SHELFCRITTER AI Digest

COMPANY

Euclyd

companytopic-noteai-hardwareinferenceeuropean-ai

Overview

Euclyd is an Eindhoven-based AI inference chip startup founded in 2024 by former ASML director Bernardo Kastrup, advised and backed by ex-ASML CEO Peter Wennink. The company is targeting NVIDIA’s Vera Rubin generation on the axis that matters most for 2026–27 inference economics — power efficiency — with an architecture that claims to process data in multiple places instead of moving it through a centralized memory stack. Euclyd is one of the highest-profile entrants in Europe’s emerging inference-chip cohort, which has raised ~$800M year-to-date (versus $4.7B for US peers per CNBC).

Timeline

  • 2026-04-18-AI-Digest — Euclyd raising at least €100 million (~$118M) to scale its first chip and develop a multi-chiplet system targeting 2028 production. Kastrup’s claim in CNBC coverage: Euclyd’s architecture can deliver 100× higher power efficiency for inference than NVIDIA’s Vera Rubin generation. The company reports four customers in negotiations — two for initial deliveries in 2027 and two in 2027. The raise lands inside a broader European inference-chip wave: Netherlands’s Axelera and the UK’s Olix have drawn over $200M combined, and the European Commission awarded its €180M / six-year / four-provider sovereign-cloud tender the same day. The structural signal is that the funding environment for credible European alternatives to NVIDIA is the best it has been since CUDA lock-in became apparent; whether the 100× claim survives commercial validation is the unresolved technical question.

Key Developments

  1. 100× Power-Efficiency Claim: Kastrup’s headline number — 100× inference power efficiency over Vera Rubin — is the boldest public claim against NVIDIA’s newest platform this cycle. The architectural basis is in-memory / distributed processing, eliminating the data-movement cost that dominates energy at inference scale. Commercial validation is still years out, but the claim has anchored the European inference-chip narrative around power-per-token rather than raw FLOPS.

  2. ASML Pedigree as Credibility Signal: Euclyd’s founding team — an ex-ASML director as CEO, an ex-ASML CEO as advisor and backer — gives the company the strongest European-semiconductor pedigree of any inference-chip startup on the continent. That lineage is central to the “Euclyd is serious” framing in the CNBC and Dutch-press coverage.

  3. Part of the European Sovereign-AI Wave: Euclyd is the most visible single company in a broader pattern: European AI-inference chip startups have raised ~$800M YTD, the European Commission’s €180M sovereign-cloud tender awarded April 17, and emerging political consensus that European AI infrastructure can’t fully depend on NVIDIA. Euclyd’s €100M raise is the proof point that the political narrative is translating into venture capital flows.

  4. 2028 Production as the Commercial Test: The multi-chiplet system targeting 2028 production is when Euclyd’s claims convert from bench numbers to customer-validated performance. Four customers in negotiations (two for 2027, two for 2028) is the commercial backbone the raise is designed to finance.