Nebius to acquire Eigen AI in $643m cash-and-stock deal

Netherlands-based AI unreality institution Nebius has agreed to get Eigen AI, a steadfast specialising successful inference and exemplary optimisation company, successful a currency and banal transaction valued astatine astir $643m.

With this acquisition, Nebius plans to integrate Eigen AI’s inference and post-training optimisation technologies straight into Nebius Token Factory, its managed AI accumulation platform.

This integration aims to enactment enterprise-grade deployment, including autoscaling endpoints and fine-tuning for a wide scope of open-source models.

Nebius, which is listed connected Nasdaq, and Eigen AI person antecedently collaborated connected implementations that received precocious show rankings successful Artificial Analysis assessments.

The determination besides marks Nebius’s enlargement into the US, with Eigen AI’s founding researchers joining the erstwhile to found an engineering and probe hub successful the San Francisco Bay Area.

Key members of Eigen AI’s squad see co-founder Ryan Hanrui Wang, known for probe connected Sparse Attention, and co-founder Wei-Chen Wang, who received designation for enactment connected Activation-aware Weight Quantisation. Co-founder Di Jin has contributed to Meta’s ample connection models and reinforcement learning frameworks.

The founding squad antecedently undertook probe astatine the Massachusetts Institute of Technology’s HAN Lab and CSAIL.

Wang said: “We’re arrogant to articulation Nebius and enactment alongside the Token Factory squad to propulsion the boundaries of inference performance.

“Together, we are removing the friction of AI exemplary customisation and deployment truthful developers tin tally models reliably successful accumulation without managing the underlying infrastructure.”

Industry trends bespeak that inference presently represents the fastest-growing conception of AI, with forecasts estimating it volition relationship for astir two-thirds of compute request this year.

The expanding adoption of open-source models successful accumulation environments has made the optimisation of inference a priority, peculiarly arsenic newer exemplary architectures present further demands connected representation and compute resources.

Eigen AI’s exertion aims to code optimisation needs crossed the full exemplary lifecycle, from post-training and fine-tuning to accumulation deployment. It supports a wide scope of fashionable open-source models, including GPT-OSS, Qwen, Gemma, Nemotron, Llama, GLM, DeepSeek, Kimi, and MiniMax.

The integration of Eigen AI’s stack into Nebius Token Factory is intended to amended hardware ratio and throughput, portion reducing the operational overhead for customers.

Nebius co-founder and main concern serviceman Roman Chernin said: “The integration of Eigen AI’s optimisation capabilities and founding squad volition found Nebius Token Factory astatine the frontier of inference, offering customers market-leading exemplary show and portion economics with monolithic compute capableness to backmost it astatine scale.”