One consequence of the resignation of Jensen Huang from the post of chief executive of NVIDIA is that it potentially puts at risk the ‘détente’ between United States President Donald Trump and Chinese President Xi Jinpingreached at the Asia-Pacific Economic Cooperation (APEC) summitin Busan in October 2025. The détente saw China lift export controls on rare earths, while the US started to take a less-restrictive approach to export controls on advanced chips than under the Biden administration, when the artificial intelligence race started in earnest.
Huang, who stepped down from NVIDIA amid a USnational-security investigationinto the diversion of its chips to China, was a major voice pushing for the lifting of such controls, based on the premise that leaving the market to China would only hurt US AI companies. Huang dubbed export controls “a failure”, as they would hand the market to China’s Huawei and speed up the creation of a non-US AI stack – the full suite of elements needed for AI systems. In the event, in December 2025, the US licensed NVIDIA H200 chipsfor approved Chinese customers, but kept the company’s most advanced Blackwell chips off limits.
When Trump, accompanied by Huang, met Xi again in Beijing in May 2026, the détente became a“constructive relationship of strategic stability”. It may now collapse as the Trump administration grapples with the consequences of Huang’s push to provide China with advanced AI hardware, both through direct exports and by allowing Chinese AI companies to use data centres with advanced NVIDIA processors elsewhere in Asia andeven in Europe.
Furthermore, the détente delivered less than both sides hoped. Huang himself said that, despite the H200 licences, NVIDIA’s share of China’s market for specialised and faster AI hardware fell from 95% to zero, as Beijing steered its companies towards Huawei by prohibiting the acquisition of NVIDIA chips exceptwhen strictly necessary(for training advanced models but not for inference). This has not stopped China from trailing the US closely in the performance of advanced AI models, thanks to theacquisition of ‘strictly necessary’ NVIDIA hardwareand model training overseas.
Political change in the US is also derailing the détente. With the Democrats set to control Congress from January, both Republican and Democrat leadersnow back the inclusion of chip-export restrictions in the 2027 defence bill. Furthermore, the Democrats, who will chair the trade committees, are already investigating the H200 licence.
Trump’s room to trade chips for concessions has thus narrowed sharply. This is bound to shape the tone of the third Trump-Xi encounter of 2026, in Shenzhen at the 19 November APEC meeting.
Beijing has warned Washingtonagainst using the Huang episode as a pretext for new technology restrictions. Beijing’s instruments of retaliation are well known: rare-earth licensing, which Beijing haspaused but not dismantled, a faster shift to Huawei’s Ascend chips and less patience with a proposed US-China AI safety dialogue, on whichlittle visible progresshas been made.
Europe is a bystander in these developments. Europe’s AI factories run on NVIDIA hardware and software, while its industry relies on Chinese rare earths and mature chips. If China reimposes controls on rare earth exports to the US, the European Union might also be caught, as was the case until the US managed to have controls lifted at the Busan summit. Furthermore, if it is confirmed that Chinese AI companies have used European data centres to train their models, it might become harder for Europe to obtain advanced US chips, and thus to increase compute in the interests of greater competitiveness.
Authors
Alicia García-Herrero
Theme
Global economy and trade
Keyword
artificial intelligencetrade policytechnologyEU-China relationstransatlantic relations
Country
United StatesChina
Language
English
