Singapore illustrates the resulting uncertainty. Its share of NVIDIA's reported revenue geography had increased rapidly before disappearing from the breakdown. Meanwhile, Chinese AI companies have been seeking access to NVIDIA computing capacity outside mainland China as restrictions on direct imports have tightened.[2, 4] The difference between where a customer is headquartered and where its computing equipment is ultimately used has become increasingly consequential.
Washington's latest restrictions target precisely that gap. On November 16, the administration moved to tighten controls on advanced processor exports to Indonesia, Malaysia, Singapore, Thailand and Brazil, requiring licenses for shipments to those destinations. It also moved against overseas data centers intending to provide Chinese companies with remote access to advanced processors and servers.[5, 6] Previously, Chinese companies unable to import the most advanced GPUs directly could still obtain computing capacity through facilities in countries including Singapore, Malaysia and Thailand.[6]
The regulatory perimeter is therefore expanding beyond the physical delivery of chips into China. It now includes overseas facilities and intermediary destinations through which Chinese customers can obtain access to American computing technology. For NVIDIA, that creates potential exposure among customers whose purchases may not appear in the reported China revenue category.[2, 4, 5, 6, 7]
The enforcement effort is moving beyond chip shipments and into the networks that enable remote access
The Megaspeed investigation gives this enforcement expansion a criminal-investigation dimension, although its chronology is important. The Department of Commerce began examining the Singapore-based company in 2024 after identifying suspicious activity involving potentially diverted semiconductors. Investigators requested records from NVIDIA in May 2025 and visited Megaspeed's premises alongside Singaporean law enforcement the following June. The Justice Department subsequently assumed responsibility for the criminal investigation.[8]
The November 17 announcement made an existing investigation public; it did not establish that the recent controversy had initiated it. Nor have authorities identified specific offenses or announced charges against Megaspeed.[8, 9] The investigation concerns alleged diversion involving a company previously led by a woman whom US officials believe had connections to Chinese intelligence and who had a relationship with NVIDIA's former chief executive.[8, 9]
The political consequences have extended beyond the investigation itself. Jensen Huang resigned on November 12 following reports of relationships that included one with a Chinese businesswoman alleged to have worked for Beijing's intelligence services.[10] His departure has not ended congressional scrutiny: Huang and NVIDIA general counsel Tim Teter agreed to transcribed interviews before three House committees, scheduled for December 1 and 2 respectively.[11, 12]
The committees want explanations of how Chinese actors obtained access to advanced NVIDIA processors and what role a potential Chinese intelligence asset played in the controversy.[11, 12, 13] Their involvement brings the company's export-control procedures directly into a national-security debate that already has support across party lines.
Congressional scrutiny is turning NVIDIA's China-sales strategy into a bipartisan policy fight
Republican and Democratic lawmakers have expressed opposition to NVIDIA's preferred policy of retaining substantial chip sales to Chinese customers. At least four legislative proposals targeting Chinese access to advanced processors are receiving congressional attention, with analysts identifying the December defense package as a more promising vehicle for their passage.[4, 14]
NVIDIA has increased its efforts to influence Washington during this period. The company disclosed $4.69 million in federal lobbying expenditure during the first three quarters of 2026, while its lobbying workforce reached 34 people, compared with seven in 2024.[14] Those expenditures have coincided with growing bipartisan resistance to the company's China-sales position. They have not prevented the emergence of multiple legislative proposals, although the available record does not establish that lobbying has changed any particular legislative outcome.
The immediate commercial disruption is more clearly defined. NVIDIA disclosed $2.4 billion in licensed Hopper-generation processor sales to China during the quarter ended October 25. The US administration had approved sales of the H200 to Chinese customers, but NVIDIA subsequently stopped shipping licensed Hopper products to China.[5, 15, 16]
Management also confirmed that these products were excluded from forward revenue guidance.[5, 15] Consequently, the discontinued shipments do not by themselves represent an additional guidance reduction beyond the adjustment NVIDIA had already made. Nor can the decision automatically be attributed to a new government prohibition: NVIDIA possessed permission to make the licensed sales but chose to discontinue them without publicly explaining the decision.[5, 16]
Licensed Hopper sales have stopped, but the wider exposure to Chinese demand remains unresolved
The more consequential commercial question concerns what happens to demand outside this discontinued product category. If Washington's expanded controls restrict third-country shipments and overseas computing access, NVIDIA could face additional limitations on serving Chinese customers through arrangements that do not appear as direct sales into China. Its reported geographic revenue figures do not provide a complete measure of that exposure.[2, 4, 5, 6, 7]
Chinese manufacturers already have a substantial domestic market position from which to compete for any displaced demand. In 2025, domestic GPU and AI-chip companies accounted for approximately 41% of China's AI accelerator server market. NVIDIA retained 55%, while AMD held 4%. Total shipments reached approximately four million units, including 1.65 million supplied by Chinese vendors.[17]
That existing supply base makes substitution commercially possible. It does not establish that the latest restrictions have already transferred NVIDIA's customers to domestic competitors. Demand affected by tighter access could instead be deferred or canceled, while overseas arrangements may continue to provide alternative access where restrictions permit. Updated shipment and market-share figures will be needed to establish which outcome is occurring.
December provides the first important legislative test. Incorporating at least one of the four proposed chip restrictions into the defense package would strengthen the case that political opposition is becoming enforceable policy. Passage without additional China-chip restrictions would weaken that interpretation.[4, 14] Implementation decisions by the Bureau of Industry and Security will subsequently establish how licensing requirements affect shipments involving the five designated countries and whether overseas data-center arrangements can continue serving Chinese customers.[6]
The decisive test will be whether tighter controls divert Chinese demand to domestic suppliers or suppress it altogether
NVIDIA's subsequent financial disclosures and Chinese accelerator shipment figures will provide the commercial test. They will show whether tighter restrictions reduce the company's China-linked sales, whether domestic manufacturers capture additional demand, and whether previously accessible overseas computing arrangements are being curtailed.
The exposure already exists beyond the number most commonly used to describe NVIDIA's China business. Its $19.68 billion reported China-region revenue sits alongside potentially substantial Chinese end-user demand attributed to other headquarters and overseas computing arrangements. Washington is now targeting the intermediary channels that separate those customers from the company's reported China category. The immediate financial loss remains unquantified, but NVIDIA's geographic reporting can no longer be treated as a reliable boundary for the commercial exposure created by US–China technology restrictions.[1, 2, 3, 4, 5, 6, 7]