The New AI Chip Export Policy to China: Strategically Incoherent and Unenforceable
from China Strategy Initiative and China Policy Accelerator
from China Strategy Initiative and China Policy Accelerator

The New AI Chip Export Policy to China: Strategically Incoherent and Unenforceable

January 14, 2026 9:22 am (EST)

Article
Current political and economic issues succinctly explained.

On January 13, the Department of Commerce published a new regulation permitting the sale of advanced AI chips to China, codifying the major policy change that President Trump announced on December 8. The regulation loosens restrictions on the export of Nvidia H200 chips, as well as the AMD MI325X and equivalent chips from other companies, which had previously been banned for export to China.

More From Our Experts

The regulation acknowledges that exporting advanced AI chips to China poses serious national security risks, while simultaneously creating a pathway to permit their sale. The result is a framework that is strategically incoherent. If implemented faithfully, it likely would block most or all exports to China, but if implemented loosely, it would fail to address any of the concerns that motivated export controls in the first place.

Key Elements of the Regulation

More on:

China Strategy Initiative

China Policy Accelerator

China

  • Revised Thresholds: The regulation allows the export of AI chips that are 13 times more powerful than what was previously permitted, allowing shipments of AI chips that are currently available and have a total processing performance (TPP) of less than 21,000, or a total DRAM bandwidth less than 6,500 GB/s.
  • Chip Volume Cap: For each product, the number of chips exported to China cannot exceed 50% of the number of chips shipped to U.S. customers for end-use in the United States.
  • No Exports for Data Centers Outside of China: The regulation maintains a “presumption of denial” policy on exports of AI chips to China-owned data centers located outside of China.
  • U.S. Supply Certification: The exporter must certify that the exports to China will not cause “any delay” in fulfilling existing or new orders from U.S. customers of any AI chips, and that global foundry capacity that would otherwise be used to produce AI chips for U.S. end-users will not be diverted to produce products for China.
  • End-Use Certification: The exporter must certify that the chips will not be used for military, intelligence, or weapons of mass destruction end-uses, that end-users in China implement “robust” know your customer (KYC) practices to prevent the chips from being used by Chinese end-users, that model weights will not be transferred, that remote access to AI algorithms won’t be provided to restricted parties, and that the user have a security plan to physically protect the chips.
     

Assessment of Impact of Regulations

A Large Volume Cap: While the regulation does put some limits on how many AI chips China can receive, the topline cap will still be very large. Analysts estimate that Nvidia sold around 2 million H200 chips to the United States (and likely an additional 2 million H100 chips), so the regulation would cap H200 sales to China at around 1 million H200s, and potentially another 1 million H100s. 1 million H200s is half the number that Chinese companies reportedly already ordered from Nvidia, but still enough to create the largest AI data center in the world (xAI’s Colossus data center in Memphis, Tennessee). Shipments of 1 million H200s would increase the total amount of AI compute installed in China in 2026 by 250% relative to if China relied solely on domestic AI chips; shipments of an additional million H100s would double that number. [1]

The fact that the regulation restricts Chinese companies from using U.S. chips in data centers outside of China is a double-edged sword. On one hand, it ensures Chinese companies will not use U.S. chips in data centers that compete with offerings from U.S. cloud providers globally. On the other, it ensures all U.S. chips will be used inside China for AI training and other sensitive use-cases. Chinese firms will likely still build data centers overseas using Chinese AI chips, which now will be less needed inside China.

More From Our Experts

Certifications Will Pose Challenges: Some of the certifications required in this rule would be difficult or impossible to make credibly. For example, the regulation borrows heavily from concepts in the GAIN AI Act, and requires exporters to certify that orders for Chinese customers will not negatively impact the supply of AI chips available for U.S. customers. Given there are severe global shortages in many elements of the AI chip supply chain – most notably high-bandwidth memory (HBM), a critical input into AI chips for which supply is inelastic – it may prove difficult for exporters to credibly certify that increasing AI chip production for China will not have any impacts on the rest of the AI chip supply chain. Nevertheless, it will be difficult, if not impossible, for Commerce to prove that any certification is knowingly false.

Furthermore, the regulation requires exporters to certify that Chinese customers conduct “robust” KYC practices to prevent the PLA from accessing the chips. However, the most likely Chinese purchasers are cloud providers such as Alibaba and Tencent, as well as AI companies such as DeepSeek – companies with documented relationships with China’s military and security services. Tencent in particular is identified by the U.S. Department of Defense as a “Chinese Military Company.” Certifying that these entities will reliably exclude PLA access contradicts publicly available information about their government ties. But again, Commerce has few means to prove that any certification is knowingly false or not being fulfilled.

More on:

China Strategy Initiative

China Policy Accelerator

China

Sets a Dangerous Precedent: While the regulation does not explicitly create a sliding scale, the policy logic—that older chips commercially available in the United States should be exportable to China at some ratio—sets a precedent that could be extremely dangerous if applied to more advanced chips. As noted previously, selling 1 million H200s to China increases China’s domestic AI compute installed this year by 250%. If the same formula in this regulation was expanded to Blackwell chips next year, permitting the sale next year of 50% of all GB300 chips that had been sold to the United States, this could green-light the sale of 2.5 million GB300s to China – which would increase the amount of AI compute China installs in 2027 by over 1000% relative to if China relied solely on Chinese AI chips.  This would not be in the strategic interest of the United States. 

Conclusion

The rule inadvertently demonstrates that there is no version of an AI chip export policy to China that is simultaneously permissive, implementable, enforceable, and protective of US national security. While the restrictions included in the rule are preferable to unconditional exports, they are still severely problematic, creating a false sense of security while authorizing very large numbers of AI chip exports to China without enforceable guardrails that prevent their misuse.  This will allow China’s AI industry to close the gap with the United States and build the largest data centers in the world, and will not stop China’s military or intelligence services from using these chips.

The strategic case for selling advanced AI chips to China has always been weak: modest commercial gains for a handful of US chip companies against substantial risks of enabling Chinese military AI capabilities, eroding US technological leadership, and increasing Chinese competition for every other U.S. technology company. This regulation, in attempting to thread an impossible needle, only demonstrates that no such needle exists.  The most effective policy remains the simplest one: denying the export of all AI chips to China.

 

 

Creative Commons
Creative Commons: Some rights reserved.
Close
This work is licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) License.
View License Detail
Close

Endnotes

2
Close

Endnotes

  1. ^ Relative to median China AI chip production estimates included in modeling published by CFR on December 15: https://www.cfr.org/article/chinas-ai-chip-deficit-why-huawei-cant-catc…

Top Stories on CFR

 

United States

Trump returned to office propelled by a seemingly isolationist promise, but the U.S. capture of Maduro illustrates the White House’s growing fondness for military intervention—revealing a striking strategic incoherence.

Artificial Intelligence (AI)

Six CFR fellows examine the challenges that lie ahead, reviewing how governance, adoption, and geopolitical competition will shape artificial intelligence and society’s engagement with this new technology.