Skip to main content
Back to News
news/AI Policy

The US cleared Nvidia chips for Chinese firms that barely exist

A New York Times investigation finds Commerce approved Nvidia H200 buyers in China whose offices are empty and whose owners tie to blacklisted funds.

Stefan Trbojevic

Stefan Trbojevic

7 November 20263 min read
LinkedIn
Frosted glass panels in a dark empty corridor, editorial illustration of empty corporate offices

The takeaway

The revenue-sharing export bargain only works if the buyers are real. Vetting failed, so vendor-side diligence and a likely congressional audit are next.

Why it matters for builders

Provenance of compute supply is becoming a compliance requirement: know the legal entity behind any H200 capacity you rent, and expect license-by-license disclosure pressure.

The US cleared Nvidia chips for Chinese firms that barely exist

The Commerce Department has signed off on roughly two dozen Chinese companies cleared to buy or assemble Nvidia's H200 accelerators. Reporters from The New York Times went looking for two of them and found empty buildings, no staff, and owners tied to funds the US government has already blacklisted.

What the reporters found

New Matrix, a Beijing company incorporated 17 months ago, holds one of the licenses. Its registered address is a hollow gray structure with wires hanging where light fixtures should be. The building's company directory does not list it, and both a receptionist and a manager at the industrial park's property office said they had no record of the firm. Corporate filings tie New Matrix's owners to Wise Road, a Chinese private equity fund placed on the US entity list in 2024 for allegedly helping Beijing acquire sensitive semiconductor assets.

It is not an isolated case. A second licensee, Maginfra, had an equally deserted office when Times reporters visited in September. Analysts cited by the paper believe Maginfra is linked to Inspur, which landed on the entity list over work with the Chinese military. The broader roster includes Baidu, BYD and Tesla's Shanghai operations alongside the shell-like entities.

Abstract render of solid and hollow glass containers scanned by a thin light line

Why the licensing bargain is being tested

Washington relaxed the H200 restrictions after Nvidia agreed to hand the US government a share of the revenue from those sales. That deal only works if the firms receiving the chips can be verified as real, independent buyers. The Times reporting suggests the vetting did not reach that bar.

"I have a ton of concerns that neither the Commerce Department or Nvidia is doing the due diligence needed to ensure that the licensing terms are being followed," Liza Tobin, a former China director at the White House National Security Council, told the paper. Nvidia spokesman John Rizzo pushed back, arguing it "makes no sense for the Chinese military to depend on American technology," while a Bureau of Industry and Security spokesman said every application is reviewed for national security risk.

Commercially the stakes look small: H200 shipments to China accounted for less than 1% of Nvidia's data center revenue in the quarter ending July. Politically they are not. These licenses are the first real test of whether a revenue-sharing export compromise can be policed, and the answer so far is that nobody can say who sits on the other end of the wire transfer.

What it means for builders

If licensing authorities cannot verify a counterparty, the burden lands on vendors and on anyone buying capacity through intermediaries. Teams sourcing H200 compute in Asia should be able to name the legal entity behind their supplier, not just the reseller. Expect audit pressure too: congressional China-competition committees now have a concrete case to demand license-by-license disclosure, and any pause or revocation would ripple across every other name on the approved list.

For anyone building agents on rented GPU capacity, the lesson is unglamorous: supply-chain provenance is becoming a compliance feature, not a procurement detail.

Share𝕏

The Automation Brief

Read 5 AI stories instead of 50.

The essential moves in AI agents, models, automation and infrastructure — filtered for builders and operators, with the part that actually matters.

No noise. Unsubscribe anytime.

Editorial notes

Reported by

Stefan Trbojevic

Edited by

n8n Lab Editorial

Published

7 November 2026

Updated

7 November 2026

AI disclosure: AI assisted with research and drafting. Factual claims are reviewed by an editor.

n8n Lab is an independent service provider. We are not affiliated with, endorsed by, or sponsored by n8n GmbH. “n8n” is a trademark of n8n GmbH and is used here only to describe the platform-specific implementation and automation services we provide.