Euclyd Details 'Craftwerk' AI Inference Architecture Behind Its Samsung-Backed Series A
The AMW Read
Adds architecture and investor detail to a Series A already tracked by AMW, and chip co-design is the article's primary subject, but the raise stays below the capital-cycle threshold.
Euclyd Details 'Craftwerk' AI Inference Architecture Behind Its Samsung-Backed Series A
Euclyd, a Netherlands-based developer of AI inference chips and systems, closed a Series A round exceeding €200 million co-led by Samsung, Somerset Capital Partners, EQT's Scaleup Europe Fund, and Innovation Industries, with participation from Denmark's export investment fund EIFO, imec.xpand, the Brabant Development Agency, and Quadri. Former ASML President and CEO Peter Wennink has joined as board chairman. Per the AI Market Watch index, the round brings Euclyd's tracked total funding to $231M, inside an index that covers roughly 5,000 companies rather than a full census.
Euclyd frames its strategy around memory and efficiency limits on inference at scale: as foundation models grow, power draw, processor-to-memory bandwidth, and data-center capital spending all compound faster than raw compute gains. Rather than a faster standalone chip, it is building a co-designed stack — custom compute silicon called craftwerk, a proprietary memory architecture, and a data-center system layer called craftwerk station, aimed at lowering the power cost of large-scale AI inference. Pairing that full-stack pitch with a chairman drawn from ASML's leadership signals investors are betting on systems-level differentiation, not a single chip generation, in an inference market increasingly stocked with alternatives to Nvidia.
For builders and investors, Euclyd's three named target markets — enterprise, sovereign compute, and hyperscale — mean it is pursuing both government buyers building national AI capacity and the hyperscale customers Nvidia already serves, an ambitious spread for a Series A-stage chip company. Funds are earmarked for agentic AI silicon, memory architecture, and data-center systems, but no shipment timeline or customer commitment appears in this announcement, leaving execution as the open question the funding itself doesn't answer.
