Insider Brief
- Nebius is partnering with Nvidia to deliver a cloud-based platform that unifies simulation, training and deployment for physical AI, targeting fragmented infrastructure and limited access to high-quality edge-case data.
- The platform combines Nebius AI cloud with Nvidia’s Physical AI Data Factory Blueprint, using managed orchestration and synthetic data generation to streamline robotics development and reduce engineering overhead across the pipeline.
- Early adopters including RoboForce, Voxel51 and Milestone Systems are using the stack to accelerate deployment cycles, scale data workflows and train advanced vision and robotics models, with the service now available in U.S. and European data centers.
Nebius is partnering with NVIDIA to build a cloud-based platform aimed at streamlining the full lifecycle of physical AI, from simulation and training to deployment.
Nebius said the offering combines Nebius’s AI cloud infrastructure with Nvidia’s Physical AI Data Factory Blueprint, an open architecture designed to generate and evaluate large volumes of robotics data. The goal is to address two persistent bottlenecks in robotics development: fragmented infrastructure across training, simulation and deployment environments, and limited access to high-quality data for rare real-world scenarios.
“Physical AI is going to be one of the defining technology shifts of this decade, and the teams building it today are being held back by infrastructure and tooling that was never designed for those workloads,” Head of Physical AI at NebiusEvan Helda said in the announcement. “Working with Nvidia, we are building the execution layer for the entire physical AI ecosystem — so that any team, anywhere, can go from idea to deployed robot at the speed the market demands.”
The Amsterdam-based company said the platform is designed to be consumed as a service, allowing teams to avoid provisioning hardware or managing distributed systems. The platform is available across Nebius data centers in the United States and Europe.
Nebius said the platform is an end-to-end environment where robotics teams can move from model development to production without stitching together separate systems. The stack integrates GPU-based training, physics-based simulation and edge deployment into a single workflow, an area where companies often spend significant engineering time on integration rather than performance improvements.
At the core of the system is Nvidia’s orchestration layer, delivered as a managed service, which coordinates workloads across the pipeline, Nebius indicated. The platform also incorporates Nvidia’s world models to generate synthetic data designed to replicate edge cases that are difficult, costly or unsafe to capture in the real world. Nebius said this synthetic data capability is intended to complement real-world datasets and improve model robustness.
The Nasdaq listed company is also extending the platform into production through managed inference and deployment services, enabling models trained in the cloud to be executed on edge systems with low latency.
What Companies are Using Nebius and Nvidia?
- RoboForce develops AI robots for unstructured outdoor environments such as construction, agriculture and energy sites, where handling edge cases is critical. Using Nvidia Cosmos models on Nebius cloud, the company reduced pipeline setup time by more than 70% and accelerated the pace of deploying new robot policies.
- Voxel51 provides data visualization, curation and annotation tools that help teams build high-quality datasets for physical AI and computer vision systems. Running its FiftyOne workflows on Nebius GPU infrastructure enables customers to process, validate and augment data at scale, shortening the path from data collection to deployment, including work with Porsche Engineering on synthetic data pipelines for autonomous driving.
- Milestone Systems is using Nebius infrastructure to train and fine-tune next-generation vision-language models based on curated video data. The company relies on Nebius for access to large GPU clusters and managed workflows to support stable, cost-efficient training of specialized AI models for real-world applications.
Nvidia’s VP of Omniverse and simulation technologies Rev Lebaredian said the next phase of computing is physical AI where systems will be “trained, tested and validated in the real world.”
“By integrating the Nvidia Physical AI Data Factory Blueprint, Nebius is enabling developers to generate physics-grounded synthetic data and build safe, robust autonomous machines at scale,” Lebaredian said.
