NVIDIA and Doosan Group are joining forces across robotics, AI factory power, and advanced circuit board materials to build the next generation of intelligent industrial systems.
There is a moment in industrial history when two forces arrive at the same frontier from opposite directions. NVIDIA built its reputation in silicon, processing data at speeds that redrew the boundaries of what machines could perceive and decide. Doosan Group spent decades building machines that actually move things in the physical world: robots on factory floors, turbines powering cities, bulldozers shaping land.
On June 7, 2026, those two trajectories met officially, as the companies announced a wide-ranging collaboration to advance physical AI, robotics, and the infrastructure that feeds the world's most powerful AI systems.
The partnership draws together four arms of Doosan Group — Doosan Robotics, Doosan Bobcat, Doosan Enerbility, and Doosan Corporation Electro-Materials BG — and pairs their industrial depth with NVIDIA's full-stack accelerated computing platforms. What emerges is not a single product announcement, but a coordinated effort to embed intelligence into machines that have historically operated on scripted instructions rather than genuine reasoning.
"Doosan Group's businesses span several layers of the AI factory ecosystem, from intelligent robotics systems to the full spectrum of large-scale power solutions and advanced electronics materials for AI data center equipment."
Teaching Robots to Think, Not Just Follow Instructions
The most immediately tangible piece of the agreement involves Doosan Robotics and its ambition to move from being a supplier of robot arms to becoming what the company calls a "full-stack AI-first robotics solution company." To get there, Doosan is integrating a suite of NVIDIA technologies into its Agentic Robot OS — an AI-powered platform designed to connect perception, reasoning, simulation, learning, and on-device inference into a single operational layer.
The tools feeding into that platform include NVIDIA Isaac Sim and Isaac Lab, open robotics frameworks that let engineers train and test robots inside high-fidelity virtual environments before a single physical component is assembled. Alongside them comes NVIDIA Cosmos, a family of open world foundation models that help robots develop a richer understanding of how the physical world behaves — how objects interact, how surfaces resist or yield, how sequences of actions unfold in time.
Key NVIDIA Technologies in the DealIsaac Sim and Isaac Lab for simulation-based training; Cosmos world foundation models; the Newton open-source physics engine; Jetson Thor for on-device edge inference; the DSX AI factory platform; MGX modular reference architecture for AI servers.
Also in the mix is the Newton physics engine, an open-source tool for physics simulation and calibration, and NVIDIA Jetson Thor, a compact computing module designed for on-device AI inference at the robot level. Together, these tools are meant to help Doosan's collaborative robots perceive more accurately, reason through ambiguity, and adapt to the unpredictable conditions of real factory environments.
The companies are targeting specific high-value industrial tasks first: depalletizing, which involves autonomously sorting and moving stacked goods, and sanding, a surface-finishing task that demands consistent pressure and motion judgment. Beyond those reference applications, they are exploring new robot form factors altogether, including dual-arm platforms and humanoid designs — machines shaped and jointed more like the workers they are meant to augment.
From Factory Floors to Open Fields: Bobcat Enters the Picture
Doosan Bobcat, the brand long associated with compact loaders and skid-steer machines humming across construction sites and farms, is also part of the collaboration. The focus here is on developing specialized world models for equipment used in construction, landscaping, agriculture, and material handling — sectors where the terrain, conditions, and tasks shift constantly and unpredictably.
The goal is to give Bobcat's machines the ability to perceive their operating environments with greater nuance, reason about changing conditions in real time, and perform tasks with increasing autonomy. NVIDIA and Doosan Bobcat are also aiming to help establish what they describe as an industry-standard ecosystem for compact autonomous equipment — an acknowledgment that the machinery governing worksites is approaching the same inflection point that automotive vehicles reached when fully driver-assisted driving entered the mainstream conversation.
Powering the Machines That Power AI
The collaboration's reach extends well past robots into the question of energy — a subject that has grown urgent as AI data centers multiply and their electricity demands strain grids. Doosan Enerbility, the group's energy infrastructure arm, is exploring how its large-scale power portfolio can support NVIDIA's AI factory deployments and the DSX AI factory platform.
That portfolio is substantial: gas turbines, steam turbines, small modular reactors, and hydrogen fuel-cell systems through the Doosan Fuel Cell division. These are not speculative concepts. They are operational or near-operational technologies that address the specific requirements of data centers: reliable power delivered without interruption, at high efficiency, around the clock.
Future collaboration could include power supply design for AI factory deployments, optimization of generation equipment, and evaluation of low-carbon power sources — including small modular reactors.
The alignment here reflects a broader industry reckoning. The computational demands of training and running large AI models have made power access a genuine bottleneck for AI infrastructure. Finding energy sources that are both abundant and low-carbon is now a strategic priority for every major technology company — and Doosan Enerbility's portfolio positions it squarely in that conversation.
The Hidden Layer: Circuit Board Materials
Less visible but no less significant is the role of Doosan Corporation Electro-Materials BG in the collaboration. This division produces copper clad laminate, or CCL — a foundational material used in printed circuit boards (PCBs) that shows up inside networking equipment, AI accelerators, and the server motherboards at the heart of every data center.
As AI servers and networking systems push toward higher performance and greater bandwidth, the quality of those underlying materials becomes increasingly consequential. High-performance CCLs must deliver low signal loss and consistent reliability under demanding thermal and electrical conditions. NVIDIA's MGX platform — a modular reference architecture that helps manufacturers design and build AI server and rack-scale infrastructure — is one environment where advanced PCB materials will matter more, not less, as performance ceilings rise.
The inclusion of Doosan's materials business in a partnership otherwise focused on software, robots, and energy underscores how deep the supply chain implications of the AI buildout actually run. The intelligence in an AI server does not begin with the chip. It begins, in part, with the substrate the chip sits on.
A Partnership That Reflects Broader Strategic Momentum
The announcement came as NVIDIA CEO Jensen Huang was engaged with South Korea's technology and industrial ecosystem — a series of engagements that have highlighted the country's outsized importance to global AI infrastructure ambitions. South Korea is home to major semiconductor manufacturers, advanced materials suppliers, and industrial conglomerates whose manufacturing expertise is directly relevant to the physical demands of scaling AI.
For Doosan Group, the collaboration represents a deliberate pivot toward AI as a core operating principle rather than an adjacent capability. The vision, as the companies describe it, is a Doosan Group-wide direction for physical AI that extends beyond any single product line — from robot arms to construction machinery to turbines — unified by a shared investment in the underlying AI platforms that NVIDIA provides.
Whether that vision produces robots that genuinely reason or merely execute more sophisticated scripts remains to be seen. But the collaboration establishes the infrastructure, the toolchain, and the institutional commitment to find out.
0 Comments