Key Takeaways
- The US Department of Commerce approved Chinese companies to handle Nvidia’s H200 artificial intelligence chips, according to an investigation by The New York Times.
- One approved company, New Matrix, was found to have an apparently empty office at its listed Beijing address.
- New Matrix was established only 17 months before the investigation and has limited publicly available information about its operations.
- Its owners reportedly have connections to Wise Road Capital, a Chinese investment firm placed on the US government’s Entity List in 2024.
- Another approved company, Maginfra, also had a reportedly deserted office and alleged links to Inspur, a Chinese technology company on the US Entity List.
- The investigation raises questions about how thoroughly Washington checks companies before allowing them to handle advanced AI processors.
- Neither the empty offices nor the reported corporate connections prove that illegal chip transfers took place.
US Approval of Nvidia H200 Sales to China Raises Questions
The United States is facing questions over its export-control process after approving Chinese companies to handle Nvidia’s advanced H200 artificial intelligence chips, including a recently established company whose listed office was found to be empty.
The investigation, originally reported by The New York Times on November 7, 2026, examined companies that received approval from the US Department of Commerce to purchase or handle Nvidia’s H200 processors. Reporters found that some of the approved businesses had little visible operational presence, while others had ownership connections to companies facing US national-security-related trade restrictions.
The most prominent example is New Matrix, a Chinese company authorised to assemble Nvidia H200 chips into AI servers. When a reporter visited its listed address in Beijing, the office appeared to be an unfinished, empty space. The company was also absent from the building directory, and the industrial park’s property-management representatives reportedly said they had no record of it.
The findings have raised questions about whether the US government and Nvidia have sufficient visibility into the companies involved in the Chinese chip supply chain.
However, the investigation does not establish that Nvidia shipped chips to an empty office, that New Matrix received any processors or that the company committed an offence. The central issue is the quality of the screening process and the identity of the businesses authorised to handle sensitive technology.
New Matrix: The Company at the Centre of the Investigation
New Matrix is the most striking example in the investigation.
The company was established approximately 17 months before the November 2026 report. It received approval to assemble Nvidia H200 processors into computers used for artificial intelligence applications.
However, its listed Beijing office did not appear to have the facilities or visible activity normally associated with a company handling sophisticated AI hardware.
According to the investigation, several details attracted attention:
- Empty office: The listed address was found to be an unfinished space rather than an operating business office.
- Missing directory listing: The company’s name did not appear in the building’s directory.
- No property-management record: Representatives at the industrial park reportedly said they had no record of the company.
- Limited online presence: Its website contained little information about its history, employees or operations.
- Recent establishment: The company had been created only 17 months earlier.
These findings do not automatically mean New Matrix is a fake company. Businesses can operate from different locations, use contractors or change addresses. However, they raise questions about how its operational capabilities and ownership structure were assessed before approval.
New Matrix’s Reported Connection to Wise Road Capital
Another concern involves the company’s ownership.
Chinese corporate records reviewed by The New York Times reportedly showed that New Matrix’s owners had connections to Wise Road Capital, a Chinese private equity firm added to the US Commerce Department’s Entity List in 2024.
The US government previously said Wise Road had helped the Chinese government acquire strategically important companies with semiconductor technology.
Two analysts who examined the corporate records and other Chinese-language information reportedly believed New Matrix could be a front for businesses associated with the investment firm.
That interpretation remains an assessment by the analysts, not an official government finding that New Matrix is a front company.
Nevertheless, the reported connection raises an important question: how effectively can US authorities identify indirect ownership relationships when reviewing companies seeking access to advanced American technology?
Maginfra Also Had a Describedly Empty Office
New Matrix was not the only company to attract attention.
The investigation also examined Maginfra, another Chinese company approved to handle Nvidia H200 chips and assemble them into AI servers.
When a New York Times reporter visited its listed address in September 2026, the office appeared deserted. The space reportedly contained a few chairs and empty desks inside a small locked room.
Workers in the building said they had not seen anyone using the office, while reception staff reportedly did not recognise the company.
Analysts interviewed by the newspaper believed Maginfra could be an offshoot of Inspur, a Chinese server manufacturer that has been placed on the US government’s Entity List over national-security concerns.
The reported connections include shared personnel and patents, as well as government ownership links.
Again, these relationships do not independently prove that Maginfra violated export rules or transferred chips to a prohibited end user. They do, however, highlight why ownership and operational checks are important when approving companies to handle advanced processors.
Other Approved Chinese Companies Have Reported Military-Linked Connections
The investigation also identified concerns involving other companies that received approval to purchase or handle Nvidia H200 processors.
Two examples were Zhejiang Yindun Cloud Technology and Shanghai Jiatang.
| Company | Reported connection | Why it attracted attention |
|---|---|---|
| New Matrix | Owners linked to Wise Road Capital | Listed office appeared empty; Wise Road is on the US Entity List |
| Maginfra | Reported links to Inspur | Listed office appeared deserted; Inspur is on the US Entity List |
| Zhejiang Yindun Cloud Technology | Connections to Range Intelligent Computing Technology Group | The group has provided computing services connected to defence-related programmes |
| Shanghai Jiatang | Originated from state-owned Datang Telecom | The wider organisation has connections to technology supplied to China’s military |
The details matter because US export controls are designed to prevent sensitive American technology from reaching restricted organisations or being used in prohibited military applications.
A company may not itself appear on a restricted list but could still have ownership, management or commercial relationships that deserve additional scrutiny.
The investigation therefore raises questions about whether screening systems are sufficiently equipped to identify complex corporate structures and indirect relationships.
What Is the Nvidia H200 Chip and Why Does China Want It?
The H200 is an advanced artificial intelligence accelerator developed by Nvidia. It belongs to the company’s Hopper architecture and is designed to support demanding AI workloads.
Unlike ordinary consumer graphics cards, data-centre GPUs such as the H200 are built to process large amounts of data simultaneously. They are particularly important for training and running large AI models.
Nvidia H200: Key Technology Details
| Feature | Details |
|---|---|
| Manufacturer | Nvidia |
| Product | H200 Tensor Core GPU |
| Architecture | Hopper |
| Primary market | AI data centres |
| Main applications | AI model training and inference |
| High-bandwidth memory | 141GB HBM3e |
| Memory bandwidth | 4.8TB/s |
The H200’s large memory capacity and high bandwidth make it valuable for handling large AI models and data-intensive workloads.
China wants access to advanced computing hardware because powerful processors are essential for developing competitive AI models, operating cloud infrastructure and expanding data-centre capacity.
However, the US government has restricted exports of certain advanced processors to China because of concerns about their potential use in military and other sensitive applications.
This makes the identity of companies receiving access to H200 chips an important national-security issue.
How the US Changed Its Nvidia China Chip Policy
The investigation comes after a shift in Washington’s approach to Nvidia’s Chinese market.
The Biden administration imposed restrictions on exports of advanced AI processors to China. These controls were designed to limit access to computing technology that could support China’s military and advanced AI capabilities.
In December 2025, President Donald Trump announced a policy change allowing Nvidia to sell its H200 processors to approved Chinese customers under specified conditions.
The new arrangement did not mean that every Chinese company could freely purchase the chips. Companies involved in the transactions still had to undergo review by the US Commerce Department.
The department’s approval process was intended to assess national-security risks associated with the businesses handling the processors.
The New Matrix investigation has now raised questions about whether that process adequately examined the companies’ physical operations, ownership structures and potential connections to restricted organisations.
The financial conditions attached to the revised export policy have also been reported differently across media coverage. The exact terms should not be confused with the separate question of which Chinese companies received individual approvals.
US Commerce Department and Nvidia Respond to the Concerns
The investigation has prompted questions about how the US government evaluates export applications.
A spokesperson for the Commerce Department’s Bureau of Industry and Security told The New York Times that the agency carefully reviews applications for national-security risks.
The department also pointed to its enforcement activity, including reported export-control penalties and additional enforcement personnel.
Nvidia, meanwhile, defended the broader policy of allowing some chip exports to China.
Company spokesperson John Rizzo argued that it made little sense for the Chinese military to depend on American technology. Nvidia has also maintained that China has been restricting imports of H200 processors as it promotes domestically developed alternatives.
These responses highlight the competing considerations behind the policy.
The US government wants to protect sensitive technology, while Nvidia wants access to a major market for its AI processors. China, meanwhile, is trying to expand its own semiconductor industry while maintaining access to computing resources.
The latest investigation adds another question to that debate: whether the companies receiving approval are being properly identified and monitored.
Why the Empty Offices Matter for US Export Controls
An empty office might appear to be a minor administrative problem. In the semiconductor industry, however, the identity of the company receiving or assembling advanced processors can have significant consequences.
AI chips are high-value products that can be incorporated into powerful computing systems. Once exported, they may be used by cloud providers, research institutions, commercial businesses or organisations involved in sensitive activities.
That is why export-control systems often examine more than a company’s registered name.
Authorities may need to understand:
- Ownership: Who ultimately controls the business?
- Management: Who makes operational and purchasing decisions?
- Facilities: Does the company have the infrastructure required to handle the products?
- Customers: Who will ultimately use the processors?
- End use: Will the technology be used for permitted commercial purposes?
- Diversion risks: Could the products be transferred to restricted organisations?
The New Matrix and Maginfra findings raise questions about how these checks were conducted.
However, a site visit showing an empty office is only one piece of evidence. A full assessment would require access to licensing records, ownership documents, customer information and the government’s internal review process.
What This Means for Nvidia and the Global AI Chip Market
China remains an important potential market for AI hardware, but Nvidia’s direct sales to Chinese customers have been limited by export restrictions and uncertainty over import approvals.
According to figures cited in the investigation, shipments of H200 chips to China-based customers accounted for less than 1% of Nvidia’s data-centre revenue in the quarter ending in July 2026.
The report also noted that Nvidia had not included China data-centre sales in its revenue projections for the following quarter.
This suggests that the immediate financial contribution from the reopened Chinese market was relatively small compared with Nvidia’s overall data-centre business.
Nevertheless, the market remains strategically important.
More scrutiny of Chinese buyers could slow the approval of future transactions, increase compliance costs and make it harder for Nvidia to plan its China business.
For Chinese AI developers, uncertainty over access to advanced American processors could also encourage further investment in domestic alternatives.
For Washington, the issue is how to maintain a balance between commercial exports and national-security restrictions.
The Bigger Picture
The Nvidia H200 approval controversy highlights a major weakness that can emerge in global technology supply chains: controlling which products can be exported is only part of the challenge. Governments must also identify the companies receiving those products, understand who owns them and verify their intended use.
The New Matrix investigation does not prove that prohibited chips reached China’s military or that the company committed wrongdoing. But the reported empty offices and ownership connections show why export-control screening must extend beyond paperwork and registered company names.
As artificial intelligence becomes increasingly important to economic and national-security competition, advanced processors will remain closely monitored. The effectiveness of export restrictions will depend not only on the rules themselves but also on how consistently authorities enforce them.
Frequently Asked Questions
1. Why did the US approve Nvidia chips for a Chinese company with an empty office?
The US Commerce Department approved New Matrix to handle or assemble Nvidia H200 chips into AI servers. The New York Times later found that its listed Beijing office appeared empty. The investigation has raised questions about the screening process, but the full internal reasoning behind the approval has not been publicly established.
2. What is New Matrix?
New Matrix is a Chinese company established approximately 17 months before the November 2026 investigation. It received approval to assemble Nvidia H200 processors into AI servers. Its listed office was found to be empty, and its owners reportedly had connections to Wise Road Capital.
3. What is Wise Road Capital?
Wise Road Capital is a Chinese private equity firm that was added to the US Commerce Department’s Entity List in 2024. Its reported connections to New Matrix’s owners have raised questions about the company’s ownership and approval.
4. What is Nvidia H200?
The Nvidia H200 is a high-performance AI accelerator based on the Hopper architecture. It features 141GB of HBM3e memory and is designed for demanding AI training and inference workloads in data centres.
5. Did Nvidia actually ship H200 chips to New Matrix?
The investigation does not establish that New Matrix received any H200 processors. Approval to handle or assemble chips is not proof that a shipment took place.
6. Why does the US restrict advanced Nvidia chip sales to China?
The US government has restricted exports of advanced AI processors because of concerns that powerful computing technology could support China’s military capabilities and other sensitive applications.
7. Is Nvidia accused of breaking export-control laws?
The reporting raises questions about the screening of approved companies, but it does not establish that Nvidia violated export-control laws or that the companies named committed an offence.
Looking Ahead
The investigation could increase pressure on the US Commerce Department to explain how it evaluates Chinese companies seeking access to advanced American processors. If further reviews identify weaknesses in ownership checks or end-user verification, Washington could introduce stricter documentation requirements, additional inspections or tighter restrictions on individual companies.
For Nvidia, the situation creates another layer of uncertainty around its China strategy. The company is trying to serve demand for AI computing while operating under export restrictions that can change with government policy. The longer-term outcome will depend on whether the US can maintain a credible approval system, whether China permits the imports and how quickly domestic Chinese chipmakers can offer competitive alternatives.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
