EUCLYD raises more than €200 million as Peter Wennink becomes chairman

EUCLYD announced a Series A worth more than €200 million on September 15, 2026, as the Dutch semiconductor-systems startup works toward specialized infrastructure for AI inference—the stage where trained models generate responses to user requests. The company also named Peter Wennink, former President and CEO of ASML, chairman of its board.

Samsung, Somerset Capital Partners, the Scaleup Europe Fund managed by EQT, and Innovation Industries co-led the round. EIFO, imec.xpand, Brabant Development Agency (BOM), and Quadri also participated.

What EUCLYD is building

EUCLYD is developing an inference-focused platform that combines compute silicon, memory architecture, and datacenter systems. Its roadmap includes craftwerk and craftwerk station CWS, with a design centered on programmable ASIC computing, processor-memory co-design, and system-level optimization.

That focus matters because inference has a different bottleneck from training. Once a model is trained, every response still consumes computing power and moves large amounts of data between processors and memory. EUCLYD is betting that designing those parts together can reduce the energy and cost required for each generated token.

The company was founded in 2024 by Bernardo Kastrup and Atul Sinha at High Tech Campus Eindhoven. Kastrup is its CEO.

Why inference is becoming a chip battleground

The AI hardware market has been built around flexible GPUs that can handle training and inference across a wide range of models. EUCLYD is taking a narrower route: it is targeting the repeated, high-volume workloads involved in serving foundation models.

That specialization could be valuable when a workload is predictable enough for efficiency to matter more than maximum flexibility. The trade-off is that a specialized architecture has to keep pace with changing models and deployment requirements, while established GPU platforms already offer broad hardware and software ecosystems.

EUCLYD's financing gives it room to expand engineering, develop its silicon and systems, and build partnerships for enterprise, sovereign, and hyperscale markets. Samsung's role is particularly relevant to that strategy because it brings semiconductor and memory expertise alongside its investment. EUCLYD has described Samsung's value as extending beyond money to engineering, supply-chain knowledge, and industrial reach.

craftwerk and craftwerk station CWS remain roadmap products

EUCLYD presents craftwerk as agentic AI silicon and craftwerk station CWS as a rack-scale AI system. The company claims that craftwerk could deliver up to 100× greater power efficiency per generated token than Nvidia Vera Rubin in a modeled inference comparison based on Meta's Llama 4 Maverick.

That figure is a company projection, not a commercial-scale performance result. EUCLYD's systems are still at the roadmap stage, and the financing announcement does not turn the comparison into a shipped product or a customer deployment.

The distinction is crucial. A theoretical efficiency advantage can make a specialized chip attractive to datacenter operators, but the real test also includes software support, model changes, reliability, production, and sustained operation at scale. Those are the conditions that will determine whether EUCLYD's architecture can move beyond an ambitious design target.

Samsung and Peter Wennink raise the European stakes

Samsung's participation places a major semiconductor company behind EUCLYD's effort, while Peter Wennink adds experience from the leadership of ASML. Wennink became chairman as part of the September 15 financing announcement and said EUCLYD could turn European strengths in semiconductors, advanced manufacturing, and systems engineering into a globally competitive AI-infrastructure platform.

The investor group also includes European deep-tech and technology funds. Alongside Samsung, the four co-leads are Somerset Capital Partners, the Scaleup Europe Fund, and Innovation Industries; EIFO, imec.xpand, BOM, and Quadri joined as participating investors.

The money is intended to accelerate EUCLYD's engineering program and strengthen the partnerships needed to build and deploy a complete system rather than a standalone processor. That is a much larger challenge than designing a chip: memory, networking, software, cooling, manufacturing, and datacenter integration all have to work together.

The roadmap points to 2028

EUCLYD is targeting physical-system rollout in 2028 and has said it aims to serve thousands of enterprise customers by 2030. The company has also been reported as targeting earlier deliveries to two prospective enterprise customers in 2027.

For now, the financing marks a major expansion of EUCLYD's effort to build inference hardware in Europe. The next decisive milestone is whether craftwerk and craftwerk station CWS can progress from roadmap designs to systems operating in real enterprise and datacenter environments.