Nvidia’s business exposure to China is facing renewed scrutiny after reports suggested that a significant share of its artificial intelligence computing revenue is connected to Chinese customers through data centres in Southeast Asia. A report by US short-selling research firm Culper Research estimated that 20%–40% of Nvidia’s data centre business could be effectively linked to China, including computing capacity accessed through overseas cloud infrastructure and technology companies.
The allegations have gained attention as the United States tightens restrictions on advanced semiconductor exports and investigates possible routes through which Nvidia chips reach Chinese customers. However, the headline figure requires an important distinction: the 20%–40% estimate relates to China-connected data centre business, not necessarily Nvidia’s total reported revenue or proven illegal chip exports. A separate review by South Korea’s SBS News found insufficient evidence to establish that indirect exports to China account for 20%–40% of Nvidia’s overall revenue.
How Nvidia’s China-Linked Business Operates
US export restrictions prevent the direct sale of certain advanced AI chips to Chinese companies. However, Chinese businesses can potentially access computing power through data centres located in other countries.
Instead of purchasing restricted Nvidia chips directly, a Chinese technology company may rent computing capacity from a cloud provider or data centre operator based in Singapore, Malaysia or another overseas market.
The infrastructure can contain Nvidia’s advanced processors, while the computing services are delivered remotely to customers.
This creates a distinction between the physical location of the chips and the location of the companies ultimately using their computing power.
According to a November 14 report by The Wall Street Journal, large data centres in Southeast Asia have become an important source of computing capacity for Chinese companies. The report examined how overseas infrastructure has allowed Chinese businesses to access advanced AI computing despite restrictions on direct chip purchases.
The arrangement has also attracted regulatory attention because authorities are concerned that overseas computing services could be used to circumvent restrictions intended to limit China’s access to advanced AI technology.
Where Did the 20%–40% Revenue Estimate Come From?
The widely circulated estimate originated in a May 13, 2026 report published by Culper Research titled NVIDIA: The China Problem.
The research firm cited comments from a former senior Nvidia employee who estimated that approximately 20%–40% of Nvidia’s data centre business remained connected to China even after US export restrictions were introduced.
The report suggested that the actual figure could be closer to the higher end of that range.
However, the estimate was based partly on an anonymous former employee’s account and did not provide independently verifiable transaction records establishing the precise scale of China-linked sales.
The figure also included different forms of business exposure, including cloud computing arrangements and other transactions that do not necessarily involve illegal exports.
This distinction matters because China-related revenue, overseas computing access and prohibited chip shipments are three different categories.
South Korea’s SBS News examined the original research report, Nvidia’s financial disclosures and US Department of Justice announcements. Its fact-check concluded that available evidence was insufficient to establish that indirect exports to China represented 20%–40% of Nvidia’s total revenue.
Nvidia’s Financial Statements Show a Different Picture
Nvidia reports geographical revenue according to the headquarters location of its direct customers.
This accounting method does not necessarily identify where the company’s chips are ultimately installed or used.
For example, if Nvidia sells processors to a data centre company headquartered in Singapore and that company provides computing services to Chinese businesses, the revenue is generally attributed to Singapore rather than China in Nvidia’s geographical reporting.
Nvidia’s financial disclosures therefore cannot independently establish the full extent of China-related end-user demand.
Its fiscal 2026 annual report showed the following geographical revenue distribution:
| Customer headquarters location | FY2026 revenue |
|---|---|
| United States | $149.62 billion |
| Taiwan | $42.35 billion |
| China, including Hong Kong | $19.68 billion |
| Other markets | $4.30 billion |
| Total | $215.94 billion |
Source: Nvidia FY2026 annual report. Figures are based on direct customer headquarters, not necessarily the final destination or use of the products.
China, including Hong Kong, accounted for approximately 9.1% of Nvidia’s reported FY2026 revenue under this classification.
The company’s subsequent FY2027 second-quarter filing reported $12.43 billion in revenue from customers headquartered in China, including Hong Kong, during the first six months of the fiscal year.
That amount represented approximately 7% of Nvidia’s total revenue of $177.84 billion during the same period.
These figures do not disprove the possibility of additional China-linked business through third-country customers. They demonstrate why reported geographical revenue and estimates of indirect end-user exposure should not be treated as interchangeable.
New Street Research Estimates a Smaller Scale of Indirect Sales
Another estimate provides a different perspective on the controversy.
New Street Research reportedly estimated Nvidia’s illicit, circumvented and offshore China-related revenue at approximately $15 billion, equivalent to around 4% of total revenue.
This estimate is considerably lower than the 20%–40% figure cited by Culper Research.
The difference reflects the fact that the estimates are measuring different things.
The larger estimate includes broader China-related business exposure, such as computing services accessed through overseas data centres. The smaller estimate focuses on revenue associated with illicit, rerouted or offshore China-related sales.
Neither estimate should be treated as an independently confirmed measurement of all Nvidia chips ultimately used in China.
US Tightens Export Rules for Overseas Data Centres
The controversy has coincided with a significant regulatory development.
On November 16, 2026, the US Department of Commerce’s Bureau of Industry and Security announced an interim final rule titled Addressing Remote Access to Advanced Semiconductors.
The new framework is designed to prevent companies in restricted countries from accessing advanced US-designed computing power through overseas data centres.
The rules introduce additional licensing requirements for advanced semiconductor exports, re-exports and transfers involving five countries:
- Singapore
- Malaysia
- Indonesia
- Thailand
- Brazil
These countries have been identified by the US government as presenting a higher-than-normal risk of diversion.
The regulation also addresses overseas data centres that provide significant computing capacity to companies headquartered in restricted jurisdictions, including China. Operators and exporters will face additional due-diligence requirements concerning customers, ownership structures and end users.
Existing data centre operators covered by the remote-access provisions will generally have 90 days to obtain the required licences.
The rule is particularly relevant to Nvidia because its advanced processors are widely deployed in large AI data centres around the world.
Why Singapore and Malaysia Are Important
Singapore and Malaysia have become important locations for data centre investment and cloud computing infrastructure.
Their geographical position, international connectivity and growing technology ecosystems have attracted global cloud providers and data centre operators.
These facilities can provide computing services to customers across several countries without requiring the customer to physically own the underlying processors.
For Chinese AI companies facing restrictions on direct access to advanced American chips, overseas computing capacity can provide an alternative route to obtaining processing power.
However, the existence of data centres in these countries does not automatically mean that their operators are violating US export controls.
The legality of individual arrangements depends on the chips involved, the parties, the destination, the end use and applicable licensing requirements.
The new US rules are intended to strengthen oversight of these arrangements.
What Does the Investigation Mean for Nvidia?
The growing scrutiny could create additional compliance obligations for Nvidia, its distributors, cloud partners and international data centre customers.
The company operates through a large network of direct and indirect customers, including distributors, original design manufacturers, original equipment manufacturers, cloud service providers and system integrators.
This makes it more complicated to track the ultimate destination and use of every processor.
The US Department of Justice has announced several investigations and enforcement actions involving alleged attempts to divert Nvidia AI chips to China through third countries.
These cases have increased pressure on regulators and technology companies to strengthen supply-chain monitoring.
However, allegations involving third-party companies do not by themselves establish that Nvidia management knowingly participated in illegal transactions.
The extent of any corporate responsibility will depend on the evidence, the transactions involved and the findings of relevant investigations.
The Bigger Picture
The Nvidia controversy highlights a major challenge in global semiconductor regulation: restricting the physical export of advanced chips does not necessarily prevent access to the computing power those chips provide.
As AI infrastructure becomes increasingly distributed across international data centres, regulators must consider not only where processors are shipped but also who ultimately uses the computing services.
For Nvidia, China represents both a significant market opportunity and a complicated regulatory risk. The company must balance international demand for its technology with increasingly strict export-control requirements. For the United States, the challenge is enforcing restrictions without creating uncertainty for legitimate cloud providers and international technology businesses.
Looking Ahead
The immediate focus will be on the implementation of the new US export rules and the investigations into alleged diversion of advanced AI chips through overseas intermediaries. More detailed information about data centre customers, ownership structures and end-user access could help regulators determine the actual scale of China-linked computing activity. The 20%–40% estimate will remain unverified unless stronger transaction-level evidence becomes available.
For Nvidia, the outcome could influence its international distribution strategy, cloud partnerships and compliance costs. Investors will need to distinguish between reported China revenue, broader China-related computing exposure and proven illegal exports when evaluating the company’s business risks. These distinctions will become increasingly important as governments expand controls over the global AI semiconductor supply chain.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.



