USA remains firm: No backing down on AI restrictions against China – NVIDIA remains on the outside

Despite direct pleas from NVIDIA CEO Jensen Huang, the US government under President Donald Trump remains unimpressed. The export restrictions on high-performance AI chips to China remain in place – even if this means severe cuts for US manufacturers. NVIDIA in particular, the company with the highest market capitalization in the AI sector, is therefore facing a difficult strategic adjustment of its business in China.

Political blockade instead of industry cooperation

Since the Biden era, the US government has been pursuing the goal of restricting China’s access to Western AI hardware with its so-called “AI diffusion” policy. Trump is now tightening this policy even further, which not only blocks the export of individual chips such as the H20 AI Accelerator, but also prohibits the use of Chinese hardware such as that from Huawei in third countries. Companies that do not comply could be in breach of US sanctions laws. Any room for negotiation? Not a chance. Even a hypothetical investment pledge of over 500 billion US dollars by NVIDIA could not change the administration’s mind. The background: there are fears that high-performance graphics chips – once in Chinese data centers – can no longer be controlled and could contribute to military use in the long term.

Huang warns against self-sabotage

Jensen Huang had publicly emphasized several times that the restrictive course would weaken rather than protect the USA. His argument: China would make itself independent in the long term with its own semiconductor solutions such as those from Huawei or the Chinese research consortium DeepSeek. He points to the enormous talent pool in China – according to Huang, over half of all globally active AI engineers are said to come from the Middle Kingdom. He sees the “AI diffusion” strategy as a missed opportunity to remain the dominant player in the global AI race with American technology and software. Instead, China could be forced to decouple itself from Western technology and establish an alternative tech stack on its own – a scenario that Huang openly describes as a “nail in the coffin” for American influence in AI.

US government maintains its restrictive course

However, the political message from Washington is clear. The Senior Policy Advisor for AI, Sriram Krishnan, made it clear that there is bipartisan agreement on limiting Chinese access to US chips. The concern: once in China, even the most modern export and licensing requirements would be almost impossible to enforce. Krishnan emphasizes that the government’s focus is on the global US AI infrastructure – from hardware to models to software. This “AI stack sovereignty” is non-negotiable.

NVIDIA must react – Blackwell light for China?

In view of the political blockade, NVIDIA is expected to prepare alternative product lines for the Chinese market. According to internal sources, a slimmed-down version of the new Blackwell architecture, specifically for regulatory non-critical scenarios, is to be presented as early as July. Technical details are scarce so far, but it is likely to be a variant with reduced interconnects, lower bandwidth and limited computing power – compliant with the new export regulations. Whether such products are economically worthwhile is questionable. China has recently been one of the most important sales markets for data-centric GPUs – especially in the training and inference area of large language models.

Strategic division instead of a global market

The USA is pursuing the goal of slowing down China’s AI ambitions through technical isolation with a firm hand. For NVIDIA and other US manufacturers, this means that they will lose access to one of the world’s most dynamic technology markets in the medium term – while China will be forced to create its own alternatives. The result: a bifurcation of the global AI market – with different hardware platforms, software stacks and geopolitical loyalties. The USA is focusing on control, China on self-sufficiency – and in the middle of it all, companies like NVIDIA are caught in the middle economically.

Source: Yahoo