Nvidia's GTC 2025 Keynote: Everything Announced in 16 Minutes

Summary

In this exciting keynote from Nvidia's GTC, the company showcases groundbreaking advancements in artificial intelligence, computer graphics, and robotics. From the evolution of GeForce GPUs, which now leverage AI to enhance performance, to the partnership with GM for self-driving cars, Nvidia emphasizes its commitment to revolutionizing industries. The presentation highlights innovations in AI-powered data centers, robotics, and Omniverse, an operating system for physical AI. Nvidia also unveils new technologies like Cosmos and Newton, showcasing the power of real-time simulation and reinforcement learning in robotics, marking a new era for AI-driven industries.

Takeaways

• 😀 GeForce continues to lead the world in graphics, with the release of the 90 and 5090, now 30% smaller and more efficient thanks to AI-powered advancements in computing.
• 😀 Nvidia's AI-driven technology has revolutionized computer graphics, making real-time 100% path-traced graphics a reality with the help of artificial intelligence.
• 😀 Nvidia has partnered with GM to build an autonomous vehicle fleet, marking a significant step in AI's role in the automotive industry, including manufacturing, design, and in-car systems.
• 😀 Nvidia is a leader in automotive safety, with its Halos initiative ensuring safety standards are deeply integrated into system development, including diversity, transparency, and explainability.
• 😀 The transition from air-cooled to liquid-cooled systems, along with disaggregated MV link architecture, allows Nvidia to scale computing power and efficiency to new heights, including a 1 exaflop computer in a single rack.
• 😀 Nvidia Dynamo is introduced as the new operating system for AI factories, which replaces traditional systems like VMware and orchestrates AI workloads in a completely new paradigm.
• 😀 New Nvidia hardware like the djx station, with capabilities like 20 petaflops, sets the standard for AI-driven computing, with multiple enterprise solutions designed for researchers and data scientists.
• 😀 The shortage of human workers is pushing the robotics industry forward, with Nvidia's Omniverse and Cosmos platforms enabling the creation of infinite virtual environments for robot training.
• 😀 Reinforcement learning and physics engines are vital for training robots, and Nvidia is now collaborating with DeepMind and Disney to create a new physics engine called Newton for fine motor skills and real-time robot simulation.
• 😀 Nvidia’s robotics progress includes the open-source release of Groot N1, expanding accessibility for developers and advancing AI applications in physical environments.

Q & A

What milestone is Nvidia celebrating in this keynote?

Nvidia is celebrating the 25th anniversary of GeForce, with the launch of the GeForce 90 and 5090 graphics cards, marking a major advancement in computer graphics technology.

How has artificial intelligence influenced the development of the GeForce 90 and 5090?

AI has played a pivotal role in the performance of these graphics cards. AI helps predict the missing pixels in real-time rendering, which enhances image accuracy and maintains temporal stability between frames.

What partnership was announced with GM in this keynote?

Nvidia announced a partnership with GM to build a self-driving car fleet, focusing on AI for manufacturing, enterprise, and in-vehicle infrastructure for autonomous vehicles.

How does Nvidia ensure safety in its autonomous vehicle technology?

Nvidia ensures safety by assessing every line of code, including system software and algorithms, for diversity, transparency, and explainability. This involves third-party safety assessments for over 7 million lines of code.

What was the significant change made in Nvidia’s data center architecture?

Nvidia transitioned from integrated MVLink systems to disaggregated MVLink, enhancing performance by using liquid cooling and reducing the number of components, which allows the creation of AI supercomputers in a single rack.

What is the Nvidia Dynamo, and what role does it play in AI operations?

The Nvidia Dynamo is an operating system designed for AI factories. It replaces traditional data center operating systems like VMware, orchestrating AI applications in data centers and enabling more efficient AI-driven processes.

What makes the new Nvidia architecture for AI supercomputing unique?

The new Nvidia architecture uses disaggregated MVLink, liquid cooling, and a scalable system that allows for one EXL flop of computation in a single rack. It offers unprecedented memory bandwidth and computational power.

How is Nvidia contributing to robotics and automation?

Nvidia is advancing robotics by developing Omniverse, an operating system for physical AI, and introducing Cosmos, a generative model to simulate infinite virtual environments for robot training. It also announced a collaboration with DeepMind and Disney to create a specialized physics engine called Newton for real-time robot training.

What challenges does the world face regarding labor, and how are robots expected to help?

The world is facing a severe shortage of human labor, with predictions of a 50 million worker gap by the end of this decade. Robots are expected to fill this gap, potentially earning $50,000 a year, creating a large new industry.

What is the significance of the partnership between DeepMind, Disney Research, and Nvidia for robotics?

This partnership resulted in the development of Newton, a real-time physics engine designed for training robots with fine motor skills, tactile feedback, and actuator control, using GPU acceleration for super real-time simulations.