Title: Physical AI leaves the laboratories: NVIDIA and its industrial partners deploy the next generation of robots
Physical artificial intelligence, which gives robots the ability to interact and act in the real world, is experiencing significant industrial acceleration. NVIDIA, in collaboration with global robotics leaders such as FANUC, ABB Robotics, YASKAWA and KUKA, is deploying advanced solutions. These partnerships build on an installed base of 2 million robots worldwide and integrate new models like Isaac GR00T N2, which enables robots to successfully complete new tasks twice as often as previous models. This evolution marks a shift from research to large-scale deployments, with major implications for work, productivity and safety in key sectors such as manufacturing, logistics, healthcare and services.
1. 2 million industrial robots integrate physical AI through NVIDIA partnerships
The integration of artificial intelligence into physical robotic systems is transforming the industry. NVIDIA has forged strategic partnerships with the four largest industrial robot manufacturers: FANUC, ABB Robotics, YASKAWA and KUKA. Together, these companies represent an installed base of more than 2 million robots worldwide. The goal of this collaboration is to integrate NVIDIA's simulation and AI technologies to accelerate the development and deployment of smarter and more autonomous robots. Companies use NVIDIA Omniverse libraries and NVIDIA Isaac simulation frameworks to create physically accurate digital twins of their production lines. These virtual environments enable the design, testing and validation of complex robotic applications before physical deployment, thus reducing costs and commissioning delays. Additionally, the integration of NVIDIA Jetson modules in robot controllers enables real-time AI inference directly at the edge, giving machines instant decision-making capability. This synergy between existing robotic hardware and NVIDIA's advanced software platforms opens the way to a new era of automation, where robots are no longer simple executors of pre-programmed tasks, but adaptive agents capable of reacting to the unexpected and continuously optimizing their performance [1].
2. Isaac GR00T N2 doubles humanoid robot performance on new tasks
At the heart of this advancement lies the development of increasingly sophisticated robotics foundation models. NVIDIA has introduced Isaac GR00T N2, a next-generation model based on DreamZero research. Designed specifically for humanoid robots, GR00T N2 is a vision-language-action (VLA) model that allows robots to understand instructions in natural language, perceive their environment through computer vision and execute complex actions. Tests show that robots equipped with GR00T N2 succeed in accomplishing new tasks in unfamiliar environments with a success rate more than twice that of previous generation VLA models. This model currently ranks first in the MolmoSpaces and RoboArena benchmarks for generalist robotic policies, demonstrating its ability to generalize learning to a wide variety of situations. Pioneer companies in humanoid robotics such as 1X, Agility, Boston Dynamics and Figure are already using NVIDIA technologies to accelerate the development of their robots. GR00T N2's ability to learn and quickly adapt to new situations is crucial for deploying robots in dynamic and unstructured environments, where flexibility is essential. This includes applications in logistics, healthcare and even exploration, where robots must face unforeseen conditions and make autonomous decisions [1].
3. Cosmos 3 generates synthetic worlds to train robots at scale
For a robot to learn to interact with the world, it needs a massive amount of training data. Collecting this data in the real world is slow, expensive and sometimes dangerous. To overcome this obstacle, NVIDIA has developed Cosmos 3, a world foundation model (WFM) platform. Cosmos 3 unifies synthetic world generation, visual reasoning and action simulation. This platform enables the creation of photorealistic and physically accurate virtual environments in which robots can be trained on millions of different scenarios. By using synthetic data, developers can expose robots to a much wider range of situations than would be possible in the real world, including rare or dangerous events. This considerably accelerates the learning process and improves the robustness and reliability of AI systems. Companies like Generalist AI and World Labs are already using Cosmos for synthetic data generation and model validation. The importance of this approach lies in its ability to create rapid feedback loops between simulation and the real world, enabling continuous improvement of robotic models. The ability to simulate complex and dangerous scenarios without physical risk is a major asset for physical AI development [1].
4. Physical AI transforms manufacturing with micron-level precision
Physical AI applications in the manufacturing sector are vast and promise significant productivity gains. Skild AI, in partnership with Foxconn, uses dual-arm manipulators equipped with AI for high-precision assembly on NVIDIA Blackwell GPU production lines, one of the most complex manufacturing tasks in the industry. These AI-guided robots can manipulate tiny components with micron-level precision, surpassing human capabilities in these repetitive and demanding tasks. Meanwhile, Lightwheel collaborates with Samsung to use NVIDIA's Newton physics engine to train assembly robots to handle complex cables in simulation, resulting in faster and more accurate assembly lines. This approach reduces assembly errors and increases production throughput. PTC has also announced a new workflow that connects its Onshape CAD platform to NVIDIA Isaac Sim, creating a seamless bridge between computer-aided design and simulation. Companies like FANUC America and Fauna Robotics use this integration to design and validate their robotic systems in digital twins before production. This enables the identification and correction of potential issues in the design phase, thus avoiding costly delays and post-implementation modifications. The impact of physical AI on manufacturing translates to improved product quality, reduced waste and increased production line flexibility [1].
5. The healthcare sector adopts simulation to validate surgical robots
Healthcare, where precision and safety are paramount, is another major application field for physical AI. CMR Surgical uses Cosmos-H simulation to train and validate the robotic intelligence of its Versius surgical system before clinical deployment. This approach ensures that the robot will behave as expected in a multitude of surgical scenarios, including the most complex and delicate ones. Simulation provides a safe environment to test and refine robot control algorithms, thus minimizing risks for patients during real interventions. Similarly, Johnson & Johnson MedTech uses workflows based on Isaac Sim and Cosmos for training and validation of systems on its Monarch platform for urology. These tools allow the simulation of urological procedures and optimization of robot movements for maximum efficiency. Medtronic is exploring the use of NVIDIA's IGX Thor robotic computing platform to ensure critical precision and functional safety in its future surgical robotics systems. IGX Thor is designed to meet the strict regulatory requirements of the medical sector, offering exceptional reliability and performance. These advances aim to improve intervention precision, reduce patient risks and improve overall healthcare efficiency, paving the way for less invasive procedures and faster recoveries [1].
6. An open ecosystem connects 15 million developers to accelerate innovation
NVIDIA has adopted an open platform strategy to catalyze innovation throughout the robotics ecosystem. The NVIDIA Inception program, which includes more than 40,000 startup members, offers privileged access to NVIDIA's physical AI technologies, as well as technical support and computing resources. This program helps young companies transform their innovative ideas into marketable products by providing them with the necessary tools and expertise. Additionally, a strategic partnership with Hugging Face has integrated the Isaac and GR00T frameworks into the LeRobot open source platform. This initiative connects the 2 million developers in NVIDIA's robotics ecosystem with the 13 million AI developers in the Hugging Face community, creating a vast network for knowledge sharing and accelerating open source robotics development. This openness is fundamental to democratizing access to these cutting-edge technologies and allowing a greater number of innovators to contribute to the field. It also promotes the creation of open standards and interoperability between different robotic platforms, which is essential for the widespread adoption of physical AI [1].
7. Human-robot collaboration redefines work in warehouses and factories
The introduction of physical AI in industrial environments does not aim to replace human workers, but rather to augment their capabilities. KION Group, a supply chain solutions company, works with NVIDIA and Accenture to develop autonomous warehouse solutions. Using Omniverse, they create large-scale warehouse digital twins to train and test fleets of autonomous forklifts based on NVIDIA Jetson. These robots can handle material transport tasks, allowing employees to focus on higher value-added activities such as inventory management or quality control. This approach not only improves operational efficiency, but also working conditions by reducing physical demands and accident risks. Similarly, WORKR integrates its AI platform with ABB industrial robots to create a robotic workforce that can be deployed by small and medium enterprises in minutes, without requiring programming knowledge. The goal is to create collaborative work environments where humans and robots work in synergy to improve efficiency and safety. This progressive transformation of work also implies the need for new skills for workers, who will need to learn to interact with and supervise these intelligent systems. Continuous training will be a key element in supporting this evolution [1].
8. Challenges and perspectives: towards ubiquitous and ethical physical AI
Despite spectacular advances, challenges remain for the widespread adoption of physical AI. The initial cost of advanced robotic systems can be a barrier for some companies, particularly SMEs. The complexity of integrating these technologies into existing infrastructures also requires investments in time and expertise. Finally, social acceptance of autonomous robots, particularly in interactive roles with humans, remains a subject of ethical and psychological debate. However, the trend is clear: physical AI is becoming ubiquitous. The massive investments by NVIDIA and its partners, combined with advances in foundation models and simulation capabilities, indicate a rapid trajectory toward smarter, more adaptable and more autonomous robots. The future will see robots capable of adapting to changing environments, learning new tasks and collaborating more fluidly with humans, thus transforming many aspects of our economy and society. It will be crucial to develop ethical and regulatory frameworks to guide this evolution, ensuring that physical AI serves the common good and respects human values. Transparency, accountability and safety will be the pillars of this new robotic era [1].
References
- [1] NVIDIA News. NVIDIA and Global Robotics Leaders Take Physical AI to the Real World. http://nvidianews.nvidia.com/news/nvidia-and-global-robotics-leaders-take-physical-ai-to-the-real-world
- Receive Journal analyses directly in your inbox.
- China's CO2 emissions fell by 0.3% in 2025. This is a modest figure. But it extends a trend that has lasted for 21 consecutive months: since March 2024, emissions from the world's largest polluter have been "stable or declining."
- Does AI destroy jobs? Two studies published in March 2026 — one by Harvard Business School, the other by Anthropic — provide the first solid empirical data. And the answer is more nuanced than public debate suggests.
- Malaria killed 608,000 people in 2022, 95% in sub-Saharan Africa and 78% children under five years old. And for the first time, a vaccine deployed at scale shows measurable results in the field.
- Want to be 20 years old today. An independent media that documents progress with rigor, without naivety or catastrophism.
- Structured reading sheets: central thesis, key arguments, limitations, and verdict.
- JdP is an independent editorial project based on data, counter-narratives and lucid optimism. Each article is sourced, nuanced and open to discussion.
- The Journal of a Progressive uses cookies to improve the reading experience and understand how the site is used. No data is collected for commercial, advertising or resale purposes. Cookies necessary for site operation are always active. Optional cookies are only activated with your explicit consent, in compliance with GDPR.