Pomegra
Eindhoven's EUCLYD closed a €200M+ Series A co-led by Samsung, with ex-ASML CEO Peter Wennink joining as chairman, to build lower-power AI inference chips targeting Nvidia's dominant position.
Key Takeaways
- Samsung co-led the €200M+ (~$231M) round alongside Somerset Capital Partners, EQT's Scaleup Europe Fund, and Innovation Industries.
- Peter Wennink, who ran ASML as President and CEO from 2013 to 2024, joins Euclyd's board as Chairman, lending rare manufacturing-ecosystem credibility to a two-year-old startup.
- Commercial hardware shipments are not projected until 2028, making this a long-horizon capital deployment by any measure.
Lead
On September 15, 2026, EUCLYD - an Eindhoven-based semiconductor systems company - announced a Series A financing round of more than €200 million, approximately $231 million at current exchange rates. Samsung, Somerset Capital Partners, the Scaleup Europe Fund managed by EQT, and Innovation Industries co-led the round. EIFO, imec.xpand, the Brabant Development Agency, and Quadri participated. Founded in 2024 by Bernardo Kastrup and Atul Sinha at Eindhoven's High Tech Campus, the company has now closed one of the largest chip startup rounds in European history before shipping a single product.
What Does EUCLYD Actually Build?
The company designs chips for AI inference - the compute stage where a trained model responds to user queries - rather than for model training, where Nvidia's H100 and B200 GPUs currently command the market. Euclyd's roadmap centers on two products: craftwerk, described as agentic AI silicon, and the CWS 32 system, positioned as a low-power exascale AI factory. Both combine programmable ASIC compute with processor-memory co-design, targeting the memory bandwidth and power consumption limits that push current GPU infrastructure toward its ceiling.
Those claims cannot be independently verified yet. Commercial hardware shipments are not expected before 2028. This round buys engineering time and talent, not near-term revenue, and investors are pricing in a multi-year development cycle.
Why Did Samsung Lead This Round?
Samsung's decision to co-lead a European chip startup's Series A reflects competitive pressure more than geography. The company operates a foundry business that competes with TSMC for advanced node manufacturing contracts. Backing a credible Nvidia alternative - one whose chips could eventually be manufactured on Samsung fabs - serves both strategic positioning and future revenue potential. Samsung has joined inference chip investments across multiple continents over the past two years as GPU shortages and power constraints push hyperscalers to evaluate alternatives.
The round's European composition carries its own significance. EQT's Scaleup Europe Fund exists partly to reduce the continent's dependence on US and Asian technology suppliers. The Brabant Development Agency's participation indicates active regional support for Eindhoven's effort to build a sovereign chip ecosystem around its High Tech Campus, the same industrial cluster that houses ASML's headquarters.
Can a Startup Break Nvidia's Grip on Inference?
That question is the only one that matters for EUCLYD's long-term prospects. Nvidia's inference business has expanded sharply alongside generative AI deployments. The economics are also visibly strained: a single rack of H100s can draw upward of 10 kilowatts, and power availability has become a hard constraint on AI data center expansion. A chip that delivers lower cost per token at lower power consumption addresses a genuine, growing problem.
The gap to close is large. Nvidia's CUDA software ecosystem represents two decades of developer tooling and switching cost that no hardware specification can simply override. US competitors including Cerebras, Groq, and SambaNova have built credible inference silicon and still hold single-digit market share. Europe has produced capable chip designers but almost no chip companies at commercial scale.
Peter Wennink's board role addresses one specific constraint: relationship capital. His 11 years leading ASML gave him unmatched visibility into European semiconductor supply chains, advanced packaging vendors, and the broader equipment ecosystem that any chip startup needs to navigate. That network shortens timelines and opens doors that balance sheets alone cannot.
Euclyd enters this race with a stronger starting capital position than most European challengers have ever had. Whether that translates into shipping silicon depends on execution between now and 2028, a period in which Nvidia will not stand still.
Outlook
EUCLYD's €200M+ Series A is a serious commitment to a serious problem. Power and cost inefficiency in AI inference infrastructure is real, measurable, and growing as a bottleneck. The investor syndicate is credible, the geographic concentration in Eindhoven is deliberate, and Wennink's presence signals that this is not a paper-thin European AI venture. The harder tests come next: tape-out execution, software ecosystem development, and customer acquisition before the inference market consolidates around one or two dominant non-Nvidia architectures. Valuation was not disclosed.
