NVIDIA’S TRUE CHINA EXPOSURE COMES UNDER SCRUTINY AS REPORTS SAY BILLIONS IN AI REVENUE MAY TRACE BACK TO CHINESE DEMAND

Nvidia’s own filings show that customers headquartered in China and Hong Kong generated $12.43 billion in revenue during the six months ended July 26, 2026, equal to roughly 7 per cent of the company’s $177.84 billion in total revenue over that period. Claims that more than 20 per cent of Nvidia’s fiscal 2026 compute revenue was ultimately driven by China come from short seller Culper Research and have not been independently confirmed, although Wall Street Journal reporting suggests Chinese demand may have generated billions more through Southeast Asian data centres and intermediaries that do not appear as China sales in Nvidia’s geographic disclosures. The issue has taken on greater significance following Jensen Huang’s resignation and growing U.S. scrutiny of companies such as Megaspeed, while Washington is tightening export controls to prevent Chinese companies from accessing advanced Nvidia computing power through overseas facilities. StreetInsider.com
Nvidia is facing intensifying scrutiny over the true scale of its business exposure to China as new reporting and earlier short-seller research raise questions about whether official revenue figures dramatically understate how much Chinese demand has contributed to the chip giant’s extraordinary growth. The company’s own financial filings show that customers headquartered in China and Hong Kong generated approximately $12.43 billion in revenue during the six months ended July 26, 2026, equivalent to roughly 7 per cent of Nvidia’s $177.84 billion in total revenue over that period, but investigators, journalists and market analysts are increasingly focused on a much broader question: how much revenue was ultimately driven by Chinese end users even when the direct customer was headquartered somewhere else? That distinction has become central to a widening political and regulatory controversy surrounding Nvidia, former chief executive Jensen Huang and the network of data centres, distributors and cloud providers that enabled Chinese technology companies to continue accessing advanced Nvidia computing power despite tightening U.S. export controls.
The most explosive estimate comes from Culper Research, a U.S. short-selling firm that disclosed in May that it was betting against Nvidia and alleged that more than 20 per cent of the company’s fiscal 2026 compute revenue remained connected to Chinese demand through a combination of Southeast Asian intermediaries, overseas data centres and alleged illegal diversion of restricted chips. Culper’s report, titled “NVIDIA: The China Problem,” argued that the market was underestimating how dependent Nvidia remained on Chinese customers even after U.S. controls sharply restricted direct sales of advanced artificial-intelligence processors to the country. Because Culper holds a short position in Nvidia, its financial incentive is aligned with a decline in Nvidia’s stock price, and its claims must therefore be treated as allegations rather than neutral findings. Independent fact-checking has also concluded that there is not yet enough public evidence to prove that 20 to 40 per cent of Nvidia’s overall revenue was secretly flowing to China, although multiple sources agree that official geographic disclosures do not necessarily capture the final destination or user of Nvidia’s chips.
Nvidia itself acknowledges that its geographic revenue disclosures can differ significantly from the ultimate destination of its products. In its quarterly filings, the company states that revenue by geographic region is classified according to the headquarters location of the direct customer, while the end user and shipping location may be entirely different. That accounting convention is crucial to understanding the controversy. A Singapore-based distributor, Taiwanese original-design manufacturer or Malaysian data-centre operator that buys Nvidia chips would normally be recorded under the headquarters location of that direct purchaser, even if the computing capacity was ultimately rented to, financed by or otherwise used by a Chinese technology company. Nvidia’s filings therefore provide a reliable measure of direct sales by customer headquarters but do not necessarily reveal the complete economic exposure to Chinese demand.
For the first half of Nvidia’s fiscal 2027, customers headquartered in the United States generated $123.84 billion in revenue, Taiwan accounted for $38.99 billion and China, including Hong Kong, contributed $12.43 billion, with another $2.57 billion attributed to other regions. Those figures show that China remains a meaningful but officially much smaller part of Nvidia’s business than it was before Washington imposed sweeping restrictions on high-end artificial-intelligence chips. Yet the enormous revenue attributed to Taiwan and the wider Asian supply chain has drawn attention because many direct buyers in those markets manufacture systems, operate cloud capacity or distribute products on behalf of clients located elsewhere. That makes it difficult to determine from financial filings alone how much of the value ultimately derives from Chinese technology groups.
The Wall Street Journal has now added new detail to that debate by reporting that former Nvidia chief executive Jensen Huang helped build a sprawling commercial ecosystem in which advanced Nvidia processors were supplied to data centres across East and Southeast Asia, allowing Chinese companies to rent access to high-performance computing infrastructure outside mainland China. According to reporting republished by The Times, Chinese firms could obtain access either through alleged smuggling of restricted processors or by renting computing power from overseas facilities in countries such as Malaysia, where massive data-centre investment has been driven in part by demand from Chinese technology companies. The report says such arrangements helped preserve access to cutting-edge Nvidia hardware even when U.S. export restrictions prevented direct shipment of particular chips into China.
That business model occupies a complicated legal and regulatory space. U.S. export controls historically focused heavily on the physical export of advanced chips and systems into China, but they did not always prevent a Chinese company from remotely accessing a powerful GPU located in another country. That distinction created an opportunity for overseas data-centre operators to buy Nvidia hardware legally and then sell cloud-computing access to customers in China. The arrangement did not necessarily violate export rules simply because the end user was Chinese, although regulators have increasingly worried that remote computing could allow restricted capabilities to reach Chinese military, intelligence or commercial users in ways that undermine the intent of the controls. The U.S. government has therefore begun tightening licensing rules for advanced chips shipped to several Southeast Asian countries, including Malaysia and Singapore, precisely because authorities are concerned about diversion and remote access.
The controversy has focused particularly heavily on Megaspeed, a Singapore-based company that has emerged as one of the most important entities in the wider Nvidia-China debate. The Wall Street Journal reported that Megaspeed arranged nearly $12 billion in Nvidia chip shipments to Chinese companies between 2024 and 2025, although available records do not establish how many of those chips were ultimately delivered, which precise models were involved or where every shipment originated. Megaspeed has said that it complies with applicable laws and is reviewing aspects of its business, while Nvidia has said it works to prevent smuggling and does not support customers that violate U.S. export restrictions. Those competing statements underscore why the case remains one of allegations, investigations and unresolved evidence rather than a settled finding that Nvidia knowingly participated in illegal diversion.
Megaspeed has attracted even more attention because of its connection to Alice Huang, a Chinese businesswoman who had a personal relationship with Jensen Huang and who became a central figure in the scandal that ultimately preceded his resignation as Nvidia’s chief executive on November 12. Reuters has reported that U.S. intelligence agencies are investigating whether Alice Huang had ties to China’s Ministry of State Security and whether she may have shared information with Beijing, while stressing that authorities have not publicly reached a final determination. Jensen Huang acknowledged having an affair with her, and Nvidia’s board opened an investigation into potential foreign-intelligence concerns before Huang stepped down as president, chief executive and board member. Colette Kress, Nvidia’s longtime chief financial officer, is serving as interim CEO while the board searches for a permanent successor.
That personal scandal has accelerated scrutiny of business decisions that previously attracted less public attention. Huang had spent years arguing that continued engagement with China was essential both to Nvidia’s commercial success and to America’s broader technological leadership. He repeatedly warned that forcing Chinese customers entirely away from U.S. technology could accelerate the growth of Chinese semiconductor alternatives and permanently reduce American influence over the global AI ecosystem. His argument was essentially strategic: selling U.S.-designed hardware to the widest possible international customer base, within the law, would keep global AI infrastructure dependent on American technology rather than pushing customers toward Huawei or other Chinese competitors. Critics now argue that this approach may also have created incentives to preserve Chinese demand through increasingly indirect channels.
The financial significance of the China market explains why the debate has become so intense. Even using Nvidia’s own direct-customer accounting, the company generated $12.43 billion from customers headquartered in China and Hong Kong during only the first six months of fiscal 2027. That represents roughly 7 per cent of global revenue, a percentage that would make China a major standalone market even without including indirect demand. Nvidia’s total revenue over the same six months was $177.84 billion, driven overwhelmingly by its Data Center business, which generated $164.27 billion. The company’s explosive growth demonstrates why every large source of demand matters, particularly at a time when investors are valuing Nvidia largely on expectations that AI infrastructure spending will continue expanding at extraordinary rates.
The figure cited in current reports that Nvidia generated approximately $12.4 billion in China-related revenue over the six months ended July 26 is therefore directly supported by Nvidia’s financial filings. What is not established is whether that $12.4 billion represents the full economic value of Chinese demand. The Wall Street Journal and Culper Research both suggest the true figure could be substantially higher because chips sold elsewhere in Asia may ultimately have served Chinese technology companies. That distinction between officially reported China revenue and economically China-driven revenue is now at the heart of the regulatory debate.
Culper’s “over 20 per cent” claim goes further than the Journal’s reporting and should be treated more cautiously. The short seller alleges that China remained responsible for more than one-fifth of Nvidia’s fiscal 2026 compute revenue even after direct restrictions sharply limited shipments into the mainland. According to Culper, Chinese demand persisted through illegal diversion and Southeast Asian intermediaries, including facilities that purchased Nvidia hardware and then made computing capacity available to Chinese customers. The firm also cited comments from former Nvidia employees and external estimates suggesting that 20 to 40 per cent of Nvidia’s data-centre demand may at times have remained functionally Chinese. Those estimates have not been independently verified, and the firm’s disclosed short position gives it a financial interest in emphasising downside risk.
The difference between direct revenue and end-user demand becomes especially important because Nvidia’s supply chain is highly international. The company does not manufacture its own most advanced chips; they are fabricated primarily by Taiwan Semiconductor Manufacturing Company and then incorporated into servers and systems by a wide range of manufacturers, original-design companies and cloud operators. A sale recorded as Taiwanese revenue may therefore represent a chip purchased by a Taiwanese company for integration into a system ultimately installed in Malaysia, Singapore, the Middle East, Europe or the United States. The financial statements were not designed to track every ultimate end user, which makes it difficult to reconstruct the true geographic distribution of computing demand from public filings alone.
Alibaba has become part of that debate because reports have linked Chinese technology companies to financing or consumption arrangements involving overseas data centres. Culper Research alleged that Megaspeed received indirect funding connected to Alibaba through a network of companies, while Wall Street Journal reporting has described Chinese technology groups as major users of computing power hosted abroad. Alibaba has not been proven to have orchestrated illegal Nvidia-chip diversion, and any such claim should remain clearly attributed. The broader issue is whether Chinese companies were legally renting cloud capacity located outside China or whether physical chips were unlawfully diverted across borders, two very different scenarios under U.S. export law.
ByteDance and other Chinese technology companies have also been cited in reports discussing overseas AI-computing demand. Chinese firms remain among the world’s largest consumers of artificial-intelligence infrastructure because they operate enormous cloud platforms, recommendation systems, e-commerce networks and generative-AI models. U.S. export restrictions therefore created a huge commercial incentive for companies to find lawful alternatives to direct chip purchases, including renting GPU capacity abroad. That demand helps explain why countries such as Malaysia have experienced a surge in data-centre construction and why U.S. officials increasingly view Southeast Asian cloud infrastructure through the lens of strategic competition with China.
The new U.S. response reflects that concern. Washington has moved toward requiring licences for advanced semiconductor exports to countries including Malaysia, Singapore and Indonesia when officials believe the hardware could ultimately be accessed by Chinese users. The objective is not simply to prevent a restricted chip from physically crossing into China but also to prevent Chinese entities from obtaining equivalent computing capability remotely. That represents a significant evolution in export-control policy because cloud computing does not respect national borders in the same way physical goods do. A Chinese AI company can potentially train models on GPUs thousands of kilometres away while never taking possession of the hardware itself.
For Nvidia, tighter controls create a strategic dilemma. The company wants to maximise global sales and maintain its dominant share of AI computing while complying with U.S. national-security restrictions. Every expansion of licensing requirements therefore risks cutting into demand, while every perceived loophole generates political pressure for even stricter rules. Huang spent years arguing that overly aggressive controls would strengthen Chinese semiconductor competitors by encouraging customers to abandon Nvidia entirely. U.S. policymakers, by contrast, increasingly argue that the most advanced AI processors have military and intelligence significance that justifies restricting access even at substantial commercial cost.
That debate has become more politically sensitive following Huang’s resignation. Before stepping down, he was one of the world’s most powerful technology executives and had developed close relationships with policymakers, customers and political leaders, including President Donald Trump. His influence over discussions about AI export policy meant that his personal scandal immediately raised questions about whether Nvidia’s lobbying and commercial interests had shaped U.S. decisions regarding China. There is no public evidence that Huang acted on behalf of Chinese intelligence or intentionally violated export laws, and allegations surrounding Alice Huang remain under investigation, but the combination of personal and business scrutiny has created a serious corporate-governance crisis for Nvidia.
Treasury Secretary Scott Bessent has publicly demanded answers from Nvidia regarding whether the company knowingly allowed advanced AI processors to reach China through shell companies or intermediaries. His intervention demonstrates how what began as a corporate scandal has expanded into a national-security investigation involving the architecture of U.S. technology controls. Bessent has said the question of whether Nvidia maintained an adequate compliance culture is more important than Huang’s personal conduct, focusing attention on whether senior executives knew how overseas customers were using the chips they purchased.
The distinction between knowing and unknowingly serving Chinese customers will be crucial. Nvidia sells hardware through complex networks of manufacturers, distributors and cloud operators. A company cannot necessarily control every eventual use of a product after a lawful sale, but export-control obligations can require companies to investigate suspicious customers, end users and transaction structures. Regulators will therefore want to know whether Nvidia identified warning signs, whether employees raised concerns internally and whether the company took reasonable measures to prevent prohibited diversion.
The company has consistently said it works to prevent smuggling and complies with U.S. regulations. Nvidia has also argued that the vast majority of its international sales are legitimate and that overseas data-centre development should not automatically be treated as an attempt to evade China restrictions. That position is important because Southeast Asia is experiencing genuine AI demand from governments, businesses and cloud companies unrelated to China, and indiscriminate restrictions could harm legitimate economic development in those markets.
Megaspeed has likewise denied knowingly violating export restrictions and said it is examining its operations amid growing scrutiny. Singapore and U.S. authorities have agreed to continue discussing the company’s role, while Malaysia has also examined reports that data-centre infrastructure could be facilitating re-export or remote access. Those investigations have not yet produced a final public finding that Megaspeed illegally transferred Nvidia chips into China.
The unresolved evidence is why the “20 per cent” claim must be used carefully. It is tempting to present the number as a confirmed measurement of Nvidia’s Chinese revenue, but that would go beyond what is publicly established. Nvidia’s official financial reporting confirms approximately 7 per cent of first-half fiscal 2027 revenue came from customers headquartered in China and Hong Kong. Culper’s estimate of more than 20 per cent refers specifically to what it believes was China-driven compute revenue in fiscal 2026, a different denominator and a broader concept of end-user demand. Comparing those percentages directly without explaining the methodology can create the misleading impression that Nvidia secretly admitted one number while reporting another.
The underlying question is nevertheless legitimate. If Chinese companies financed or consumed billions of dollars of Nvidia-powered computing through overseas entities, then Nvidia’s economic dependence on China could be considerably greater than its headline geographic revenue table suggests. That would matter for investors because future export restrictions could remove not only direct Chinese sales but also significant demand currently recorded in Singapore, Taiwan, Malaysia or other jurisdictions.
It would also matter for Nvidia’s valuation. The company has become one of the world’s most valuable corporations because investors expect AI infrastructure spending to grow for years. Any evidence that a meaningful percentage of that demand depends on a market U.S. regulators are actively trying to restrict could create significant earnings uncertainty. Conversely, if Culper’s estimate proves exaggerated and most Asian demand is genuinely independent of China, the regulatory risk could be smaller than current headlines suggest.
This tension is particularly important as Nvidia prepares to report financial results under new leadership. Investors will scrutinise geographic disclosures, customer concentration and forward guidance for any indication that tighter export controls are beginning to reduce orders from Asia. The company’s August quarter already demonstrated its extraordinary scale: $96.2 billion in quarterly revenue, including $89 billion from Data Center, with total revenue more than doubling year over year. Even a multibillion-dollar reduction in China-linked demand could therefore be absorbed in absolute terms, but it could still affect the growth rates that underpin Nvidia’s market valuation.
The China issue also has implications for the global AI race. The United States is attempting to preserve a technological advantage by limiting China’s access to the most advanced GPUs, while China is investing heavily in domestic alternatives. Huawei and other Chinese companies have made progress developing AI accelerators, but Nvidia remains widely regarded as the industry leader because of both its hardware performance and its CUDA software ecosystem. If China gains continued access to Nvidia computing through overseas cloud infrastructure, U.S. policymakers may conclude that existing export controls are ineffective. If Washington closes those pathways completely, Chinese companies may accelerate investment in domestic alternatives.
That strategic trade-off was at the centre of Huang’s longstanding argument for maintaining commercial engagement. He repeatedly warned that excluding Chinese developers from Nvidia’s ecosystem could encourage the creation of a parallel AI stack based on Chinese hardware and software. Supporters of stricter controls respond that preserving Nvidia’s market share cannot take priority over preventing a strategic rival from accessing computing power that can be used for military applications, surveillance, cyber operations or advanced scientific research.
The controversy therefore extends far beyond one company’s revenue mix. It touches the fundamental question of whether a global technology supply chain can be divided along geopolitical lines when computing power can be rented remotely across borders. Traditional export controls were built around physical objects moving from one country to another. Artificial-intelligence infrastructure increasingly operates as a service, meaning the economic benefit of a chip can cross borders without the chip itself moving.
That reality explains why Southeast Asian data centres have become so strategically important. Malaysia, Singapore and other countries are building enormous facilities because AI workloads require extraordinary amounts of power, cooling and specialised hardware. These projects attract billions of dollars in legitimate investment, but their geographic proximity to China and connections to Chinese technology companies have made them a focus of U.S. enforcement.
Any attempt to determine Nvidia’s true China exposure must therefore distinguish at least three categories: direct sales to customers headquartered in China, indirect sales through intermediaries whose products ultimately enter China, and overseas computing capacity rented by Chinese companies while the hardware remains outside the country. Only the first category appears clearly in Nvidia’s geographic revenue disclosure. The second may involve export-control concerns depending on the facts, while the third may be legal under some rules but increasingly restricted under newer policies.
The $12.43 billion figure reported by Nvidia belongs to the first category. Culper’s estimate attempts to capture all three. That is why the numbers are so different.
The allegations surrounding Megaspeed could eventually provide regulators with more concrete evidence. If investigators obtain shipping records, customer contracts, financing documents and cloud-usage agreements, they may be able to determine how much Nvidia hardware was ultimately connected to Chinese customers and whether any transactions were structured to bypass restrictions. Until that evidence is public, estimates of the total China-driven revenue remain uncertain.
The role of Alibaba will also be closely watched because of the company’s scale and importance within China’s cloud-computing industry. If evidence shows that Alibaba or other Chinese technology groups financed overseas data-centre purchases specifically to secure access to restricted Nvidia hardware, regulators could view the arrangements as an attempt to circumvent U.S. policy even if the chips never physically entered China. If instead the companies simply rented lawful cloud capacity in foreign jurisdictions, the legal analysis could be different.
For investors, the uncertainty itself is a risk. Nvidia’s share price has already been volatile since Huang’s resignation and the broader scandal erupted, and the company has lost hundreds of billions of dollars in market value during the period of intense scrutiny. Some technology leaders and investors have called for Huang to return, while Nvidia’s board has said it is not considering reinstating him and is proceeding with a search for a permanent chief executive.
President Trump has said he would not object to Huang returning as chief executive, demonstrating the unusual political significance of Nvidia’s leadership. At the same time, administration officials including Bessent have indicated that the company may need an outside leader and stronger governance following the scandal. Those conflicting views reflect Nvidia’s extraordinary importance to both U.S. markets and national technology policy.
The board now faces several challenges simultaneously: finding a permanent chief executive, maintaining customer confidence, reassuring investors about corporate governance and responding to investigations into China-related sales. Interim CEO Colette Kress is a veteran Nvidia executive who has overseen the company’s finances for more than a decade, but the transition comes at a moment when Nvidia is under more external scrutiny than at any point during the AI boom.
The company’s financial strength gives it substantial room to navigate that turmoil. Its revenue and profits remain extraordinary, and demand for AI accelerators from U.S. hyperscalers, cloud providers and frontier-model developers continues to expand. The question is not whether Nvidia has a viable business without China; it clearly does. The more difficult question is how much future growth has quietly depended on Chinese demand that could disappear if Washington closes overseas routes.
That is why the current investigation matters even if the 20 per cent figure ultimately proves too high. A smaller China dependence could still represent tens of billions of dollars because Nvidia has become such a large company. Even a 10 per cent exposure at current revenue levels would be economically significant and could affect supply-chain planning, earnings expectations and shareholder returns.
The issue also illustrates why revenue geography can be misleading in a highly globalised semiconductor industry. A chip may be designed in California, fabricated in Taiwan, packaged elsewhere in Asia, sold to a Singapore company, installed in a Malaysian data centre and used remotely by a customer in Shanghai. Traditional accounting may record that transaction as Singapore or Taiwan revenue, while regulators and investors may reasonably view the ultimate economic demand as Chinese.
Nvidia is not alone in facing that challenge. Semiconductor companies, cloud providers and equipment manufacturers increasingly operate through supply chains that make it difficult to associate a dollar of revenue with a single country. Export-control enforcement therefore increasingly depends on end-use verification rather than simple destination addresses.
The coming months could produce a much clearer picture. Congressional committees are examining Nvidia’s China dealings, U.S. agencies are discussing stricter rules, Singapore and Malaysia are reviewing relevant companies and transactions, and Nvidia is under pressure to provide additional information about its compliance processes. Those investigations could either substantiate the claim that Chinese demand was far larger than disclosed geographically or demonstrate that many of the headline estimates overstate the true exposure.
Until that evidence emerges, the strongest verified conclusion is narrower than the most dramatic social-media claims. Nvidia officially generated $12.43 billion from customers headquartered in China and Hong Kong during the first six months of fiscal 2027, about 7 per cent of global revenue. Independent reporting indicates that additional billions may have been connected to Chinese companies through overseas data centres and intermediaries, but the widely circulated claim that more than 20 per cent of Nvidia’s revenue was ultimately generated by China is based principally on Culper Research’s analysis and should not be presented as an established company figure.
The broader issue, however, is becoming increasingly difficult for Nvidia to avoid. The company spent years building the world’s dominant AI-computing ecosystem, and Chinese demand was an important part of that global expansion. Washington is now attempting to divide that ecosystem along national-security lines, while the technology itself increasingly operates through international cloud infrastructure that makes geographic separation difficult.
Jensen Huang’s departure has removed the executive most closely associated with Nvidia’s strategy of preserving access to China, but it has not resolved the commercial and strategic questions he leaves behind. Nvidia’s new leadership will have to decide how aggressively to pursue international markets, how much compliance oversight to impose on overseas customers and how to reassure Washington that its extraordinary global reach does not undermine U.S. export policy.
For policymakers, the controversy may force an even more fundamental rethink of chip controls. If advanced AI capability can simply be rented from a data centre in another country, blocking physical exports into China may no longer be enough. The result could be a new generation of rules governing not only where chips are shipped but who is allowed to use the computing power they generate.
For investors, that transition could determine whether China remains a major hidden source of Nvidia demand or becomes increasingly inaccessible. The answer will affect not only Nvidia’s revenue but the broader economics of the global AI boom.
The current evidence therefore points to a company whose officially reported China sales are already worth billions, whose indirect exposure may be substantially larger and whose future access to that demand is now becoming one of the most important regulatory and financial questions facing the semiconductor industry.


