Nebius expands physical AI strategy with robotics programme and $643 million Eigen AI acquisition
Images: Nebius
Nebius has launched a new Physical AI Living Lab for British and European robotics start-ups while completing its $643 million acquisition of California-based inference and model optimisation company Eigen AI. The two developments arrived within days of one another and point to a company that is extending its reach beyond cloud computing and into the software and tools used to deploy artificial intelligence in the real world.
On 9 June, Nebius unveiled the Physical AI Living Lab, a six-month programme developed with Nvidia to support robotics companies building autonomous machines and other physical AI applications. A week later, the company announced the completion of its acquisition of Eigen AI, following regulatory approval and the formal closing of the transaction on 10 June.
The announcements come at a time when attention is shifting from the creation of increasingly capable AI models towards the infrastructure required to run them efficiently and deploy them at scale.
From Yandex spin-out to AI infrastructure provider
Nebius emerged from the restructuring of Yandex N.V., the former Dutch parent company of Russian technology group Yandex. Following the divestment of its Russian businesses in 2024, the company rebranded as Nebius Group and refocused on artificial intelligence infrastructure, cloud services and data centres serving international markets.
Since then, the company has positioned itself as a specialist AI cloud provider rather than a general-purpose cloud platform. Its infrastructure is designed specifically for organisations training and deploying AI models.
That strategy has helped Nebius attract significant investor interest during a period of rapid growth in demand for AI computing capacity. Yet the company's latest moves suggest management sees the next opportunity extending beyond access to graphics processing units, or GPUs.
Increasingly, value is being created in the layers that sit above the hardware itself.
Why AI inference is becoming a competitive advantage
Training AI models has dominated industry headlines over the past two years. Running those models efficiently once they are deployed is becoming an equally important challenge.
This stage is known as inference. Every time a user interacts with a chatbot, generates an image or submits a request to an AI application, the model performs inference.
As AI systems move from experimentation into production environments, the economics of inference become increasingly important. Faster and more efficient inference can reduce infrastructure costs while improving performance.
That is the area Eigen AI was created to address.
Founded by researchers from the Massachusetts Institute of Technology's HAN Lab and Computer Science and Artificial Intelligence Laboratory, or CSAIL, Eigen AI built a reputation for improving the performance and efficiency of large AI models. The company's founders include Ryan Hanrui Wang, Wei-Chen Wang and Di Jin, all of whom have contributed to influential research in model optimisation, post-training and inference technologies.
Wei-Chen Wang received the MLSys 2024 Best Paper Award for Activation-Aware Weight Quantisation, or AWQ, a technique that has become widely adopted for running AI models more efficiently. Ryan Hanrui Wang's work on Sparse Attention is among the most cited contributions to high-performance AI computing in recent years.
The acquisition gives Nebius far more than a software product.
It brings a team of researchers whose work already underpins many modern AI deployment techniques. It also establishes a Nebius engineering and research presence in the San Francisco Bay Area, placing the company closer to a significant portion of the global AI ecosystem.
Several industry observers have described the deal as part of Nebius' evolution from a so-called "neocloud" provider that rents computing capacity into a more integrated AI platform combining infrastructure, optimisation and deployment capabilities. That interpretation aligns with Nebius' plans to integrate Eigen's technology directly into its Token Factory managed inference platform.
Building infrastructure for robotics and physical AI
While the Eigen acquisition addresses the software layer of AI deployment, the Physical AI Living Lab focuses on a different challenge.
The programme is designed to support companies developing intelligent machines that interact directly with the physical world.
Physical AI refers to artificial intelligence systems embedded in robots, autonomous vehicles, industrial equipment and other machines capable of sensing, interpreting and responding to real-world environments.
Building such systems requires more than training a model.
Developers must generate synthetic data, create simulations, test behaviour in virtual environments and evaluate performance under a wide range of operating conditions. These activities demand significant computing resources and specialised software tools.
Nebius says the Living Lab will provide participating start-ups with access to Nvidia technologies alongside Nebius cloud infrastructure and engineering support. The first cohort is expected to begin in September. Participants will work with engineers from both companies throughout the six-month programme.
Europe's growing investment in robotics and physical AI
The launch comes as robotics and embodied AI attract increasing attention from investors, researchers and policymakers across Europe.
The continent already possesses many of the ingredients required to build competitive robotics companies. European universities continue to produce influential robotics research, while countries including Germany, the Netherlands, France and Sweden maintain strong industrial engineering traditions.
The challenge has often been commercialisation and scale.
Training modern AI systems requires substantial computing resources. For early-stage robotics companies, access to those resources can become a significant barrier to development.
That issue has become more pressing as robotics developers incorporate increasingly sophisticated AI models into their systems.
European policymakers have also become more vocal about the strategic importance of AI infrastructure, semiconductors and advanced computing capabilities. Physical AI sits at the intersection of all three.
Against that backdrop, initiatives such as the Physical AI Living Lab represent an effort to strengthen the supporting infrastructure available to European robotics companies rather than focusing solely on individual start-ups.
Nvidia's expanding role
The programme also highlights Nvidia's growing influence across the AI ecosystem.
The company remains the dominant supplier of GPUs used for AI training and deployment, but its ambitions now extend well beyond hardware. Nvidia has built software frameworks, simulation platforms and robotics development environments that are increasingly embedded in AI workflows.
For robotics developers, access to these tools can simplify development and testing. For Nvidia, partnerships with infrastructure providers such as Nebius create additional pathways for adoption.
At the same time, Nvidia's dominance underscores the concentration that exists within parts of the AI industry. Many infrastructure providers remain heavily dependent on Nvidia hardware and software, limiting the number of alternative ecosystems available to developers.
That dependence has become an increasingly important topic as governments and companies seek greater technological resilience.
Beyond renting compute
Viewed separately, the Physical AI Living Lab and the Eigen AI acquisition address different parts of the AI stack.
One is aimed at robotics developers building physical systems. The other focuses on making AI models faster and more efficient once they are deployed.
Together, they provide a clearer picture of where Nebius intends to compete.
The company continues to invest in computing infrastructure, but it is also expanding into optimisation, deployment and developer tooling. Those capabilities move Nebius closer to becoming a full-stack AI platform rather than a provider of raw computing capacity alone.
The first Physical AI Living Lab cohort is scheduled to begin in September. By then, Eigen AI's optimisation technology will already be integrated into the Nebius platform.
Both developments reflect a common challenge facing the industry. As artificial intelligence moves beyond chat interfaces and into robotics, autonomous systems, factories and vehicles, the infrastructure required to train, optimise and deploy AI is becoming just as important as the models themselves.
Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.