The China-based DeepSeek AI model has shaken up the tech world by delivering high performance at a fraction of the cost of models from US giants, sparking interest in investment opportunities for smaller European players. Experts suggest that DeepSeek’s breakthrough signals the potential for Europe to compete without relying on massive funding.
DeepSeek’s large language model (LLM) has “made a mockery” of the belief that $1 trillion is necessary to develop the next level of artificial general intelligence (AGI), according to Neil Lawrence, a machine learning professor at the University of Cambridge in the UK. His comments came in response to the US’s announcement last week of the $500 billion “Stargate” program to build AI infrastructure, led by OpenAI, the creator of ChatGPT.
While much of the funding for Stargate will be used for data centres filled with the latest AI chips, highlighting the overwhelming resources needed to stay at the cutting edge, DeepSeek’s claim to have built a model with similar capabilities for just $5.6 million has upended these assumptions.
The technology promises “models that are more efficient and less hungry for GPUs, energy, and cash,” said Laurent Daudet, CEO of the French AI firm LightOn. He pointed out that Europe’s ability to innovate without needing a $500 billion budget could have a significant impact.
This disruption has affected the stock prices of AI-related companies, including Nvidia. “It shows that competition is fierce and there will be a price war,” said Nicolas Gaudemet, AI chief at consulting firm Onepoint. “An additional provider will bring prices down.”
– Open Source Innovation –
Cambridge’s Lawrence expressed disappointment that DeepSeek didn’t emerge from Europe, given its vast AI talent pool. However, he noted that the company’s innovative approach offers a glimpse of what’s to come. "It is very encouraging for Europe," he said, adding that there will likely be more companies like DeepSeek.
Lawrence praised DeepSeek’s open-source methodology, where the project’s core components are shared with the wider AI community, allowing others to build on them. This is especially beneficial for European companies like France’s Mistral, according to Gaudemet. “They can reuse DeepSeek’s models to train their own and stay competitive,” he said, although he warned that the competition is not just between the US and Europe, as China’s progress is also noteworthy.
Jonas Andrulis, CEO of German AI firm Aleph Alpha, noted that his company has already integrated DeepSeek’s capabilities into their offerings. However, he emphasized that the future of AI models lies in meaningful innovation beyond just copying what others have done.
– Frugal Models and Local Adaptation –
The rise of more affordable AI could lead to models tailored for local markets and individual business needs, rather than relying on massive, centralized systems. “There’s a future for more frugal models that can perform just as well, especially for business needs,” Daudet said. He described his company’s role as providing the framework to make AI models usable and secure for businesses.
Gaudemet noted that European firms have a unique advantage when it comes to security and data privacy. “Europe’s companies can position themselves as offering a collaborative, fair alternative to the US and China,” he said. Andrulis agreed, pointing out that many countries feel uneasy about becoming overly dependent on either of those two power blocs.
Cambridge’s Lawrence echoed this sentiment, emphasizing that Europe has “some of the strongest researchers in the world” and that local talent could drive innovation without massive investments from outside forces. He urged that Europe should focus on its strengths, without being overly swayed by the latest developments from US firms like OpenAI.
– ASML’s Positive Outlook –
The head of Dutch chipmaker ASML, which makes the equipment powering the chip industry, welcomed DeepSeek’s arrival, as it addresses two major challenges facing the AI sector: high costs and excessive energy consumption.
ASML CEO Christophe Fouquet stated that the cheaper AI solutions created by companies like DeepSeek could expand AI’s reach into more applications, increasing demand for chips. “Lower cost means more chips,” he explained. Fouquet also predicted that more companies like DeepSeek will emerge in the coming months, further intensifying competition, particularly in software.
Despite a dip in after-tax profits, ASML reported a strong 2024, with sales totalling 28.3 billion euros ($29.5 billion), and a positive outlook for the future. Fouquet said the AI boom is only beginning and presents a significant opportunity for the company.
ASML has left its sales forecast for 2025 unchanged, projecting sales between 30 and 35 billion euros. Long-term, the company expects sales to reach between 44 and 60 billion euros by 2030, with AI playing a pivotal role in that growth.
blogpay
Total Comments: 0