Nebius Agrees to Acquire Eigen AI, Strengthening Nebius Token Factory as a Frontier Inference Platform

Nebius, the ultimate cloud for AI innovators, led by Arkady Volozh, and the team, has agreed to acquire Eigen AI, a leading inference and model optimization company, led by Ryan Wang, Wei-Chen Wang, Di Jin, and the team.

The acquisition strengthens Nebius Token Factory by combining its optimization stack with Nebius’s global compute and AI cloud, while adding advanced inference research talent.

Following the closing, Eigen AI’s inference and optimization layers will be integrated into Nebius Token Factory, strengthening its autoscaling and fine-tuning capabilities. Joint implementations have already ranked among the fastest on Artificial Analysis.

The acquisition also accelerates Nebius’s US expansion, with Eigen AI’s founding team joining to establish an engineering and research presence in the San Francisco Bay Area.

Roman Chernin, co-founder and Chief Business Officer of Nebius, said:

“We are operating in a capacity-constrained environment where AI builders require both optimized inference and infrastructure at scale. The integration of Eigen AI’s capabilities and team positions Nebius Token Factory at the forefront of inference, delivering strong model performance and unit economics, supported by large-scale compute capacity.”

Eigen AI’s founding team brings deep expertise from research shaping modern inference deployment. Co-founders Ryan Hanrui Wang and Wei-Chen Wang are alumni of MIT HAN Lab, led by Song Han, a pioneer in AI computing and model efficiency.

Ryan Hanrui Wang’s Sparse Attention (SpAtten) work is the most-cited HPCA paper since 2020, while Wei-Chen Wang received the MLSys 2024 Best Paper Award for Activation-aware Weight Quantization (AWQ), now a standard for 4-bit model serving in production.

Co-founder Di Jin, an MIT CSAIL PhD, brings expertise in post-training and large-scale model alignment, with contributions to Llama 3 and Llama 4, and co-authorship of the CGPO RLHF framework.

Ryan Hanrui Wang, co-founder and CEO of Eigen AI, said:

“We’re joining Nebius to push the boundaries of inference performance alongside the Token Factory team. With a strong engineering culture and a world-class AI cloud, the combined platform is designed to remove friction in model customization and deployment, enabling reliable production use without the need to manage underlying infrastructure.”

Inference is now the fastest-growing segment of AI, expected to account for roughly two-thirds of compute demand this year, alongside rising adoption of open-source models. As more workloads move into production, the system optimization layer is emerging as critical infrastructure.

Efficient production inference is inherently complex, requiring deep expertise across the full execution stack, from model representation and GPU kernel execution to real-time workload scheduling.

Open-source models often ship unoptimized, while newer architectures like MoE and long-context models add complexity across memory and compute, challenges most teams cannot handle in-house.

Eigen AI addresses this with full-stack optimization across the model lifecycle, from post-training to production inference, supporting major open-source models.

By integrating Eigen AI’s optimization layer into Nebius Token Factory, this bottleneck is removed across the lifecycle. System-, model-, and kernel-level techniques are designed to extract higher performance from hardware out of the box, delivering greater throughput and lower cost per inference without added engineering overhead.

As a result, Nebius Token Factory customers benefit from faster time to production, improved unit economics, and quicker adoption of new models, while existing Eigen AI customers gain access to Nebius’s global AI infrastructure and platform capabilities.

The transaction consideration will be paid through a mix of cash and Nebius Class A shares, valued at approximately $643 million at signing based on Nebius’s 30-day weighted average stock price, subject to adjustments. Closing is expected in the coming weeks, pending customary conditions including antitrust clearance.