Never, even when every little is a reason, I'm still never stop trying. Strangers over... dreams went quiet, lost the life, but I learnt to glow in the shadows. Talk to the mirror like a coach, get back up, all she wrote. So I run, every beat in my chest, but this don't accept any regret. Not even when you see in my eyes. Never stop reaching, even when my hands fall. Every little is a reason, I'm still trying. If I fall tonight, on my side, like a fighter for my life, singing one more time, one more time. I'm never gonna stop trying, even when they keep... you can see the fear in my eyes. Never stop trying, even when my strength... every little is a reason. Never stop, never quit. Now on my...
Face the sun. Rise up, light the fire, the fire... freed my spirit, so catch me in the... and up, oh, the fire, oh, oh, oh, I'm the... I'm the...
Write your name on my life. This is where I stay. I can't wait to see... I can't wait for that first kiss. What if... because I can see. You say you always smile, even when you see my face, just to let go... to see this first kiss. Because I can't see this. What if all my daydreams, what if I believe on the outside, and I waited all this time? Can't wait for this. I can't wait to look... first kiss, whatever... I can't wait to see...
Up the back, every sideways look when the race runs. I carried every... now I'm louder than the flame, you can write it on the page. Everything changed, was made today, from every little heartbreak to all of... I almost walked away, and I turned it into a name. Remember this place when the crowds stand up, and a legend was made today. A legend was made today. Hands up when I cross that line, flash bulbs in my open eyes. But the loudest thing I heard was my own voice saying 'right'. All the times I came up played back... now just the growing steps I took to reach this day. I carried every... now I'm the... you can write it on the page... everything, a legend was made today, from every little... every heartbreak took all the... almost away, and I remember this face, this place, when the crowds stood up... was made today. A legend was made today.
Yeah, you feel that? You are... just keep, just keep it up, super... you just keep a few things, everybody, for the newsroom people. I don't know how you go, you deserve it. Put in work for someone. You get what you want when you... when you say you can have the world. If you're on the edge of who you are, the mistake is up, you are nowhere. Don't you see your chance? Don't let look too good on your resume when you are... You just keep getting, just keep getting better. You just keep getting better, just keep getting better. Don't you get it, you are, you just... don't you got a number for the top.
Is everything? I don't know, how many minutes do we have? Midnight, lights on the floor, I heard people don't dance anymore when it comes. Ignore people you like, they never seen real before, but you don't, you don't even... you see your chance, don't hesitate. Look too good on your resume when that plays... you are, you just keep, you just keep... Yeah.
Coffee stain on my shirt, drink, playlist, thinking, is this... oh, I'm worse... up my screens since the back fits my... keep your eyes wide now. I can't look away, you're gonna say something... the first kiss... today, heart behind home, like praying you in the air... like... didn't think the world could... but the day when I saw you say something...
Maybe strange, maybe you... for this time. Maybe it's just... I believe you can see something... to the sky... be something... a full tank, sun burns as you laugh, boy, this is your big plans...
My girl, that's the first... We ain't even hit that, we ain't even killed that. If you think this is a party, you ain't seen nothing. Got it back, got suns, got your hand in mine, kiss my baby, this is just... you ain't seen nothing, you ain't seen all... shirts for a blanket that you love... you said this is like the end. Now this is the preview. We ain't even hit, we ain't even... if you think you know, I'll make you my home, you ain't seen nothing. We just got back in... this is the first. You seen nothing, you seen nothing... little boots by the front, jersey with my last name. If you think this love is crazy, you ain't seen the best of this, you ain't seen nothing. We just went from the back to the front porch to a crowded church, and every year, every sweet wind that drifts... this is just the first. You ain't seen nothing, you ain't seen nothing.
You seen for days thinking I don't wake up... I don't need a love, put a smile on your face, up everybody. All I see is blue skies on my horizon, no better place... freedom... so only focus on the coming... this belief, energy up rolling and we won't turn back.
Hello Berlin. Hello Berlin. Welcome to GTC, great to be here. Before we start, our show hosts, smart and adorable, smart and adorable, they spoke about AI infrastructure, sovereign AI, AI safety and physical AI. They covered an enormous amount of ground this morning, and we are going to take it one step further.
It's interesting: what do Revolut, Spotify, Merck, Vodafone, Zara, Porsche, Nestlé, and HSBC have in common? Well, they are all European companies, and all working with NVIDIA on AI. Everybody, everybody here is activated on the AI ecosystem and working towards the AI economy. We've been working here in Europe now for 33 years. You don't... NV1, NVIDIA's first chip, NV2, and NV3, which became RIVA 128, the chip that saved our company really and launched our company, were all built in Europe. Our first partner was SGS-Thomson; we manufactured a chip in Crolles, France, assembled it in Malta, and shipped it around the world. Our first largest customer, the largest customer in the GTC days in the beginning, was a company called MediaMarkt, you know them. MediaMarkt had a spectacular idea about building PCs, as you remember, and they were our largest customer. It was Elsa, a German company, that brought NVIDIA into the workstation industry in the year 2000. Now, NVIDIA is in every single workstation all over Europe — cars, everything from styling, design, engineering, simulation, all the way through to even inside the car, with a European company, Audi, in the year 2005.
GTC here. This year, this last year, we started a new journey to building a new industry here in Europe. Europe... all over Europe, this last year alone, as I journeyed from France, the UK, to Sweden and now here, we announced the investment of 14 companies, 12,000 partners, all getting so big now. After decades, after decades of advancement in the computer industry, we are now at the beginnings of a new industrial revolution.
This image was generated by a computer; it was prompted by me. Nice simple prompt: 'AI powering a new industrial revolution', generated this. I gave it a few partners that I wanted to highlight, companies that we're working with, and this is an extraordinary thing. In the mind of the AI, this is what it means to build a new industry: the generation of tokens, the production of intelligence, which integrates into the industry, new industrial... all new tech stuff.
Show how we build something, and what we can build, how we build something and what we can build. I think in every single way, artificial intelligence will affect everything we build, and in fact enable us to build new things in this way, because everyone will use the computer. New industrial revolution, and at the very beginnings of that question is: what is AI? And this is a big deal. If you want to be part of creating this future, understanding what is AI is vitally important.
And this is really a simple framework. We were early to observe that, in fact, AI is a manufacturing industry. AI is not just software technology. It is software technology, but not just. AI, of course, runs on a computer, but the way that computer operates is fundamental. In relation... a simple way to think: an AI layer cake. Every year, advanced computers are AI-oriented. A layer cake for the AI economy: you don't want to be consuming, you want to be producing the AI. We could do all of that here in Europe, and in fact, we've already started.
Europe is building the AI economy, and it's a layer cake represented here. We've got great partnerships, and when you look across this layer cake, you could recognize that, in fact, it is a new industry because there's growth at every single layer. We're creating jobs, creating growth at an extraordinary rate. Companies here in Europe: Schneider, Siemens, Electric — incredible — Siemens Energy doing incredibly well around the world.
We are the compute layer, compute... Asia region and beyond. Every single telco in the future, and I predict this: every single telco in the future with 100,000 sites representing 100 gigs of distributed telco base stations. And data centers will in the future evolve into a distributed AI network. So, the AI infrastructure layer is being built out, which is then, above that, the AI models. It's really important to recognize that long before the AI models emerge, the infrastructure has to exist. Just as before you can have an internet, you have to have the networking. Just as you could enjoy electricity, you have to have electrified houses, electrified power plants. And the models are the next layer. Also interesting to observe that the model layer is not just language models; it represents language, biology, chemicals, physics, robotic actions, animation, and it's models of all sizes.
These are great... into the AI... new startups, so many here. Europe's AI economy is starting and really accelerating, it's great to see. At the foundation of it, where we contribute, most people think it's GPUs. It is true: we, in 1993, started building a graphics chip, which emerged into the GPU, the programmable shading GPU. Over time, we added acceleration to that, reinvented computing as we know it, which then brought forth artificial intelligence.
But NVIDIA today is really a full-stack, end-to-end, complete AI factory platform. The breadth of technology: switches... connect the GPU, the Vera BlueField-4 compute that we introduced — and I'll say a few words about that later — to Spectrum-X, which is our Ethernet solution, making NVIDIA the largest networking company in the world today, to the Grace decode accelerator generation for token generation, to Vera BlueField-4 storage acceleration — say a few words on that in just a second — and then Vera RTX compute — and I'll mention that in just a second. But this now represents the breadth of technology necessary to create the AI factory. But nothing works unless you...
These AI systems are gigantic. A 1 gigawatt data center, a 1 gigawatt AI factory, about 50 billion in power... all of the compute, networking, storage, power, cooling, all connected together, multiple acres of AI factory. Just turning it on and operating it is a marvel in itself. The magnificent amount of software necessary to energize this factory, to put it to use and produce tokens, is an extraordinary feat.
NVIDIA offers the entire full-stack software. On top of that are what we call the CUDA-X libraries, the algorithms of our company. What NVIDIA truly is, is an algorithm company that designs silicon to accelerate them, of course. One of them is cuDNN, what many researchers recognize as the library that brought forth, that started the big bang of modern AI. cuDNN. cuDNN is to AI what SQL is to the world's database industry. It's a domain-specific language, a domain-specific library, just as OpenGL and DirectX enabled real-time computer graphics.
An AI company... 360... those libraries are in fact crown jewels of our company: from computational lithography, signal processing, image processing, data processing, particles, fluids, sparse solvers — incredible breadth of libraries that many of you in the audience have connected into. This is a new market; this is connecting you into algorithms that enable you to do computation at incredible speeds.
And then finally, on the right-hand side, on that side, the models. This is the new software, the new engine of new applications, and it sits on top of and collaborates with the CUDA-X libraries.
These models, generation... Pascal, it started with the Pascal era, what is called the DGX-1 system. DGX-1 brought to the world the world's first AI supercomputer. I introduced it 10 years ago and delivered the first one to OpenAI. A computation journey from Pascal, to Volta, to Ampere, then to — many of you who heard about us for the very first time — Hopper, the current generation, Blackwell, and then now Vera Rubin in full production. Generations of architectures... AI per million. If you take a step back, think about what Moore's Law would have done: in 10 years' time, the first 5 years would have improved by a factor of 10, the second 5 years by another factor of 10 — 100x improvement in 10 years' time. In the same period, accelerated computing... accelerated AI by far more.
This is what we build and Europe's AI economy. Now each generation: in the beginning Pascal was GPUs, Volta was GPUs, Ampere was GPUs, and Hopper was GPUs. We invented the new DGX form factor, then it was Blackwell that introduced to the world the first rack-scale computing. And I'll show... rack-scale computing, turning one entire system into one giant computer, one GPU. And now with Vera Rubin, we're extending it to processing beyond models, beyond large language models, to now processing the entire agent.
Now, we'll refer back to this slide later, but remember this: the agent has to think. The agent has to think, and that's where the model comes in: perception, reasoning, planning, and using tools. All of that runs on the large language model, and the models are designed to do that. But the agent also has what is called a harness — basically the exoskeleton of the digital agent, just as we have the brain, the body, the exoskeleton. The agent skeleton orchestrates and allows the agent to actually engage memory: active memory, working memory, as well as long-term memory, and activating tools. Then the third part is a computer that processes knowledge, designed for data processing. And then, lastly, a computer dedicated to CUDA-X libraries and to run tools. And these computers now represent the entire systems.
So Rubin is in full production, nice to kind of see it. This is the first computer in the world that we created: no cables, no wires, completely 100% liquid cooled, and just really delightful, really delightful to be able to build it. And this is one Vera computer. So this has two GPUs. Yep. What used to take, in the case of Grace Blackwell, 2 hours to assemble one node — and when you see it you'll know what I mean: cables and wires everywhere, 80% liquid cooled. Now Vera is 100% liquid, and it takes less than a few minutes to assemble it. And so, incredible, incredible achievement. Of course, this translates to higher cycle times in manufacturing, much better reliability, and of course a lot better performance. This is the NVLink switch tray, and what this does is it connects 72 of these GPUs. There are four of them in here right now. It connects 72 GPUs into one gigantic system.
Thank you. Thank you. And this is our new Vera computer. It's okay, go easy. Let her... give her a chance. Everybody has a chance to say something. But today, let... No, no, take just... I'll talk to you later. I'll talk to you guys later.
Thank you. All important issues, all important issues. But let me get through my keynote. We have... this auditorium is packed. We have 100 other watch parties around the world, 100 watch parties. 15,000 people are watching us right now. And so I'll get through the keynote. There's some really, really fun things I want to show you in the end, okay? So you just have to... I have to torture you through some education first, and then the entertainment is coming.
This is our brand new... this is our brand new Vera computer. The Vera CPU is an extraordinary CPU beast: eight instructions per clock. It has the world's first LPDDR memory system, so that we can have high performance, very high bandwidth, as well as extremely excellent energy efficiency. The on-chip fabric is the fastest of any chip in the world. Processing is off the charts, and this is designed for doing exactly that. This is our Vera computer. And this computer here, this computer here is Spectrum-X. This is the world's first co-packaged optics, and you can see it here. I'll show you...
Some of this area falls into the ground, and some of this area you can walk on. I'm just not sure which. All right, so if I accidentally disappear, you guys know us — it's not a stunt, it's an accident. And so you could see here, these are the lasers, and these lasers come up through optics and connect directly to our chip, and it connects using a process technology we invented with TSMC called COUPE. And so silicon photonics are directly attached right onto our chip.
Otherwise, what you would see is all of this would be PC board, and the PC board's transmission capability is optical — and we have extremely high rate, extremely high quality data that comes directly onto the chip, and this goes directly into what would otherwise be a transceiver, without going through a PCB, and the transceiver goes right into the optics. So, this is the world's first CPO in complete volume production: 512 ports at 200 gigabits per second, the world's most advanced Ethernet switch, and we're really proud of it. Thank you.
And so what that gives us is the ability to process the thoughts of the large language model, process the agent harness, run data processing for its knowledge, and then accelerated computing for all of the tools that represent essentially the agent workload. And I'll show you more in a second. While Vera is in full production and the performance is off the charts, every single generation, as I mentioned, through extreme co-design, we're able to achieve performances that are well beyond reason, well beyond what Moore's Law would have dictated. And as you know, Moore's Law has really largely ended, and we need these extreme designs.
Vera is another step on top of Blackwell. Blackwell was about 30 times faster than Hopper in doing large language model processing, and now Vera is another 30 times greater than Blackwell — Vera in the full factory.
What do you think? So in 2016, we introduced the world's first AI supercomputer. It was one node, one GPU node. Then 3 years ago, we introduced the Grace Blackwell NVL72 rack-scale computer. Both times, in the case of NVL72, we took the entire rack that would have been assembled on-site and we assembled it inside the supply chain. We essentially manufactured supercomputers at scale. It is the reason why we are able to move such complex manufacturing... 72... producing computing at this scale is unheard of, ever unimaginable.
Well, even then, when you bring these systems on-site... and 17,000 meters of wiring to connect Spectrum-X for the scale-out computing. 17 km, 17,000 meters. It's a crazy number, and so 17 km, incredible numbers. Now suppose we were to do that all inside the supply chain and run it... it is functional. The amount of cabling and on-site assembly is going to be greatly reduced. But there is something much, much more. In fact, what is going to happen here is that by moving this into the supply chain, the amount of idle computing... the amount of idle computing...
Each gigawatt data center is 50 billion. The revenue generation you expect is tens of billions per year, and so we just said it's 50 billion per year of revenue generation that comes out of that infrastructure. 40 months... billion, billion of revenue, billion of revenue will make anybody anxious. And so by moving this into the supply chain, the speed by which we can turn on — from installation, arrival of computing to production of tokens — the transition...
People, data centers... velocity goes up. The wasted computing, the wasted CapEx will be reduced, the business becomes more profitable. This NVLink 72 is now into a large cabinet system, we call it DGX. So this is the next click of NVIDIA computing. Now, what you make and how you make it makes you special. Just as the work you know makes you special: you design machines, you operate factories, you discover medicine, you have extraordinary craft.
That craft, that specialty makes you special. What makes us special are really two things. The first thing is that our architecture is the only accelerated computing architecture for this purpose. We don't specialize in just one algorithm, for example transformers; we specialize across anything that could be parallelized — it could be fluids and particles, it could be quantum, it could be quantum chemistry, it could be image processing, it could be data processing, structured vector data processing, processing of all kinds. And because we've been working on it for so long, and because our architecture has... numbers...
...accelerated, grown enormously: our install base, our general-purpose architecture, our dedication to this architecture, and our collaboration with all of the world's industries has made our architecture literally everywhere. The second thing is how we do it: the architecture, the stack, the complete AI factory. There are many companies in the world who want to design their own chips, but most of the companies in the world just want to deploy the AI infrastructure so that they could run their business on top of it. The ability to design a chip, build a supercomputer, write all the software is really out of reach.
Computing, AI factories, and it exposes us to markets that others aren't. We are able to reach many companies here for this very reason. If I go around this room with CSPs, this is where most people think the vast majority is today. As you can see, I mentioned a whole bunch of companies who are not those. So CSPs remain very large businesses for the world's computing, but there are AI-native companies that are starting all over the world.
There are so many that are here, neoclouds, because every region wants to participate in the AI economy, every region wants to ensure that the valuable resources that they have, the energy that they have, is dedicated to companies in their region... here... companies building systems, and ODMs who are building computers for the world's enterprise AI labs. Of course, we know... last year, all we spoke about was OpenAI. Today, we speak about OpenAI, Anthropic, but we also talk about the incredible companies...
For Europe to benefit from AI and create an AI ecosystem and AI economy, you need all the end markets, the end industries to adopt AI and to succeed with AI. The industry, because building cars, manufacturing, robotics, new drugs, health, financial services — the industry, all these different industries have become AI companies, every single one of them. We have a unique ability through what I just mentioned about CUDA and the way that we build computers, we have a unique ability to literally... on one architecture for many markets.
And because infrastructure is so expensive, we need to make sure that they are as productive as possible, as soon as possible, that they are fungible — that you can use it for every workload from the beginning of your AI journey: data processing, to training, post-training, inference, and agentic processing, every single step of the AI journey, to every type of AI, chemicals and physics to languages, to every single industry and the algorithms that they use. And so NVIDIA's architecture is fungible, and as a result of that, continuous innovation, continuous updating of software. Because the install base is so great, we have such an incredibly rich ecosystem of software that the architecture that stays the foundation... that is the infrastructure.
Now, what do we do with it? This last year, we saw a transition: AI is now productive. And one of the most important times was in the beginning of the year when we spoke about Open-CLAW, which introduced the concept of an open-source agent, and really started what we all see across the world today. Agents think, reason, iterate, use tools, and as a result, consume an enormous amount of tokens. But as a result, they become useful. So more agents are being developed; they're more useful, and therefore they're used by more people. And because they have to use a lot of tokens to think, the amount of token consumption has gone through the roof. GPU demand has really gone through the roof.
Now, what processing is this? This is an agent. These are the elements of an agent. The first part are the language models. These are the tokens that are running on top of the GPUs and generating tokens that are consumed by the agent. What is it? I call it the exoskeleton: that Open-CLAW harness sits inside a sandbox, and that...
The sandbox isolates it from the rest of the system. This is a very important concept. Back there are things there: harness, the sandbox that the harness sits in, and two more things: knowledge and tools. Models, sandbox, harness, knowledge, and tools — these are the five elements of an agent. Now, before I talk about how to make the agent useful, the most important thing is we have to make the agent secure. Agents are autonomous; autonomous systems access resources, networks, access tools, and we need to make sure the privileges that we provide this autonomous system are minimal and that we constantly monitor it. We introduced a fantastic new product, a new platform, and it's called OpenShell. OpenShell came just in and base monitor before you are preparing your AI agent to be useful. They can plan, use tools, access data, and together the agent is useful. It could delete anything, your SSH keys.
Please, don't just hope; hope isn't security. Safety isn't an aspiration, it's technology and engineering. NVIDIA OpenShell and Centry put that philosophy into practice. You write a policy for what each agent can touch, and OpenShell enforces it at the granularity of file access, web access. Formal verification, policy change against permission ceilings, so agents never gain more access than intended. The early internet had viruses everywhere until browsers enforced strict limits on what a web page could do. OpenShell does the same for agents: enforcement outside the agent, not limiting the agent — agent isolation in a sandbox. Out network requires rules.
It checks policy if not permitted. Keys stay outside with the supervisor and are only on the way to approved endpoints. A compromised agent can't leak a key it never had. The gateway agent proposed policies against OpenShell. Combines Linux kernel isolation with runtime policy enforcement. NVIDIA Centry adds an optional layer of protection outside the host itself, away from the agent running on NVIDIA BlueField-4. Centry monitors agent activity and enforces security policy on agent data. Verify, detect, investigate. And OpenShell is open source under Apache 2.0. Runs everywhere your agent does, edge to cloud. Agents are only as useful as the access you give them. OpenShell and Centry put that access under control — that's how we advance capabilities and safety.
Agents are autonomous systems; we have to create extremely secure technology so that we could develop these agents and deploy these agents safely. Now it's incredible: after announcing OpenShell, just about every major company in the world — from the frontier labs, computer company, platform company, everything — really fantastic. And the reason for that is because this technology has arrived at exactly the right time, at a time when we want to deploy agents all over the world. Now remember, this isn't just about the services using OpenShell as they deploy agents; this is really about all of the world's companies also adopting OpenShell so that you could receive agents, use agents.
Those interactions require different levels of privilege to the agent, and every time the agent is launched, the privileges are different, and privilege is the policy enforcement. NVIDIA partner. Agents.
There, about the first part. The first part is how do you use these harnesses — Open-CLAW, OpenShell, OpenHarness. You could create your own harness for deep agents or from Python. There are a whole bunch of different ways you could create agent harnesses that sit inside the OpenShell secure harness, okay? And that's the sandbox that it sits in, the secure sandbox. And then next, let's talk about models. You should use all models of all kinds in the future. It won't just be one model that runs everything. All the models to do your work — the work you do, fairly generic artificial general intelligence is going to do. Incredible. However, the work you want to do captures too much of your proprietary knowledge, too much of your specialty, and you want to continuously enhance that intelligence, compute, models as well.
By the end of the session, you will see that, in fact, building your own AI is not that hard. It starts with, of course, the models we've been dedicated to advancing: open models, to open up new opportunities and enable people to build their own custom AI. Nemotron is for languages, Cosmos is our world foundation model. Grus for human robotics — we are the frontier. Alpamo for autonomous navigation — we are the frontier. Earth-2 for quantum classical computing — we are the frontier. And of course Earth-2 physics AI — we are the frontier. So in all of these different areas, we offer you all the models and we're continuously working on it. We've been working on it for a decade to work on this forever; we completely open the data that goes along with it so that you know the provenance of the model that we created, as well as the scripts necessary for you to reproduce it.
So not only do you have the open weights, you have from us the entire recipe and the data to recreate the open models, so that you could know where it came from and have complete trust in it. Well, these models with AI and the enterprise ecosystem that work you could find. Data into sovereign intelligence. Here is an example: AI21 helping turn Nemotron into a model that exceeds frontier capabilities. Freebase10 coded rapid trajectory. Harvey again using Nemotron, post-trained it and turned and created their own frontier-level capability. Here is Scale working with SkyDance, again doing the same thing. It goes on and on and on. The point being that you could start with models, and with your own data post-train it for your use case and achieve, of course, super intelligence. That is a very sensible thing to do, and all the technology exists today. Here's an example of Palantir with Tampa General Hospital — the list goes on and on. We have a rich ecosystem of partners that can help you, and they are incredible companies, and they can work with you, you know, from around the world.
Have here mentioned all of it — I love, I love, love all. And so, first, agent harness running in a secure sandbox: OpenShell. Second, Nemotron large language models. The first two components of what you need, but ultimately what turns these agents special — this is called the agent knowledge foundation that we put together: libraries and models that are completely world-class, allowing you to turn your company's data. The data could be in assets, could be in unstructured format and documents, so that you need an AI to understand how to read these PDFs and documents. Could be dates... we have libraries and tools necessary for you to extract meaning and understanding from your company data, one at a time. Nemotron parsers, completely world-class, top of leaderboard. RAG, taking a query and associating it with the knowledge inside the vector database, top of leaderboard.
Pieces it brings together, ranked based on the most relevant to you, to be able to predict from your structured database what is likely to happen. And so, this is, for example, if you were to receive a notice that one of your suppliers — un-documentation, orders, completely inside structured database, for example — and in order to predict what is going to happen to all of these orders, who are going to be short, the amount of delay, you use it to make that prediction. And then, in order to figure out what to replace it with, you might use vector database search and go and find an engineering document of a replacement component. Okay. So all of those different libraries and data processing capabilities now make for you your knowledge foundation. You've now armed your AI with proprietary data, proprietary knowledge. Now the last part I'll talk about is ultimately tools. Tools: accelerated simulators, data processing engines, domains. Used in scientific discovery in all the different fields of science, they are used in engineering, used in research, used in designing, building, creative arts. Libraries.
Math is beautiful. CUDA-X libraries — these are the crown jewels of our company. And the reason for that is this: when you're building an AI assistant — it could be a lab assistant, it could be a research assistant, it could be a designer, somebody who is building — agent. Work are the tools, and we now have agent to reason, understand information, reason to be able to plan and iterate and use tools until it could get the job done. But it still comes down to the tools.
Could be dynamics, could be chemistry, could be process, could be sequencing — all these tools run on that one computer, and that one computer speeds up the processing of it compared to just running it natively on a CPU by 10, 50, 100x. Agent tools trade-off. So this is the magic, magic behind the agents. Ultimately, the agents have to use tools. Now of course, in one of the first fields of agentic use is software programming, and therefore a compiler or Python is all it really needs. But in the future, as we have these tools do more and more and more and more useful work, we're going to need to give them more tools. And here are some examples. Here is sciences: they're creating a lab assistant and they're using NVIDIA BioNeMo models that are designed for understanding proteins and understanding cells and generating proteins and chemicals in drug discovery. They're using BioNeMo and CUDA-X tools. This is Seon's factory building a factory agent that uses NVIDIA Cosmos, our world foundation model that understands the physical world, and CUDA-X tools. And here's an example of creating a virtual companion that uses NVIDIA Nemotron language and tools.
The most intelligent scientist still needs a laboratory. The best engineer still needs simulators and solvers. An AI agent is no different. An agent brings together five building blocks. The model perceives, reasons, and the harness iterates the work. Knowledge foundation gives the agent context from your company's structured and unstructured data. Tools let the agent simulate, experiment, and act. NVIDIA OpenShell controls what agents can access and execute. Start with a frontier closed model or customize an open model. With NVIDIA's agent toolkit, any company can turn its knowledge and expertise into a custom super agent. Look at examples for LSC: the goal is design therapy for B-cell lymphoma; their agent is built on NVIDIA, turning that data into experimental in the computational dry lab. Agents use the NVIDIA BioNeMo agent toolkit to call models for protein design and RNA optimization, running millions of experiments for the most promising candidates. AI science autonomous. Cosmos makes hands. To AI design and experiment the loop.
At an electronic factory, AirLine, the goal is to bring new products into production. Process preparation agents use NVIDIA Cosmos to generate virtual prototypes and synthetic data in an NVIDIA factory operations blueprint, use orders, supplies, and lookups for production. Digital Twin Composer uses NVIDIA Omniverse libraries to validate those plans with physics simulation and real-time factory data before changes reach the factory floor.
With these systems, an engineer needs to adapt a Volkswagen Golf door design for the Aura, and Leo uses NVIDIA to reason over engineering requirements. The harness coordinates the specialized sub-agents and preserves context across planning, design, and simulation. Aura uses NVIDIA to plan the work around the vehicle dimension, weight limits, and other engineering requirements. Leo design agent selects a compatible golf design and uses it for the door. Leo simulation agent uses physics accelerated by NVIDIA X-libraries to test the vibration problem. This goes to the design agent for refinement. The engineer reviews the refined designs and selects the final one. Different goals, different expertise. The same building blocks: open models, CUDA-X accelerated tools, and OpenShell secure runtime, brought together with the NVIDIA agent toolkit.
Great. Some people said that AI here is going to be the end of software, it's going to be the end of tools companies. It's exactly the opposite. AI is here. There will be more users of these tools that we created over time than ever before. AI companies optimizing tools and crowd and enterprise company out there are all building agents, and we're just delighted to be working with them. Building agents is not that hard. It's got five components: it's got the model, the harness — both of them have closed and open versions — the OpenShell secure runtime, and then most importantly the knowledge and tools, and those knowledge and tools belong to you. That's your expertise. And so, all of this is now possible to do, and I'll show you more in just a second. But before I do that, let's talk about physical AI.
That's really the next wave of AI — AI that understands the physical world, can reason about, operate tools. When people think about physicals, they think robotics. One of my favorite physical world applications is the radio cell, one cell towers. And the reason for that is this: it uses, instead of cameras, instead of speakers and microphones, it uses radio waves to sense the world. And it reasons about it, it processes information, reasons about it, and interacts with the world using beamforming. So, in fact, this is an application of technology that is just perfectly suited for AI. AI will revolutionize the way we do telecommunications. It will enhance our spectrum efficiency, as saying all is nearly 100% computation we thought necessary in the past, and now that we apply AI to, is a million-to-one difference.
And so, we're going to bring to the telecommunications network an enormous step up in spectrum efficiency. But beyond that, we're also going to revolutionize all of these cell towers. There's 100,000 sites around the world, around the western world. Data structures that future distributed AI infrastructure modeled with companies here, and this last week after working on this for nearly years, NVIDIA made the first commercial phone call on the T-Mobile network with Nokia, and so now CUDA makes phone calls the traditional way. Then, once you can do that, then we can layer AI on top of it, and so the revolution is going to begin. This is a really exciting opportunity for us.
The other physical AI application that I'm super excited about is essentially robotic radiotherapy. The tumor is changing in shape. The physical body is changing while you're in therapy. The robot — this radio adaptive radiotherapy system — has to perceive the changing environment, understand where the treatment is, and precisely radiate accordingly. Essentially robotic. Right, go ahead. This is a form of robotic surgery.
Now, of course, the highest volume, the most near-term robotics application or physical AI application that everybody recognized has been something we've been working on over. Now, look, car companies, part of the ride-hailing fleet today. We have ride-hailing and part of Uber, and here in Europe we have Bolt, and we have Grab and Lyft and so on, so forth, and all these different ride-hailing services. In the future, there will be robo-taxis that will be connected to those ride-hailing services. Won from, and building systems, expensive, particularly when standard architecture: 14 different cameras, N-radar systems, three lidars, ten ultrasonic — this cocoon of a shell of sensors powered by dual NVIDIA Orin systems, computing system. We, in the yard, and now designed into one car fleet after another. On top of it, AV software companies, and we are in fact one of them as well, to build an autonomous robo-taxi capability, which then is connected into the ride-hailing services.
This ecosystem is being built. The world travels about a trillion... taxi, this is one giant opportunity fragment. Creating industry standard Hyperion and working with entire ecosystem, we can create essentially a giant virtual fleet that everybody can connect into. Today, I'm super proud to announce that four more brands, four more cars are going to, lines are going to come in. The VW van — I just can't imagine anything more iconic. The JLR brand new car, the Volvo fleet, as well as Mercedes Next Generation are going to be also Hyperion compatible and be part of this network, this ride-hailing network robot fleet. So, welcome Hyperion.
The next major industry, the next great industry — this one is likely to be extremely large — is humanoid robotics. There's a reason why robotics is hard to do, because all the different work is rather diverse. And so, building a special configuration for each one of the diverse tasks segments the robotics. Quickly, personal computers so quickly got so advanced so quickly is because those standardized configurations allow many customers to use it, the volume to therefore be derived. That volume drives the technology flywheel, and it gets faster, better, incredibly. And so, we need as much standard configuration as possible. Now, humanoid robotics obviously has ability. It probably isn't perfect for everything, but likely to work for many things. As a result, this is likely going to be the largest volume robotics configuration. Now if we could just create the AI that allows it to learn very quickly for small and medium businesses who don't have an entire robotics team to teach and program — the robot can now show the robot what needs to be done. The robot, because it has general capabilities and general intelligence, would be able to learn and adapt quickly, and so this is likely to be the next major technology platform. We're working with so many different companies; I recognize many of them here in Europe. This is your great opportunity. No country, no region... Europe region companies here are good at manufacturing; this is the high volume industry, and you could see there are so many companies that are now part of it, and we're delighted to be partnered with.
Now in order to create a physical AI system, you need three computers. The first computer is the one we've been talking about and everybody knows quite well, which is training the AI model. The second computer is where we have the AI learn how to do its skills. You can't just have the AI digital dome in in you, all the libraries, all done inside a digital twin world. That digital twin world is called Omniverse. Omniverse is a library connected into all kinds of platforms and tools and Cadence and so forth, and so you could see these digital twin systems being created — the second computer. The third computer is the robotics computer. That computer integrates — in the case of the self-driving car, the robot taxi, we call it Hyperion. Hyperion consists of the computer itself, the functional safety system, the active safety system, an operating system designed for physical AI safety. We need to do exactly the same for robotics systems.
So that every single time you start to develop research and start to work on your own robotics systems, you don't have to start from scratch to cobble up all of the software and computing systems and the sensors and all the mechanical systems — that alone would take several years. And so we decided to build a reference robot. This is the NVIDIA Agile One in partnership. This will be available first quarter of year and will come assembled with computer inside, software integrated, and be able to perform quite well right out of the bat. And so, this is a great starting point and will be the flywheel by which we can accelerate development of robotics. Let me end with speed.
In the PC revolution, it was 1993, and the PC revolution was just about to start. Remember, Windows 95 kicked it into turbocharge. We came since then, invented the GPU, programmable shading, real-time ray tracing. We added CUDA to the computing model and helped ignite the modern AI revolution. And so, years ago, we started to think: what can we do to reimagine the personal computer for this new era, where the personal computer is not just your tool, your personal computer tool, but your personal agent or your personal assistant? And the way you interact with this computer would be essentially a companion, somebody who helps you do work. And so, this computer needs to have exactly the five things that I mentioned today. It needs to have a model — it could be running locally or could be running in the cloud. It needs a harness. It needs a secure runtime, OpenShell. And it needs access, because it's available on your PC, to knowledge and tools. And the good news is, because NVIDIA has been accelerating tools for so long...
GL, NVIDIA has ever worked on this computer. And this is years of efforts. Microsoft, we have created the RTX Spark, and this is a brand new line of computers. This, literally thousands of engineering years went into this. We've been secretly working on it for four years, created a chip called the N1X. In laptops and workstations, this is one, one active in the year 2016. I delivered one petaflops and 128 GB of memory to a small company called OpenAI. The only difference, of course, is that I can hold this like this, and DGX-1 took two people to carry upstairs. And so, this is the brand new RTX Spark PC. And if you would like to have a companion workstation that runs... station Windows and 20 petaflops and 754 GB of memory — incredible. And of course there's also the desktop version. All of these are now available, and I think the DGX Station with RTX Spark, DGX Station with Windows, will be available at the end of this year.
Now the question is, so far I just talked about the five components of an agent, and now the question is: how do you build it? How do you build an agent that runs, let's say, on one of those computers, or runs in a cloud, or runs somewhere? Well, the answer is really easy, and you do it this way. The way you build an agent is you ask the agent to... So this is the prompt that is going to run the next application. And what you're about to see is a robotics application. And while you're watching this demo, just remember, there is an agent running on this personal computer. Demo.
Two ideas: agents — how do you build it and run it, and what you can use it to do — and the second is physical AI computing model.
