bytevyte
bytevyte
Language
EN DE ES FR PT IT
ai-beats —

Nebius Strengthens AI Infrastructure with $643 Million Acquisition of Eigen AI

AI inference optimization

Nebius has entered into a definitive agreement to acquire Eigen AI for approximately $643 million, a move aimed at enhancing its full-stack AI inference optimization capabilities. The transaction involves a combination of $98 million in cash and 3.8 million shares of Nebius stock. This acquisition, announced this week, signals a significant expansion for the Dutch-based technology firm as it seeks to scale its managed inference services for open-source models.

The deal integrates Eigen AI’s specialized optimization stack into the Token Factory platform, which Nebius launched earlier this year. By combining these technologies, the company aims to maximize token throughput per GPU, addressing the growing demand for efficient large-scale model deployment. The integration focuses on several advanced techniques, including Activation-aware Weight Quantization (AWQ), sparse attention, and custom CUDA kernels.

Strategic Integration and Performance Gains

The technical synergy between the two entities has already demonstrated measurable results in AI inference optimization. Collaborative efforts have produced optimized versions of prominent open-source models such as Llama, DeepSeek, and Qwen. These versions have reached output speeds of up to 911 tokens per second, placing them at the top of industry performance benchmarks. This level of efficiency is critical for enterprises looking to reduce the latency and cost of running high-performance AI applications.

Beyond software integration, the acquisition brings elite research talent to the Nebius team. The founders of Eigen AI are alumni of MIT’s HAN Lab and are recognized for their contributions to model efficiency. Ryan Hanrui Wang is a specialist in sparse attention, while Wei-Chen Wang is the developer of the AWQ method. Additionally, Di Jin has extensive experience in the post-training processes for Meta’s Llama 3 and Llama 4 models.

Expanding Global Engineering Presence

As part of the agreement, the Eigen AI leadership team will establish a new engineering hub for Nebius in the San Francisco Bay Area. This expansion provides the company with a direct presence in a primary global center for AI development, facilitating closer collaboration with the broader research community. The new office will focus on further refining the Token Factory managed inference platform and developing new post-training quantization methods.

The acquisition is expected to close within the coming weeks, subject to standard regulatory approvals. For Nebius, which originated from the international assets of Yandex, this investment represents a clear commitment to becoming a dominant player in the AI infrastructure market. By securing proprietary optimization techniques and top-tier talent, the company is positioning itself to compete directly with established cloud providers in the specialized field of AI model serving.

While we strive for accuracy, bytevyte can make mistakes. Users are advised to verify all information independently. We accept no liability for errors or omissions.

✔Human Verified

Share