EUCLYD, FMC team up on sovereign AI infrastructure

EUCLYD and Ferroelectric Memory Company said Tuesday they are collaborating to build secure, sovereign AI infrastructure for Europe, with early integration work underway and commercial products targeted from 2029. The partnership combines compute and persistent memory technology as power limits, data movement costs and data sovereignty reshape AI data center strategy.

Why it matters: - EUCLYD and Ferroelectric Memory Company are trying to address two of AI infrastructure’s biggest constraints: power consumption and memory bottlenecks. - The partnership is aimed at European AI sovereignty, with a stack designed to keep sensitive data local and reduce reliance on public cloud infrastructure. - The companies say the approach could also fit regulated sectors such as defence, banking and healthcare.

What happened: - EUCLYD and FMC announced a strategic technology and business collaboration during the AI Infra Summit 2026 at the Santa Clara Convention Center. - The companies said early integration work is already underway. - Both companies are targeting commercial products from 2029. - EUCLYD is based in Eindhoven, the Netherlands. - FMC is headquartered in Dresden, Germany.

The details: - The collaboration will align compute, memory and system architectures. - EUCLYD’s compute architecture will be paired with FMC’s persistent nvDRAM+ technology. - FMC also develops high density Cache+ memory. - The companies said the goal is to move data closer to compute for better efficiency, lower operating costs and improved performance. - The partnership will explore AI architectures for model state, KV-cache, checkpointing, recovery and selective persistent working data. - The companies say tighter integration of compute and persistent memory should support more power efficient, scalable, resilient AI infrastructure. - EUCLYD and FMC have raised more than €300 million combined to date. - EUCLYD develops next-generation AI compute and AI factory infrastructure. - FMC develops advanced memory technologies designed to address performance, capacity, persistence and power constraints in AI systems. - EUCLYD’s website - EUCLYD on LinkedIn

Between the lines: - Power has become a major constraint for AI infrastructure as data centers scale. - The release cites McKinsey projections that by 2028, AI inference will overtake training as the largest single source of data center power use. - The release also says nearly half of global data center projects due to complete this year face delays tied to power supply limits. - It adds that more than $130 billion in U.S. projects were blocked or delayed in the first quarter of 2026, largely because of energy costs and local opposition to new data center construction. - EUCLYD and FMC are positioning European hardware as a way to compete on both performance and sovereignty.

What's next: - The companies will pursue joint commercial opportunities to bring the combined architecture to market. - EUCLYD and FMC will continue integration work on persistent memory-enabled AI system designs. - The 2029 target suggests the collaboration is still in an early build phase, with product development and commercialization ahead.