European AI chip startups Axelera AI and Euclyd have secured major milestones in their bid to challenge Nvidia's dominance in the AI inference market. Eindhoven-based Euclyd raised €200 million ($231 million) in a Series A round co-led by Samsung Electronics and the EU's Scaleup Europe Fund, appointing former ASML CEO Peter Wennink as Chairman. Meanwhile, Axelera AI launched its second-generation 'Europa' chips and secured multi-million-dollar supply contracts for EU-backed AI factories.
Euclyd Series A funding round
- ▪Euclyd's Series A funding round was co-led by Samsung Electronics, Somerset Capital Partners, Innovation Industries, and the European Union's Scaleup Europe Fund managed by EQT.
- ▪Dutch chipmaking startup Euclyd announced a Series A financing round of more than €200 million, or approximately $231 million, on September 15, 2026.
- ▪Peter Wennink, the former CEO of semiconductor equipment manufacturer ASML, joined Euclyd's board as Chairman in September 2026.
- ▪Euclyd plans to use its Series A capital to expand its engineering team, accelerate its silicon and systems roadmap, and prepare for commercial deployment.
Axelera Europa chip launch
- ▪Dutch chipmaking startup Axelera AI launched 'Europa', its second generation of chips designed for heavier workloads like corporate servers, on September 15, 2026.
- ▪Axelera AI's Europa chips are designed to be compatible with specified products from Dell and Supermicro.
- ▪Axelera AI plans to develop a future chiplet architecture named 'Titania' to target use in data centres and supercomputers.
Axelera AI factory contracts
- ▪Axelera AI CEO Fabrizio Del Maffeo stated that the company has signed deals worth tens of millions of dollars and has over 600 customers using its chips.
- ▪Axelera AI is pursuing $1.5 billion in potential sales, though Fabrizio Del Maffeo noted this figure does not represent finalized orders or revenue.
- ▪Axelera AI signed multiple contracts to supply chips to AI factories, which are data or scientific computing centres devoted to training and running artificial intelligence software.
- ▪Axelera AI is working with Dell and systems integrator E4 as a supplier for the EU-backed IT4LIA project in Italy and the MeluXina project in Luxembourg.
CRAFTWERK power efficiency architecture
- ▪Euclyd claims its CRAFTWERK system delivers roughly 100 times greater power efficiency per generated token than Nvidia's Vera Rubin chips on comparable large language model workloads.
- ▪Euclyd projects that its CRAFTWERK STATION CWS 32 rack-scale system will deliver over one exaflop of compute performance and 8,000 terabytes per second of bandwidth at 125kW power draw.
- ▪Euclyd is developing CRAFTWERK, a custom AI inference chip system that integrates 16,384 custom SIMD processors directly with Ultra Bandwidth Memory in a single System-in-Package.
European AI chip sovereignty
- ▪The European Union is revamping its network of supercomputing centres into AI factories to narrow the technological gap with the United States and China.
- ▪Privately-held Axelera AI has raised more than $450 million since its founding in Eindhoven in 2021, including a $250 million funding round in February 2026.
- ▪The European Union's €5 billion Scaleup Europe Fund, established on August 4, 2026, made its second major investment by co-leading Euclyd's Series A round.
AI inference market competition
- ▪Samsung Electronics is co-leading Euclyd's funding round and manufacturing Euclyd's first CRAFTWERK silicon in South Korea, hedging its position as an HBM4 supplier for Nvidia.
- ▪European AI chip startups, including Axelera AI, Euclyd, France's VSORA, Britain's Fractile, and Spain's Semidynamics, are competing in the AI inference market.
- ▪Euclyd aims to begin shipping its chip systems in 2028 and serve thousands of corporate customers by 2030, targeting secure, self-hosted AI inference.
Debatable claims
- ▪The EU should prioritize funding domestic AI chip startups
- ▪The EU should convert its supercomputing centers into AI factories
- ▪The emerging custom AI silicon market challenges Nvidia's general-purpose GPU dominance in AI workloads
Story comments
Loading comments…