The New York Times reported on November 7, 2026 that the US Commerce Department cleared Chinese companies to buy or assemble Nvidia H200 artificial intelligence chips even when some licensees had little visible presence at their listed addresses. The investigation centers on New Matrix, a firm created about 17 months earlier and approved this year to assemble H200 chips into AI servers. Times reporters found its listed office in an industrial park on the northwest edge of Beijing empty, with dangling wires and no listing in the building directory; property managers said they had no record of the company.
Chinese corporate records reviewed by the Times link New Matrix's owners to Wise Road, a private equity fund the United States added to the Commerce Entity List in 2024 for helping Beijing acquire strategically sensitive chip companies. Analysts who examined the records told the paper they believed New Matrix may be a front for firms tied to that fund. New Matrix did not respond to requests for comment. The paper says New Matrix is the second H200 licensee whose listed address appeared deserted; Maginfra, approved as a server assembler rather than an end buyer, also looked unused when a reporter visited in September, and analysts tied it to Inspur, which is on the Entity List for military work.
The Times places the clearances in a wider pattern. It says Commerce has approved roughly two dozen Chinese firms to buy or handle H200s, including Baidu, BYD and Tesla Shanghai, with quotas of tens of thousands of chips each that have not been fully filled. Reuters had earlier reported licenses for Alibaba, Tencent, ByteDance and ZTE Kangxun Telecom. Other licensees named by the Times, Zhejiang Yindun Cloud Technology and Shanghai Jiatang, are described as having ownership or operational links to Chinese military suppliers or state telecom entities that serve defense customers.
Policy context matters for the AI desk. The Biden administration had tightened AI chip export limits to China; President Trump later allowed H200 sales after lobbying by Nvidia CEO Jensen Huang, with chips routed through the United States so Washington can collect a cut of the sales (reported as a 25 percent tariff on value). Nvidia says China H200 shipments were under 1 percent of data-center revenue in the quarter ended July and that it is not assuming China sales in the current-quarter outlook. Nvidia spokesman John Rizzo said it makes no sense for the Chinese military to depend on American technology and that limited H200 sales still support US jobs. A Bureau of Industry and Security spokesman said the agency carefully reviews applications and cited $324 million in 2025 export-violation penalties plus more enforcement hiring.
Independent coverage arrived the same day. Terrence O'Brien at The Verge summarized the Times findings under the headline that the US let Nvidia sell chips to an empty office in China. Quartz, The Jerusalem Post, FourWeekMBA and other outlets also relayed the New Matrix and Maginfra empty-office details and Entity List links. The story is distinct from earlier Nvidia China export reporting (including prior Maginfra and Megaspeed coverage): the new development is the Times's November 7 finding that Commerce-cleared H200 handlers include shell-like firms with Entity List and military-supplier ties.
For AI infrastructure and policy desks, the report matters because export licensing is the main safeguard once H200 sales to China are allowed. If firms with empty offices and blacklisted ownership links can pass that review, the enforcement of Trump's H200 bargain is in question even while actual China-bound volumes remain small and Beijing still promotes domestic chips.