Samsung Electronics is backing Dutch semiconductor startup Euclyd with a more than €200 million ($231 million) round of Series A to help the company develop a new type of processor architecture designed to make artificial intelligence inference more energy-efficient.
Samsung was joined in leading the funding round by Somerset Capital Partners, EQT’s Scaleup Europe Fund and Innovation Industries, with participation from EIFO, imec.xpand, the Brabant Development Agency and Quadri.
With the funding comes some serious experience in the form of Peter Wennink, former president and CEO of ASML, The Dutch giant that makes all of the manufacturing equipment used by TSMC, Intel, and other fabrication companies. Wennink joins the company as chairman of the board.
Wennink’s appointment gives Euclyd another connection to Europe’s semiconductor ecosystem, particularly as the region seeks to build more advanced semiconductor and AI infrastructure companies.
The startup is initially targeting AI inference — the stage at which trained models process queries and generate responses — rather than the training of large AI models. Inference generally requires less compute per operation than training, but its power consumption can scale significantly as people make more queries against a model. So inference power efficiency has become an important factor and there are several startups trying to make power efficient inferencing chips.
Euclyd describes its approach as an attempt to overcome an “efficiency wall” facing AI infrastructure. Its technology combines programmable ASIC computing, processor-memory co-design and system-level optimization rather than relying on conventional GPU architectures.
At the center of Euclyd’s development roadmap are its Craftwerk AI silicon and Craftwerk Station (CWS) systems. The company says the platform is designed to deliver high levels of AI performance while substantially reducing power consumption.
Euclyd plans to use the new financing to expand its engineering operations, accelerate development of its silicon and systems roadmaps and prepare for commercial deployment. It should be noted that Euclyd has not yet demonstrated the technology at commercial scale, so its claims should be taken with a grain of salt.
The company expects to begin shipping physical chip systems in 2028 and has said it intends to serve thousands of companies by 2030. Its planned products are aimed at enterprise, sovereign and hyperscale customers that want to run AI inference closer to where data is generated or processed.
