Samsung Co-Leads $231M Round for AI Chip Startup Euclyd, Wins OpenAI Foundry Role

Samsung is repositioning from a component supplier into an active architect of the AI hardware and model stack, and three moves this quarter make the strategy legible. The newest is co-leading a $231 million Series A — alongside Somerset Capital — for Euclyd, a Dutch startup building lower-cost inference accelerators aimed squarely at NVIDIA's dominance in the inference layer.

The more consequential win is a custom-chip foundry role with OpenAI, positioning Samsung's fabs to manufacture bespoke AI silicon for the largest model maker — a direct challenge to TSMC's grip on advanced AI chip production. Together with Samsung's lead investment in Mistral's record €3 billion Series D at a €21B+ valuation, the pattern is a memory giant buying its way into the highest-value parts of the AI supply chain: models, custom silicon, and inference startups.

The logic is defensive and offensive at once. Samsung's HBM memory already rides every NVIDIA GPU, but that leaves it captive to NVIDIA's roadmap and margins; backing Euclyd and winning OpenAI's foundry business diversifies its exposure and captures more of the value it currently supplies as a commodity.

The caveats are execution and timing: inference-accelerator startups have a graveyard of predecessors, and displacing NVIDIA requires software ecosystem parity Euclyd doesn't yet have. The OpenAI foundry role is the higher-conviction bet. Watch whether Samsung's foundry can deliver competitive yields on OpenAI's custom designs — that, more than the Euclyd stake, determines whether this strategy reshapes the inference layer.