European Startups Challenge Nvidia's Dominance in AI Chip Market

European chip startups are increasingly positioning themselves as competitors to Nvidia in the artificial intelligence (AI) chip market. Notably, the Dutch company Euclyd, founded by former ASML director Bernardo Kastrup, is seeking to raise at least 100 million euros ($118 million) to scale its technology. Euclyd aims to develop AI chips that promise 100 times higher power efficiency for AI inference compared to Nvidia's latest Vera Rubin chips. Other European startups, including the U.K.'s Optalysys and Fractile, as well as France's Arago, are also pursuing significant funding rounds, with Optalysys planning to raise over $100 million later this year.

CNBC

Euclyd's innovative approach involves a multi-chiplet system designed to process data more efficiently than traditional GPUs, which often struggle with data movement through memory stacks. Kastrup asserts that Euclyd's technology will reduce the energy, cost, and physical footprint of AI data center infrastructure. The startup is currently negotiating with four potential customers, with plans to begin supplying two of them as early as next year. Meanwhile, Olix is developing photonics-based processors, which utilize light for data movement and computation, targeting initial customers in the near future.

CNBC

Despite the promising developments, European startups face significant challenges. Patrick Schneider-Sikorsky from the NATO Innovation Fund highlights the lengthy development timelines and the need for a more mature foundry ecosystem in Europe. Additionally, Axelera CEO Fabrizio Del Maffeo notes that European governments remain conservative in their investments in new technologies and lack a centralized funding body akin to the U.S. Defense Advanced Research Projects Agency (DARPA). Furthermore, fragmented labor laws across Europe complicate talent recruitment, and the funding gap is evident, with European AI chip startups raising $800 million in 2026 compared to $4.7 billion for U.S. counterparts.

CNBC

Kastrup emphasized the efficiency of Euclyd's chips, stating, "Inference is dominant now, and the existing GPU architecture wasn't built for it in ways that matter most at scale." He also mentioned the company's goal to produce a more advanced multi-chiplet system by 2028. Meanwhile, Hinrikus from Olix remarked on the limitations of current electronic chip architectures, asserting, "The heat [current chips] generate is becoming a major issue."

CNBC

While Euclyd and other startups are making strides, the effectiveness of their technologies at scale remains unproven. Euclyd's systems have not yet been deployed commercially, and the overall funding landscape for European startups continues to lag behind that of the United States.

CNBC

• “Inference is dominant now, and the existing GPU architecture wasn't built for it in ways that matter most at scale,” — Bernardo Kastrup, Founder of Euclyd

CNBC

• “The heat [current chips] generate is becoming a major issue. We strongly believe that the photonics platforms will be the next paradigm,” — Taavet Hinrikus, Partner at Plural

CNBC

As European startups continue to innovate and seek funding, the competition with established players like Nvidia is set to intensify, potentially reshaping the landscape of the AI chip market.

CNBC