NVIDIA CEO Jensen Huang Delivers Major AI Keynote in Taiwan | AC1W

Added: 2026-06-02

[00:00:00]the language of chip design.

[00:00:02]RTL must be verified in simulation.

[00:00:05]A single bug can delay a chip by months.

[00:00:08]At Nvidia, thousands of engineers, billions of compute hours per year, millions of tests [music] written, run, and debugged. A cycle that takes teams weeks.

[00:00:18]To compress [music] this cycle, Cadence and Nvidia built a design verification agent.

[00:00:23]Codex orchestrates the process.

[00:00:25]Cadence ChipStack launches the RTL [music] verification loop.

[00:00:29]Powered by NeemaTron and secured by Nvidia [music] Open Shell.

[00:00:33]Calling on expert sub-agents in RTL generation, test bench creation, regression testing, and debug.

[00:00:41]The system drives itself.

[00:00:43]The ChipStack agents [music] run hundreds of simulations with Cadence Xcelium, formal verification with Jasper.

[00:00:50]Design flaws revealed, bugs in the code fixed.

[00:00:54]What once took weeks, now takes [music] hours.

[00:00:57]Verification cycles over 40 times faster.

[00:01:01]Together, Nvidia and Cadence [music] are reinventing chip design with AI agents.

[00:01:08]>> From weeks from weeks to hours.

[00:01:12]From weeks to hours. From weeks to hours. Nvidia has thousands of chip designers. We are going to hire hundreds of thousands of Cadence super agents that work with us, so that we can accelerate our company, so that we can be even more ambitious, create even more amazing things, run even faster. You saw earlier that the toolkit with models, harness, tools.

[00:01:41]The tools in this case are Cadence simulators and verifiers, formal verification systems. It is the reason why we're working with Cadence so hard to accelerate all of their tools on CUDA.

[00:01:53]Because the agents are impatient. The agents want the answer immediately.

[00:01:58]And so models harnesses accelerated CUDA accelerated libraries and tools and then the runtime.

[00:02:08]What you saw just now is all of that coming together.

[00:02:12]Now, one of the things that it starts with is a great model that Cadence could modify and tune to be expert at the Cadence workflow at the Cadence expertise so that they could create super agents that are proprietary to Cadence with their proprietary knowledge.

[00:02:32]They have to start with an excellent model. We call it NeMoTron.

[00:02:36]Nvidia's dedicated to build open models for the world so that all of you, all of us could create our own agents.

[00:02:45]Today, we're announcing the NeMoTron 3 Ultra.

[00:02:52]Yep, our next open model. And it is smart.

[00:02:58]>> [applause] >> The NeMoTron models not only give you the model we give you all the data that we used to train the model and because we have a coalition of incredible partners, you can see all of our partners down here.

[00:03:16]We work together contribute data to each other.

[00:03:21]NeMoTron is trained on one of the largest suites of long-running reasoning models, long-running tool task-solving, tool-using data sets in the world.

[00:03:33]Because of all of our great partnerships all of this from the model the training script and the data made all completely available to you. This is open models at its best.

[00:03:47]The best open model system policies in the the Simple goal is so that you can take all of it, add to it, make it even better, make it yours.

[00:03:58]NeMo Tron 3 Ultra is five times faster.

[00:04:04]This is the world's first model based on a hybrid architecture of SSM state space models with mixture of experts. The architecture is incredibly fast. We made it fast so that you could think fast. When you think fast, you can think longer at the same cost.

[00:04:22]So, five times faster.

[00:04:25]It is also 30% cheaper.

[00:04:29]30% lower cost to run in total flops and total inference time than even the most cost-effective in the world.

[00:04:37]We're comparing against the world's best open models.

[00:04:40]Frontier smart.

[00:04:43]Five times faster.

[00:04:45]30 30% cheaper.

[00:04:47]Completely open.

[00:04:49]We're completely dedicated to this. This is now NeMo Tron 3.

[00:04:53]We're currently working on NeMo Tron 4.

[00:04:56]So, this entire toolkit from models, harnesses, tools and skills, and runtimes is the reason why every enterprise company in the world has the ability now to create their own agents just like Cadence did with their super agents. And we're working with so many companies.

[00:05:18]Cadence and CrowdStrike and Dassault and Palantir, SAP and ServiceNow.

[00:05:23]People were always said, "Jensen, the agents are going to disrupt these markets."

[00:05:29]I said completely opposite and you can now see it.

[00:05:33]Agents is going to create the largest opportunity ever for my partners and friends.

[00:05:39]And we have the NeMo the the NVIDIA agentic toolkit for enterprise AI to help them.

[00:05:47]So, there you go.

[00:05:51]>> [applause] >> First, Vera Rubin in full production.

[00:05:57]Two, Vera CPU. CPU built for a new generation for agents.

[00:06:03]And three, NVIDIA's enterprise AI toolkits, so that every enterprise and every enterprise software company can build agents.

[00:06:21]>> [applause] >> My relationship with you started here.

[00:06:29]And many of you, many of you, many of my friends and partners here in Taiwan, your companies started here.

[00:06:37]This is in a lot of ways the beginning of the modern computer industry, 40 years now.

[00:06:45]NVIDIA is 33 years old.

[00:06:47]The PC industry was already starting to get to peak Windows 1 and Windows 2 and Apple Apple 1 and Apple 2 and by the time that we came along, Windows 3.1 was the PC.

[00:07:02]And as you know, Windows 95 made PC personal.

[00:07:06]It took PC from enterprises, com- companies, and made it into a consumer electronics device.

[00:07:14]Everybody should have one, and everybody does.

[00:07:17]This is the beginning.

[00:07:18]This computing platform did several things incredibly smart.

[00:07:24]Windows was not just disaggregated, as you know.

[00:07:27]Windows was properly abstracted. It was architected just right.

[00:07:32]Systems BIOSes, open chipsets, the operating system with drivers, drivers that could be connected and installed at runtime.

[00:07:45]And an abstraction layer with a multimedia API that was that opened up the PC to what we all know today.

[00:07:56]Each one of these elements were essential in making the PC so popular.

[00:08:02]40 years later, Microsoft and Nvidia are going to reinvent the PC.

[00:08:10]This is going to be the new PC.

[00:08:13]Now, tomorrow night tomorrow night, I think it's tomorrow night our time, but I'm going to be with Satya. We're going to talk a lot more about the work that we're doing together.

[00:08:23]Microsoft and Nvidia over the last 3 years, it took this long to completely reinvent how the PC's going to work so that we could be ready for this moment.

[00:08:34]As I mentioned earlier, that compute pattern called the agent is going to run in AI clouds.

[00:08:42]It's going to run inside enterprises.

[00:08:45]It is also going to run on your PC.

[00:08:49]What's going to happen to that PC when it has an autonomous agent?

[00:08:53]An agent that's helping you, that understands you.

[00:08:56]You could talk to it. It could look at you. You could ask it to read files, go help you do some research.

[00:09:05]It could do a lot more that I'll show you.

[00:09:08]But the new operating system is of course the old operating system plus large language models.

[00:09:16]Large language models in a lot of ways is the modern version of DirectX.

[00:09:22]It has of course input and output, understands prompts, it understands computer vision, it can generate video, it can generate sounds. It is the modern extension, the intelligence extension of the PC, of a computer.

[00:09:37]On top of that, the application, as I mentioned before, is going to be replaced by now an agentic runtime.

[00:09:46]And that is the modern application, an agent.

[00:09:49]Let's now take a look at what it can do.

[00:09:54]>> It started with a spark.

[00:09:56]An idea to reimagine the PC for the first time in 40 years for the age of AI.

[00:10:05]What becomes of our personal computer in [music] a world of agents?

[00:10:10]Agents running natively, connected to models, [music] local or in the cloud. Our personal AI, sandboxed for [music] security, running continuously, getting work done.

[00:10:23]The chips and the OS must evolve.

[00:10:27]Introducing RTX Spark. Everything we've learned over 33 years distilled into one chip.

[00:10:36]Blackwell RTX [music] GPU with 6,144 CUDA cores, 1 petaflop [music] of AI performance, a custom 20-core Grace CPU built in partnership with MediaTek, fused by NVLink, 128 [music] GB of unified memory, TSMC 3 nanometer process, 70 billion [music] transistors.

[00:11:04]And in close collaboration with Microsoft, a Windows platform for agents.

[00:11:10]We're reinventing the personal computer.

[00:11:14]For creating, for gaming, for agents.

[00:11:22]This is the dawn [music] of a new personal computing revolution.

[00:11:26]And it starts with NVIDIA RTX Spark.

[00:11:36]>> [applause] [applause] >> Here it is.

[00:11:47]Of course, I got to show you the most beautiful part, which is video games.

[00:11:52]It is It's also the closest to our heart.

[00:11:55]This is Forza.

[00:11:57]This is 007, by the way.

[00:12:00]The new 007 game, I'm looking forward to playing it.

[00:12:03]I look a little bit like him.

[00:12:06]Ladies and gentlemen, Nvidia's RTX Spark laptops. Now, >> [applause] >> thank you.

[00:12:22]I have too many things in my pocket.

[00:12:28]Okay.

[00:12:29]All right. This is the most amazing chip the world has ever built.

[00:12:33]This is the N1X that we built in partnership with MediaTek. I think I saw I saw Rick earlier. This is N1X. This is a beautiful chip. And this is This is a a a chip that frankly would take 33 years to build. And the reason for that is because 100% of Nvidia's software stack runs here.

[00:12:56]If you want to run uh uh digital biology, no problem. If you want to do seismic processing, no problem. You want to astrophysics, no problem. Everything associated with CUDA, all the physics, all the biology, all the genomics, all the AI, no problem. All the computer graphics, no problem.

[00:13:13]Every single application Nvidia has ever created, and every single application that Windows has ever run.

[00:13:23]Microsoft and Nvidia meticulously optimized everything so that this computer literally runs everything the world has ever created plus it now runs agents.

[00:13:37]An incredible [snorts] computer. I'm so proud of it.

[00:13:42]>> [applause] >> Okay.

[00:13:47]Now, I want you to keep that in mind in the next video I just I'm going to show you.

[00:13:53]Just imagine everything here is going to run on your PC. Now, that computer could have a local Nemetron 3 Ultra model or Nemetron 3 super model or it could have a Claude code or Codex or some other model in the cloud or something on the network and it's going to it's going to work and do something amazing. Let's play it.

[00:14:18]>> Every house [music] starts as an idea.

[00:14:21]Getting from idea to design takes a myriad of tools, expertise, and a lot of time.

[00:14:30]Now, an [music] agent running locally on RTX Spark can help me design a house using the tools on my laptop with an open shell sandbox running the Hermes harness >> [music] >> connected to Claude Sonnet in the cloud.

[00:14:43]I select the site, share my concept sketches, and mood [music] board of styles to inspire my design and the prompt, a text description of the requirements, and the design intent.

[00:14:58]My agent goes to work using [music] the tools on my laptop. It opens Rhino and starts modeling the site, shaping terrain, setbacks, >> [music] >> and the building envelope.

[00:15:09]Then it proposes building forms optimized for [music] cost, comfort, and quality.

[00:15:16]With the form defined, my agent generates the interior [music] layout.

[00:15:20]Walls, circulation, rooms begin to take shape. I jump in whenever I want [music] to adjust, to change.

[00:15:30]Doors, windows, and structural elements are placed [music] automatically. My agent detects its own mistakes and fixes them.

[00:15:42]When I approve, the agent exports the model from Rhino into [music] Blender.

[00:15:46]Materials and object properties transfer with the design context [music] intact.

[00:15:52]I fine-tune the materials, get the look just right. Then, I pick the shots.

[00:15:58]Blender renders [music] the house. My agent, using generative AI with the Flux 2 model, makes them photoreal. Multiple viewpoints, lighting conditions. What was once a complex [music] workflow is now guided and simplified by my agent, working with me on RTX Spark.

[00:16:16]Design at the speed of imagination.

[00:16:25]>> [applause] >> PC in the world of agents.

[00:16:28]The developers are so excited by this is an incredible computer. All of the acceleration, all the software capabilities is associated with it, working with every developer to make it incredible for all of you.

[00:16:41]The next one, Adobe.

[00:16:44]Incredible tool suite, of course, used by tens of millions of people around the world. They have re-engineered the architecture, the core of Adobe Photoshop and Premiere, and they'll release it for RTX Spark. It is twice as fast. It's already fast.

[00:16:59]Now, it's going to be twice as fast. And it's also designed to be agent-friendly.

[00:17:05]With its MCP server, it can now interact with agents on your laptop.

[00:17:10]The number of customers, the number of partners that are so excited to bring RTX RTX Spark to the market is just incredible.

[00:17:19]You know, this is the first across the lineup of PC reinvention for 40 years.

[00:17:27]And I'm just so happy that all of you and the ecosystem around the world has joined us.

[00:17:33]This is basically everybody. Everybody will support RTX Spark and will be building incredibly smart and powerful and beautiful laptops with all of us.

[00:17:43]Thank you very much.

[00:17:46]>> [applause] >> But that's not all. That's not all.

[00:17:54]RTX Spark is a reinvention of laptop.

[00:17:58]But in fact, Microsoft and video is reinventing all of PC. And today we're announcing a whole new line.

[00:18:07]Three revolutionary Windows machines covering desktop, laptop, and workstations. All 100% Windows compatible, 100% CUDA, 100% NVIDIA AI Tensor Core.

[00:18:23]Everything that runs that you see that runs on NVIDIA in all these different platforms around the world runs here.

[00:18:30]This is the first completely re-engineered, reinvented line of PCs that has happened in 40 years.

[00:18:39]Now, what's really amazing is this. So, this is this is the RTX Spark laptop.

[00:18:44]This is the desktop.

[00:18:48]So, this one's from MSI.

[00:18:50]Joseph, this one's yours.

[00:18:52]Okay.

[00:18:53]Look how beautiful it is. This agent could run 24/7.

[00:18:58]Meter free.

[00:19:00]And you could download your agent. You could raise your laps >> dryer, your water cooler, your water heater, your everything, whatever you want. Your security system, all connected to this, and this becomes your personal AI, your personal AI agent.

[00:19:35]And it gets smarter and smarter and smarter over time because today we have Nemotron 3 Ultra.

[00:19:41]Tomorrow we have Nemotron 4, and then Nemotron 5, Nemotron 6, and we just keep getting it smarter and smarter and smarter.

[00:19:47]And meanwhile, this is sitting at home helping you do things. If you want to book a travel, no problem.

[00:19:53]And if you if you want an incredible system, this is a DGX Station for Windows compatible with Windows once everything in Windows, and and it has 768 GB of memory.

[00:20:14]And so you could run a trillion parameter model.

[00:20:18]This is unbelievable. 20 petaflops, 8 TB per second of memory bandwidth, and this sits by your desk.

[00:20:28]You basically if you're a developer of large language models, you're a developer of agents, having this sit by your desk gives you all the compute you need, and then when you deploy it, you put it into the cloud.

[00:20:41]Now, there's something that if you look at this and think about this, something is happening here.

[00:20:48]Remember, 15, 20 years ago, we used to have an idea called a phone.

[00:20:56]Today, we have an idea called a PC.

[00:21:01]Today, when you think about your phone, the one thing you don't do with it is make phone calls.

[00:21:10]You do just about everything else.

[00:21:13]And so that phone means something very different to you than a phone of the past.

[00:21:19]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 where do you launch applications click and type.

[00:21:33]And this PC is going to be completely different.

[00:21:37]Here's my theory.

[00:21:39]I can totally imagine just as every house today has a home theater or many houses have home theaters, big TVs lawn mowers dishwashers.

[00:21:53]I could totally imagine that someday there's actually an AI supercomputer in your house.

[00:21:59]And it's running all of your agents, it's running all of your assistants. And they're doing all kinds of things for you all the time.

[00:22:07]And you have to have it in your house just like you have a home theater in your house, you have stereos in your house, you have game consoles in your house.

[00:22:15]You want to assist AI agent computers running in your house.

[00:22:19]And these in time becomes a lot more like R2-D2 to you.

[00:22:26]It becomes more like C-3PO to you than it feels like a PC to you.

[00:22:33]There is no question this reinvention of the computer is as big of a deal as the reinvention of the phone into what we now know as the smartphone. And so this is the beginning of that journey. This is the beginning of a new line. And so we have a road map for this. This is a brand new product family for us.

[00:22:53]Every single generation of architecture we will have a desktop a laptop, a workstation, and then a desktop, a laptop, and workstation. And the thing that I'm just incredibly pleased, incredibly honored, is that 100% of the world's PC industry has joined us to reinvent the PC. A new line, a new beginning. Thank you.

[00:23:19]>> [applause] [applause] >> As you know, agentic AI is just a digital robot.

[00:23:39]It understands, it reasons, it plans, and it acts and uses tools.

[00:23:46]Agentic AI is going to run across all of these computers.

[00:23:51]And you've seen me talk about each and every one of these over time.

[00:23:54]We're working on humanoid robotics computers, robotics computers of all kinds. We're working on self-driving car computers. We're working on satellites.

[00:24:04]You have GeForce, which is has tensor cores. I just talked about a whole new line of PCs.

[00:24:09]Agriculture equipment, manufacturing equipment, heavy industry equipment will all be agentic.

[00:24:16]You'll even have a little agentic helper for yourself.

[00:24:21]Even your base stations, the radio stations of the future, are going to be agentic.

[00:24:26]Understanding traffic and thinking about how to coordinate with the other base stations so that you could use as little energy as possible, increase the utilization, the efficiency of the spectral efficiency.

[00:24:41]And so, everything will run agents.

[00:24:43]Today, NVIDIA is largely in the center.

[00:24:47]But I am pretty certain that there will be tens of billions, hundreds of billions over time, of agentic systems, agentic computers that are going to be running around the world. The biggest problem is data.

[00:25:02]In the case of language models, all the English and all the language that we have on the internet that we trained on was from the perspective of us. We wrote it and we're reading it.

[00:25:13]However, in order to create a data for AI robotics, it has to be in the perception, the perspective of the robot. And most of the world's video data is from a third person, not first person.

[00:25:28]And so, agentic systems, robotic systems, physical AI, the data is the hardest problem.

[00:25:36]You've seen us move up this ladder. We started with teleoperations, which is basically human demonstration. This is no different than the big breakthrough of reinforcement learning human feedback.

[00:25:48]This then we use simulation. This is where Omniverse comes in. This is no different than reinforcement learning verif- verifiable rewards. Okay? And so, we use these systems to bootstrap the AI model, the physical AI model.

[00:26:06]Eventually, we're able to learn from third per- third person, reprojecting it into first person. And now, eventually, through bootstrapping, we have a world foundation model that can understand the physical world from any perspective you want. Third third person, first person, outside in, inside out, doesn't matter.

[00:26:29]This is a big breakthrough indeed. And today, we are announcing Cosmos 3.

[00:26:38]Cosmos 3 is the frontier of physical AI.

[00:26:43]We are at the frontier with language models. There are so many people working on it. However, in physical AI, we are absolutely the world's best. I am so proud of the team for doing this. This is the foundation model for all of your work. 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 there in that involves physical world, you now have a companion, a Cosmos 3 that can understand and reason.

[00:27:16]It can generate.

[00:27:17]It can simulate in the loop. It can even be the policy itself. It is on the top of leaderboards all over the all over the world. I am incredibly proud of Cosmos, and today we're announcing Cosmos 3. Let's take a look.

[00:27:32]>> The real world [music] is infinite and a