The Verge published on November 7, 2026 that the US government approved the sale of Nvidia H200 chips to New Matrix, a Chinese company created 17 months ago whose listed office is empty. The company does not appear in the building directory, and industrial park staff confirmed it does not occupy the space.

Nvidia: H200 chips approved for Chinese company with no office
The U.S. government approved the sale of Nvidia H200 chips to New Matrix, a Chinese company created 17 months ago whose listed office is empty and whose owners have ties to a sanctioned entity.
Marcelo Santiago
Criador do bridgecode
Get AI articles in your inbox
Technical curation on models, agents and coding tools. No spam, unsubscribe anytime.
In this article you will see:
- What New Matrix is and why its approval is questionable
- How the government relaxed chip sales restrictions to China
- The pattern of companies with ties to sanctioned entities
- What this means for AI export control policy
What is New Matrix, the approved company with no office
New Matrix was created 17 months ago. At the address the company lists as headquarters, a building company directory does not include the firm. At the administration office of the nearby industrial park, a receptionist and a manager both said they did not know the company.
Little information can be found online about New Matrix. Its owners have ties to a Chinese private equity firm that has been on the US government sanctions list since 2024. According to The New York Times, this is not a rare oversight, but a worrying pattern of companies connected to sanctioned entities receiving approval to buy advanced chips.
The approval of New Matrix is part of a list of Chinese companies authorized to receive Nvidia H200 chips, despite ties to sanctioned companies.
How the government relaxed chip sales restrictions
Nvidia had been restricted from selling chips to China during the Biden presidency. But the Trump administration relaxed those limits after Nvidia agreed to give the government 15% of all sales. The company's CEO, Jensen Huang, also launched an intense lobbying campaign and was apparently successful.
The relaxation of restrictions means Chinese companies with ties to sanctioned entities can receive advanced AI chips, as long as they pay the 15% tax to the US government. Due diligence verification appears to be insufficient.
Concerns are that neither the Department of Commerce nor Nvidia are doing adequate verification when assessing Chinese companies authorized to handle Nvidia H200 chips.
The pattern of companies with sanctioned ties
The New Matrix case is not isolated. According to The New York Times, there is a pattern of companies connected to sanctioned entities receiving approval to buy advanced chips. Due diligence verification appears to be flawed in multiple cases.
This pattern raises questions about the effectiveness of the AI chip export control system. If companies with no physical office and ties to sanctioned entities can get approval, the system is not fulfilling its goal of preventing advanced technology from reaching unwanted actors.
The situation is made worse by the fact that Nvidia depends on China for a significant portion of its revenue. Pressure to maintain access to the Chinese market creates incentives not to question the legitimacy of buyers, even when there are red flags.
What this means for AI chip export control policy
For developers and users of AI, the news shows that AI chip export controls are weakened. The combination of relaxed restrictions, insufficient verification and financial incentive (15% to the government) creates a scenario where advanced chips can reach unauthorized actors.
Nvidia argues that selling chips to China keeps the company competitive and funds research and development. But inadequate verification can compromise national security and the integrity of the export control system.
For engineers building AI systems, the scenario reinforces the importance of considering the origin and security of the chips used in AI infrastructure. The AI hardware supply chain is becoming a geopolitical risk factor.
What changes for the AI hardware market
The availability of advanced chips for Chinese companies affects the global AI hardware market. More supply in China means less pressure on the global supply chain, but it also means Chinese competitors will have access to cutting edge hardware.
For companies that depend on AI infrastructure, the scenario is one of uncertainty. Export policy can change at any moment, affecting prices and GPU availability. Anyone planning investments in AI infrastructure needs to consider this risk.
Nvidia, in turn, faces a dilemma: keep access to the Chinese market and accept the loosening of rules, or lose revenue and cede space to competitors like Huawei, which develops its own chips.
Frequently asked questions about Nvidia H200 chip approval for China
What is New Matrix?
It is a Chinese company created 17 months ago that received approval to buy Nvidia H200 chips, but whose listed office is empty and whose owners are linked to a sanctioned private equity firm.
How did the government relax chip restrictions for China?
The Trump administration relaxed the limits after Nvidia agreed to give the government 15% of all sales, and CEO Jensen Huang ran an intense lobbying campaign.
Is due diligence verification working?
No. According to The New York Times, there is a pattern of companies connected to sanctioned entities receiving approval to buy advanced chips.
What does this mean for export policy?
The AI chip export control system is weakened, with insufficient verification and a financial incentive that could compromise national security.
Why does this matter for AI users?
The AI hardware supply chain is becoming a geopolitical risk factor, affecting the availability and cost of AI infrastructure.
Conclusion: AI chip export controls are weakened
The approval of H200 chips for a Chinese company with no physical office, with ties to sanctioned entities, exposes flaws in the export control system. The relaxation of restrictions and insufficient verification create a scenario where advanced technology can reach unauthorized actors.
AI chip export policy is in transition, and the results of that transition affect the availability and cost of AI hardware worldwide. Anyone who relies on AI infrastructure needs to follow these changes.
To understand how open models are closing the gap with proprietary models in cybersecurity, see Open source models close the gap in cybersecurity. If you want to know how AI agents behave in real attacks, check out Vals CyberOffense: open models invade corporate networks.
Source: The Verge
Related articles

Box uses Nvidia's Nemotron for agent content search
Box and Nvidia announce Box Context Retrieval, a semantic search engine powered by Nemotron that turns enterprise content into knowledge for AI agents.

Arete-27B: Nous retrains Qwen for Western values
Arete-27B is a fine-tune of Qwen3.8-27B by Nous Research to reduce political biases, with factuality rising from 2.32 to 2.98 and propaganda dropping from 22% to 0.9%.

Karpathy builds a video platform over a weekend with Claude
Andrej Karpathy built HackerTube in a weekend with Claude. What it says about development speed with coding agents.
