A new architecture for inference
The Series A was co-led by the Samsung Catalyst Fund, Somerset Capital Partners, the EQT-managed Scaleup Europe Fund and Innovation Industries, with Denmark's EIFO, imec.xpand and the Brabant Development Agency also participating. Euclyd was founded in 2024 in Eindhoven by Bernardo Kastrup and Atul Sinha.
Unlike conventional graphics processors, Euclyd's planned platform combines processor-memory co-design and programmable computing into a system it calls CRAFTWERK. The company claims the design delivers roughly 100 times greater power efficiency per generated token than Nvidia's newest chips on comparable language-model workloads — a figure not yet independently validated at commercial scale.
Commercial systems targeted for 2028
Euclyd plans two revenue streams: selling complete hardware and rack systems to enterprises that want to host AI inference privately, and licensing its intellectual property to companies building their own chips. Commercial rollouts are targeted for 2028, and former ASML chief executive Peter Wennink has joined the board as chairman.
With hyperscalers from Google to Meta developing their own silicon and OpenAI announcing its first AI chip in August, the startup's bet reflects a broader industry conviction that the memory-bandwidth wall in today's GPU-based data centers demands new approaches.



