European AI infrastructure company Nebius has acquired US based startup Eigen AI in a deal worth approximately $643 million, combining advanced compute infrastructure with cutting edge inference optimisation technology. The move reflects growing competition in the AI infrastructure space, where performance, efficiency, and cost are becoming critical differentiators.
Expanding AI Infrastructure Capabilities
Headquartered in Amsterdam, Nebius operates large scale data centres equipped with high performance GPUs, providing compute power to enterprises and AI companies. The company also offers specialised software tools that allow customers to deploy and manage AI applications more efficiently.
Nebius has already secured major contracts with global technology leaders such as Meta and Microsoft, highlighting its role as a key provider of AI infrastructure.
The acquisition of Eigen AI is expected to strengthen Nebius’s capabilities by enhancing how AI models are executed in real world environments.
Focus on AI Inference
A central focus of the deal is improving AI inference, the stage where trained models are applied to real data to generate outputs. Inference is becoming the fastest growing segment of AI workloads and is projected to account for a significant share of global compute demand.
Eigen AI specialises in optimising this process, helping models operate more efficiently by improving how they process data. Its technology increases the yield from tokens, the fundamental units of data used by AI systems, enabling better performance while reducing computational costs.
By lowering the cost per inference, the technology makes AI applications more accessible and scalable for enterprise users.
Integrating Optimisation into the Stack
Nebius plans to integrate Eigen AI’s optimisation layer directly into its platform, particularly within its Token Factory system. This integration is designed to address common bottlenecks in AI inference, including challenges related to memory, routing, and compute efficiency.
The result is expected to be higher throughput and improved performance without requiring additional engineering effort from customers. This streamlined approach allows organisations to deploy AI models more quickly and adapt to new technologies with greater ease.
Acquiring Talent and Expertise
In addition to the technology, the acquisition brings a team of around 20 researchers and engineers into Nebius. The Eigen AI team is recognised for its expertise in model efficiency and inference optimisation, making it a valuable addition to Nebius’s capabilities.
Co founders Ryan Hanrui Wang and Wei Chen Wang are alumni of the HAN Lab at MIT, led by Song Han, a prominent figure in AI computing research.
The team will establish a research and engineering presence in the San Francisco Bay Area, strengthening Nebius’s footprint in one of the world’s leading technology hubs.
Driving Efficiency and Adoption
By combining infrastructure and optimisation, Nebius aims to deliver a more efficient AI stack for its customers. The integration of Eigen AI’s technology is expected to reduce costs, accelerate deployment timelines, and improve overall system performance.
For enterprises, this means faster time to production and the ability to adopt new AI models more quickly. It also enhances unit economics, making large scale AI deployments more viable.
Positioning in a Competitive Market
The acquisition comes as competition intensifies in the AI infrastructure market, with companies seeking to differentiate through performance and cost efficiency. As demand for AI continues to grow, the ability to optimise inference will play a key role in determining market leaders.
With the addition of Eigen AI’s technology and talent, Nebius is positioning itself to address these challenges and strengthen its offering across the AI lifecycle.
By integrating advanced optimisation into its infrastructure, the company is aiming to deliver a more powerful and cost effective platform for the next generation of AI applications.