Nvidia has introduced Cosmos 3 Edge, a 4-billion-parameter AI model aimed at robots and vision systems that need to interpret and act on the physical world locally. The announcement, made in Tokyo on July 15 and expanded during CEO Jensen Huang’s Japan visit on July 16, pairs the model with a broader effort to put Nvidia’s physical-AI stack into Japanese factories, infrastructure and robotics projects.
As reported by CNBC and confirmed in Nvidia’s announcement, Cosmos 3 Edge is designed for on-device vision reasoning and robot-policy deployment on Nvidia Jetson edge hardware. Rather than sending every camera feed or sensor event to a cloud service, the company says the model can help systems understand scenes, reason about them in real time and generate actions locally.
Cosmos 3 Edge is part of Nvidia’s Cosmos 3 family of “world models,” which are intended to work with text, images, video, sensor data and actions. Nvidia says it is built on its Nemotron architecture and can be adapted for specific robots, vehicles and environments.
The immediate target is industrial and embedded hardware, including Jetson systems and Nvidia’s newly announced T2000 and T3000 modules. Nvidia also says the model can be deployed on RTX GPUs and DGX systems, potentially giving developers a route from workstation development to edge deployment.
For Windows users, this is not a consumer AI release or a new Copilot-style feature. The relevant angle is the growing use of RTX-equipped Windows workstations as development machines for simulation, model fine-tuning, video analytics and robotics workflows before workloads move onto Jetson devices or data-center infrastructure.
Fujitsu is exploring a collaborative control platform with FANUC, Yaskawa and Kawasaki Heavy Industries. Nvidia says the platform will combine its Cosmos models with Isaac robotics software, Omniverse libraries and its Newton physics engine to support digital twins, simulation-to-real testing and pre-deployment validation.
The Associated Press reported that Fujitsu, FANUC, Yaskawa and Kawasaki did not provide a timetable for robots reaching daily use, although they described a first phase of the collaboration as arriving later this year. That distinction matters: this is an ecosystem and development commitment, not a shipping fleet of general-purpose industrial robots.
The practical result is more Nvidia software moving from AI training into edge inference and industrial automation, with Japan serving as a major proving ground for the stack.
As reported by CNBC and confirmed in Nvidia’s announcement, Cosmos 3 Edge is designed for on-device vision reasoning and robot-policy deployment on Nvidia Jetson edge hardware. Rather than sending every camera feed or sensor event to a cloud service, the company says the model can help systems understand scenes, reason about them in real time and generate actions locally.
A model for edge robots, not Windows desktops
Cosmos 3 Edge is part of Nvidia’s Cosmos 3 family of “world models,” which are intended to work with text, images, video, sensor data and actions. Nvidia says it is built on its Nemotron architecture and can be adapted for specific robots, vehicles and environments.The immediate target is industrial and embedded hardware, including Jetson systems and Nvidia’s newly announced T2000 and T3000 modules. Nvidia also says the model can be deployed on RTX GPUs and DGX systems, potentially giving developers a route from workstation development to edge deployment.
For Windows users, this is not a consumer AI release or a new Copilot-style feature. The relevant angle is the growing use of RTX-equipped Windows workstations as development machines for simulation, model fine-tuning, video analytics and robotics workflows before workloads move onto Jetson devices or data-center infrastructure.
Japan coalition broadens Nvidia’s industrial push
Nvidia is expanding its Cosmos Coalition in Japan, with companies including Fujitsu, Hitachi, Kawasaki Heavy Industries, FANUC, NEC, SoftBank, Sony, Yaskawa Electric and Kubota indicating they intend to join. Nvidia’s stated objective is to build open frontier models for physical AI and use them in manufacturing, mobility, construction, agriculture, logistics and smart-building deployments.Fujitsu is exploring a collaborative control platform with FANUC, Yaskawa and Kawasaki Heavy Industries. Nvidia says the platform will combine its Cosmos models with Isaac robotics software, Omniverse libraries and its Newton physics engine to support digital twins, simulation-to-real testing and pre-deployment validation.
The Associated Press reported that Fujitsu, FANUC, Yaskawa and Kawasaki did not provide a timetable for robots reaching daily use, although they described a first phase of the collaboration as arriving later this year. That distinction matters: this is an ecosystem and development commitment, not a shipping fleet of general-purpose industrial robots.
New Metropolis tools accompany the model
Nvidia also introduced updated Metropolis libraries for building video-analysis agents. The package includes VSS Blueprint 3.2, DeepStream 9.1, TAO 7 and Physical AI Data Factory tools. Nvidia claims the additions can speed development and operation of Cosmos-based video systems, though that performance figure is the company’s own estimate.The practical result is more Nvidia software moving from AI training into edge inference and industrial automation, with Japan serving as a major proving ground for the stack.
References
- Primary source: CNBC
Published: 2026-07-16T11:15:12+00:00
Loading…
www.cnbc.com - Related coverage: nvidianews.nvidia.com
Loading…
nvidianews.nvidia.com - Related coverage: blogs.nvidia.com
NVIDIA and Japan Bring Full-Stack AI and Robotics to Every Industry | NVIDIA Blog
NVIDIA and its partners in Japan are this week showcasing the AI ecosystem's latest advancements. Check back here for updates.blogs.nvidia.com - Related coverage: bizinsider.org
Loading…
www.bizinsider.org - Related coverage: investor.nvidia.com
Loading…
investor.nvidia.com - Related coverage: internewscast.com
Loading…
internewscast.com