NVIDIA and Doosan Group Announce Expansive Physical AI Collaboration
NVIDIA and South Korea's Doosan Group have significantly expanded their partnership, aiming to integrate physical AI into heavy industries ranging from robotics and construction equipment to power generation and advanced materials. According to a recent NVIDIA blog post, the collaboration will leverage NVIDIA's full-stack accelerated computing platforms—including the Omniverse, Isaac, and Metropolis frameworks—with Doosan’s industrial automation, energy, and electronics capabilities.
What the Collaboration Covers
The partnership spans four key subsidiaries: Doosan Robotics, Doosan Bobcat, Doosan Enerbility, and Doosan Corporation Electro-Materials BG. Each branch will tap into different NVIDIA technologies:
- Doosan Robotics will use NVIDIA Isaac Sim and Omniverse to simulate and deploy collaborative robots (cobots) for manufacturing and logistics, targeting a 30% reduction in deployment time.
- Doosan Bobcat (construction equipment) will integrate NVIDIA Metropolis for computer vision and autonomous navigation in compact equipment, enabling safer job site automation.
- Doosan Enerbility will apply NVIDIA’s AI factory blueprint to optimize power plant operations and renewable energy forecasting, aiming to lower energy costs by up to 15%.
- Doosan Electro-Materials will use AI-driven materials discovery powered by NVIDIA’s CUDA and cuDNN to accelerate development of copper foil and semiconductor substrates.
Why This Matters for AI and Industrial Automation
This deal signals a shift from AI as a software-only layer to AI embedded directly into physical machinery. For developers, the announcement means deeper integration of NVIDIA’s Isaac and Omniverse tools into real industrial stack. Doosan’s move to standardize on NVIDIA platforms could set a precedent for other conglomerates, especially in Asia, where heavy industry is ripe for AI-driven efficiency gains.
The collaboration also highlights the growing concept of "AI factories"—facilities where AI models are trained and deployed at the edge alongside production machinery. NVIDIA’s reference architecture for such factories, previously demonstrated with partners like Foxconn, now gets a concrete implementation with Doosan’s power and robotics divisions.
Technical Implications for Developers
For robotics developers, the use of NVIDIA Isaac Sim means Doosan will likely release simulation assets and URDF files for their cobots, enabling third-party development and testing before deployment. Similarly, the integration with NVIDIA Metropolis for Bobcat equipment suggests that developers working on autonomous heavy machinery will have access to pretrained vision models tailored for construction environments.
Doosan Enerbility’s adoption of AI factory infrastructure also means that energy sector AI developers will see standardized APIs for digital twin creation and reinforcement learning-based control optimization. NVIDIA’s Omniverse Cloud will likely serve as the backbone for these simulations.
What This Means for Businesses and the Industry
From a business perspective, this collaboration positions Doosan as a leader in physical AI adoption among South Korean chaebols. It also gives NVIDIA a strong foothold in the energy and construction sectors, which have historically been slower to adopt AI than tech or automotive industries.
Competitors like Siemens (with Xcelerator) and ABB (with Ability platform) may face pressure to deepen their AI integrations, as NVIDIA’s ecosystem becomes more attractive due to the combined strengths of Omniverse, Isaac, and Metropolis. For mid-market automation firms, the message is clear: partnerships with full-stack AI providers are becoming a necessity to remain competitive.
Practical Steps for Developers and Technical Leaders
- Start with simulation: Use NVIDIA Isaac Sim or Omniverse to prototype robot deployments before hardware arrives, aligning with Doosan’s approach.
- Leverage pretrained models: NVIDIA Metropolis offers models specifically for construction and manufacturing; evaluate them for your own computer vision needs.
- Plan for AI factory architecture: If your organization operates heavy equipment or power plants, consider adopting NVIDIA’s reference design for edge AI clusters.
- Watch for open-source releases: Doosan may contribute simulation models or datasets; check their GitHub and NVIDIA’s NGC catalog.
The Bigger Picture: Physical AI Enters the Mainstream
This announcement is part of a broader trend where AI moves from data centers into the real world. With NVIDIA’s market cap already reflecting enthusiasm for AI chips, the real value in 2026 lies in vertical applications. Doosan’s diversified portfolio provides a testbed for physical AI across multiple domains simultaneously.
Developers who gain early experience with the tools announced—Omniverse for digital twins, Isaac for robotics simulation, and Metropolis for vision AI—will have a significant advantage as more industrial players follow this blueprint.
The partnership also underscores the importance of sovereign AI infrastructure: Doosan’s AI factories in Korea will use locally trained models, keeping sensitive industrial data onshore. This could become a selling point for other countries seeking to balance AI adoption with data sovereignty.
Source: NVIDIA Blog. This article was produced with AI assistance and reviewed for accuracy. Editorial standards.
