Nebius rolls out “Physical AI Living Lab” to help UK and European robotics startups build real-world systems using NVIDIA tools

Nebius, an AI cloud company, said it is launching a “Physical AI Living Lab” aimed at robotics startups in the United Kingdom and across Europe. The program is set to run for six months and is meant to help small teams develop and test physical AI systems by combining NVIDIA development tools with Nebius’s cloud infrastructure.

According to the announcement, the core problem the Living Lab targets is not building a basic machine learning model, but bridging the gap between software-like testing and safe, reliable performance in the physical world. Physical AI, Nebius and NVIDIA contend, depends on large-scale simulation, synthetic data, and accelerated compute, resources that many early-stage robotics companies cannot easily assemble on their own.

The lab is described as a way to remove that bottleneck by providing the “same class” of tooling and computing used to build physical AI at scale. The intent is to let founders move from simulation to real-world deployment faster, rather than spending months assembling infrastructure or iterating with limited training resources.

Nebius said the Physical AI Living Lab is built on the company’s collaboration with NVIDIA to create a cloud platform for robotics and physical AI development. The announcement also frames the effort as extendable, saying Nebius and NVIDIA intend to roll the program out to other regions over time and add new startup cohorts as it grows.

For robotics startups, the practical value of that access is in the training loop. Simulation and synthetic data are commonly used to generate large volumes of varied environments and scenarios without the cost and risk of repeated physical trials. Accelerated compute, meanwhile, is what turns that data and simulation workload into trainable models quickly enough to support rapid experimentation.

The announcement does not provide details on eligibility criteria, geographic coverage by country within Europe, or how many startup teams will be selected for the first cohort. It also does not specify what exact NVIDIA software and hardware components the lab will include, beyond referring generally to NVIDIA’s physical AI development tools and compute.

Nebius has previously emphasized building and expanding AI infrastructure outside the lab setting through partnerships and capacity agreements, suggesting it views cloud delivery as a competitive advantage. In that context, the Living Lab can be read as both a product and a pipeline, offering startups a fast route to production readiness while building familiarity with Nebius’s platform among the next generation of robotics developers.

Still, key questions remain unanswered in the announcement. It does not outline what, if any, commercial terms startups face during the program, whether Nebius plans to charge for compute beyond a trial allocation, or how outcomes will be measured. There is also no disclosed timeline for broader European rollout milestones beyond an intention to extend the initiative over time.

Investors and industry watchers will likely focus next on cohort announcements, the specific technical stack Nebius and NVIDIA provide, and whether the program results in pilots that move beyond early prototypes. If Nebius publishes technical summaries or case studies, those could clarify how physical AI development workflows change when teams start with large-scale simulation and compute instead of cobbling together limited resources.