Nebius, the AI cloud infrastructure company listed on Nasdaq, has launched the Physical AI Living Lab, a six-month program giving robotics startups in the UK and Europe access to NVIDIA’s physical AI software tools alongside Nebius’s cloud-based AI computing platform. The initiative is designed to address the compute, simulation, and synthetic data bottlenecks that prevent early-stage robotics companies from advancing their physical AI applications at the pace their development timelines require.
Applications will be reviewed through the NVIDIA Inception program, with the first group of selected startups expected to begin in September 2026. The initial phase will run on Nebius’s UK infrastructure, powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
What Participants Receive
Startups accepted into the program will have access to a range of NVIDIA technologies: OSMO for workload management, Cosmos world foundation models for environment simulation, Isaac Sim and Isaac Lab for robot simulation, testing, and policy training. Synthetic data generation capabilities will be delivered through Voxel51’s FiftyOne platform, integrated with Cosmos world foundation models.
Engineering teams from both Nebius and NVIDIA will work directly with participating startups throughout the program, providing technical guidance alongside the compute and software access.
“Most robotics teams can build a strong model – the bottleneck is getting the simulation, synthetic data, and compute in place to take it further,” said Evan Helda, Head of Physical AI at Nebius.
The Problem Being Addressed
Large-scale simulation and synthetic data generation are prerequisites for training physical AI systems at the capability levels that commercial deployment requires. Building that infrastructure independently is prohibitively expensive for early-stage startups – the GPU clusters required for meaningful simulation workloads represent capital expenditure that most seed or Series A companies cannot justify before they have commercial revenue.
The Living Lab addresses this by providing temporary access to the infrastructure on a program basis, allowing startups to run the experiments and generate the training data that determine whether their underlying approach is commercially viable – before requiring them to build or lease that infrastructure permanently.
Anthony Hills, Director for UK and Ireland at NVIDIA, said the initiative bridges the gap between innovation in the UK robotics sector and the development of commercially viable physical AI solutions, by making large-scale cloud training resources and NVIDIA simulation technologies more affordable for early-stage companies.
Expansion Plans
The Physical AI Living Lab builds on an existing partnership between Nebius and NVIDIA focused on cloud infrastructure for robotics workloads. Both companies indicated plans to run additional cohorts and expand the program into other geographic markets. The geographic starting point – UK and Europe – aligns with the broader European physical AI ecosystem development effort reflected in Google DeepMind’s concurrent European Robotics Accelerator, also launched this week.
