US expands AI chip export controls to seven nations to block China

Closing the data center loophole

The United States is preparing to expand its semiconductor export controls to prevent Chinese artificial intelligence companies from accessing high-end hardware. According to a report by Chosunbiz, the U.S. Commerce Department's Bureau of Industry and Security is finalizing rules to target seven specific countries. These nations include Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway, and South Africa.

The primary goal of this expansion is to stop a specific workaround used by Chinese firms. These companies have reportedly been using data centers in third countries to train AI models on advanced semiconductors that are otherwise subject to U.S. export bans. By moving the computing power to Southeast Asia and other regions, these firms could bypass direct trade restrictions.

The new regulations are expected to be announced as early as this month. Once they take effect, the seven countries will need explicit approval from the Commerce Department to import high-end hardware. This includes Nvidia's Blackwell-class chips and other advanced AI semiconductors that are critical for training large-scale models.

New oversight and reporting rules

Under the proposed rules, any entity in the seven designated countries seeking to import advanced AI chips must report exactly who will receive the hardware. They are also required to disclose how the chips will be used in their specific operations. This creates a layer of ongoing oversight to ensure that the technology does not end up in the hands of restricted actors.

The government is also looking to limit remote access from countries currently under arms embargos, such as China and Russia. Operators of data centers in the new target countries will be required to report the specific countries of their customers and the total scale of the computing being performed. This is designed to prevent a situation where a chip is physically located in a friendly nation but controlled remotely by a restricted entity.

There is a specific carve-out for data centers that are run by U.S. companies. These facilities may be exempt from the new licensing requirements if they meet certain conditions. This distinction is intended to allow U.S. firms to continue their global operations while still preventing unauthorized access by foreign competitors.

Political pressure and Nvidia

The push for tighter controls follows significant pressure from hard-liners in the U.S. government who want to limit Chinese influence in AI. This movement has also been linked to a recent scandal involving Nvidia CEO Jensen Huang. Reports suggest that a firm where a woman linked to Huang worked as an executive may have been involved in routing chips to China.

The firm involved, Megaspeed, is a Singaporean company suspected of operating a data center in Malaysia. It is alleged that this facility was used to route Nvidia's advanced AI chips to Chinese companies. This situation has drawn sharp criticism from members of Congress toward the Commerce Department for its handling of the investigation into the firm.

The Information reported that the Commerce Department faced scrutiny after it reportedly halted its probe into Megaspeed following Nvidia's involvement. This has led to a perception that the company's political influence in Washington may have waned. Lawmakers are now attempting to fold three separate bills into the National Defense Authorization Act to further strengthen these export controls.

Expanding the existing framework

These new measures do not exist in a vacuum but rather build upon existing trade restrictions. The United States already requires Commerce Department approval for advanced AI chip imports in several other countries, including Saudi Arabia and Vietnam. These existing rules already include conditions such as limits on remote access to ensure the technology is used appropriately.

The addition of seven more countries represents a significant widening of the scope of these controls. By including nations like Norway and Brazil alongside Southeast Asian countries, the U.S. is casting a wider net. The goal is to ensure that no region becomes a convenient

What comes next for the industry

The tech industry will need to prepare for a more complex licensing environment as these rules take effect. Companies operating in the seven target countries will need to establish clear reporting mechanisms to comply with the new oversight requirements. This will likely involve more rigorous audits of customer bases and computing usage scales.

For Chinese AI firms, the move significantly complicates their ability to acquire the high-performance computing power needed for advanced research. They will have to find alternative ways to train their models without relying on U.S.-made semiconductors. This could slow the pace of their development relative to other global competitors.

The final version of the rules, expected soon, will provide the definitive roadmap for these restrictions. Investors and tech companies will be watching closely to see how the Commerce Department balances these security goals with the needs of U.S. companies operating abroad. The outcome will determine how much friction exists in the global AI hardware supply chain.