.

A spokesperson for the Commerce department declined to comment. Huawei did not immediately return a request for comment.

Reuters reported on Friday that the US Chamber of Commerce, a powerful advocacy group for American businesses, warned its members in an email last week that a new round of export controls targeting China would be arriving “prior to the Thanksgiving break," though that timing now appears to have been pushed back by a few days.

“China is firmly opposed to the US overstretching the concept of national security, abusing export control measures and making malicious attempts to block and suppress China,” Mao Ning, a spokesperson for China’s foreign ministry, said at a regularly scheduled press conference earlier this week in response to the anticipated controls.

The limits on China’s access to the high-bandwidth memory seem aimed at slowing the country’s efforts to develop domestic chips capable of training very large and powerful AI models. The new restrictions are expected to block access to HMB3, one of the people told WIRED, the latest and most advanced version of the technology, and impose some limits on access to the previous generation, known as HMB2.

The US government has been imposing similar export controls on China aimed at limiting its ability to mint advanced silicon for years, but the controls apparently didn’t stop Huawei from developing competitive chips for training large AI models.

The Chinese tech giant, which was temporarily crippled by US sanctions half a decade ago, sent samples of its latest AI training chip, called Ascend, to customers this September, according to the South China Morning Post. Companies testing Ascend reportedly include ByteDance, the Chinese parent of TikTok, which is said to be training a large model primarily using Ascend. Baidu, which makes China’s leading search engine and has developed autonomous driving systems, recently placed an order for Huawei’s chips in a shift away from US chip giant Nvidia, according to Reuters. (Nvidia declined to comment.)

Export restrictions aimed at curbing China’s AI sector began under the first Trump administration. In 2019, several up-and-coming Chinese AI firms were added to the entity list, meaning that US firms, including chipmakers like Nvidia, would be required to get a special license to do business with them. This was followed by restrictions on sales of chips made with US technology to Huawei, China’s dominant telco and a leading smartphone manufacturer.

The Biden administration ratcheted up the controls in October 2022, limiting exports to China of cutting-edge GPU chips, including those made by Nvidia, a move aimed at curbing any Chinese company’s ability to train the most powerful AI models. The rules were tightened a year later to close loopholes that still allowed Chinese firms to access some advanced chips.

It can be tricky to gauge the impact of US chip sanctions, and some experts question whether the controls are spurring China to make more rapid advances in chipmaking itself, reducing its reliance on American companies.

In late 2023, Huawei unveiled the Mate 60, a smartphone featuring an advanced chip from the Chinese chipmaker SMIC. The announcement caused a stir in Washington, because it suggested that SMIC had made substantial progress in advancing its own manufacturing techniques. (Further analysis indicated that Huawei and SMIC were still reliant on foreign suppliers.)

But a report published this week by the Center for Strategic and International Studies, a Washington, DC-based think tank, argued that the Chinese government had already begun ramping up investment in domestic chipmaking before the US government began clamping down on the country’s access to advanced semiconductors. It also noted that China has made bigger strides in sectors not subject to export controls, such as solar cell and electric vehicle manufacturing.

Will Knight is a senior writer for WIRED, covering artificial intelligence. He writes the AI Lab newsletter, a weekly dispatch from beyond the cutting edge of AI—sign up here. He was previously a senior editor at MIT Technology Review, where he wrote about fundamental advances in AI and China’s AI… Read more
Senior Writer

Louise Matsakis is a senior business editor at WIRED. She was previously deputy news editor at Semafor, a senior editor at Rest of World, and a staff writer at WIRED. An investigation she cowrote on the Chinese fast-fashion giant Shein won the 2022 Society of Publishers in Asia award for… Read more
Senior Business Editor

Read More

Why the US Government Banned Investments in Some Chinese AI Startups

The Biden administration targeted Chinese companies developing frontier AI models, but Donald Trump could take a more sweeping approach.
Zeyi Yang

A New Benchmark for the Risks of AI

MLCommons provides benchmarks that test the abilities of AI systems. It wants to measure the bad side of AI next.
Will Knight

OpenAI Is Working With Anduril to Supply the US Military With AI

The ChatGPT maker is the latest AI giant to reveal it’s working with the defense industry, following similar announcements by Meta and Anthropic.
Will Knight

The US Patent and Trademark Office Banned Staff From Using Generative AI

The agency dedicated to protecting new innovations prohibited almost all internal use of GenAI tools, though employees can still participate in controlled experiments.
Reece Rogers

OpenAI Poaches 3 Top Engineers From DeepMind

The new hires, all experts in computer vision, are the latest AI researchers to jump to a direct competitor in an intensively competitive talent market.
Reece Rogers

Nvidia Says Its Blackwell Chip Is Fine, Nothing to See Here

Chip production delays and a rumored overheating issue haven’t slowed down Nvidia, which reported another quarter of blockbuster earnings and said Blackwells are now in the hands of Microsoft and OpenAI.
Lauren Goode

How Best to Use ChatGPT, Gemini, and Other AI Tools? Our AI Expert Answers Your Questions

If you missed our live, subscriber-only Q&A with WIRED’s AI columnist Reece Rogers, you can watch this replay of the livestream.
Michael Calore

Amazon Is Building a Mega AI Supercomputer With Anthropic

At its Re:Invent conference, Amazon also announced new tools to help customers build generative AI programs, including one that checks whether a chatbot’s outputs are accurate or not.
Will Knight

This Website Shows How Much Google’s AI Can Glean From Your Photos

A photo sharing startup founded by an ex-Google engineer found a clever way to turn Google’s tech against itself.
Paresh Dave

More Spyware, Fewer Rules: What Trump’s Return Means for US Cybersecurity

Experts expect Donald Trump’s next administration to relax cybersecurity rules on businesses, abandon concerns around human rights, and take an aggressive stance against the cyber armies of US adversaries.
Eric Geller

Parallel Parking Is Stressful. Most Drivers Still Won’t Let the Robots Take Over

In the Western world, some automated driving features seem to have fallen out of favor. Not true in China—and the reasons why offer a preview of auto industry changes to come.
Aarian Marshall

Yes, That Viral LinkedIn Post You Read Was Probably AI-Generated

A new analysis estimates that over half of longer English-language posts on LinkedIn are AI-generated, indicating the platform’s embrace of AI tools has been a success.
Kate Knibbs

*****
Credit belongs to : www.wired.com

Check Also

Nvidia’s $3,000 ‘Personal AI Supercomputer’ Will Let You Ditch the Data Center

Will Knight Business Jan 6, 2025 11:24 PM Nvidia’s $3,000 ‘Personal AI Supercomputer’ Will Let …