New U.S. restrictions on chip sales aim to hamper China’s AI efforts.
What’s new: The U.S. government published sweeping limits on sales of processors that involve U.S. designs and technology to Chinese businesses. U.S. officials stated that the restrictions are meant to prevent China from militarizing AI.
New rules: The rules block sales of certain processors as well as U.S.-made equipment used to design and manufacture them. This includes high-end graphics processing units (GPUs) and other processors optimized for machine learning.
- The rules apply to chips capable of processing and interconnection speeds on par with Nvidia’s flagship A100 GPU, which is designed to be used in data centers. (Nvidia supplies 95 percent of China’s AI chips.) The less-capable chips typically found in personal computers and video game consoles are not restricted.
- The restrictions prohibit sales to Chinese companies of advanced chips produced using U.S.-made software and hardware as well as sales of the equipment itself. This goes for companies anywhere in the world.
- They also bar U.S. citizens and permanent residents from supporting development or manufacturing of advanced chips without permission from the U.S. government.
China’s response: A spokesperson for China’s foreign ministry accused the U.S. of abusing export-control measures to target Chinese firms, stating that it would hinder global cooperation and supply chains.
Behind the news: The restrictions initially came to light in September, when Nvidia and AMD independently alerted shareholders that the U.S. had imposed controls on their most advanced products. However, their details became publicly available only last week. They represent a significant escalation of earlier U.S. efforts to thwart China’s ambitions in advanced technology.
- In May 2020, the U.S. required foreign chipmakers that use U.S. equipment to obtain permission to do business with the Chinese tech giant Huawei.
- In 2019, the government blocked U.S. firms from selling equipment to Huawei and 114 of its affiliates.
- In 2015, the country barred Intel from selling high-end chips to the Chinese military.
Why it matters: China has announced its ambition to become the global leader in AI by 2030, and this requires access to cutting-edge processing power. The most advanced chips are manufactured in Taiwan and South Korea using chip-fabrication equipment made by U.S. companies, and the leading chip designers and makers of chip-design software reside in the U.S. This gives U.S. authorities a tight grip on other countries’ ability to buy and make chips. China’s effort to build domestic capacity to produce advanced semiconductors — which are hampered by the sheer difficulty and expense of etching features on silicon measured in nanometers — now faces additional hardware, software, business, and talent hurdles.
We’re thinking: International cooperation has been essential to recent progress in AI. As barriers rise between the U.S. and China, the AI community must navigate a world where geography will have a much bigger impact on access to ideas and resources.