Nebius launches ‘Physical AI Living Lab’ with NVIDIA to help UK and European robotics startups scale

Nebius Group, an AI cloud provider listed on Nasdaq as NBIS, has launched the Physical AI Living Lab with NVIDIA, positioning the effort as a fast-track program for robotics startups across the UK and Europe. The company says the initiative is designed to help smaller teams build and validate “physical AI” applications, which require robotics models to work not just in software simulation but also in physical environments.

Nebius describes physical AI as a workflow that depends on large-scale simulation, synthetic data, and accelerated compute. In its announcement, the company argues that early-stage robotics teams often cannot assemble all three resources on their own, creating a bottleneck between developing an initial model and deploying it in the real world.

The program is set up as a six-month cohort. Participating founders receive access to NVIDIA’s physical AI development tools running on Nebius AI Cloud, along with direct time with Nebius engineers. Nebius said the intent is to give teams “the full NVIDIA physical AI stack” on its infrastructure, so founders can spend more time on building robots and less time assembling the surrounding simulation and compute environment.

Nebius also framed the lab as a feedback loop that benefits both sides. In remarks quoted in the announcement, the company said working closely with startup teams helps Nebius “run physical AI” more effectively, and that it wants to replicate that approach as it adds future cohorts.

The announcement further states that the Physical AI Living Lab builds on an ongoing Nebius-NVIDIA collaboration to create a cloud platform for robotics and physical AI. Nebius said it plans to extend the program to other regions over time and bring in additional cohorts as demand grows.

From a broader industry angle, the Physical AI Living Lab highlights how compute and tooling are becoming strategic for robotics. Training robotics policies and validating them for real-world deployment typically involves iterative simulation, domain randomization, and large datasets. As NVIDIA has increasingly emphasized physical AI development as distinct from purely virtual AI, companies like Nebius are trying to make those stacks easier for startups to access without building their own infrastructure from scratch.

Nebius did not disclose, in the announcement and accompanying coverage, details such as the number of startups selected per cohort, the specific NVIDIA components included, the expected commercialization timeline for participating teams, or whether Nebius will provide any follow-on production or customer deployment support beyond the lab period.

For robotics startups watching the program, the practical question will be whether the lab meaningfully reduces the time and cost to move from simulation to deployment. For Nebius and NVIDIA, what to watch next is the scale of participation, any published outcomes from cohorts, and whether the companies broaden the model across additional European markets or internationally.