Nebius acquires Eigen AI for $643M as inference optimisation becomes AI infrastructure’s most valuable layer

Nebius, an AI cloud company, has agreed to acquire Eigen AI, a 20-person MIT spinout focused on inference and model optimization, in a deal valued at approximately $643 million. Eigen AI’s technology specializes in maximizing the efficiency of tokens processed per GPU, directly improving throughput and reducing costs for AI model inference. The acquisition is a strategic move by Nebius to enhance its Token Factory, a managed inference platform, by integrating Eigen AI’s advanced optimization stack with Nebius’s growing global compute capacity.

Eigen AI’s founding team, renowned for their influential contributions to industry-standard optimization techniques, will join Nebius to establish an engineering and research presence in the San Francisco Bay Area. This integration is poised to position Nebius at the forefront of AI inference infrastructure, enabling customers to achieve superior performance and cost-efficiency from the same underlying hardware.

Nebius is part of the emerging “neocloud” segment, which offers AI-optimized cloud infrastructure with faster deployment and lower overhead compared to traditional hyperscalers like AWS, Microsoft Azure, and Google Cloud. Nebius recently tripled its Nvidia GPU capacity, deploying Nvidia’s latest H200 chips, and has invested heavily in data centers across Europe. By acquiring Eigen AI, Nebius aims to differentiate itself in the competitive AI cloud market by enhancing inference speed and lowering per-token prices, critical factors as demand for inference-optimized AI infrastructure surges globally.