Nebius acquires Eigen AI to boost its Token Factory inference platform

Nebius has acquired Inferize, a move aimed at squeezing more work out of its GPUs and cutting the cost of running AI models on its Token Factory platform. What Nebius is buying, and why According to Nebius, the Inferize deal targets two things: better GPU utilization and lower inference costs on Token Factory. GPU utilization measures how much of a chip’s capacity is doing useful work at any given moment. Idle GPU capacity means every percentage point recovered means more tokens served from the same hardware, which drives down the cost of each response. Token Factory is Nebius’s production inference platform. It serves open-source AI models with a 99.9% uptime service-level agreement and carries compliance certifications including SOC 2, ISO, and HIPAA. A pattern of buying optimization talent Inferize is not the only optimization specialist Nebius has pursued for Token Factory. On May 1, 2026, the Amsterdam-based company, which trades on Nasdaq under the ticker NBIS, announced plans to acquire Eigen AI, a Bay Area inference and model optimization firm whose legal name is MagicByte, Inc. That deal was valued at approximately $643 million, paid in a mix of cash and Class A shares. It...