Nebius celebrates startups from model-builders to meal-assembly robots at 2026 Physical AI Awards
November 18, 2026
4 mins to read
Nebius named the winners of the second annual Nebius Physical AI Awards, a showcase of the leading startups taking physical AI from technical experimentation to real-world deployment.
The 2026 winners are Archetype AI, Outsight, Luminary, Chef Robotics, and LanceDB — startups putting physical AI to work in airports, food factories, aerospace design, and the data pipelines behind them.
In its second year, the program expanded beyond physical AI applications to cover the full stack, from models and perception to simulation, deployed systems, and orchestration.
“Demand for physical AI is growing across industries, and startups are the ones building it, ” said Evan Helda, Head of Physical AI at Nebius. “Getting it into the field takes progress at every layer: models, perception, simulation and data, deployed systems, and the tools to run them. These awards recognize the teams advancing each of those layers, while giving them Nebius AI Cloud for Physical AI and our Workbench framework, so we can work together on what deployment at scale takes.”
The 2026 applicant pool of 363 startups shows physical AI maturing in both capital and capability. Compared with the 2025 awards, nearly three times as many applicants had raised $10M to $25M, and submissions from startups with $25M+ raised more than doubled.
Overall, judges observed that physical AI startups are increasingly focusing on specific industries, from manufacturing and logistics to construction, food production, and agriculture. Among established applicants, the strongest concentration sat in systems and deployment, with a focus on reliability, fleet orchestration, edge inference, and integration with existing operations. More teams also collect, clean, and test real-world robot data, and many now treat simulation as core engineering that shortens development cycles.
Representatives from Nebius and NVIDIA teamed up with operators and investors building the category, including Accel, Lightspeed, Northzone, Voxel51, Foxglove and Encord, to review submissions from startups with at least $1M raised across North America, EMEA, and APJ. Judges scored each application on technical innovation, real-world implementation, market potential, and responsible design, criteria developed to match what production demands.
Each winning startup receives $150,000 in Nebius AI Cloud compute credits, accelerated by NVIDIA AI Infrastructure. To accelerate their growth and real-world production speed, category winners also receive technical support through hands-on collaboration with Nebius and NVIDIA engineers, executive mentorship, and introductions across the NVIDIA and investor ecosystem. They will use agents to build their physical AI workflows on Nebius Physical AI Workbench, an open-source framework for composing and running physical AI development workflows on Nebius.
Nebius congratulates all finalists and the following winning physical AI startups by category:
Physical AI Models:Archetype AI's Newton foundation model learns from sensor data to detect, predict, and act on machine behavior.
Perception & Spatial Intelligence:Outsight's spatial AI software tracks people and vehicles in real time using 3D LiDAR, capturing no personal data.
Simulation & Synthetic Data:Luminary gives engineering teams physics AI models that predict how designs behave in seconds instead of hours.
Systems & Deployment:Chef Robotics puts AI-powered meal assembly robots on food manufacturing lines across the US, Canada, and Europe.
Software, Tooling & Orchestration:LanceDB's multimodal lakehouse speeds the data curation and feature engineering behind large-scale AI model training.
Physical AI is still in its early development, and the startups building it today will shape how the field matures. Nebius builds alongside them, from the cloud they train on to the tools they ship with.
As a Nebius for Startups initiative, the second annual Nebius Physical AI Awards extends the organization’s standing commitment to help AI-native companies build faster at every stage, with credits, connections and engineering support from first experiments through serving production workloads.
NVIDIA and Nebius are expanding the AI Builder Program into Europe with $10 million in compute grants across Nebius AI Cloud and Token Factory, founder funding opportunities through Accel and our broader European VC network, and hands-on community programs. Applications are open.
Nebius Academy and NVIDIA are launching an education collaboration at GTC Berlin, featuring: NVIDIA Deep Learning Institute (DLI) open-source courses hosted on Nebius Academy starting with Fundamentals of Deep Learning, co-developed courses in agentic and physical AI that run on Nebius, and Nebius GPU compute behind NVIDIA DLI’s sold-out full-day workshop at GTC.
Customers using Nebius Physical AI Workbench report up to 40% less engineering time with AI agents operating robotics workflows directly. Built in collaboration with NVIDIA and open source on GitHub, it runs data generation, simulation, training, and evaluation in one place. This release adds NVIDIA Cosmos Data Factory and a sim-to-real workflow co-designed with Sereact for Zalando returns.