European chip startups are gaining momentum and seeking significant funding to develop alternatives to Nvidia’s dominant Graphics Processing Units (GPUs). These companies aim to capitalize on the artificial intelligence (AI) boom, focusing on more efficient ways to handle AI inference – the process of using trained AI models.
Dutch company Euclyd, co-founded by former ASML director Bernardo Kastrup and advised by ex-ASML CEO Peter Wennink, is currently in talks with investors for a funding round of at least 100 million euros ($118 million). Euclyd claims its AI chips can achieve 100 times higher power efficiency for inference compared to Nvidia’s latest Vera Rubin chips.
Across the U.K., Optalysys plans a fundraise exceeding $100 million later this year, while British company Fractile and France’s Arago are also reportedly seeking nine-figure investments. Earlier in 2026, investors injected over $200 million into the Netherlands’ Axelera and the U.K.’s Olix, signaling strong confidence in the European sector.
Nvidia rapidly became the world’s most valuable company as its GPUs, originally designed for gaming, proved highly effective for training AI models. However, the focus is now shifting to AI inference, where new European startups believe they can offer more efficient solutions.
Patrick Schneider-Sikorsky, director at the Nato Innovation Fund (NIF), an investor in Fractile, noted, “Inference is dominant now, and the existing GPU architecture wasn’t built for it in ways that matter most at scale.” He also pointed to geopolitical factors like U.S. export controls and the desire for European “sovereign compute imperative” as reasons for increased investment in homegrown silicon.
Euclyd, founded in 2024, has already secured a seed round of under 10 million euros. It is developing chip systems with a different architecture from GPUs, aiming to process data in multiple locations to boost efficiency for AI inference. The company believes its silicon systems will reduce energy, cost, and the physical footprint of AI data centers. While its systems are not yet commercially proven at scale, Euclyd has developed an AI inference chip and plans a faster multi-chiplet system by 2028. It is negotiating with four potential customers, hoping to begin supplying two next year and two the year after.
Olix, another startup, is developing photonics-based processors that use light to move and compute data. Taavet Hinrikus, a partner at Plural and an investor in Olix, stated that Olix also aims for initial customers next year, currently in its research and development phase. Hinrikus highlighted that the electronic architecture of current chips, including GPUs, is nearing its limits in terms of miniaturization and heat generation, making photonics a promising “next paradigm.”
Nvidia remains highly competitive, investing over $18 billion in research and development in its most recent fiscal year. It acquired assets from AI inference startup Groq for $20 billion in December and invested $4 billion in two photonics technology companies in March.
