Nvidia’s Computex 2026 Keynote in Less Than 12 Minutes

[00:00:00] Welcome to the stage, NVIDIA founder and CEO, Jensen Wang. [00:00:11] Welcome to GTC Taiwan. [00:00:17] The wave of AI was agentic AI. [00:00:20] And today, we can say that agentic AI has arrived, that useful AI has arrived. [00:00:26] People talk about AI reducing jobs, complete nonsense. [00:00:31] It's causing more software engineers to be hired. [00:00:34] And the reason for that is very simple. [00:00:36] So let me take a quick moment and show you what I'm talking about. [00:00:39] Inside, this is an agent. [00:00:42] It's an agent application. [00:00:45] It is agent which consists of a large language model or many sitting inside a harness. [00:00:53] And that harness helps it, orchestrates it, to do productive work. [00:00:58] This entire system is called an agent. [00:01:02] This is now the new computing pattern. [00:01:04] Whereas we used to launch an application, click and type. [00:01:11] We now replace that with explaining to the AI what we want, our intent, and the AI generates the code or uses tools and produce the necessary output. [00:01:24] It is the ultimate disaggregated and distributed computing model. [00:01:32] Today's agents are relatively simple users of tools. [00:01:38] Tomorrow, they're going to be very sophisticated users of tools. [00:01:41] They solve some of the most important problems the world knows. [00:01:45] The memory system of AIs is going to cause the storage system to be completely revolutionized. [00:01:51] It's precisely the reason we built our next generation. [00:01:57] Verirubin. [00:01:59] Verirubin is not one chip. [00:02:01] This entire thing is Verirubin. [00:02:04] From end to end. [00:02:06] It has GPUs. [00:02:09] Verirubin, NVLink 72. [00:02:11] Each one of these systems would be a complete revolution in itself. [00:02:16] Verirubin is the most ambitious endeavor in the history of our company. [00:02:20] A long time ago, NVIDIA used to be a GPU company. [00:02:24] But over the years, we've evolved to become a systems company. [00:02:29] You're looking here now for the most complex systems, most complex and ground-up system ever designed. [00:02:37] But ultimately, our customers, our partners, don't want to buy computers. [00:02:44] They want to build AI factories. [00:02:46] Which is the reason why NVIDIA has really started to transform ourselves yet again. [00:02:52] We are building a full-stack system so that our customers could build amazing AI infrastructure. [00:02:59] I'm so happy. [00:03:01] Here I am standing in front of you. [00:03:04] Verirubin is in full production. [00:03:10] This is Verirubin. [00:03:16] Verirubin NVLink 72. [00:03:18] This is Verirubin CPU rack. [00:03:21] 256 CPUs, all liquid-cooled. [00:03:25] This is the Verirubin Bluefield storage processing system. [00:03:30] No cables, no hoses, no fans. [00:03:35] There's a PCB in the middle, which connects both sides. [00:03:39] What used to take two hours, now takes five minutes. [00:03:42] A CPU the world has never seen before. [00:03:45] We call it Verirub. [00:03:47] This is a CPU for agents. [00:03:51] All the CPUs of the past, we built for humans. [00:03:56] This CPU is built for agents. [00:03:58] Now, inside our system, it's used for three different ways. [00:04:02] The first way, of course, is Verirubin for thinking. [00:04:09] And inside the Verirubin rack, there are already two CPUs. [00:04:14] The Verirubin compute is used for the harness, the orchestration of the AI models, tool use, accessing the database. [00:04:24] And the data servers are right here, Verirubin Bluefield, the fastest storage system the world has ever made. [00:04:32] This is the first CPU that's been built a long time, that is literally at radical limits, with a fabric that connects all of the CPU cores that is Speedolite. [00:04:43] 3.6 terabytes per second. [00:04:46] NVIDIA VERA is the CPU built for the agentic loop. [00:04:50] VERA is the CPU for the age of agents. [00:04:56] The takeaway here is that this is the application pattern. [00:04:59] This is the computing pattern of the next decade. [00:05:03] And so we have the NVIDIA agent toolkit for enterprise AI. [00:05:09] You can run agents from anybody. [00:05:13] You could run Cloud Code, incredible agent, Codex, incredible agent. [00:05:18] You could run it inside this harness called Open Shell, which will be highly secure for your inside the enterprise. [00:05:25] NVIDIA is dedicated to build open models for the world. [00:05:29] So that all of you, all of us, could create our own agents. [00:05:34] Today we're announcing the NemoTron 3 Ultra. [00:05:41] Yep. [00:05:42] Our next open model. [00:05:44] And it is smart. [00:05:45] All of this from the model, the training script, and the data made completely available to you. [00:05:54] Microsoft and NVIDIA are going to reinvent the PC. [00:05:58] Introducing RTX Spark. [00:06:01] Everything we've learned over 33 years, distilled into one chip. [00:06:08] Blackwell RTX GPU with 6,144 CUDA cores. [00:06:13] One petaflop of AI performance. [00:06:16] A custom 20-core Grace CPU built in partnership with MediaTek. [00:06:22] Refused by MVLink. [00:06:25] 128 gigabytes of unified memory. [00:06:28] And in close collaboration with Microsoft, a Windows platform for agents. [00:06:34] We're reinventing the personal computer. [00:06:37] Here it is. [00:06:40] Of course, I got to show you the most beautiful part, which is video games. [00:06:46] It is also the closest to our heart. [00:06:48] This is the most amazing chip the world has ever built. [00:06:51] Every single application NVIDIA has ever created. [00:06:55] And every single application that Windows has ever run. [00:07:00] Microsoft and NVIDIA meticulously optimized everything so that this computer literally runs everything the world has ever created. [00:07:11] Plus, it now runs agents. [00:07:14] This is the first completely re-engineered, reinvented line of PCs that has happened in 40 years. [00:07:23] Everything in Windows. [00:07:25] And it has 768 gigabytes of memory. [00:07:31] And so you could run a trillion parameter model. [00:07:35] This is unbelievable. [00:07:36] 20 petaflops. [00:07:38] 8 terabytes per second of memory bandwidth. [00:07:41] And this sits by your desk. [00:07:43] 15, 20 years ago, we used to have an idea called a phone. [00:07:49] Today, we have an idea called a PC. [00:07:53] Today, when you think about your phone, the one thing you don't do with it is make phone calls. [00:08:02] You do just about everything else. [00:08:05] And so that phone means something very different to you than a phone of the past. [00:08:12] I am certain what's going to happen here is that the PC 10 years from now and the PC that you think about today, a tool, whether you launch applications, click and type. [00:08:26] And this PC is going to be completely different. [00:08:29] You want to assist AI agent computers running in your house. [00:08:33] And these, in time, becomes a lot more like R2D2 to you. [00:08:41] It becomes more like C3PO to you. [00:08:43] We will have a desktop, a laptop, a workstation, and then a desktop, a laptop, and workstation. [00:08:51] And the thing that I am just incredibly pleased, incredibly honored, is that 100% of the world's PC industry has joined us to reinvent the PC. [00:09:02] A new line, a new beginning. [00:09:05] And today, we're announcing Cosmos 3, the foundation model for all of your work. [00:09:09] Whenever you want to create a robot, whenever you want to create a factory robot or a robot that works in a factory, any kind of robot that involves physical world, you now have a companion. [00:09:22] Cosmos watches the physical world, understands what's happening, describing scenes, and flagging what matters. [00:09:31] As a world model, Cosmos generates physics-accurate synthetic video from an image, text, or video. [00:09:39] Cosmos, the foundation for developers of the age of physical AI. [00:09:45] Today, we're announcing Alpa Mile 2, an open model for self-driving cars. [00:09:52] We're working with car companies across the world. [00:09:56] If you look at these brands that have signed up for the NVIDIA Hyperion, that are building NVIDIA Hyperion cars, this represents about 80% of the world's cars. [00:10:09] Hey Mercedes, let's go to my favorite sandwich shop. [00:10:13] Routing to your destination. [00:10:16] Lane is clear, pulling out to start drive. [00:10:19] Nudge left due to the stationary lead vehicle ahead, blocking our lane. [00:10:24] Slow down to stop at the stop sign, controlling the intersection. [00:10:27] Alpa Mile. [00:10:29] The world's first reasoning autonomous vehicle. [00:10:34] The NVIDIA Isaac Groot is our human or robotic stack. [00:10:38] Data generation, simulation, the runtime, including the operating system. [00:10:47] This represents Groot platform, the Isaac Groot platform. [00:10:53] Every one of our systems, as you can see, the exact same pattern, whether it's a genetic system for the cloud, a genetic system for the PC, a robotic system for a self-driving car, a robotic system for a human or robot, all the same. [00:11:07] We need a reference platform for robotic systems. [00:11:11] These robotic systems are so complicated, so many motors, so many sensors, so fragile, and yet we need to have a way to deliver these reference platforms, just like we do with PCs and DGXs and clouds and self-driving cars. [00:11:28] We now are going to do it for robots. [00:11:30] NVIDIA Isaac Groot, an open development platform for humanoid robots, open models, simulation and training libraries and data generators, plus the robot computer, fully pipe clean, ready to go in hours. [00:11:46] NVIDIA Isaac Groot: Have a great Computex. [00:11:49] NVIDIA Isaac Groot: Thanks for an amazing year. [00:11:52] NVIDIA Isaac Groot: Thank you for all your friendship and support. [00:11:55] NVIDIA Isaac Groot: Thank you. [00:11:55] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care. [00:11:56] NVIDIA Isaac Groot: Take care.