Euclyd, a Dutch AI chipmaker founded in 2024, has closed a $231 million Series A round led by Samsung to develop inference processors designed to undercut Nvidia's dominance in data center AI workloads.
The round, which exceeded 200 million euros, included Somerset Capital Partners, the Scaleup Europe Fund managed by EQT, and Innovation Industries.
Euclyd is designing a system-on-chip architecture optimized specifically for inference—the lower-margin, compute-intensive phase where trained models run in production. The company claims its processor and memory design reduces energy consumption and total cost of ownership compared to GPU-based inference, a growing expense for enterprises running large language models at scale.
CEO Bernardo Kastrup said Samsung's value extends beyond capital. The chipmaker brings memory manufacturing expertise, supply chain relationships, and systems integration knowledge critical for competing against Nvidia's vertically integrated advantage.
The company is pursuing a two-pronged business model: selling proprietary hardware and rack systems to enterprises seeking on-premises inference, and licensing its chip architecture to other manufacturers.
Euclyd joins a crowded field. Google, Amazon Web Services, Meta, and OpenAI are all developing custom silicon for AI workloads. OpenAI announced its "Jalapeño" inference chip in August, claiming efficiency gains over general-purpose GPUs.
The challenge: Euclyd has no commercial deployments at scale. The company targets 2028 for initial rollout and projects serving thousands of enterprise customers by 2030—a timeline that gives Nvidia nearly four years to entrench custom silicon of its own.
