The world of artificial intelligence is on the brink of a major shift. This was the message from Jensen Huang, the head of Nvidia, during a recent event. He spoke to a massive audience about the rapid advancements in AI and what lies ahead.

Huang highlighted the growing demand for GPUs from major cloud service providers. He predicted that Nvidia's data center infrastructure revenue could reach $1 trillion by 2028. This is a bold claim, but it reflects the increasing importance of AI in various industries.

One of the most anticipated parts of Huang's talk was the unveiling of Nvidia's next-generation graphics architectures. These include Blackwell Ultra and Vera Rubin, named after a renowned astronomer. Blackwell Ultra is set to launch in the second half of 2025, with Rubin AI chip following in late 2026. Rubin Ultra will make its debut in 2027.

Huang's speech lasted over two hours, during which he discussed the remarkable progress in AI over the past decade. He noted that AI has evolved from basic perception and computer vision to generative AI, and now to agentic AI. This type of AI can understand context, interpret requests, and generate answers, fundamentally changing how computing is done.

Looking ahead, Huang believes the next big thing in AI is robotics. He explained that physical AI can understand concepts like friction, inertia, cause and effect, and object permanence. This opens up new opportunities for innovation and development.

A key aspect of Huang's announcements was the use of synthetic data generation for model training. AI needs digital experiences to learn, and synthetic data can provide this at a much faster rate than human training. This is a significant breakthrough in reinforcement learning.

Huang also introduced Isaac GR00T N1, an open-source foundation model designed to help develop humanoid robots. This model will be paired with an updated Cosmos AI model to create simulated training data for robots.

The Cosmos series of AI models, introduced earlier this year, can generate cost-efficient, photo-realistic video for training robots and other automated services. This technology is much cheaper than traditional training methods, such as having cars record road experiences or people teaching robots repetitive tasks.

General Motors plans to use Nvidia's technology in its new fleet of self-driving cars. The two companies will collaborate to build custom AI systems using Omniverse and Cosmos for training AI manufacturing models.

Huang also unveiled the Halos system, an AI solution focused on automotive safety, particularly for autonomous driving. He claimed that Nvidia is the first company to have every line of code safety assessed.

Towards the end of his talk, Huang introduced Newton, an open-source physics engine for robotics simulation. This project is being developed with Google DeepMind and Disney Research. A small robot named Blue joined him on stage, demonstrating its ability to follow commands.

Huang's message was clear: the age of generalist robotics is here. This is an exciting time for AI, with many new possibilities on the horizon.

The Future of AI: Nvidia's Big Reveal

Actions