“We Are Manufacturing in America Again!” – Jensen Huang Unveils the AI Factory Revolution | AI14

We are manufacturing in America again. It is incredible. The first thing that President Trump asked me for is bring manufacturing back, bring manufacturing back because it's necessary for national security, bring manufacturing back because we want the jobs, we want that part of the economy. And nine months later, nine months later, we are now manufacturing in full production Blackwell in Arizona. Extreme Blackwell, GB200, Grace Blackwell, NVLink72, extreme co-design gives us 10x generationally. It's utterly incredible. Now, the part that's really incredible is this. This is the first AI supercomputer we made. This is in 2016 when I delivered it to a startup in San Francisco, which turned out to have been OpenAI. This was the computer. And in order to create that computer, we designed one chip. We designed one new chip. In order for us to do co-design now, look at all of the chips we have to do. This is what it takes. You're not going to take one chip and make a computer 10x faster. That's not going to happen. The way to make computers 10x faster that we can keep increasing the performance exponentially, we can keep driving costs down exponentially, is extreme co-design and working on all these different chips at the same time. We now have Ruben back home. This is Ruben. This is the Vera Ruben and Ruben. Ladies and gentlemen, Ruben. This is our third generation NVLink72 rack scale computer. Third generation. GB200 was the first one. All of our partners around the world, I know how hard you guys worked. It was insanely hard. It was insanely hard to do. Second generation, so much smoother. And this generation, look at this. Completely cable-less. Completely cable-less. And this is, this is all back in the lab now. This is the next generation Ruben. While we're shipping GB300s, we're preparing Ruben to be in production, you know, this time next year, maybe slightly earlier. And so, every single year, we are going to come up with the most extreme co-design system so that we can keep driving up performance and keep driving down the token generation cost. Look at this. This is just an incredibly beautiful computer. Now, so this is amazing. This is 100 petaflops. I know it doesn't mean anything. 100 petaflops. But compared to the DGX-1 I delivered to OpenAI 10 years ago, 9 years ago, it's 100 times the performance right here versus 100 times of that supercomputer. 100 times, 100 of those, let's see, 100 of those would be like 25 of these racks all replaced by this one thing. One Vera Ruben. Okay, so this is, this is the compute tray, and this is a Vera Ruben superchip. Okay, and this is the compute tray, this up right here. It's incredibly easy to install, just flip these things open, shove it in, even I could do it. Okay, and this is the Vera Ruben compute tray. If you decide you wanted to add a special processor, we've added another processor. It's called a context processor because the amount of context that we give AIs are larger and larger. We want it to read a whole bunch of PDFs before it answers a question. We want it to read a whole bunch of archive papers, watch a whole bunch of videos, go learn all this before you answer a question for me. All of that context processing could be added. And so you see on the bottom, 8 ConnectX 9 new SuperNICs. You have CX, you have CPXs, 8 of them. You have Bluefield 4, this new data processor, 2 Vera CPUs, and 4 Ruben packages or 8 Ruben GPUs. All of that in this one node, completely cable-less, 100% liquid-cooled. And then this new processor, I won't talk too much about it today, I don't have enough time, but this is completely revolutionary. And the reason for that is because your AIs need to have more and more memory. You're interacting with it more, you want it to remember our last conversation, everything that you've learned on my behalf, please don't forget it when I come back next time. And so all of that memory is going to create this thing called KV caching, and that KV caching, retrieving it, you might have noticed, every time you go into your AIs these days, it takes longer and longer to refresh and retrieve all of the previous conversations. And the reason for that is we need a revolutionary new processor, and that's called Bluefield 4. Next is the ConnectX switch, excuse me, the NVLink switch, which is right here. This is the NVLink switch, this is what makes it possible for us to connect all of the computers together, and this switch is now several times the bandwidth of the entire world's peak internet traffic. And so that spine is going to communicate and carry all of that data simultaneously to all the GPUs. On top of that, this is the Spectrum X switch, and this Ethernet switch was designed so that all of the processors could talk to each other at the same time and not gum up the network. Gum up the network, that's very technical. Okay, so these three combine, and then this is the Quantum switch. This is for InfiniBand, this is Ethernet, we don't care what language you would like to use, whatever standard you like to use, we have great scale-out fabrics for you, whether it's InfiniBand or Quantum or Spectrum. Ethernet, this one uses silicon photonics and is completely co-packaged options. Basically, the laser comes right up to the silicon and connects it to our chips. Okay, so this is the Spectrum X Ethernet. And so now let's talk about, thank you, oh, this is what it looks like. This is a rack. This is two and a half, this is two tons, 1.5 million parts, and the spine, this spine right here, carries the entire internet traffic in one second. Same speed, moves across all of these different processors, 100% liquid cool, all for the, you know, fastest token generation rate in the world. Okay, so that's what a rack looks like. Now, that's one rack. A gigawatt data center would have, you know, call it, let's see, 16 racks would be 1,000, and then 500 of those, so whatever 500 times 16. And so call it 9,000 of these, 8,000 of these would be a one gigawatt data center. And so that's a future AI factory. Now, we used, as you notice, NVIDIA started out by designing chips, and then we started to design systems, and we designed AI supercomputers. Now we're designing entire AI factories. Every single time we move out and we integrate more of the problem to solve, we come up with better solutions. We now build entire AI factories. This is going, this AI factory is what we're building for Vera Rubin, and we created a technology that makes it possible for all of our partners to integrate and Vertiv deliver prefabricated modules, factory built, tested, and ready to plug in. This shrinks build time significantly, achieving faster time. Millions of dollars in additional revenue per year. So