Nebius Group acquires Inferize to optimize GPU utilization

Nebius Group has acquired Inferize, an inference optimization startup, to improve GPU utilization and reduce infrastructure costs within its managed production-inference platform, Nebius Token Factory.

The acquisition specifically targets the issue of “cold starts,” which occur when delays in loading model weights or spinning up new capacity leave expensive GPUs sitting idle. By integrating Inferize’s technology, Nebius aims to make its inference capacity more elastic, allowing resources to scale more closely with real-time demand and reducing the need to maintain costly spare capacity.

This marks the third acquisition integrated into the Token Factory platform, following previous additions from Eigen AI and Clarifai. The Inferize engineering team will join Nebius to work on capacity-layer elasticity and system responsiveness.