Enterprise buyers and the startup gain a major strategic backer after Samsung joined a €200 million Series A in Euclyd, accelerating the Dutch company's effort to build inference hardware that does not use graphics processing units. The round was co-led by Somerset Capital Partners, the Scaleup Europe Fund managed by EQT, and Innovation Industries, the company said.
Founded in 2024, Euclyd is designing a silicon and system architecture that replaces GPU processors and memory layouts, targeting inference workloads for large-scale language and foundation models. CEO Bernardo Kastrup told CNBC the company expects its silicon systems for foundational models to cut the energy demands and operating costs of AI data center infrastructure, though the systems have not yet been proven at commercial scale.
Euclyd plans two revenue streams: selling hardware and complete rack systems to enterprises seeking secure, self-hosted inference, and licensing intellectual property to firms building their own processors. Kastrup framed the investment as more than cash, noting Samsung's industrial strengths in systems and memory. “AI is becoming a foundation of economic growth, scientific discovery and national competitiveness, but its potential will remain constrained unless we fundamentally change the infrastructure beneath it,” he said. He added that Samsung brings more than capital, noting, “They are one of the biggest memory manufacturers in the world. They do a lot of engineering, they know a lot about systems, they know the supply chain, they have a huge network.”
The deal lands amid active moves to diversify AI hardware beyond Nvidia GPUs, which dominate high-end training and inference. Hyperscalers and startups are pursuing proprietary processors: OpenAI announced its Jalapeño chip in August, and Google, AWS and Meta continue in-house silicon programs. Euclyd aims to begin shipping physical chip systems in 2028 and to serve thousands of enterprise customers by 2030, targets the CEO provided.
Samsung’s capital and manufacturing expertise could shorten the runway from prototype silicon to deployed racks, but the immediate test will be real-world scale deployments and whether Euclyd’s designs deliver the promised energy and cost gains. The company now has a multi-year product timetable and a clearer path to production, while investors and incumbents will watch whether non‑GPU approaches can take measurable share from entrenched GPU supply chains.
