
The News
US President Donald Trump called it a “wake up call,” while tech investor Marc Andreessen said it was a “Sputnik moment,” but in Europe, the sudden emergence of Chinese AI startup DeepSeek’s R1 chatbot was greeted rather differently.
R1 has ripped what was once considered a truism of the industry away — that lots of money and the most-advanced chips are needed to make an AI that can compete with industry leaders like OpenAI’s ChatGPT and Anthropic’s Claude. In turn, global markets have been spooked, as investors reconsider whether the US leads on AI, and how much it should cost.
Across the Atlantic, meanwhile, the reaction has been more optimistic, with tech leaders eschewing US alarmism in favor of a more nuanced view of what the model means for innovation on the continent.
In this article:

The View From Europe
The fact that DeepSeek’s low-cost, high-performing chatbot is challenging US AI models that cost billions of dollars to build offers hope to resource-strapped European tech firms, industry experts told Semafor — meaning that Europe can still hope to be a key player in AI innovation.
“We’re so fixated on the geopolitical story,” Andreas Cleve, CEO and co-founder of Denmark-based AI healthcare company Corti, referring to US-China tech rivalry.
Europe’s role in developing specialized AI often gets overlooked, he added, stressing that European AI firms “should be excited about the fact that [DeepSeek’s success] shows we’re not in the final inning at all.”
European startups “have historically excelled at building focused, efficient solutions rather than chasing scale at all costs,” said Muj Choudhery, co-founder of UK company RocketPhone.ai. DeepSeek “suggests there’s room for strategic players who can execute well without massive capital outlays.”
The UK — the third-largest AI market globally after the US and China, according to the US International Trade Administration — earlier this month published an AI Opportunities Action Plan that emphasized the need for more computing power to stay competitive, including plans to commission a new supercomputer.
But DeepSeek demonstrates “that AI leadership isn’t just about computational scale — efficiency and innovation are powerful advantages,” Zoi Roupakia, Policy and AI Research Lead for the Institute for Manufacturing at the University of Cambridge, told Semafor.
In countries like the UK, which has a vibrant research and startup scene but lacks the deep pockets of US tech giants, there is a need to balance infrastructure investment with innovation to succeed, she said.

Room for Disagreement
Other experts were more skeptical of the theory that highly-capable AI models need less compute: Mahdi Yahya, the founder of UK-based AI cloud infrastructure firm Ori, told Semafor that takeaway was “completely overblown.”
DeepSeek’s model relies on a training technique known as “model-driven reinforcement learning” — where the AI generates tasks for itself to learn from — that also uses intensive amounts of computing power, he said.
The “one major benefit,” Yahya said, is that R1’s method of refining itself after its initial training is far less dependent on user-generated data than other leading-edge models: “This could blow open the data bottleneck that is currently holding the industry back.”
Still, European Union regulation over AI could “throw a spanner in the works” for European firms that want to test or even build on DeepSeek’s open-source model, said Alexandra Ebert, Chief AI and Data Democratization Officer at Austrian startup MOSTLY AI.
Open-source models theoretically allow any developer anywhere to build on another’s work, but the lack of transparency over the data used to train DeepSeek could contravene the EU’s “hazy” regulations, she said.

Notable
- The EU has long been focused on upgrading and scaling its AI infrastructure, but a digital economy expert argued in an October essay for Bruegel that taxpayer money might be better spent on using existing open-source models to develop “specialised AI applications at a much lower cost.”
- DeepSeek’s success doesn’t mean that US export controls on chips have failed, because the company wasn’t significantly more resource-constrained than US AI firms, Anthropic’s CEO argued: Its engineers are simply “very talented,” and “show why China is a serious competitor to the US.”
Paige Bruton contributed reporting.