Nebius Group agreed to acquire Eigen AI for about $643 million in cash and stock, a deal that adds model optimization and inference technology to its Token Factory managed inference platform. The move is meant to help Nebius offer customers a more efficient and scalable way to run AI models as demand for compute-hungry applications keeps building.
The acquisition also gives Eigen AI’s team a home in Nebius’s first engineering and research hub in the San Francisco Bay Area. For Nebius, which focuses on AI infrastructure, the deal plugs directly into its core business by adding a post-training optimization layer on top of Token Factory and pushing the company further toward full-stack AI infrastructure.
That matters because Nebius is already scaling contracts for companies like Meta while competing with Amazon, Microsoft and Google in a market where large clients want high-throughput inference on Nvidia hardware. Eigen AI is built around squeezing more tokens and better performance out of each GPU, which makes it a natural fit for Nebius’s effort to sell efficiency at scale.
The tradeoff is that the purchase adds to Nebius’s data center and GPU spending at a time when analysts have flagged heavy capital expenditure and rising debt as key risks. Nebius is betting that deeper control over the AI stack will help it win larger workloads, but the bill for building that stack is only getting bigger.
For now, the deal sharpens the company’s pitch: not just more compute, but better use of every chip it buys. The next question is whether customers will pay for that promise fast enough to keep the spending from outrunning the business.
