Nebius teams with Nvidia to build cloud for robotics and physical AI

Nebius, the AI cloud firm, has introduced it’s collaborating with Nvidia to speed up bodily AI improvement with an end-to-end platform purpose-built for the total robotics lifecycle, from simulation and coaching to real-world deployment at scale.

Combining Nebius’s international AI cloud infrastructure with the Nvidia Bodily AI Knowledge Manufacturing unit Blueprint, an open reference structure for enormous knowledge technology and analysis, Nebius will present robotics builders and enterprises an agent-driven surroundings that addresses the 2 elementary boundaries to bodily AI at scale: infrastructure and tooling fragmentation, and the shortage of high-quality coaching knowledge for uncommon, unpredictable situations that decide real-world success.

“Bodily AI goes to be one of many defining know-how shifts of this decade, and the groups constructing it immediately are being held again by infrastructure and tooling that was by no means designed for these workloads, ” stated Evan Helda, head of bodily AI at Nebius.

“Working with Nvidia, we’re constructing the execution layer for the complete bodily AI ecosystem – in order that any workforce, anyplace, can go from thought to deployed robotic on the pace the market calls for.”

“Bodily AI is the subsequent section of computing – the place intelligence is skilled, examined and validated in simulation earlier than it operates in the actual world, ” stated Rev Lebaredian, VP of Omniverse and simulation applied sciences at Nvidia.

“That calls for tightly built-in methods connecting large-scale AI coaching with bodily correct simulation to create a steady knowledge flywheel. By integrating the Nvidia Bodily AI Knowledge Manufacturing unit Blueprint, Nebius is enabling builders to generate physics-grounded artificial knowledge and construct secure, strong autonomous machines at scale.”

Fixing bodily AI’s three-computer downside

Constructing bodily AI at scale means working throughout three distinct environments – large-scale GPU coaching, simulation testing, and edge deployment – every with its personal infrastructure and tooling. Engineering groups routinely spend 30-40 p.c of their time on integration work quite than bettering robotic behaviour.

Actual-world coaching knowledge compounds the problem: it’s costly and harmful to gather, inconsistent throughout firms, and by no means adequate to cowl the long-tail edge circumstances that decide whether or not a robotic succeeds or fails within the subject.

The Nebius cloud answer for bodily AI addresses each challenges. Nvidia OSMO – delivered as an easy-to-consume managed service – supplies unified, agentic orchestration throughout the complete pipeline. Nvidia Cosmos open world basis fashions generate large-scale, physics-consistent artificial knowledge that bridges the hole that real-world assortment can not shut.

The entire stack runs on Nebius AI Cloud – purpose-built infrastructure combining Nvidia RTX PRO 6000 Blackwell Server Version GPUs, high-throughput object storage, built-in knowledge administration and labelling, serverless options and managed inference instantly inside the platform – so groups can eat it as a service, with out having to provision clusters or handle integrations.

Past large-scale simulation and coaching, Nebius extends the robotics lifecycle into manufacturing with serverless and managed inference providers, together with Nebius Token Manufacturing unit, enabling groups to deploy and scale skilled insurance policies with low-latency execution from cloud to edge.

The result’s an entire managed bodily AI runtime, from artificial knowledge technology to real-world inference, delivered by means of a tightly-integrated platform that may be consumed as a service.

Main bodily AI firms constructing the long run with Nebius and Nvidia

RoboForce builds AI robots for unstructured out of doors environments – photo voltaic farms, building websites, agricultural fields – the place encountering uncommon edge circumstances is a every day actuality. Utilizing Nvidia Cosmos open world basis fashions on the Nebius cloud, RoboForce lower pipeline setup time by greater than 70% and considerably accelerated the speed at which new insurance policies attain manufacturing.

“Guide handoffs between knowledge technology, simulation, and coaching means our GPUs can sit idle – costing us each money and time,” stated Calvin Zhou, co-founder of RoboForce. “Utilizing OSMO agentic orchestration, our engineers can push a single configuration file and run the complete pipeline end-to-end.

“We’re producing hundreds of situation variations with Nvidia Cosmos on Nebius AI Cloud, powering our AI knowledge flywheel and accelerating the event of our robotic basis mannequin. This permits us to push hardened robotic fashions straight to the sting and lower our iteration cycles from weeks to days.”

Voxel51, a bodily AI knowledge platform and key know-how associate of Nebius, supplies highly effective knowledge visualization, curation, annotation, and evaluation capabilities for groups to construct high-quality datasets for mannequin coaching and simulations.

By working FiftyOne workflows on Nebius GPU clusters, Voxel51 prospects can curate, increase, and quality-check visible datasets at scale – decreasing the time between knowledge assortment and mannequin deployment.

“Knowledge is the most important determinant of pc imaginative and prescient success. As imaginative and prescient AI methods turn out to be extra succesful, the limiting issue is now not algorithmic innovation however the high quality, protection, and observability of the information used to coach fashions, ” stated Brian Moore, CEO and co-founder of Voxel51.

“Nebius offers our customers the compute infrastructure for working advanced knowledge duties corresponding to auto-labeling and producing novel scenes on the pace and scale wanted by bodily AI methods.”

Along with Nebius cloud for bodily AI and Nvidia applied sciences, Voxel51 is delivering an artificial knowledge technology pipeline for its buyer, Porsche, to speed up autonomous driving knowledge augmentation workflows.

Milestone Programs, a worldwide chief in clever video administration software program and the corporate behind the Hafnia platform for pc imaginative and prescient, chosen Nebius to fine-tune its next-generation Imaginative and prescient-Language Fashions (VLMs).

Milestone curates real-world video footage into compliant, annotated coaching knowledge, then makes use of it to fine-tune Nvidia Cosmos Purpose into extremely correct, use-case particular VLMs. For this computationally intensive work Nebius supplies sustained entry to massive GPU clusters, high-throughput knowledge pipelines, and managed workflow orchestration that retains coaching runs steady and cost-efficient.

“We evaluated a number of cloud suppliers, and Nebius provided the perfect mixture of GPU availability, price-performance, and hands-on engineering help for our bodily AI and VLM coaching workloads, ” stated Edward Mauser, director of Hafnia at Milestone Programs.

“We selected Nebius not only for their tech, but additionally for his or her dedication to knowledge sovereignty – guaranteeing that European prospects’ knowledge can stay inside Europe.”