EUCLYD Raises €200 Million to Break the AI Efficiency Wall With More Efficient Chips

EUCLYD has raised more than €200 million in a Series A round to develop low‑power AI chips and energy‑efficient data centres designed to tackle what the company calls the “efficiency wall” in artificial intelligence. The Netherlands‑based startup said the round was co‑led by Samsung, Somerset Capital Partners, EQT’s Scaleup Europe Fund and Innovation Industries, with additional participation from EIFO, imec.xpand, the Brabant Development Agency and Quadri. Peter Wennink, the former President & CEO of ASML, joins EUCLYD as Chairman of the Board.

"Artificial intelligence is becoming a foundation of economic growth, scientific discovery, and national competitiveness," said EUCLYD CEO Bernardo Kastrup. "Its potential will remain constrained unless we fundamentally change the infrastructure beneath it." The new capital will be used to hire engineers, accelerate chip development and ready the company to sell to governments and enterprise customers.

What EUCLYD is building

EUCLYD, founded by Bernardo Kastrup and Atul Sinha and based at the High Tech Campus in Eindhoven, is pitching a different route to AI scale: redesigning the relationship between processors and memory rather than simply adding raw compute. The company’s two headline products are Craftwerk, described as the first "agentic AI silicon" chip, and Craftwerk Station (CWS), billed as "the lowest‑power exascale AI factory" on the market. EUCLYD says this co‑design approach reduces wasted energy in the communication between chip and memory — the central bottleneck in running large AI models.

• Series A funding: more than €200 million
• Co‑lead investors: Samsung, Somerset Capital Partners, EQT’s Scaleup Europe Fund, Innovation Industries
• Other backers: EIFO, imec.xpand, Brabant Development Agency, Quadri
• Chairman: Peter Wennink, former ASML CEO
• Location: Eindhoven, Netherlands (High Tech Campus)

Strategic backers have framed the investment as a pivot in the AI race toward infrastructure efficiency. "The next big fight in AI won't be about who has the smartest model but about who can run it most efficiently," said Samsung Senior Vice‑President Dede Goldschmidt. Ted Persson of EQT added that Europe already has the engineering talent needed, and that EUCLYD "is combining that with ambitions that go way beyond just the local market." Peter Wennink invoked European manufacturing and chip expertise as a competitive advantage: Europe "just needs to turn that into a real global AI platform," he said.

Context and implications

The release stresses that current AI scaling trends — more power, more memory, more chips — make running large models increasingly expensive, effectively limiting advanced AI to well‑funded tech firms. EUCLYD’s stated mission, framed as "Abundant Intelligence," aims to democratise access by lowering the energy and cost per inference. If Craftwerk and CWS deliver the promised reductions in power consumption, organisations from startups to public institutions could run larger, more sophisticated models without the current level of infrastructure expense.

EUCLYD plans to channel the funding into expanding its engineering team, speeding chip design and development cycles, and moving toward commercial availability for enterprise and government customers. The technical challenge is substantial: solving memory bandwidth and processor co‑design at scale requires both novel silicon architectures and manufacturing partnerships.

Outlook

Investors’ willingness to commit over €200 million signals belief that the next phase of AI competition will hinge on hardware efficiency as much as model performance. For EUCLYD, success will depend on moving from prototype to production, securing manufacturing capacity, and proving real‑world energy and cost savings against incumbent GPU‑centric solutions. If it can do that, the company could materially shift who can afford to run large AI workloads — potentially widening access to advanced AI capabilities beyond the largest tech players.