Euclyd, a two-year-old Eindhoven startup with no commercial silicon, just closed a Series A of more than €200 million co-led by Samsung, Somerset Capital Partners, the EQT-managed Scaleup Europe Fund, and Innovation Industries1,3,4. ANALYSIS The round's size and its backers' profiles mark a new phase in the inference-chip challenge to Nvidia: industrial incumbents with supply-chain leverage, not just venture capital, are now writing the checks.
Why it matters
The AI chip market is projected to grow from roughly $100 billion this year to more than $400 billion by 2030. Nvidia still controls an estimated 80% of AI training chip demand. ◆ That concentration has made inference, the workload where a trained model actually answers a prompt, the most contested frontier for alternatives. Euclyd's pitch is that GPUs are the wrong tool for inference because they spend most of their energy moving data between memory and compute cores. Its CRAFTWERK chip packs 16,384 custom processors that work directly on data in memory. A rack-scale version, CRAFTWERK Station, combines 32 of those chips and targets one exaflop of compute by 2028. Euclyd claims up to 100 times the power efficiency of Nvidia's latest Vera Rubin chips, modelled on Meta's Llama 4 Maverick, though the company itself said that projection has not been tested at commercial scale.
The big picture
Euclyd is not the only startup drawing large checks for inference silicon. Etched's valuation rose from $5 billion in June to $21 billion by August. Axelera AI raised more than $250 million in February for edge inference. OpenAI announced in August that its first AI chip, the Jalapeño, had entered development. ◆ What distinguishes Euclyd's round is the identity of its lead backer: Samsung is one of the biggest memory manufacturers in the world, and Euclyd's architecture depends on tight coupling between compute and memory. Samsung can help Euclyd beyond money because it has engineering, systems, supply-chain, and network expertise. The appointment of Peter Wennink, the former president and CEO of ASML, as chairman reinforces the industrial logic. Wennink spent years at the center of the semiconductor equipment supply chain; his presence signals to prospective customers and fabrication partners that Euclyd's roadmap has been vetted by someone who understands what it takes to move from design to volume production.
The round also doubled what Euclyd first sought. The startup originally targeted at least €100 million. "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. This financing accelerates our mission to make intelligence abundant through greater efficiency, lower cost per token, and broader access to advanced AI," said Bernardo Kastrup, founder and CEO of Euclyd.
Between the lines
Euclyd is trying to own the whole stack: chip, memory, and the datacenter systems around them. Its platform pairs programmable ASIC compute with processor-memory co-design and system-level optimisation. ◆ That full-stack ambition explains both the round's size and its risk. Euclyd's systems have yet to be proven at scale in commercial deployments. The company is in talks with four prospective customers and hopes to start supplying two of them in 2027. It aims to begin rolling out physical chip systems in 2028 and to serve thousands of enterprise customers by 2030.
Euclyd is targeting two revenue streams: selling hardware and physical rack systems to enterprise customers for secure self-hosted AI inference, and selling its intellectual property to companies developing their own chips. ◆ The IP-licensing path is notable because it positions Euclyd less as a pure hardware vendor and more as an architecture provider, a model closer to Arm's than to Nvidia's.
The new capital will grow Euclyd's engineering team, speed up its silicon and systems roadmap, and prepare it for enterprise, sovereign, and hyperscale customers. Additional backing came from EIFO, Denmark's export and investment fund, imec.xpand, the Brabant Development Agency, and Quadri. ◆ The participation of sovereign-adjacent funds from Denmark and the Netherlands underscores a European policy appetite for homegrown AI infrastructure that does not depend on a single U.S. supplier.
What's next
Euclyd's founder, Kastrup, previously co-founded Silicon Hive, a parallel-processor company Intel acquired in 2011. He holds PhDs in both philosophy and computer engineering. ◆ His track record gives the venture a credibility floor, but the gap between a funded design and commercial silicon remains wide. Euclyd hopes to start supplying two prospective customers in 2027. Euclyd's own acknowledgment that its efficiency claims have not been tested at commercial scale sets the benchmark for the next milestone: first silicon in customer hands.