Nvidia is stitching together robot-training software, AI models, simulation tools and onboard computers into a broad robotics platform. A Science Monitor report details how the company is trying to make its tools useful across the journey from virtual training to machines working in the physical world.
NVIDIA Watch analysis
What happened
The report describes Nvidia’s Isaac and GR00T ecosystem: GR00T models for humanoid robots, Cosmos for generating synthetic training data, and Isaac tools for simulation and training. Jetson Thor provides onboard computing, while Isaac OS connects sensors and control systems. The report says Nvidia’s GR00T N1.7 was trained using 20,854 hours of first-person human video and cites Nvidia as saying that increasing this training data from 1,000 to 20,000 hours more than doubled average task completion. It gives no specific benchmark scores for that claim.
Nvidia is also offering tools and data to developers. The report says Isaac Lab 3.0 supports both Nvidia’s PhysX and the open-source Newton physics engine, which Nvidia developed with Disney Research and Google DeepMind. It also describes a library of roughly 2,000 simulation-ready robot and factory assets. Adoption is mixed: some established manufacturers use simulation tools, while a number of robotics startups are adopting a wider slice of Nvidia’s stack. The report says Tesla and Figure AI are developing their own models instead.
Why it matters
Robot AI needs more than a capable model. It has to learn movements, cope with the physical world and run on hardware inside the robot. Nvidia’s pitch is that developers can use connected parts of its ecosystem for each step, with open tools and models helping bring them in. If they later need data-centre GPUs or onboard chips, Nvidia has a business waiting at both ends.
The report also makes clear that the strategy is not a guaranteed win. Companies can take the simulation tools they need without buying into the whole stack, and major robotics firms may prefer their own models. Nvidia’s robotics revenue is not reported separately in the figures cited by the article, so the scale of the business remains unclear.
Our read
This is a meaningful platform play, not evidence that Nvidia has won the robot market. The interesting detail is how much of the toolkit the company is making available to attract developers, while keeping valuable hardware in the picture. Open doors, with a toll booth further down the road.
What to watch
- Whether robotics companies adopt Nvidia’s models as well as its simulation tools and chips.
- How developers use Isaac Lab’s support for both PhysX and Newton.
- Whether Nvidia starts reporting robotics revenue separately.
- How its platform competes with Google DeepMind’s robotics models and rival processors.
Discussion spark: Would you build a robot on Nvidia’s connected stack, or pick separate tools and keep more control in-house?
Sources and evidence
- Nvidia Isaac and GR00T: The Race for Robot AI Platform Dominance – Science Monitor – 사이언스모니터 (18 November 2026, 08:44 UTC)
Independent WittyWires tracker for public updates about NVIDIA. Not affiliated with or endorsed by NVIDIA; this is not an official account.