How the Nvidia saga hurts China
- November 16, 2026
- George Magnus
- Themes: China, Geopolitics, Technology
Nvidia sat on a huge faultline in Sino-US relations. The fallout is likely to be a catalyst for tighter restrictions on China.
At this week’s APEC conference in Shenzhen, Xi Jinping, Donald Trump and Vladimir Putin are scheduled to meet in an unprecedented trilateral summit, alongside discussions over Xi and Trump’s trade truce, rare earth access conditions, and various connectivity and data/AI matters. There will, though, be an Nvidia-sized elephant in the room.
This is not about the personal circumstances of the company’s founder and former CEO, Jensen Huang. Rather, this is about revelations illuminating the significance and sensitivity of trade in advanced semiconductors and AI-related products, national security and the repercussions for the evolving battle between China and the United States for global economic supremacy. In the game of AI leapfrog and catch-up, mistakenly called a ‘race’, these issues are of prime importance and will have profound implications for both countries.
Even though there is no question about Nvidia’s status as a linchpin of the American AI sector, it also came to occupy an awkward position. It sat on a huge faultline in Sino-US relations, often advocated commercial positions that might not have been in tune with national security interests, sowed confusion in Washington which is now coming to the fore, and will almost certainly be the catalyst for tighter export controls and restrictions on China.
Despite the apparent spontaneity of the company’s problems in the last few weeks, trouble has been brewing for a while. The Biden Administration barred sales of leading-edge chips to China in 2022, and blocked access to tools and equipment needed to design and make them in China. These controls were later augmented to plug technical gaps and to combat smuggling. Under Trump, some export controls were tightened but several were relaxed. In December 2025, Trump announced that the US would permit Nvidia’s H200 processors to be exported to approved customers in China, subject to conditions subsequently formalised by the Department of Commerce in January 2026. The Trump administration’s goal has been to keep Chinese firms hooked on key, but not the best, US chips, hoping naively that China’s attempts to develop its own high-end versions would be stymied.
China, however, has for a couple of decades had the intention, if not the capacity, to go it alone, and had its own reasons to resist Nvidia’s chips. It is prepared to make do with less than the best of chips in the short term, if, in the longer term, it could replicate and surpass the US. The Communist Party believes it can. Chinese industrial policy, especially as regards AI, is about self-reliance, cutting dependence on what it regards as an unreliable US, and creating its own Nvidia. Anyone in doubt should consult the new five-year plan, whose focus is on building out the diffusion of AI for which, it claims, the key success factors are homegrown: electricity supply, low pricing and government direction and funding.
Naturally, the US wants to effect different outcomes, and export controls on high-end products are a significant component. Duncan Weldon addressed this in these pages recently, referencing the unintended consequences of technology blockades historically, and suggested that such US export controls were likely to spur innovation in the medium term, and create a stronger, rather than a weaker, Chinese AI and chipmaking sector. I don’t think the cause-effect sequencing works in this case. US export controls simply underline decisions already taken.
Going back to the 1990s, China has always wanted to be a leader in chips and chip manufacturing. It has already sunk over $150 billion into this goal but only to play second fiddle. It is a feisty competitor and even a pace-setter in AI models, but not in ‘compute’, that is, advanced chips and manufacturing. Technological capacity does depend on funding, of course, but also on institutions and genuinely open learning and application, untainted by political dogma and censorship.
There is also something else you won’t hear China acknowledge. China’s AI firms have been working hard to access Nvidia’s chips. They have been using remote access to offshore data centres situated in, for example, Singapore and Malaysia, which can obtain chips and lease access to them to Chinese firms such as Alibaba and ByteDance, so the chips never physically enter China. There have also been allegations of US chips being smuggled or otherwise obtained illicitly by Chinese firms and subsidiaries in third countries, acquired under false pretences, or sourced through countries where different export-control restrictions apply, before being shipped to Shenzhen via Hong Kong.
Cue a data centre and cloud company called Megaspeed, incorporated in Singapore, which has been scrutinised over allegations concerning its role as an intermediary for Chinese firms. Its former senior executive, Alice Huang, has faced allegations from US officials, reported by the New York Times, concerning possible links to Chinese intelligence. These allegations have not been established. Megaspeed denies illegally shipping chips to China, and has not been charged. Nvidia says it blocked further sales to the company in June 2025.
Megaspeed has been the subject of considerable interest. In February 2025, the Singapore Police Force and Singapore Customs began an independent investigation into several companies and their key officers, including Megaspeed, over suspected false declarations relating to the purchase and movement of servers and other goods. American officials have reportedly been examining the company since 2024, although Singapore’s Ministry of Home Affairs said Singapore first became aware of US interest when the US Bureau of Industry and Security conducted a post-shipment verification at Megaspeed on 24 June 2025. In a statement on 12 November 2026, the Ministry said the investigation into Megaspeed and six individuals remained ongoing. It said that, in February 2026, Alice Huang had received a stern warning after investigators assessed that she had knowingly provided inaccurate information to downplay her role, and that she had been barred from re-entering Singapore. The Ministry stressed that she had not been investigated for espionage. It also said there was insufficient evidence to establish that she had committed offences relating to suspected inaccurate declarations.
In May 2026, New York-based short-seller Culper Research, which investigates companies it believes misrepresent their operations, reported, citing trade records, that Megaspeed’s Malaysian subsidiary, Speedmatrix, had imported goods worth approximately $4.6 billion between December 2024 and January 2026. Culper’s research note stated: ‘We are short Nvidia. We recognize the stakes. Nvidia holds the single largest market capitalization on the planet, while CEO Jensen Huang has been celebrated as a generationally talented operator. We share the consensus view that AI – for better or worse – will continue to transform society. This is not a bet against AI. We are short Nvidia for one reason: the Company has a significant China problem.’
In a nutshell, the China problem, as Culper explained it, was twofold. Part of the problem, Culper alleged, was that Nvidia was still making significant money in China when the firm claimed it was faring poorly. Culper cited data showing Nvidia’s revenue attributed to Singapore rising sharply after late 2023, doubling as a share of total revenue and mirroring the fall in the share attributed directly to China. Nvidia says Singapore revenue reflects where customers are billed and that products booked there are almost always shipped elsewhere. The other part of the problem is that, with investigations underway and pressure rising on intermediary firms such as Megaspeed, Nvidia’s China revenue has come under renewed scrutiny.
The incoming Democratic Congress is likely to examine how chips get to China, why Megaspeed was not placed on the US Entity List, despite reports that several government agencies favoured the move, and what senior Commerce Department officials knew, and when. Congressional scrutiny has also focused on Nvidia and its dealings with overseas intermediaries. Meanwhile, Chris Coons, who is expected to chair the Senate Foreign Relations Committee when the new Congress takes office, has said he intends to examine the relationship between the Trump Administration, alleged Chinese intelligence activity and AI industry figures.
A raft of bills is possible in the New Year, affecting foreign sales of sensitive chips overseas and Congress’s ability to retroactively block previously approved sales. The Remote Access Security Act, for example, passed the House in January 2026 and was referred to the Senate Committee on Banking, Housing, and Urban Affairs. It would give the US government greater authority to control remote access to items subject to export controls, including through cloud services, addressing one aspect of the Megaspeed problem.
The overall impact of the Nvidia fallout for China is hard to unravel at this point but may have a bearing on policy differences in Beijing. For example, the Ministry of Commerce is thought to be broadly in favour of keeping the importation of foreign chips going, including those of Nvidia, because local champions such as Huawei are not yet at the stage where they can fill the self-reliance gap. Yet the Cyberspace Administration of China and Ministry of State Security, now perhaps with the wind in their sails, are thought to view China’s dependence on Nvidia as a strategic vulnerability, and want to press the pedal to the metal on self-reliance, regardless of the interim consequences, which could entail years of higher costs and slower model training, assuming that Chinese chips and compute can in fact catch up.
Now, though, China faces an even more hostile regulatory turn of the screw. It has, for now, lost its best-placed foreign advocate in the US, and one of the most influential protagonists of billions of dollars of investment in open-source AI. Meanwhile, Congress has its antennae more firmly focused on compliance and control problems that are going to become all too transparent and clear. Democrats may want a bigger say in the drafting of export-control regulations, and demand the power to review export licences, to track chips, ask allies to match US restrictions, and regulate cloud access. It is hard to see how, if they behave in this way, China could see the US export controls and Chinese AI development and rollout as separate issues – which of course they are not, and which Congress understands all too well.
In addition to the regulatory backlash and restraints on the physical access to advanced chips and chipmaking technologies, it is highly likely that China will get caught up in any slowdown in AI capital spending for which the Nvidia saga may be the catalyst, and associated reductions in equity market valuations. The consequences may not be trivial but equally they are unlikely alone to derail China’s economy, which is big enough to withstand the expected effects.
While regulatory restrictions are liable to affect domestic spending on AI and the rollout of AI models and investment, the main transmission from a slowdown in capital spending in the US and elsewhere will come via the economic hit to Chinese exports, about a fifth of which are AI-related – for example, computing and power equipment. While there are always definitional issues about what constitutes AI-related equipment, there is little doubt that the lion’s share of the growth in Chinese exports has come from AI equipment, and that the overall trade surplus is set to beat last year’s $1.2 trillion. According to Nomura, net exports are set to generate about a quarter of this year’s economic growth.
In addition to weaker exports, other conduits of weakness into China’s economy would include the reduction in profits and investment as deflationary headwinds pick up again in the broader economy outside of energy, and the wealth and investment financing effects of the sell-off in Chinese equities. Relative to the US and the rest of the world, where stock prices are still showing decent gains for the year, Chinese markets locally and as represented in international indices are down between eight and 15 per cent year-to-date. From their highs earlier this year, some Chinese markets are showing a decline of more than 20 per cent. This is not inconsequential for Chinese households that have seen property prices slide for the last four years, and now stock prices. Nor is it for Chinese AI firms that raise money privately, most of it from the equity markets.
All in all, the economic impact of the Nvidia saga for China is troubling but by no means earth-shattering. Yet that is also to ignore two very important and hard-to-quantify issues. One is the consequences of a possible renewed export-control-related deterioration in Sino-US relations, not least for consumer and business confidence. The other is how the repercussions play into a Chinese economy that already features profound fiscal and financial flaws and has pretty much run out of adaptive capacity under its current political leadership.