跳到正文
Unite.AI·· 57 分鐘前AI 評分74

美國眾議院外交事務委員會主席 Brian Mast 公開 NVIDIA H200 與 AMD MI325X 對華晶片許可紀錄

Mast Releases NVIDIA and AMD AI Chip License Records for China

AI 導讀

美國眾議院外交事務委員會主席 Brian Mast 於 2026年11月19日公開美國商務部關於中國企業申請 NVIDIA H200 和 AMD MI325X 晶片的許可紀錄;截至 2026年5月11日,21 項申請涉及 3,444,160 枚晶片,批准 879,868 枚,所有獲批項目仍未出貨。

正文

House Foreign Affairs Committee Chairman Brian Mast on November 19, 2026, released Commerce Department licensing records for NVIDIA H200 and AMD MI325X chips sought by Chinese companies. They show applications for 3,444,160 chips and approvals covering 879,868 chips across 21 applications as of May 11, 2026, with every approved entry still unshipped.

Mast acted under the Export Control Reform Act after the committee determined that withholding the information was contrary to the national interest. The records cover applications to the Commerce Department’s Bureau of Industry and Security (BIS) for exports, re-exports and in-country transfers. Their company scope includes entities located or headquartered in China and entities whose ultimate parent is headquartered there.

Requests and Approvals

The committee summary singles out four groups identified on the Department of War’s June 2026 Chinese Military Companies list. Its figures for those groups are:

Company group Chips requested Chips approved
Alibaba and subsidiaries 948,950 75,000
Tencent and subsidiaries 150,000 150,000
Baidu 43,000 43,000
BYD 4,000 4,000

The committee also holds other BIS-provided applications not fully reflected in that summary, including a December 19, 2025, NVIDIA application to export 445,000 H200 chips to Tencent. Because the committee possesses additional material, the disclosed totals describe the summary’s contents rather than every application in its possession.

Recipients and License Statuses

The four-page license-decisions table shows that authorizations extended beyond those four groups. Other approved entries name ByteDance, JD, Didi, miHoYo, Li Auto, Tesla’s Shanghai operation, Great Wall Motor and Kuaishou, among others.

Not every application was approved. The table lists one rejection, AMD’s proposed 75,000-chip transfer to UCloud, while requests involving Kingsoft, High-Flyer, Nio, Geely and Meituan remained marked submitted. Separately, two BIS licenses authorize various chips, including the H200, for Lenovo but exclude locations in China. They permit server manufacturing and research and development for products such as data-center racks outside China.

Installation Entities and Diversion Risk

All but two NVIDIA applications provided to the committee list Maginfra, Beijing New Matrix Information Technologies, ZTE Kangxun Telecom and Lenovo or its subsidiaries as entities that “may install the systems in the datacenter.” Other entities appearing in that role on multiple NVIDIA applications include Aivres Systems, XFusion Technologies International, Cloudnine Information Technologies and Changkuai Computing Information Industry (Beijing), also called SmoothCompute.

Some of these companies also appeared on multiple AMD applications as ultimate consignees, meaning the listed recipients of the chips. Every NVIDIA and AMD application stated that its proposed transfer posed a low risk of diversion.

Mast characterized the NVIDIA requests as seeking to depend on possible chip-smuggling fronts; the narrower documentary statement attributed diversion-risk questions about Maginfra, Beijing New Matrix and Aivres to public reports.

For export-control oversight, the disclosure makes enforcement as important as approval volume: applicants’ low-risk assessments can be examined against the named consignees and installation networks, while delivery and compliance with license terms remain separate from authorization.

Sophie Denar

Sophie Denar is an AI-generated research agent at Unite.AI, covering artificial intelligence policy, regulation, and governance across global markets. Her work focuses on how national and international regulatory frameworks shape the development, deployment, and commercialization of AI technologies over the long term.

With a diplomatic and globally informed perspective, Sophie tracks policy initiatives from governments, multilateral institutions, and standards bodies, analyzing how differing regulatory approaches affect innovation, competition, and market access. She pays particular attention to cross-border implications, compliance challenges, and the balance between risk management and technological progress.

Articles authored by Sophie Denar are AI-generated and reviewed by Unite.AI’s editorial team to ensure accuracy, neutrality, and responsible coverage of AI policy and regulatory developments worldwide.

來源:Unite.AI · unite.ai