Nvidia Unleashes the Future of AI with Humanoid Robots and Next-Gen Infrastructure

From GPUs to General-Purpose AI Robotics

At Computex 2025, Nvidia signaled a bold expansion beyond its dominant role in AI chips by debuting a full-stack platform for humanoid robotics. CEO Jensen Huang introduced Project GR00T, a new foundational model designed to power intelligent humanoid robots capable of performing complex tasks, adapting to human environments, and learning from minimal input. These robots are built to interact with the real world in dynamic ways—walking, talking, manipulating objects, and collaborating with humans. The company also presented a robotics reference design called Jetson Thor, powered by the next-gen Thor SoC, merging AI computing and sensor fusion at unprecedented levels. This signals Nvidia’s commitment to becoming not just a toolmaker for AI but a leader in embodied AI systems that could redefine labor, logistics, and home assistance.

Custom AI Factories Built For the AI Gold Rush

Complementing its robotics announcement, Nvidia unveiled Blackwell Ultra chips and the new Rubin AI platform, targeting enterprises looking to build custom AI factories. These platforms are optimized for generative AI workloads and boast improved memory bandwidth, energy efficiency, and massive compute capabilities. Recognizing that every major tech firm is betting on tailor-made models, Nvidia also introduced new data center infrastructure tools including cabling, cooling systems, and simulation software—all aimed at helping customers scale AI training with reduced costs and complexity. The emphasis was clear: Nvidia wants to own not just the chip but the entire AI hardware and software pipeline. With partners like AWS, Google, and Dell already on board, the runway for Nvidia’s infrastructure ambitions looks long and profitable