The U.S. has introduced strict new regulations on chip exports, significantly impacting China's AI industry.

The U.S. Department of Commerce recently announced new chip export control regulations, further tightening restrictions on chip exports to China. The new rules change the assessment criteria, primarily using computing power as a measure instead of the previous bandwidth parameters. This change affects almost all high-performance chips.

It is worth noting that even NVIDIA's consumer-grade gaming graphics card, the RTX 4090, has not been spared. Although its compute power and memory width product is only around 2700, its average chip area capability happens to fall within the judgment range of the second-highest performance chips on the export control list. This means that NVIDIA's sales of graphics cards to gamers will also be restricted. According to reports, U.S. officials have stated that even the strongest consumer-grade chip orders for laptops need to notify the Department of Commerce.

What is even more surprising is that even chips with performance slightly below regulatory standards need to notify the government before export. This regulation actually covers all high-performance chips, including AI chips that were previously developed specifically for the Chinese market, such as NVIDIA's A800, H800, L40S, as well as AMD's MI250 and Intel's Gaudi2, which has not yet been officially used in China.

Despite market expectations, the severity of the new regulations is still shocking. According to unconfirmed reports, Nvidia urgently notified customers interested in purchasing the H800 to place orders before the new rules take effect. This news has raised widespread concerns in the AI field.

The new regulations introduced by the U.S. Department of Commerce are clearly well-prepared. In addition to explicitly using computing power as a standard, the new rules also set more obstacles regarding the list of restricted export countries and export licenses for semiconductor manufacturing equipment. Furthermore, the U.S. plans to initiate a process to restrict China's access to cloud computing power, which will impact the world's largest cloud service providers (mostly U.S. companies).

Overall, this new regulation takes a dual approach, both fully preventing China from obtaining chips from third countries and stopping China from acquiring the capability to manufacture advanced chips. This policy reflects the extreme shift in the U.S. attitude towards China. It is worth noting that during the policy's development, U.S. media frequently reported on news related to Chinese chips, and Commerce Secretary Raimondo even expressed "deep concern" about Huawei's new smartphone, all of which seem to set the stage for the introduction of the new policy.

Although there is a 30-day public notice period for the new regulations, no significant changes are expected. Raimondo further stated that the regulatory rules will be updated at least once a year, indicating that the U.S. intends to implement this policy in the long term.

There are differing views in the industry and among politicians regarding this strict new regulation. The Semiconductor Industry Association (SIA), which represents 99% of U.S. chip companies, believes that overly broad unilateral controls could harm the U.S. semiconductor ecosystem and may lead overseas customers to turn to other suppliers. However, some members of Congress argue that the existing policy is not strong enough and call for continued vigilance to prevent evasion of restrictions.

It is worth noting that semiconductor companies seem to have failed to exert influence during this policy-making process. The business plans of companies like Intel and Nvidia may therefore be significantly affected. Nvidia candidly stated in regulatory documents that while it can assist customers in applying for export licenses, it cannot guarantee that these licenses will be approved or processed in a timely manner.

This policy has sparked reflections on globalization and the division of labor in the industrial chain. Some scholars refer to the US approach as "weaponization of interdependence" or "weaponized dependence," pointing out that countries controlling key nodes may achieve strategic objectives through a generalized concept of national security.

For China, enhancing domestic chip capabilities has become a necessary task. Although domestically produced alternative chips, such as Huawei's Ascend ecosystem, have gained certain strengths, overall they still struggle to meet the market's demand for computing power. However, China's strong performance in the field of large models indicates that while chip issues may slow down the pace of development, they will not completely hinder progress in the field of artificial intelligence.

Ultimately, this challenge may become a catalyst for China's technological development, as the ancient saying goes: Whatever hinders you will ultimately make you stronger.