DeepSeek defies America’s AI supremacy

AMERICA’S superinflated tech stocks had seemed due for a correction for months, but the trigger has come from an unexpected source. The latest large language model (LLM) from China’s artificial intelligence (AI) startup DeepSeek may not be quite a “Sputnik moment”. There are parallels, even so, with the early space race, when ingenuity helped Soviet engineers keep pace with and sometimes surpass US rivals, despite their relative lack of computing power and cutting-edge technologies. The US went on to win the Moon race, and establish a sustainable edge in space. But DeepSeek’s breakthrough upturns the assumptions that have underpinned US tech valuations, of an unassailable supremacy in AI that would be extended by spending billions of dollars on chips and infrastructure.

DeepSeek’s achievement is to have developed an LLM that AI experts say achieves a performance similar to US rivals OpenAI and Meta, but claims to use far fewer – and less advanced — Nvidia chips, and to have been trained for a fraction of the cost. Some of its assertions are yet to be verified. If they are true, however, DeepSeek represents a potentially formidable competitor.

First, like Meta of the US but unlike OpenAI or Google’s Gemini, it is open-source – ready to share the recipe for its secret sauce rather than keep it locked away in hope of extracting maximum financial gain. That makes it appealing for developers to use and build on. Second, it can be put together on much more of a shoestring budget and with much less computing power. This explains Monday’s plunging stock prices not just of frontline tech companies, but of those that make chip equipment and supply electrical hardware for data centres.

The big US companies are focused on a particular model of scaling AI by throwing vast amounts of capital, data and computing power at the problem. First Mistral – a French AI startup – and now DeepSeek have demonstrated that companies with fewer resources may end up with smarter and more efficient models. The increasing efficiency of these models is also likely to spur increased demand for their use.

DeepSeek’s advance highlights, too, that China is managing to make technological leaps in AI despite export controls introduced by the Biden administration intended to deprive it of both the most powerful chips and the advanced tools needed to make them. Chinese AI startups have been compelled to find inventive ways of extracting the most juice from the chips they do have. Far from stifling Chinese innovation, Washington may have stimulated it. And the success of often domestically trained Chinese engineers in increasing efficiency and finding workarounds raises questions over whether the technological “moat” established by high-spending US groups such as Meta, Google, OpenAI and Anthropic is as wide and impregnable as they had thought.

Many advanced democracies will be wary of a Chinese government – seen in many ways as inimical to Western interests – potentially acquiring leadership in the most transformative technology of our era. Yet some political leaders elsewhere, along with many consumers and developers, may welcome a market that is less dominated by a handful of American companies.

The open question now is not necessarily who will develop the best AI models, but who can apply them best to real-world tasks. Kai-Fu Lee, a Chinese AI pioneer, has long argued that China excels on the application front even if it may lag behind in infrastructure. That was before the Chinese startup world was squeezed by the political clampdown on tech entrepreneurs and the surge of investment in US AI startups. But after DeepSeek’s achievement, it looks a much more even game.