NVIDIA and Doosan Group Expand Collaboration Across Physical AI, Robotics, and AI Factory Infrastructure

NVIDIA and Doosan Group announced an expanded collaboration on June 7, 2026, covering physical AI systems, industrial robotics, power infrastructure for AI factories, and substrate materials for AI accelerator hardware. The partnership spans four Doosan business units and integrates NVIDIA's full physical AI stack with Doosan's manufacturing, energy, and materials capabilities.

What's New

Robotics (Doosan Robotics): Doosan Robotics is building what it describes as an "Agentic Robot OS"—an AI-powered platform connecting perception, reasoning, simulation, learning, and on-device inference. The stack integrates NVIDIA Isaac Sim (simulation), Isaac Lab (reinforcement learning), Cosmos foundation models (world modeling), Newton physics engine, and Jetson Thor (edge inference). Initial industrial targets are depalletizing and sanding tasks; the roadmap extends to dual-arm and humanoid robot form factors. Doosan Bobcat will separately explore integrating NVIDIA physical AI into construction and agricultural equipment.

AI factory power (Doosan Enerbility): Doosan Enerbility—which produces gas turbines, steam turbines, small modular reactors, and hydrogen fuel-cell systems—is exploring how its large-scale power infrastructure portfolio can support NVIDIA AI factory installations as power demand grows.

PCB materials (Doosan Electro-Materials): Doosan Corporation Electro-Materials produces copper clad laminate (CCL) for printed circuit boards used in networking equipment, AI accelerators, and AI server motherboards, where low signal loss and high reliability are critical.

Context

The announcement coincided with NVIDIA CEO Jensen Huang's Korea visit in early June 2026, which included meetings on sovereign AI infrastructure and robotics manufacturing. Huang has repeatedly identified robotics as the next major AI sector—a framing echoed in the scope of this collaboration.

NVIDIA's physical AI stack has been accumulating components over the past two years: Jetson Thor for edge inference, Cosmos as a world foundation model, Isaac Lab for sim-to-real training, and Newton for physics simulation. The Doosan collaboration applies the full stack across a committed industrial partner spanning multiple verticals simultaneously.

Why It Matters

The scope goes beyond a typical hardware partnership. By covering robots, power generation, and accelerator materials in a single announcement, NVIDIA is extending its presence across the complete physical stack of AI infrastructure—not just the chip layer.

For enterprise robotics teams, the near-term signal is industrial applications: depalletizing and sanding are high-volume, repetitive tasks with clear ROI potential where NVIDIA's simulation-to-production pipeline is mature enough for near-term deployments.

For AI infrastructure planners, the power and materials dimensions of the collaboration reflect a growing recognition that scaling physical AI requires solving energy supply and silicon substrate constraints—not just model quality.