Nvidia CEO Jensen Huang announces next-generation Rubin GPUs at GTC 2025.
Nvidia founder Jensen Huang kicked off the company's artificial intelligence developer conference on Tuesday, describing AI as going through "an inflection point." At GTC 2025 — dubbed the "Super Bowl of AI" — Huang revealed that demand for GPUs from the top four cloud service providers is surging and predicted that Nvidia's data center infrastructure revenue will hit $1 trillion by 2028. He also announced the company's next-generation graphics architectures: Blackwell Ultra, slated for the second half of 2025; the Vera Rubin AI chip, expected to launch in late 2026; and Rubin Ultra, scheduled for 2027.
Over the course of a talk lasting more than two hours, Huang outlined the "extraordinary progress" AI has made in the past decade. "In 10 years, he said, AI graduated from perception and 'computer vision' to generative AI, and now to agentic AI — or AI that has the ability to reason." Huang emphasized that "AI understands the context, understands what we're asking. Understands the meaning of our request. It now generates answers. Fundamentally changed how computing is done." The next wave of AI, he said, is already happening: robotics, fueled by so-called "physical AI" that can understand concepts like friction, inertia, cause and effect, and object permanence.
A critical breakthrough enabling this shift is synthetic data generation — using AI or computer-created data for model training. "There's only so much data and so much human demonstration we can perform," Huang said. "This is the big breakthrough in the last couple of years: reinforcement learning." Nvidia's technology can assist with this type of learning as AI works step by step to solve problems. To that end, Huang announced Isaac GR00T N1, an open-source foundation model designed to assist in developing humanoid robots, to be paired with an updated Cosmos AI model for generating simulated training data. Benjamin Lee, a professor of electrical and systems engineering at the University of Pennsylvania, noted that while training in the real world is time-consuming and expensive, providing an open-source platform could broaden access: "More researchers could start playing with this synthetic data — not just big players in the industry but also academic researchers."
Huang introduced the Cosmos series of AI models, which generate cost-efficient photo-realistic video for training robots, working with Nvidia's Omniverse physics simulation tool to create more realistic training environments. U.S. car maker General Motors plans to integrate Nvidia technology in its new fleet of self-driving cars, with the two companies building custom AI systems using both Omniverse and Cosmos to train AI manufacturing models. Huang also unveiled the Halos system, an AI solution for automotive safety, particularly autonomous driving, saying, "We're the first company in the world, I believe, to have every line of code safety assessed." The company announced Newton, an open-source physics engine for robotics simulation being developed with Google DeepMind and Disney Research. A small robot named Blue then joined Huang on stage and followed his commands, as he concluded: "The age of generalist robotics is here."