WOYM: Nvidia 2024 GTC De-Brief & Highlights Reel

[00:00:00] Welcome to Non Consensus Investing. I'm Ram Ahluwalia, your host and CIO at Lumida Wealth, where we specialize in the craft of alternative investments. At Lumida, we help guide clients through the intricacies of managing substantial wealth so they don't have to shoulder the burden alone. Through this podcast, we draw back the curtain to reveal the strategies employed by the best in the business for their high net worth clients so that you too can invest beyond the ordinary.

Hey folks, we're going to do what's on your mind on the NVIDIA GTC conference that came out yesterday. I'm going to do this solo, my partner Justin's out on holiday this week, but we're going to focus on select clips. That are posted here on the Lumida wealth page, by the way, in Lumida wealth, we're focusing on taking the best highlights and clips of what we think is pretty meaningful content, slicing it and putting it on here.

So it's worth a follow. So what I'm going to do is just play the clip and then give my quick commentary [00:01:00] and take on it. And then towards the end, I'll just share some broader thoughts on AI semiconductors, disillusionment hypothesis versus what's real. The timing of AGI. And some sub sector themes within semiconductors, I think are going to be pretty interesting that people are not talking about.

All right. So with that, let's start with the first clip. Five things. First, a new industrial revolution. Every data center should be accelerated. A trillion dollars worth of installed data centers will become modernized over the next several years. Second, second, the computer of this revolution. The computer of this generation, generative AI, trillion parameters, Blackwell.

Third, new computer creates new types of software. New type of software should be distributed in a new way. So that it can be an endpoint in the cloud and easy to use, but still allow you to take it with you. Because it is your intelligence. Your intelligence should be packaged up in a way that allows you to take it with you.

We call them NIMS. These NIMS are going to help you create a new type of application for the future. Not one that you wrote completely from scratch, but you're going to integrate them, like teams, between NIMS, the AI technology, the [00:02:00] tools, NEMO, and the infrastructure, DGX Cloud, in our AI foundry to help you create proprietary applications, proprietary chatbots.

And then lastly, everything that moves in the future will be robotic. You're not going to be the only one. And these robotic systems, whether they are humanoid, AMRs, self driving cars, forklifts, manipulating arms, They will all need one thing. They need a platform, a digital platform, a digital twin platform.

And we call that Omniverse, the operating system of the robotics world. Five things. First, a new industrial revolution. Alright, so he dropped quite a bit there. He gets into some of these topics later, so I'm not going to focus on all of them here. But I did want to get the main point across, is that the basic thesis he's saying is, look, we've got all these legacy data centers that are built on CPUs.

CPUs. which do linear sequential processing, and now the world needs GPUs that can do parallel processing to power AI, and there's a big rebuild. That's point one. This plays right into the semiconductor CapEx receiver thesis we have, and our basic point is that the best customers in the world that you could ever have as a company, [00:03:00] those are governments because they've got big budgets, they can borrow, they can spend.

Big tech, those are the folks like Google, Microsoft, Meta, and others, and then three well funded startups, including a long tail of other startups. Those are the customers of NVIDIA. It's high customer quality, number one. Number two, it's a top priority for them. It's the stakes are so high, they have to spend.

And the gain to being the market leader so high that in general, they're price insensitive. They'll complain about the price, but they'll buy. So that's the main message I think coming out of this video. The other one that stuck out for me was that he's saying, look, the compute needs to go with you. What he's doing there is he's making a dig against these so called hyperscalers or cloud AI companies.

Such as like Microsoft, right? Microsoft, their vision of AI is that it's on premises, it's secure on a Microsoft cloud, NVIDIA saying, no, it's local, it's portable. [00:04:00] It goes with you or it's on prem. I believe most of the sovereigns are going to want on prem because they're not going to trust. A U. S.

based company that's subject to U. S. law. So that's a key distinction to play here because NVIDIA's customers are in a way like their competitors, right? Microsoft is building its own GPU, so is Google, but they need NVIDIA now. And, these are two different combating visions of the world. So if you can figure out which world will evolve, that can, hopefully position you better.

So let's go check out the next video now. Transistors. This is the first time two guys have abutted like this together in such a way that the two chip, the two guys think it's one chip. There's 10 terabytes of data between it, 10 terabytes per second. And so this is the, this is a fully functioning board and I'll just be careful here.

This right here is, I don't know. Ten billion dollars. The second one's five. [00:05:00] It gets cheaper after that. Any customers in the audience? It's okay. What's amazing is this computer is the first of its kind where this much computation, first of all, fits into this small of a place. Second, it's memory coherent.

They feel like they're just one big happy family working on one application together. And so everything is coherent within it. Oh, this is the most advanced GPU in the world in production today. This is Hopper. This is Hopper. Hopper changed the world. This is Blackwell. It's okay, Hopper. You're very good.

Good boy, 208 billion transistors. This is the first time two dyes have a budget like this together in such a way that the two chip, the two dyes, think it's one chip. There's 10 terabytes of data between it, 10 terabytes per second. And so this is the, this is a fully functioning board and I'll just be careful here.

This right here is, I don't know. Look, I think the main highlights there are, one, just his presentation style and storytelling. It does remind you of Steve Jobs. He's very intentional, of course. He's got a whole comms team around him. He's wearing the same, black leather jacket each and every time.

He's got a nice [00:06:00] humor about him. And so we're going to hear a lot more about, Jensen, the best CEOs are great storytellers. He's done a good job of taking this complex abstract technology and making it concrete and real by literally showing you the technology in hand. Not an easy thing to do and you can't, demo hardware.

So let's go to the next clip here. Now, this is really about Moore's law. Yes, the token generation, 5X, the inference capability of Hopper seems like enough. But why stop there? The answer is it's not enough. And Decided to scale it over the course of the last eight years, we've increased computation by 1000 times.

Remember back in the good old days of Moore's law, it was two X. All right. That's wild. This is one of the reasons why I moved up my timetable for AGI. We'll come back to AGI at the end. Let me repeat that. He's saying there's a 1000 X increase in AI compute last eight years. This is much faster than Moore's law.

They're moving at a pace of doubling AI compute in six [00:07:00] months. 10x every five years, 100 times every 10 years. In the last eight years, we've gone 1, 000 times. We have two more years to go. The future is generative, which is the reason why this is a brand new industry. There's another data point we shared, that the first GPU delivery to OpenAI had 0.

16 petaflops. Capability. Now they're going to be delivering exoflops and exoflop is a thousand petaflops, there's a handful of exoflop supercomputers globally. They're going to start manufacturing that scale. So just to show you the pace of the hardware velocity and innovation, pretty dramatic. This chip is just an incredible chip.

We call it the MV link switch. It's 50 billion transistors. It's almost the size of Hopper all by itself. This switch chip has four NV links in it, each 1. 8 terabytes per second, and it has computation in it, as I mentioned. What is this chip for? If we were to build such a chip, we can have every single GPU talk to [00:08:00] every other GPU at full speed at the same time.

That's insane. It doesn't even make sense. But if you could do that, if you can find a way to do that, and build a system to do that's cost effective. That's cost effective. How incredible would it be? That we could have all these GPUs connect over a coherent link, so that they effectively are one giant GPU.

And as a result, you can build a system that looks like this. Now this system, It's insane. This is one DGX. The first one I delivered to OpenAI was 0. 17. You can round it up to 0. 2, it won't make any difference. But, and this is now 720 petaflops, almost an exaflop for training, and the world's first one exaflops machine in one rack.

Just so you know, there are only a couple, two, three exaflops machines on the planet as we speak. And so this is an exaflops AI system in one single rack. Let's take a look at the back of it. So this is what makes it possible. That's the back, that's the back, the DGX MV link spine. 30 terabytes per second goes through the back of that chassis.

That is more than the aggregate bandwidth of the internet. So we could basically send everything to everybody within a second. [00:09:00] That's wild, right? To send everything on the internet within a second. And then the technical innovation. To have these GPUs communicate with each other at the speed of light and act.

Like one coherent chip. There's perhaps no more singular technology that represents the advancement of human civilization than the GPU chip that NVIDIA has created. So let's keep going. So very powerful, right? The Exaflop. And it's still not fast enough. So we built another chip. This chip is just an incredible chip.

We call it the NVLink switch. It's 50 billion transistors. It's almost the size of Hopper all by itself. This switch chip has four. All right. That was a misfire. Let's go to the next one here. We did that one. And AI Foundry. We will do for you and the industry on AI what TSMC does for us building chips. So we go to TSMC with our big ideas, they manufacture it, we take it with us.

And so exactly the same thing here. AI Foundry, and the three pillars are the NIMS, NEMO Microservice, and DGX Cloud. The other thing that you could teach the NIMS to do is to understand your proprietary information. Remember, inside our company, the vast majority of our [00:10:00] data is not in the cloud, it's inside our company.

It's been sitting there. The enterprise IT industry is sitting on a goldmine. It's a goldmine because they have so much understanding of, The way work is done, they have all these amazing tools that have been created over the years, and they're sitting on a lot of data. If they could take that goldmine and turn them into co pilots, these co pilots could help us do things.

And so just about every IT franchise, IT platform in the world, that has valuable tools that people use, is sitting on a goldmine for co pilots. And they would like to build their own co pilots and their own chatbots. And we're announcing that NVIDIA AI Foundry is working with some of the world's great companies.

SAP generates 87 percent of the world's global commerce. Basically the world runs on SAP, we run on SAP. NVIDIA and SAP are building SAP Jewel. NVIDIA Co pilots using NVIDIA NEMO and DGX Cloud. ServiceNow, they run 85 percent of the world's Fortune 500 companies, run their people and customer service operations on ServiceNow.

‍