Key Highlights
- Nebius (NBIS) has entered into a definitive agreement to purchase Eigen AI, a leading inference optimization firm, for roughly $643 million through a combination of cash and Class A shares.
- The acquired technology will enhance Nebius Token Factory, the company’s enterprise-focused managed inference solution.
- MIT HAN Lab alumni forming Eigen’s core team will launch Nebius’s inaugural Bay Area engineering center.
- Collaborative optimization work between both organizations has already achieved top rankings on Artificial Analysis performance benchmarks.
- NBIS shares climbed 8.51% following the announcement, reaching $150.00, recovering from a 6.07% weekly decline.
On May 1, 2026, Nebius (NBIS) revealed its intention to purchase Eigen AI for approximately $643 million. The acquisition will be structured as a combination of cash and Nebius Class A shares, with the stock component valued according to the company’s 30-day volume-weighted average price at the time of signing. Shares of NBIS surged 8.51% following the announcement, reaching $150.00.
The deal is anticipated to finalize in the coming weeks, subject to regulatory approval and customary closing requirements.
Eigen AI specializes in inference acceleration and model optimization. The company enables AI development teams to deploy open-source models with superior speed and cost efficiency in live environments, eliminating the need for internal optimization infrastructure.
Nebius intends to incorporate Eigen AI’s technology into Token Factory, its comprehensive platform offering autoscaling API endpoints and fine-tuning capabilities for prominent open-source models such as Llama, DeepSeek, Qwen, Gemma, and additional frameworks.
The partnership between the two organizations predates this acquisition. Prior to the deal announcement, they collaborated on optimized implementations that achieved leading positions on Artificial Analysis, a prominent benchmarking resource in the AI community.
Eigen AI’s Technical Capabilities and Team
Eigen AI emerged from MIT’s HAN Lab, with co-founders Ryan Hanrui Wang and Wei-Chen Wang bringing breakthrough research in AI deployment optimization.
Ryan developed Sparse Attention (SpAtten), which has become the most frequently cited HPCA paper published since 2020. Wei-Chen created Activation-aware Weight Quantization (AWQ), which earned the MLSys 2024 Best Paper Award and has become the industry standard for 4-bit model deployment.
Co-founder Di Jin earned his PhD from MIT CSAIL and played a key role in developing Meta’s Llama 3 and Llama 4 post-training methodologies. His work also includes co-authoring the CGPO reinforcement learning from human feedback system.
Upon completion of the transaction, the team will establish operations in the San Francisco Bay Area, marking Nebius’s first engineering and research facility in the United States.
The Growing Inference Computing Landscape
Inference has emerged as the most rapidly expanding segment within AI computing infrastructure. Current projections indicate it will account for approximately two-thirds of overall AI compute requirements in 2026.
Efficient inference execution presents significant technical hurdles. It requires expertise in model representation, GPU kernel optimization, and dynamic workload management — capabilities that most organizations lack internally.
Open-source models present additional complexity, as they generally lack optimization out of the box. Advanced architectures including Mixture-of-Experts and Compressed Sparse Attention create specific challenges related to memory utilization and computational efficiency that demand specialized expertise.
Eigen AI delivers comprehensive optimization spanning post-training, fine-tuning, and production inference for all leading open-source models. The company’s kernel-level and model-level optimization techniques are engineered to maximize hardware performance without requiring additional engineering resources.
