February 11, 2025 - 4:00pm

The US and the UK have today refused to sign a declaration on “inclusive and sustainable” AI at a summit in Paris. US Vice President JD Vance stated that America was withholding its signature because of Europe’s “excessive regulation” of the technology. The British delegation refused to specify why it has withdrawn its support, but it is likely tied to a potential deal emerging between the US and the UK to prevent the former imposing tariffs on the latter.

The chaos in Paris reflects deeper disruptions across the AI landscape. A few weeks ago, a Chinese company called DeepSeek released AI technology that massively undercut American equivalents — with some analysts putting the cost of using the company’s model at around 6% the price of its US alternative. DeepSeek claimed that the new platform used significantly less computer power, which led to a sharp downturn in the stock price of Nvidia, the American company that provides the graphics processing unit (GPU) chips needed to power the technology.

Yesterday, Elon Musk made an unsolicited bid to buy OpenAI, one of the leaders in the American AI space and a company which he co-founded almost a decade ago. Musk has seemingly pinned the blame for OpenAI’s failures on current CEO and fellow co-founder Sam Altman, calling him a “swindler” on social media. Altman has led the organisation in a more commercial direction, partnering it with large firms in order to raise capital. Musk believes that the company should be more open-sourced and has suggested partnering it with Tesla. That DeepSeek is an open-source platform no doubt leads the X owner to think that his assessment has been vindicated.

What is Musk’s plan if he were to acquire OpenAI? Despite its undeniable efficiency and low cost, the entrepreneur has raised doubts about DeepSeek’s abilities, questioning whether it was really trained on a relatively small number of GPUs. The argument is that since China is restricted from buying these GPUs, its tech leaders have an incentive to mislead the public about how many they possess. Yet it is not clear whether this is the correct interpretation. Though there are some consultancy firms, such as SemiAnalysis, which make this claim, how they arrive at their estimates is less than obvious. And even if their methodology is robust, their results remain speculative.

The situation in Silicon Valley circles is one of confusion. Perhaps Musk is right, and the Chinese are hiding the real number of GPUs in their possession. But perhaps he is not, and DeepSeek really does represent a quantum leap in the training of AI platforms. If he is correct then the Chinese advance, while impressive, is not industry-destroying. If he is wrong, however, the initial hype around DeepSeek is largely justified and the entire industry may need a rethink.

This brings us back to the summit in Paris. Since the Americans are no longer clear about the realities of the AI sector, they may feel that it is unwise to bind themselves to any firm position. Over the next year or two, the impact of technologies such as DeepSeek will become more apparent — as will the structure of OpenAI and Silicon Valley’s approach to artificial intelligence more generally. At this point, it might be easier for the United States to articulate its vision. Until then, however, it seems that the new Trump administration believes the wisest move is to make no move at all.


Philip Pilkington is a macroeconomist and investment professional, and the author of The Reformation in Economics

philippilk