The Trump administration authorised licences covering more than 879,000 advanced artificial intelligence (AI) chips for Chinese companies between December 2025 and May 2026, according to export-control documents released by the US House Foreign Affairs Committee. The approvals covered processors from Nvidia and Advanced Micro Devices (AMD), including Nvidia’s H200 and AMD’s comparable MI325X chips. The disclosure has intensified scrutiny of Washington’s technology-export policy and raised questions about how the United States balances commercial opportunities with national security concerns.

The companies had sought permission to export approximately 3.4 million AI chips to China during the six-month period, but US authorities approved licences for only a portion of the requested volume. The approved allocations included 150,000 chips for Tencent, 75,000 for Alibaba, 43,000 for Baidu and 4,000 for BYD. Importantly, the approvals do not establish that every authorised chip was physically delivered. Nvidia said it had stopped shipping H200 processors to China despite receiving some licences, while AMD said it complies with applicable US export-control laws.

Key Takeaways

  • The Trump administration approved licences covering more than 879,000 AI chips for Chinese customers between December 2025 and May 2026.
  • Companies applied to export approximately 3.4 million chips, meaning approvals covered only part of the requested volume.
  • Tencent received approval for 150,000 chips, Alibaba for 75,000, Baidu for 43,000 and BYD for 4,000.
  • The licences covered Nvidia H200 and AMD MI325X processors.
  • The House Foreign Affairs Committee released the licensing information under its authority to disclose export-control information in the national interest.
  • The approvals were issued under a January 2026 policy that allowed case-by-case consideration subject to additional security requirements.
  • Nvidia said it had subsequently stopped H200 shipments to China, while AMD said it follows US export-control requirements.
  • The disclosure comes as Washington introduces additional restrictions on indirect access to American AI computing technology.

US Approved More Than 879,000 AI Chip Licences for China

The US Commerce Department authorised licences covering more than 879,000 advanced AI processors for Chinese companies during the six months from December 2025 to May 2026.

The figures were made public by the House Foreign Affairs Committee, which released information about licence applications submitted by Nvidia and AMD.

The companies had requested permission to export approximately 3.4 million processors to China. The approved quantity was substantially smaller than the total requested.

The processors included Nvidia’s H200 and AMD’s MI325X, both designed for demanding AI computing workloads.

Export licensing detailReported figure
Period coveredDecember 2025–May 2026
Total chips requestedApproximately 3.4 million
Approved chip licencesMore than 879,000
Companies involvedNvidia and AMD
Nvidia processorH200
AMD processorMI325X

The disclosure provides a clearer picture of how US export policy was applied after the Trump administration introduced a more flexible licensing approach for certain advanced processors.

However, the number of approved licences should not be confused with the number of chips that actually reached Chinese customers. Licensing approval is permission to proceed with an export under specified conditions, not proof that a shipment was completed.

Alibaba, Tencent, Baidu and BYD Received Chip Approvals

The committee’s documents identified several major Chinese companies among the recipients of approved chip licences.

Tencent received approval for 150,000 processors, making it the largest recipient among the four companies highlighted in the Business Standard report.

Alibaba received approval for 75,000 chips, while Baidu received 43,000 and electric-vehicle manufacturer BYD received 4,000.

Chinese companyApproved chip quantity
Tencent150,000
Alibaba75,000
Baidu43,000
BYD4,000
Total272,000

These four allocations account for 272,000 of the more than 879,000 approved chips. The remaining approvals involved other Chinese customers and licence applications.

The records also showed that Alibaba had requested permission to purchase approximately 950,000 chips, but received approval for only 75,000.

The committee highlighted the companies’ connections to China’s technology and industrial sectors. It also noted that the companies appeared on Pentagon lists identifying businesses alleged to have links with China’s military.

There is an important timing distinction: the status of individual companies on those lists changed during the period. The inclusion of a company on a Pentagon list does not, by itself, establish that a particular export licence violated US law or that the company used the chips for military purposes.

Why Did the Trump Administration Allow AI Chip Exports to China?

The approvals followed a change in US export-control policy.

Under the Biden administration, restrictions on advanced AI processors generally applied a presumption of denial to exports to China. This approach limited Chinese companies’ access to some of the most capable American AI computing hardware.

The Trump administration subsequently moved towards allowing case-by-case consideration of certain less advanced processors, including Nvidia’s H200 and AMD’s MI325X.

A rule published by the US Bureau of Industry and Security (BIS) in the Federal Register on January 15, 2026, established the revised licensing framework.

Instead of automatically rejecting every application covered by the previous policy, the new framework allowed applications to be assessed individually if specified conditions were satisfied.

The administration’s stated approach was to permit certain commercial exports while maintaining national security safeguards.

Conditions Attached to Chip Export Approvals

The January 2026 rule introduced several requirements for eligible exports.

These included:

  • Domestic supply protection: Aggregate shipments of the covered products to China and Macau could not exceed 50% of the corresponding product shipments to US customers for end use in the United States.
  • Customer verification: Exporters and recipients had to maintain rigorous procedures to identify customers and prevent unauthorised access.
  • Independent testing: Each shipment had to undergo third-party testing in the United States to verify its technical capabilities and specifications.
  • End-user restrictions: Transactions remained subject to restrictions involving prohibited parties and unauthorised military end uses.
  • Case-by-case review: US authorities retained the ability to approve or reject applications based on the information provided.

These conditions were intended to preserve US access to advanced computing hardware while limiting the possibility that exported processors could be diverted or used in prohibited applications.

Why Are Nvidia H200 and AMD MI325X Chips Important?

The H200 and MI325X are high-performance processors designed for demanding computing applications, including artificial intelligence model training and inference.

AI model training requires large amounts of computing power to process data and adjust the parameters of machine-learning systems. Inference involves running trained models to generate responses, analyse information and perform other tasks.

Specialised processors can significantly influence how quickly and efficiently companies perform these operations.

Nvidia H200

<AsyncImage query=”NVIDIA H200 Tensor Core GPU accelerator data center product” aspectRatio=”16:9″ width=”100%” />

The Nvidia H200 belongs to the company’s Hopper generation of AI accelerators. It is designed for data-centre workloads and offers high-bandwidth memory suited to large AI models.

AMD MI325X

<AsyncImage query=”AMD Instinct MI325X accelerator GPU data center product” aspectRatio=”16:9″ width=”100%” />

AMD’s MI325X is part of its Instinct accelerator family. It is designed to compete in the market for high-performance AI and high-performance computing infrastructure.

Access to these processors can help technology companies build and operate more powerful AI systems. This is one reason why the export of advanced computing hardware has become an important issue in US-China technology relations.

House Panel Releases Export Data in the National Interest

The licensing information was released by the House Foreign Affairs Committee, which oversees aspects of US foreign policy and export-control matters.

The committee invoked provisions of the Export Control Reform Act to disclose information about the licence applications.

The law allows the committee to make certain export-control information public if it determines that withholding the information would be contrary to the national interest.

According to the released records, the Bureau of Industry and Security had shared information about approved and rejected applications involving Nvidia H200 and comparable processors with Congress.

The disclosure provides lawmakers with a more detailed view of how the Commerce Department applied its licensing policy.

It also creates a basis for congressional scrutiny of the department’s decisions, particularly where Chinese customers have been identified by US authorities as potential national security concerns.

The central question is not simply how many licences were approved, but whether the safeguards attached to those approvals were sufficient to prevent sensitive technology from reaching restricted users or being diverted.

How Have Nvidia and AMD Responded?

Nvidia said the licences were subject to strict conditions and limited volumes.

The company also stated that it had stopped shipping H200 processors to China despite receiving some export approvals.

This distinction is important because an export licence does not compel a company to complete a sale. Companies may decide not to proceed with shipments because of commercial considerations, regulatory uncertainty or changes in their export policies.

AMD said it complies with applicable US export-control laws and regulations and maintains processes to ensure compliance across its global operations.

The statements from both companies address their compliance positions, but they do not independently establish how many approved processors were ultimately delivered to Chinese customers.

The released licensing information also does not establish that the companies or their customers violated export-control rules.

Washington Faces Questions Over AI Chip Export Controls

The disclosure has intensified debate in Washington about whether the United States should allow Chinese companies to purchase certain American AI processors.

Supporters of allowing controlled exports argue that US companies should retain access to international markets where sales can be permitted under strict safeguards. They also argue that maintaining commercial relationships can support American technology companies and their ability to invest in future products.

Critics are concerned that advanced processors could strengthen China’s AI capabilities or potentially support military applications. They have questioned whether licensing conditions and end-user verification are sufficient to prevent diversion.

The debate involves competing policy objectives:

Policy considerationMain concern
US national securityPreventing advanced computing technology from supporting restricted military or intelligence activities
Commercial competitivenessMaintaining overseas markets for American chipmakers
Export enforcementPreventing unauthorised resale, diversion and indirect access
AI developmentControlling access to computing resources used to train advanced models
International relationsManaging technology restrictions alongside broader US-China trade relations

The licensing data have therefore become part of a larger debate about the role of export controls in the global AI industry.

US Introduces Additional Restrictions on Indirect Access to AI Chips

The latest disclosure comes as the Trump administration has introduced additional measures to restrict China’s access to American AI computing resources through overseas data centres.

The new rules announced in November 2026 require certain overseas data centres serving companies in restricted countries, including China, to identify their end users.

They also introduce additional licensing requirements for certain chip shipments involving countries such as Malaysia, Singapore, Thailand, Indonesia and Brazil, which US authorities have identified as higher-risk locations for potential diversion.

The measures address a growing concern: Chinese companies may be able to access computing power remotely even when they cannot directly purchase restricted processors.

For example, a company could potentially rent computing capacity from a data centre located outside China. If that facility uses American AI chips, the customer may still obtain access to the underlying computing resources.

US policymakers are increasingly examining whether export controls should cover not only the physical movement of processors but also access to computing power through overseas infrastructure.

The new measures indicate that Washington is expanding its attention from direct chip exports to the wider international AI infrastructure network.

What the Chip Approvals Mean for China’s AI Industry

Access to advanced processors is a major factor in the development of large AI models.

Chinese technology companies such as Alibaba, Tencent and Baidu operate substantial cloud-computing and AI businesses. Their products include cloud services, AI assistants, enterprise software and model-development platforms.

Advanced processors can help these businesses train models, run inference workloads and expand the computing capacity available to customers.

However, access to imported processors is only one part of the AI infrastructure equation. Data-centre capacity, electricity, networking equipment, software optimisation and access to technical talent also influence the performance and cost of AI systems.

The approvals therefore represented a potential source of additional computing capacity for Chinese businesses, but the actual impact depends on completed deliveries and the use of the processors.

China is also investing in domestic semiconductor development to reduce its dependence on foreign suppliers. Export restrictions and uncertainty around future access to American processors may increase the incentive for Chinese companies to develop alternative hardware and software ecosystems.

For Nvidia and AMD, the restrictions create a complicated commercial environment. China represents a significant technology market, but access to it remains subject to US national security policy.

The Bigger Picture

The approval of more than 879,000 AI chip licences shows how US export policy has evolved from broad restrictions towards a system that allows selected transactions under specific conditions. The House committee’s disclosure has made the scale of that licensing activity public and highlighted the tension between commercial opportunities for American chipmakers and concerns about China’s access to advanced computing technology.

The controversy also demonstrates that controlling AI technology is becoming more complicated. Restricting physical chip exports is only one part of the challenge when companies can access computing resources through international data centres. Washington’s newer restrictions on overseas infrastructure suggest that future export-control policy will increasingly focus on who can use AI computing power, where it is located and how access is monitored.

Looking Ahead

Congressional scrutiny of the Commerce Department’s export-licensing decisions is likely to continue, with lawmakers examining how applications were assessed and whether the security requirements were effectively enforced. The distinction between approved licences and actual deliveries will remain important as policymakers evaluate the consequences of the decisions. Nvidia’s statement that it has stopped H200 shipments to China also means that the commercial outcome cannot be measured simply by the number of licences granted.

For the global semiconductor industry, the issue is part of a wider shift in how governments regulate access to advanced computing technology. Nvidia and AMD must navigate changing export requirements, while Chinese technology companies face uncertainty over their ability to secure foreign processors and access overseas computing capacity. The future direction of US-China chip trade will depend on regulatory decisions, enforcement activity and the broader relationship between the two countries.

Get the day’s top stories in your inbox

One concise email. No spam, unsubscribe anytime.