The irruption of DeepSeek, a Chinese model of frugal, open-source artificial intelligence (AI), heralds a new stage in the spread of these technologies, an expansion of the market favourable to new players and a diversification of applications.
DeepSeek’s irruption on the AI scene
On 20 January 2025, DeepSeek’s AI model R1 went live and, in less than a week, it surpassed ChatGPT (OpenAI’s previously leading model) to become the momentarily most downloaded app on Apple’s App Store in the US. The app was developed by DeepSeek, a company founded by Liang Wenfeng, an engineer who previously co-founded High-Flyer, an AI-powered quantitative trading hedge fund. The model demonstrates advanced reasoning capabilities and performs comparably to OpenAI’s models from a few months ago. But it differs in that its development costs are much lower, with its creator reporting a budget of around six million US dollars, compared with the costs quoted for ChatGPT-4.0, which range from 100 million to one billion US dollars. What’s more, it is open source, which will make it easier for users to develop. Currently, the online version is free, and prices for professional use are significantly lower than those of the American competition.
Politically, the news was a bombshell, especially as it came shortly after President Trump announced the Stargate project, which would provide US$500 billion to ensure US pre-eminence in AI. On the subject of technology policy, Marc Andreessen,[1] a Silicon Valley specialist and vet
