Nebius Group, an AI infrastructure and cloud computing company, announced it has acquired Inferize, a company focused on inference optimization. As AI workloads shift from training to running models in production at scale, inference cost and latency have become the more immediate economic pressure point for most AI companies, ahead of training cost.
Nebius's acquisition fits a broader pattern among AI infrastructure providers in 2026: training-focused compute providers are increasingly acquiring inference-specialized teams, recognizing that the economics of serving a model to millions of users differ enough from training it that the two problems increasingly require distinct technical expertise.