The engineers who build digital twins of factories and robots have paid for the software for years. Starting this week, they do not have to. NVIDIA said it is making Omniverse, its digital twin platform, available free of charge for production use, eliminating the enterprise license fee that covered commercial deployments, according to StorageReview.
The change is small in accounting terms and large in strategy. Omniverse is the software layer where NVIDIA customers build simulated versions of the physical world: a BMW factory floor, a robot arm, a warehouse full of autonomous vehicles. Under the old model, companies paid a per-seat license to run those simulations in production. Under the new one, the software is free, and NVIDIA collects its money where it always makes it, on the hardware.
The logic is a familiar one in the chip industry. NVIDIA built its dominance on CUDA, the programming framework it gives away, because every CUDA program needs an NVIDIA chip to run. Omniverse is the same play at a higher level: free software that only runs well on NVIDIA hardware, generating demand for the GPUs underneath. A factory that builds a digital twin in Omniverse needs accelerated computing to run it, and that computing comes from NVIDIA.
The strategy reflects a shift in how NVIDIA sees its business. The company’s revenue is still dominated by data center chips sold for AI training, but executives have described the long game as building an ecosystem around those chips: simulation, robotics, industrial software and the developers who tie them together. Free access to Omniverse lowers the barrier for that ecosystem, and it locks industrial and robotics customers into NVIDIA’s software stack as they scale.
The industrial market is the immediate target. Manufacturers have used Omniverse to simulate production lines before they are built, to train robots in virtual environments and to test autonomous vehicles without putting them on the road. Customers including BMW, Mercedes-Benz and Siemens have used the platform for factory planning, and robotics companies have trained fleets in simulation before deploying them in the real world. Removing the license fee removes a line item from every one of those budgets.
Analysts said the move is aimed squarely at the competition. Siemens sells its own digital twin software, and Unity and Epic Games offer simulation platforms that run on general-purpose hardware. NVIDIA’s answer is price and performance: free software, and simulation built on physics engines and AI that run fastest on NVIDIA chips. “They are giving away the razor to sell the blades,” one analyst who follows industrial software said.
The free tier is the latest chapter in a product with a long run-up. NVIDIA introduced Omniverse in 2020 as a platform for connecting 3D design tools, then spent years building out its industrial capabilities: physics simulation, photorealistic rendering and the ability to mirror a factory in real time. It launched Omniverse Enterprise in 2022 with per-seat pricing aimed at large manufacturers, and the software found its footing in the automotive and robotics industries before the current AI boom expanded its uses.
The change also reflects where the money has moved. The industrial customers who once bought Omniverse licenses are now buying AI infrastructure in much larger quantities, from training systems to the edge computers that run robots on factory floors. NVIDIA has concluded, according to people familiar with the company’s thinking, that the software license is a rounding error compared with the hardware that each digital twin project requires, and that removing the license removes a reason for customers to hesitate.
The robotics angle may matter most. Training a robot in the real world is slow and dangerous; training it in simulation is fast and cheap, but only if the simulation faithfully models physics. Omniverse, combined with NVIDIA’s Isaac platform for robotics development, has become the standard tool for sim-to-real training, in which robots learn in virtual worlds and transfer the skills to hardware. Every robot trained that way carries NVIDIA software in its development path.
The economics work because the hardware margin is where NVIDIA lives. Enterprise software licenses are a small business for the company next to its data center chip sales, and giving them up costs little in revenue. What matters is that each Omniverse deployment, each simulated factory and each trained robot fleet, creates demand for NVIDIA GPUs that compounds over time.
The move also signals confidence in the platform’s hold. If NVIDIA believed customers would switch to competitors once they had built their digital twins, charging for the software would be a way to keep them. Instead, the company is betting that the ecosystem itself is the moat: the more simulations, models and robots built on Omniverse, the harder it is for customers to leave.
For industrial customers, the decision is straightforwardly good news. Digital twin projects that required budget approval for software licenses can now proceed without them, and the total cost of industrial AI deployments falls. The question they face is the one NVIDIA is counting on them not to ask: having built their digital operations on free NVIDIA software, how much will it cost to run anywhere else?
