European chip startups are positioning themselves for significant funding rounds as they aim to carve out a niche in the booming artificial intelligence market, offering alternative technologies to the dominant graphics processing units (GPUs) from industry titan Nvidia. These emerging players are focused on optimizing AI inference, the crucial stage where trained AI models are deployed to make predictions, and believe their specialized architectures can offer substantial efficiency gains.
Dutch company Euclyd, which boasts backing from a former CEO of ASML, a leading semiconductor equipment manufacturer, is reportedly in advanced discussions with investors for a funding round projected to exceed 100 million euros ($118 million). Similarly, U.K.-based Optalysys is reportedly planning a fundraising effort of over $100 million later this year. Reports also indicate that U.K. firm Fractile and France’s Arago are in the process of securing nine-figure funding rounds. While Fractile declined to comment, Arago did not respond to inquiries. This wave of investment is already evident, with over $200 million funneled into the Netherlands’ Axelera and the U.K.’s Olix in the early part of 2026.
Nvidia has ascended to become the world’s most valuable company largely due to the repurposing of its GPUs, initially designed for gaming, for the computationally intensive task of training AI models. However, the industry’s focus is now shifting towards the most efficient methods for AI inference. While Nvidia is actively developing its own semiconductor solutions for this purpose, a new generation of European startups is emerging, asserting that their proprietary technologies can achieve superior efficiency.
“Inference is now the dominant workload, and the existing GPU architecture wasn’t fundamentally designed for it in ways that matter most at scale,” explained Patrick Schneider-Sikorsky, director at the Nato Innovation Fund (NIF), an investor in Fractile. “The geopolitical tailwinds are undeniable, with U.S. export controls, concentration risk around key foundries like TSMC, and a genuine European imperative for sovereign compute capabilities all driving capital towards homegrown silicon solutions.”
### ASML Alumni Lead the Charge
Euclyd is at the forefront of this European push, developing AI chips designed for systems that the company claims can deliver up to 100 times greater power efficiency for inference compared to Nvidia’s latest generation of chips. The Dutch startup, founded in 2024 by former ASML director Bernardo Kastrup and advised and invested in by former ASML CEO Peter Wennink, has already secured a seed round of less than 10 million euros. The company is now seeking substantial new capital to scale its technology and begin supplying its initial customer base.
Kastrup elaborated that Euclyd is building chip systems to function as alternatives to GPUs but with a fundamentally different architectural approach. While traditional GPUs expend significant time and energy moving data between memory and processing units, Euclyd’s chips are engineered to process data in parallel across multiple locations, a design principle Kastrup believes will dramatically enhance inference efficiency. The company’s silicon systems are anticipated to reduce the energy consumption, cost, and physical footprint of AI data center infrastructure, particularly for foundational models. However, a key challenge remains: Euclyd’s systems have yet to achieve large-scale commercial deployment with partners, unlike Nvidia’s established offerings.
Euclyd is actively addressing this challenge. The company has already developed a functional chip for AI inference and is currently working on a multi-chiplet system, projected for production by 2028, which is expected to offer significantly faster processing speeds than its current iteration. Kastrup revealed that Euclyd is engaged in negotiations with four prospective customers, with the ambition to commence supply to two of them next year and the remaining two the year after.
Another player in this space, Olix, which is developing photonics-based processors for AI, is also targeting initial customer engagements next year, though it remains in the research and development phase, according to Taavet Hinrikus, a partner at Plural, an investor in the company. Photonics processors leverage light for data transmission and, in some cases, for computational tasks. Hinrikus indicated that Olix aims to serve a broad spectrum of customers requiring inference services, including hyperscalers and government entities.
The fundamental electronic architecture of current chips, including GPUs, is reportedly reaching its physical limits in terms of miniaturization. Chip manufacturers strive to shrink processors to accommodate more components on silicon wafers and improve the economic viability of operating complex systems. “The heat generated by current chips is becoming a major bottleneck,” Hinrikus observed. “We strongly believe that photonic platforms represent the next significant paradigm shift in computing.”
Nvidia, meanwhile, is not standing still. The chip behemoth invested over $18 billion in research and development during its most recent full financial year, ending January 2026. In a significant move in December, Nvidia reportedly acquired assets from AI inference startup Groq for approximately $20 billion. Furthermore, in March, the company announced a $4 billion investment in two firms specializing in photonics technology, underscoring its commitment to exploring next-generation semiconductor solutions.
### European Startups Face Enduring Hurdles
Despite the burgeoning innovation, European startups in this sector continue to grapple with significant challenges. “Chip development timelines are inherently long, the journey from design finalization to volume deployment is arduous, and Europe’s foundry ecosystem still requires substantial maturation,” commented Schneider-Sikorsky of the NIF.
Fabrizio Del Maffeo, CEO of Axelera, highlighted that European governments often exhibit a conservative approach to investing in products from nascent companies. He noted the absence of a European equivalent to DARPA, the U.S. Department of Defense agency renowned for funding groundbreaking research and startups. Moreover, Europe lacks robust mechanisms to stimulate demand for locally manufactured products, and fragmented labor laws across national borders impede the efficient recruitment of European talent.
According to Dealroom data, European AI chip startups have collectively raised $800 million in 2026, a stark contrast to the $4.7 billion secured by their U.S. counterparts. In the U.S., Cerebras Systems alone secured $1 billion in February, while MatX, Ayar Labs, and Etched have each raised $500 million rounds this year.
Nevertheless, the increasing focus on AI inference and the pursuit of alternatives to Nvidia are attracting growing interest from investors towards European startups. “We are observing this trend in deal flow and in our ongoing dialogues with founders in this sector,” stated Carlos Espinal, managing partner at Seedcamp, a firm that has backed chip startup Vaire Computing. “This is no longer a niche bet; it is fundamentally reshaping how the industry perceives and plans for AI infrastructure.”
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/20755.html
