Nvidia CEO Jenson Huang delivered his annual keynote address at GTC 2025 AT THE Sap Center in San Jose, California. Watch all the highlights from Nvidia's live event.
0:00 Intro
0:55 GeForce RTX 5090
1:48 AI Real-Time CGI Rendering
2:31 GM & Nvidia Autonomous Vehicle Partnership
4:37 Blackwell GPU Data Centers
8:21 Nvidia Blackwell System
9:03 Nvidia Dynamo
10:33 DGX Station
11:32 Nvidia Robotics
12:18 Nvidia Omniverse
14:39 Nvidia Partners with Disney and Google DeepMind
Subscribe to CNET on YouTube: https://www.youtube.com/cnet
Never miss a deal again! See CNET’s browser extension 👉 https://bit.ly/3lO7sOU
Check out CNET’s Amazon Storefront: https://www.amazon.com/shop/cnet
Follow us on TikTok: https://www.tiktok.com/@cnetdotcom
Follow us on Instagram: https://www.instagram.com/cnet/
Follow us on Bluesky: https://bsky.app/profile/cnet.com
Follow us on X: https://www.x.com/cnet
Like us on Facebook: https://www.facebook.com/cnet
CNET's AI Atlas: https://www.cnet.com/ai-atlas/
Visit CNET.com: https://www.cnet.com/
#nvidia #gtc #jensenhuang #ai #robot #robotics
0:00 Intro
0:55 GeForce RTX 5090
1:48 AI Real-Time CGI Rendering
2:31 GM & Nvidia Autonomous Vehicle Partnership
4:37 Blackwell GPU Data Centers
8:21 Nvidia Blackwell System
9:03 Nvidia Dynamo
10:33 DGX Station
11:32 Nvidia Robotics
12:18 Nvidia Omniverse
14:39 Nvidia Partners with Disney and Google DeepMind
Subscribe to CNET on YouTube: https://www.youtube.com/cnet
Never miss a deal again! See CNET’s browser extension 👉 https://bit.ly/3lO7sOU
Check out CNET’s Amazon Storefront: https://www.amazon.com/shop/cnet
Follow us on TikTok: https://www.tiktok.com/@cnetdotcom
Follow us on Instagram: https://www.instagram.com/cnet/
Follow us on Bluesky: https://bsky.app/profile/cnet.com
Follow us on X: https://www.x.com/cnet
Like us on Facebook: https://www.facebook.com/cnet
CNET's AI Atlas: https://www.cnet.com/ai-atlas/
Visit CNET.com: https://www.cnet.com/
#nvidia #gtc #jensenhuang #ai #robot #robotics
Category
😹
FunTranscript
00:00Welcome to GTC. What an amazing year. We wanted to do this at NVIDIA. So through the magic of
00:17artificial intelligence, we're going to bring you to NVIDIA's headquarters. I think I'm
00:24bringing you to NVIDIA's headquarters. What do you think? This is where we work. What an
00:42amazing year it was. And we have a lot of incredible things to talk about. And I just want
00:47you to know that I'm up here without a net. There are no scripts. There's no teleprompter. And
00:53I've got a lot of things to cover. So let's get started. GTC started with GeForce. It all
01:00started with GeForce. And today I have here a GeForce 5090. And 5090, unbelievably, 25 years
01:12later, 25 years after we started working on GeForce, GeForce is sold out all over the world.
01:18This is the 5090, the Blackwell generation. And comparing it to the 4090, it's 30% smaller in
01:28volume. It's 30% better at dissipating energy and incredible performance. Hard to even
01:38compare. And the reason for that is because of artificial intelligence. GeForce brought CUDA to
01:44the world. CUDA enabled AI. And AI has now come back to revolutionize computer graphics. What
01:53you're looking at is realtime computer graphics. 100% path traced. For every pixel that's
02:00rendered, artificial intelligence predicts the other 15. Think about this for a second. For every
02:06pixel that we mathematically rendered, artificial intelligence inferred the other 15. And it has
02:15to do so with so much precision that the image looks right. And it's temporally accurate.
02:22Meaning that from frame to frame to frame, going forward or backwards because it's
02:26computer graphics, it has to stay temporally stable. Incredible. Today I'm super excited to
02:32announce that GM has selected NVIDIA to partner with them to build their future self-driving car
02:41fleet. The time for autonomous vehicles has arrived. And we're looking forward to building
02:55with GM AI in all three areas. AI for manufacturing so they can revolutionize the way they
03:02manufacture. AI for enterprise so they can revolutionize the way they work. Design cars and
03:08simulate cars. And then also AI for in the car. So AI infrastructure for GM, partnering with GM
03:16and building with GM their AI. So I'm super excited about that. One of the areas that I'm
03:22deeply proud of and it rarely gets any attention is safety. Automotive safety. It's called
03:28halos. In our company it's called halos. Safety requires technology from silicon to system
03:42software, the algorithms, the methodologies, everything from diversity to ensuring diversity,
03:49monitoring and transparency, explainability. All of these different philosophies have to be
04:04deeply ingrained into every single part of how you develop the system and the software. We're
04:10the first company in the world I believe to have every line of code safety assessed. 7 million
04:16lines of code safety assessed. Our chip, our system, our system software and our algorithms
04:22are safety assessed by third parties that crawl through every line of code to ensure that it is
04:32designed to ensure diversity, transparency and explainability. Let's talk about data centers.
04:47That's not bad, huh? Blackwell is in full production and this is what it looks like. It's
04:56an incredible, incredible, you know, for people, for us, this is a sight of beauty. Would you
05:02agree? How is this not beautiful? How is this not beautiful? Well, this is a big deal because
05:16we made a fundamental transition in computer architecture. The thing we had to do was scale
05:25up first. This is the way we scaled up. I'm not going to lift this. This is 70 pounds. This is
05:30the last generation system architecture called HGX. This revolutionized computing as we know
05:38it. This revolutionized artificial intelligence. This is eight GPUs, each one of them is kind of
05:45like this. Okay? This is two GPUs, two Blackwell GPUs in one Blackwell package. Two Blackwell
05:55GPUs in one Blackwell package. There are eight of these underneath this. Okay? And this
06:04connects into what we call MV link 8. This then connects to a CPU shelf like that. So there's
06:11dual CPUs and that sits on top and we connect it over PCI express and then many of these get
06:20connected with InfiniBand which turns into what is an AI supercomputer. This is the way it was
06:28in the past. We need to disaggregate the MV link system and take it out. This is the MV link
06:34system. This is an MV link switch. This is the highest performance switch the world has ever
06:41made. This makes it possible for every GPU to talk to every GPU at exactly the same time at
06:48full bandwidth. This is the MV link switch. We disaggregated it. We took it out and we put it
06:56in the center of the chassis. So there's all the 18 of these switches in nine different racks.
07:05Nine different switch trays we call them. And then the switches are disaggregated. The compute
07:13is now sitting in here. This is equivalent to these two things in compute. What's amazing is
07:20this is completely liquid cooled and by liquid cooling it we can compress all of these compute
07:27nodes into one rack. This is the big change of the entire industry. All of you in the audience,
07:34I know how many of you are here, I want to thank you for making this fundamental shift from
07:40integrated MV link to disaggregated MV link, from air cooled to liquid cooled, from 60,000
07:47components per computer or so to 600,000 components per rack. 120 kilowatts fully liquid
07:53cooled and as a result we have a one exaflops computer in one rack. Isn't it incredible? The
07:59way to solve this problem is to disaggregate it as I described into the MV link switch tray.
08:29This is the most extreme scale up the world has ever done. The amount of computation that's
08:43possible here. The memory bandwidth, 570 terabytes per second. Everything in this machine is now in
08:49T's. Everything is a trillion. And you have an exaflops which is a million trillion floating
09:02point operations per second. Today we're announcing the NVIDIA Dynamo. NVIDIA Dynamo does all
09:08that. It is essentially the operating system of an AI factory. Whereas in the past, in the way
09:23that we ran data centers, our operating system would be something like VM ware and we would
09:28orchestrate, and we still do, you know, we're a big user, we would orchestrate a whole bunch of
09:33different enterprise applications running on top of our enterprise IT. But in the future, the
09:42application is not enterprise IT, it's agents. And the operating system is not something like VM
09:48ware, it's something like Dynamo. And this operating system is running on top of not a data
09:53center, but on top of an AI factory. Now, we call it Dynamo for a good reason. As you know, the
10:00Dynamo was the first instrument that started the last industrial revolution. The industrial
10:07revolution of energy. Water comes in, electricity comes out. It's pretty fantastic. You know, water
10:13comes in, you light it on fire, turn it to steam, and what comes out is this invisible thing that's
10:19incredibly valuable. It took another 80 years to go to alternating current, but Dynamo. Dynamo is
10:26where it all started. So we decided to call this operating system, this piece of software,
10:30insanely complicated software, the NVIDIA Dynamo. This is what a PC should look like. 20 petaflops.
10:41Unbelievable. 72 CPU cores, chip to chip interface, HBM memory, and just in case, some PCI express
10:47slots for your GeForce. Okay? So this is called DGX Station. DGX Spark and DGX Station are
11:01going to be available by all of the OEMs. HP, Dell, Lenovo, Asus, it's going to be manufactured
11:09for data scientists and researchers all over the world. This is the computer of the age of AI.
11:17This is what computers should look like, and this is what computers will run in the future. And we
11:21have a whole lineup for enterprise now, from little tiny one to workstation ones, the server ones to
11:29supercomputer ones, and these will be available by all of our partners. So let's go talk about
11:34robotics, shall we? Let's talk about robots. Well, the time has come, the time has come for
11:40robots. Robots have the benefit of being able to interact with the physical world and do things
11:52that otherwise digital information cannot. We know very clearly that the world has severe shortage
11:59of human laborers, human workers. By the end of this decade, the world is going to be at least 50
12:06million workers short. We'd be more than delighted to pay them each $50,000 to come to work. We're
12:12probably going to have to pay robots $50,000 a year to come to work. So this is going to be a very,
12:17very large industry. We created a system called Omniverse. It's our operating system for physical
12:22AIs. You've heard me talk about Omniverse for a long time. We added two technologies to it. Today
12:29I'm going to show you two things. One of them is so that we could scale AI with generative capabilities
12:38and generative model that understand the physical world. We call it Cosmos. Using Omniverse to
12:46condition Cosmos and using Cosmos to generate an infinite number of environments allows us to create
12:52data that is grounded, grounded, controlled by us, and yet be systematically infinite at the same
13:05time. So you see Omniverse, we use candy colors to give you an example of us controlling the robot in
13:14the scenario perfectly, and yet Cosmos can create all these virtual environments. The second thing
13:22is reinforcement learning. Just as we were talking about earlier, one of the incredible scaling
13:28capabilities of language models today is reinforcement learning, verifiable rewards. The question is
13:34what's the verifiable rewards in robotics? And as we know very well, it's the laws of physics.
13:42Verifiable physics rewards. So we need an incredible physics engine. Well, most physics engines
13:49have been designed for a variety of reasons. They can be designed because we want to use it for
13:55large machineries or maybe we design it for virtual worlds, video games and such, but we need a
14:01physics engine that is designed for very fine-grained rigid and soft bodies, designed for being able
14:11to train tactile feedback and fine motor skills and actuator controls, and designed for
14:19performance. We need it to be GPU accelerated so that these virtual worlds could live in super
14:25linear time, super real time, and train these AI models incredibly fast, and we need it to be
14:32integrated harmoniously into a framework that is used by roboticists all over the world. And so
14:40today we're announcing something really, really special. It is a partnership of three companies.
14:48DeepMind, Disney Research, and NVIDIA, and we call it Newton.
14:58Let's take a look at Newton.
15:08Tell me that wasn't amazing.
15:12Hey, Blue. How are you doing?
15:15How do you like your new physics engine? You like it, huh? Yeah, I bet. I know.
15:22Tactile feedback, rigid body, soft body simulation, super real time.
15:30Can you imagine just now what you were looking at is complete real time simulation?
15:36This is how we're going to train robots in the future. Just so you know, Blue has two computers.
15:42Two NVIDIA computers inside. Look how smart you are. Yes, you're smart.
15:52Okay. All right. Hey, Blue, listen. How about let's take them home. Let's finish this keynote.
15:59Our robotics has been making enormous progress. And today we're announcing that Groot N1 is
16:05open sourced. Well, have a great GTC. Thank you. Hey, Blue. Let's go home. Good job.