NVIDIA and Doosan Group Partner on Physical AI Infrastructure

🔑 Enhanced Key Takeaways

•The partnership aims to train NVIDIA's physical AI technology with Doosan's extensive industrial data to create customized Foundation Models (FMs) optimized for Doosan's specific business areas like construction machinery, power generation, and robotics.
•Doosan Robotics is integrating NVIDIA Isaac Sim and Isaac Lab open robotics frameworks, NVIDIA Cosmos open world foundation models, the open-source Newton physics engine, and NVIDIA Jetson Thor to enhance its Agentic Robot OS.
•Doosan Enerbility will leverage NVIDIA's platforms to develop AI-driven solutions for power generation, including optimizing gas turbines and small modular reactors (SMRs) for the energy demands of AI data centers.
•Doosan Corporation Electro-Materials BG, a supplier of high-speed copper clad laminates (CCL) for AI accelerators, will collaborate on materials for AI data centers, including developing high-performance copper foil to ensure signal integrity and performance stability.
•The collaboration includes a roadmap for Doosan Robotics to unveil intelligent robot solutions powered by the Agentic Operating System in 2027 and industrial humanoid robots in 2028.

🛠️ Technical Deep Dive

NVIDIA's Core Platforms: The collaboration utilizes NVIDIA's full-stack accelerated computing platforms, including the NVIDIA DSX AI factory platform, NVIDIA MGX, and NVIDIA's physical AI stack.
Robotics Integration: Doosan Robotics is integrating NVIDIA Isaac Sim and Isaac Lab (open robotics frameworks), NVIDIA Cosmos (open world foundation models), the open-source Newton physics engine, and NVIDIA Jetson Thor.
Jetson Thor Specifications: NVIDIA Jetson Thor is powered by an NVIDIA Blackwell GPU and 128 GB of memory, delivering up to 2070 FP4 TFLOPS of AI compute within a 130W power envelope. It offers up to 7.5x higher AI compute and 3.5x better energy efficiency compared to Jetson AGX Orin.
Agentic Robot OS: Doosan Robotics is developing an Agentic Robot OS, an AI-powered platform designed to connect perception, reasoning, simulation, learning, and on-device inference, enabling robots to better perceive, reason, and act in complex and dynamic environments.
Simulation-to-Real Workflows: The partnership will employ simulation-to-real workflows, physics calibration, and AI reasoning to enhance the adaptability and task specialization of collaborative robots.
Doosan Enerbility's AI Solutions: Doosan Enerbility applies AI-based non-destructive testing (NDT) solutions to analyze radiographic testing (RT) images for defect identification and utilizes Digital Twin technology with IoT and AI for performance monitoring and anomaly detection in power plants.
Doosan Electro-Materials BG Contributions: This unit focuses on high-speed copper clad laminate (CCL) solutions, which are critical core materials for multilayer PCB structures in AI accelerators, ensuring signal integrity. They are also developing hyper very low profile (HVLP) copper foil to minimize signal loss in high-speed data transmission for AI networks.

🔮 Future ImplicationsAI analysis grounded in cited sources

Doosan Group will transform into an AI-first industrial solutions provider, moving beyond traditional manufacturing.

The deep integration of NVIDIA's full-stack AI into Doosan's diverse business units (robotics, construction equipment, power generation, materials) and the development of customized foundation models and an Agentic Robot OS indicate a strategic shift towards AI-driven products and services.

The partnership will accelerate the development and commercialization of industrial humanoid robots by Doosan Robotics.

A specific roadmap has been established to launch intelligent robot solutions in 2027 and industrial humanoid robots in 2028, leveraging NVIDIA's AI infrastructure and Doosan's hardware capabilities.

Doosan's role in the AI supply chain will expand to include critical materials and energy solutions for AI data centers.

Doosan Enerbility is focusing on power solutions (gas turbines, SMRs) for energy-intensive AI data centers, and Doosan Electro-Materials BG is developing advanced CCL and copper foil for AI accelerators, positioning Doosan as a multi-faceted contributor to the AI ecosystem.

⏳ Timeline

2025-09

Doosan Group executives visited NVIDIA and other tech leaders in Silicon Valley to explore AI innovation and physical AI applications.

2025-10

Doosan and NVIDIA announced a strategic partnership to accelerate physical AI integration across Doosan's construction equipment, power generation, and robotics businesses.

2025-12

Doosan Group announced its participation in CES 2026, showcasing technologies for clean energy, jobsite automation, and intelligent machine innovation for the AI era.

2026-01

Doosan Bobcat unveiled AI-powered technologies at CES 2026, including Jobsite Companion, the industry's first on-machine AI feature for compact construction equipment. Doosan Enerbility showcased energy solutions for AI infrastructure.

2026-03

Lotte Energy Materials and Doosan Electronics BG partnered to jointly develop high-performance copper foil for advanced PCBs used in AI data centers.

2026-04

Madison Huang (NVIDIA) visited Doosan Robotics to discuss collaboration on humanoid robots and the Agentic Robot Operating System. Doosan Corporation announced establishing a CCL production base in Thailand.