EUCLYD raises €200m to tackle AI inference efficiency in Europe

The Eindhoven chip startup's Series A targets the power and cost constraints throttling foundation model deployment at scale.

By Disrupts Editorial Team Mon, 21 September 2026 AI level 2 of 5

AI level 2 of 5: drafted by our AI editorial assistant from the company's announcement; fact-checked, edited and signed off by Mark Walker, Editorial Director. What the levels mean

A brightly lit data center aisle extends into the distance, flanked by parallel rows of dark server racks with blinking colorful indicator lights.

EUCLYD, a semiconductor systems company headquartered at High Tech Campus Eindhoven, has closed a Series A financing round of more than €200 million, with co-leads spanning Samsung, Somerset Capital Partners, EQT's Scaleup Europe Fund, and Innovation Industries. The raise positions Europe's deep semiconductor heritage directly against the infrastructure bottleneck that now defines the next phase of the AI buildout: not model intelligence, but the cost and power required to run it.

The round draws in a notable coalition of public and private capital, including EIFO (Denmark's national export credit and venture agency), imec.xpand (the venture arm of Europe's premier semiconductor R&D institution), the Brabant Development Agency, and London-New York firm Quadri. Peter Wennink, who led ASML through its ascent as the world's monopoly supplier of extreme ultraviolet lithography machines, joins as Chairman of the Board, a signal that EUCLYD's ambitions extend well beyond another fabless chip startup.

Reframing the AI infrastructure problem

EUCLYD's proposition is that the AI industry has optimised relentlessly for model capability while the infrastructure layer has become an increasingly expensive liability. The company describes an "efficiency wall" in AI inference: as foundation models grow in capability, deploying them at scale demands disproportionately more power, memory bandwidth, and physical infrastructure. Its platform attempts to address all three simultaneously through what it calls craftwerk, described as the world's first agentic AI silicon, paired with craftwerk station CWS, a system it positions as the world's lowest-power exascale AI compute facility.

The underlying architecture combines programmable ASIC compute (application-specific chips that trade flexibility for efficiency), processor-memory co-design, and system-level optimisation. The memory bandwidth constraint is a genuine and well-documented problem in the industry: most GPU-based inference systems spend a disproportionate share of their energy budget moving data between processing and memory units rather than performing computation. EUCLYD is betting that co-designing those two elements from the ground up can rewrite the economics of inference.

"AI is becoming a foundation of economic growth, scientific discovery, and national competitiveness," said Bernardo Kastrup, EUCLYD's founder and CEO, "but its potential will remain constrained unless we fundamentally change the infrastructure beneath it."

The European sovereign AI angle

The capital structure of this round carries its own geopolitical signal. EQT's Scaleup Europe Fund, targeting €5 billion, is explicitly a vehicle for bringing public and private capital together behind European deep-tech champions. EIFO's participation extends the sovereign dimension: Denmark's national promotional bank is effectively deploying state balance sheet into a Dutch chip systems startup, treating AI infrastructure as a matter of national competitiveness rather than purely commercial return.

That framing matters at the macro level. The United States is still absorbing the long-term implications of the CHIPS Act, and Taiwan's TSMC remains the world's dominant advanced-node foundry. Europe has so far competed on equipment (ASML) and research infrastructure (imec) rather than on systems-level AI compute. EUCLYD's pitch is that European engineering depth, the semiconductor cluster around Eindhoven, the ASML talent network, imec's research pipeline, can be assembled into a globally competitive AI infrastructure platform rather than exported piecemeal to US and Asian hyperscalers.

For cross-sector capital allocators, the second-order read is significant. If AI inference cost falls substantially, the bottleneck shifts: life sciences and drug discovery workflows that currently cannot afford continuous foundation model inference at scale become viable; defence and intelligence agencies constrained by off-grid or low-power deployment environments gain new options; and the economics of sovereign AI cloud infrastructure in mid-tier economies change materially. EUCLYD's investors appear to be pricing in all three. The company says it is targeting enterprise, sovereign, and hyperscale AI markets simultaneously, which is an unusually broad commercial mandate for a startup still in the silicon development phase.

The financing will fund expansion of EUCLYD's engineering organisation and accelerate its silicon and systems roadmap toward commercial deployment. No timeline for tape-out or product launch was disclosed.