In a groundbreaking development that could revolutionize the robotics industry, NVIDIA has unveiled its comprehensive cloud-to-robot computing platforms designed to accelerate the development of humanoid robots. The announcement, made during CEO Jensen Huang’s keynote at Computex 2025, positions NVIDIA at the forefront of what many industry experts are calling “physical AI” – the fusion of artificial intelligence with robotic bodies capable of interacting with the physical world.

The GR00T Revolution

Central to NVIDIA’s robotics push is Project GR00T (Generalist Robot 00 Technology), a foundation model specifically designed for humanoid robots. As highlighted by Fingerlakes1, this technology aims to power general-purpose robots capable of understanding natural language and mimicking human movement, representing a significant leap toward robots that can learn and adapt in real-world environments.

“The age of generalist robotics has arrived with breakthroughs in mechatronics, physical AI, and embedded computing,” NVIDIA stated in materials released alongside the announcement, noting that these advances come “just in time as labor shortages limit worldwide industrial growth.”

What makes GR00T particularly revolutionary is its approach to addressing one of the most persistent challenges in robotics development: data scarcity. Traditional robot training methods rely heavily on human demonstrations, which aren’t scalable due to time constraints. NVIDIA’s solution, branded as “GR00T Dreams,” offers a blueprint for large-scale synthetic trajectory data generation using a process the company calls “real-to-real data workflow.”

How Robot Brains “Dream”

The GR00T Dreams process begins with developers fine-tuning NVIDIA’s Cosmos physical AI world foundation models using a limited set of human demonstrations captured through teleoperation of specific tasks. Once trained, these models can generate “dreams” – simulated future world states – when prompted with new images and instructions.

What’s particularly noteworthy about this approach is that developers can prompt the model using entirely new action words without capturing additional teleoperation data. The system then evaluates these dreams, selecting the highest quality simulations for training purposes.

But NVIDIA’s innovation goes beyond generating 2D simulations. The GR00T Dreams blueprint converts these dream videos into 3D action trajectories that robots can learn from, effectively allowing “a small team of human demonstrators to now do the work of thousands,” according to NVIDIA.

Infrastructure to Support the Vision

Supporting these advanced AI models requires equally advanced computing infrastructure, which NVIDIA has delivered through expansions to its Blackwell platform. As detailed during Computex, the GB200 NVL72 server rack incorporates 72 Blackwell GPUs and 36 Grace CPUs, connected with NVLink switch fabric – hardware specifically designed to support the trillion-parameter AI models required for sophisticated robotics applications.

NVIDIA is also providing RTX PRO 6000 Blackwell workstations and RTX PRO servers to support robot simulation and training, creating an end-to-end solution for robotics developers.

Industry Implications

The timing of NVIDIA’s robotics push coincides with growing interest in humanoid robots across industries seeking to address labor shortages. By providing both the foundation models and the computing infrastructure needed to train and deploy these robots, NVIDIA is positioning itself as an essential player in what could become a transformative industry.

Jensen Huang’s presentation at Computex 2025 also hinted at the company’s next-generation Rubin platform, scheduled for 2026, which will incorporate new AI chips, memory technology, and interconnects – suggesting that NVIDIA views its robotics initiatives as a long-term strategic priority.

As companies worldwide race to develop practical humanoid robots for industrial and commercial applications, NVIDIA’s comprehensive approach to solving the data challenge could prove decisive in determining which robotics platforms ultimately succeed in the marketplace.