Nebius has agreed to acquire Eigen AI in a move designed to strengthen Nebius Token Factory, its managed inference platform for production artificial intelligence workloads. Announced from Amsterdam on May 1, 2026, the agreement brings together Nebius’s global AI cloud capacity with Eigen AI’s model optimization and inference software, creating a combined offering aimed at customers deploying models at scale. The transaction also reflects a broader shift in the AI infrastructure market, where performance after training has become as commercially important as access to raw compute.
Strategic Focus on Production Inference
The acquisition is centered on inference, the stage where trained AI models are used to generate outputs for real applications, agents, and enterprise services. Nebius said Eigen AI’s inference and post-training optimization layers will be integrated directly into Token Factory, which provides autoscaling endpoints and fine-tuning pipelines for major open-source models. The companies have already collaborated on optimized implementations that ranked among the fastest on Artificial Analysis, giving Nebius a performance-based rationale for bringing Eigen AI’s software into its platform.
The Technology and Team Behind the Deal
Eigen AI brings a full-stack approach that spans post-training, fine-tuning, model-level tuning, GPU kernel work, and production inference optimization across model families including Llama, Qwen, DeepSeek, Gemma, Nemotron, GLM, Kimi, MiniMax, and GPT-OSS. Its founding team includes Ryan Hanrui Wang and Wei-Chen Wang, alumni of MIT’s HAN Lab, along with Di Jin, an MIT CSAIL PhD whose background includes post-training and large-scale model alignment. Nebius said the team will help establish the company’s engineering and research presence in the San Francisco Bay Area, adding US-based technical depth to its existing AI research and development capabilities.
Market Context and Deal Terms
The transaction is valued at approximately $643 million, payable through a combination of cash and Nebius Class A shares, with the final amount subject to customary adjustments. Nebius expects the deal to close in the coming weeks, pending customary conditions including antitrust clearance, and existing Eigen AI customers are expected to gain access to Nebius’s wider infrastructure and platform capabilities. Media coverage framed the purchase as a major bet on the inference layer, with Investors Business Daily noting that AI cloud providers are racing to make Nvidia-based infrastructure more efficient for model builders and application developers.
Industry Significance
The deal lands as inference demand rises with the spread of production AI systems, where every improvement in speed, utilization, and cost per output can affect margins. Nebius argues that open-source models often require substantial optimization before they can run efficiently in production, especially as newer architectures introduce memory, routing, scheduling, and long-context challenges. By embedding Eigen AI’s optimization stack into Token Factory, the company is trying to reduce that burden for customers that want flexibility from open-source models without building specialized infrastructure teams.
For Nebius, acquiring Eigen AI is less about adding another product line and more about deepening the technical layer that determines inference speed, cost, reliability, and developer experience. The deal could give Token Factory customers improved throughput, stronger unit economics, and faster access to optimized open-source models without requiring them to manage the underlying infrastructure themselves. If completed as planned, the acquisition would mark a significant step in Nebius’s effort to build a full-stack AI cloud platform for production-grade AI applications and expand its presence in the United States.