Nvidia Part III: The Dawn of the AI Era (2022-2023)

0:00 You like my Bucks t shirt? I love your Bucks t shirt. I went for the first time, what, two weeks ago when I was down for meeting at Benchmark and the nostalgia in there is just unbelievable. I can't believe you hadn't been before. I know Jensen is a Denny's guy, but

0:14 I feel like he would meet us at Bucks if we asked him. Or at the very least, we should figure out some NVIDIA memorabilia to get on the wallet books. Totally. Fit right in. All right, let's do it. Let's do it. Who got the truth?

0:29 Is it you, is it you, is it you Who got the truth now? Is it you, is it you, is it you Me down Straight! Another story on the way

0:42 Welcome to season thirteen, episode three of Acquired, the podcast about great technology companies and the stories and playbooks behind them. I'm Ben Gilbert. I'm David Rosenthal. And we are your hosts. Today we tell a story that we thought we had already finished.

0:58 NVIDIA. But the last eighteen months have been so insane listeners that it warranted an entire episode on its own. So today is a part three for us with NVIDIA telling the story of the AI revolution, how we got here. And why it's happening now, starting all the way down at the level of atoms and silicon.

1:18 So here's something crazy that I did a transcript search on to see if it was true. In our April twenty twenty two episodes, we never once said the word. Generative. That is how fast things have changed. Unbelievable. Totally crazy.

1:32 And the timing of all of this AI stuff in the world Is Unbelievably coincidental and uh very favorable. So recall back to eighteen months ago, throughout twenty twenty two. We all watched financial markets from public equities to early stage startups to real estate.

1:50 Just fall off a cliff. due to rapid rise in interest rates. The crypto and web three bubble burst. Banks fail. It seemed like the whole tech economy and potentially a lot with it.

2:02 was heading into a long winter. Including NVIDIA. Including NVIDIA, who had that massive inventory write off for what they thought was over ordering. Yeah. Wow, how things have changed. Yeah.

2:16 But by the fall of twenty twenty two, right when everything looked the absolute bleakest. a breakthrough technology finally became useful after years in research labs. large language models or LLMs. built on the innovative transformer machine learning mechanism burst onto the scene. First with OpenAI's chat GPT.

2:37 which became the fastest app in history to a hundred million active users. and then quickly followed by Microsoft, Google, and seemingly every other company. In November of twenty twenty two, AI definitely had its Netscape moment. And time will tell, but it may have even been its iPhone moment. That is definitely what Jensen believes. Yep.

2:59 Well, today we'll explore exactly how this breakthrough came to be, the individuals behind it, and of course why the entire thing has happened on top of NVIDIA's hardware. And software. If you want to make sure you know every time there's a new episode, go sign up at acquire.fm slash email. You'll also get access to two things that we aren't putting anywhere else.

3:19 One, a clue as to what the next episode will be. And two follow ups from previous episodes from things that we learned after release. You can come talk about this episode with us after listening at acquire.fm slash slack. If you want more of David and I, check out our interview show, ACQ2. Our next few episodes are about AI, with CEOs leading the way in this world we are talking about today. And a great interview with Doug DeMiro.

3:44 Where uh we wanted to talk about a lot more than just Porsche with him. But uh, you know, we only had eleven hours or whatever we had in Doug's garage. So a lot of the uh car industry chat and learning about Doug and his journey and his business. we saved for ACQ too, so go check it out. One final announcement. Many of you have been wondering, and we've been getting a lot of emails, when will those hats be back in stock while They're back. For a limited time, you can get an ACQ embroidered hat.

4:12 at acquire.fm slash store, go put your order in before they uh Go back into the Disney vault forever. This is great. I can finally get Jenny one of her own, so she stops stealing mine. Yes. Well, without further ado, this show is not investment advice. David and I may have investments in the companies we discuss, and this show is for informational and entertainment purposes only. David.

4:35 History and facts. Oh. Man. Well, on the one hand, we only have eighteen months to talk about. Except that I know you're not gonna start eighteen months ago.

4:43 On the other hand, we have decades and decades of foundational research. to cover. So when I was starting my research, I went to the natural first place, which was our old episodes from April twenty twenty two. And I was listening to them and I got to the end of the second one. And uh

4:59 Man, I had forgotten about this, I think. Jensen maybe wishes we all had forgotten about this. in one of NVIDIA's earning slides in twenty twenty one. They put up their total addressable market and they said they had a one trillion dollar TAM. And the way that they calculated this was

5:14 that they were gonna serve customers who provided a hundred trillion dollars worth of industry and they were gonna capture just one percent of it. And there was some stuff on the slide that was fairly speculative, you know, like autonomous vehicles and the omniverse and I think robotics were a big part of it. And the argument is basically like, Well, cars plus

5:36 Factories. Plus All these things added together is a hundred trillion, and we can just take one percent of that,'cause surely their compute will amount to one percent of that. Which I'm not arguing is wrong, but But it is a very blunt way to analyze that market.

5:51 Yeah, it's usually not the right way to um think about starting a startup, you know, oh if we can just get one percent of this big market, blah, blah, blah. It's the toppiest down way I can think of to size a market. So You Ben. rightly so called this out at the end of N video Part two, and you're like

6:08 You know I think to justify where NVIDIA is trading. At the moment. You kinda actually gotta believe that. All of this is gonna happen and happen soon. Autonomous cars, robotics, everything.

6:20 Yeah, importantly, I felt like the way for them to become worth what they were worth at that time literally had to be to power all of this hardware in the physical world. Yep. I kinda can't believe that I said this because it was Unintentional and uninformed, but I was kinda grasping at straws trying to play devil's advocate for you.

6:39 And we just spent most of that whole episode. Talking about how Machine learning Powered by NVIDIA. ended up having this incredibly valuable use case.

6:52 Which was powering social media feed recommenders. And that Facebook and Google had grown bigger than anyone ever imagined on the internet with Those feed recommendations. And NVIDIA was powering all of it.

7:05 And so I just sort of Idly proposed. Well Maybe. But what if

7:11 You don't actually need to believe any of that. To still think that NVIDIA could be worth a trillion dollars. What if Maybe, just maybe. The internet.

7:22 and software and the digital world. are gonna keep growing and there will be a new foundational layer that NVIDIA can power. Is that possible? And I think we were both like Yeah, I don't know. Let's end the episode.

7:36 Yeah, sure. We shrugged it off and we were like, All right, carve outs. But the crazy thing is that Of course, at least in this time frame. Most things on Jensen's trillion dollar TAM slide have not come to pass. But that crazy question.

7:49 Just might have come to pass. And from NVIDIA's revenue and earnings standpoint. Definitely has. It's just wild. All right, so how did we get here?

7:59 Let's rewind until The story. So Back in twenty twelve. There was the

8:06 Big Bang moment of artificial intelligence, or as it was more humbly referred to back then. Machine learning. And that was Alex Knett. We talked a lot about this on the last episode. It was three researchers from the University of Toronto. who submitted the Alexnet algorithm.

8:23 to the Image Net Computer Science competition. No image net. was a competition where you would look at a set of fourteen million Images. that had been hand labeled with what the pictures were of, like of a strawberry or a cat or a dog or whatever.

8:39 And David, you were telling me it's the largest ever use of mechanical Turk up to that point was to label the Image Net data set. Yeah. Until this competition and until Alex Net there was no machine learning algorithm that could accurately label images. So thousands of people on Mechanical Turk.

8:57 Got paid. However much two bucks an hour to Label these images. Yeah, and if I'm remembering from our episode, basically what happened is the AlexNet team Did

9:07 way better than anybody else had ever done. The complete step changed better. I think the error rate went from mislabeling images twenty five percent of the time to suddenly only mislabeling them fifteen percent of the time. And that was like a huge leap over the tiny incremental progress that had been made along the way. You're spot on. And the way that they did it. And what completely changed the fortunes of

9:31 the internet, of Google, of Facebook, and certainly of NVIDIA. Was They actually used old algorithms. A branch of computer science and artificial intelligence called neural networks. specifically convolutional neural networks.

9:45 Which had been around since the sixties. But They were really computationally intensive to train. And so nobody thought it would be practical. to actually train and use these things, at least not

9:59 Any time soon or in our lifetimes. And what these guys from Toronto did Is they went out probably to their local best buyer equivalent. In Canada. They bought two

10:10 G Force GTX five eighties. Which were the top of the line cards at the time. And they wrote their algorithm, their convolutional neural network, in CUDA In NVIDIA's software development platform for GPUs.

10:24 And By God, they trained this thing. on like a thousand dollars worth of consumer grade hardware. And basically the algorithm that other people had been trying over the years just wasn't massively parallel the way that uh graphics card sort of enables. So if you

10:39 actually can consume the full compute. of a graphics card. Then perhaps you could run some unique novel algorithm and do it on, you know, a fraction of the time and expense that it would take in these supercomputer laboratories. Yeah, everybody before was trying to run these things on CPUs.

10:56 CPUs are awesome. But they only execute one instruction at a time. GPUs, on the other hand, execute hundreds or thousands of instructions at a time. So GPUs, NVIDIA, graphics cards, accelerated computing, but

11:12 Jensen and the company likes to call this. You can really think of it like a giant Archimedes lever. Whatever advances are happening in Moore's Law and the number of transistors on a chip. If you have an algorithm that can run in parallel Which is not all problem spaces, but many can.

11:30 Then you can Basically lever up Moore's Law by hundreds of times or thousands of times or Today tens of thousands of times. And execute something a lot faster than you otherwise could.

11:42 And it's so interesting that There was this first market called graphics that was obviously parallel. Where Every pixel on a screen. is not sequentially dependent on the pixel next to it. It literally can be computed independently and output to the screen. So you have

11:59 However many tens of thousands or now hundreds of thousands of pixels on a screen that can all actually be done in parallel. And little did NVIDIA realize, of course, that. AI and crypto and all this other linear algebra matrix math based things that turned into accelerated computing, pulling things off the CPU and putting them on GPU and other parallel processors. was an entire new frontier of other applications that could use the very same technology they had pioneer for graphics. Yeah, it was pretty useful stuff. And this Alex Knight moment. And these three researchers from Toronto.

12:33 kicked off, you know, Jensen calls it, and he's absolutely right, the big bang moment. Four. Yeah. So David. The last time we told this story in full

12:43 We talked about this team from Toronto. We did not follow what this team of three went on to do afterwards. Yeah. So basically what we said was It turned out that a natural consequence of what these guys were doing was Oh.

12:57 Actually you can use this to surface The next one. post in a social media feed on like an Instagram feed or the YouTube feed or something like that. And that unlocked billions and billions of value. And those guys and everybody else working in the field, they all got scooped up by Google and Facebook.

13:13 Well That's true. And then as a consequence of that, Google and Facebook started buying a lot of NVIDIA GPUs. But Turns out there's also another chapter to that story that we completely skipped over. And it starts with the question you asked, Ben. Who are these people?

13:29 Yes. So The three people who made up the AlexNet team. Were of course. Alex Kushevsky, who was a PhD student.

13:37 Under His faculty advisor. The legendary computer science professor Jeff. Hint it.

13:44 I have an amazing piece of trivia about Jeff Hinton. Hm. Do you know Who his great Great grandparents were. No, I have no idea.

13:54 He is the great great grandson of George and Mary Boole. You know, like Boolean algebra and Boolean logic. This guy was born to be a computer science researcher. Oh my God. Right. Foundational stuff for computation and computer science. I also didn't know there were people named Bool, that that's where that came from. That's hilarious. Yeah. You know, the and or ex or nor operators. That comes from George and Mary. Wild.

14:21 So He's the faculty advisor. And then there was a third person on the team. Alex's fellow PhD student in this lab. One Ilya Sutziver.

14:33 And if you know where we're going with this, you are probably jumping up and down right now in your seat. Iliad is the Co founder and current chief scientist of Open AI. Yes.

14:44 So after AlexNet. Alex. Jeff and Ilya. Do the very natural thing. They start a company.

14:51 I don't know what they were doing in the company, but uh it made sense to start one. And whatever they did, it was gonna get acquired real fast. By Google. Within six months. So They get scooped up by Google.

15:03 They joined A bunch of other academics and researchers that Google has been Monopolizing, really, in the field. Three specifically. Greg Carado, Jeff Dean, and Andrew Ng.

15:15 the famous Stanford Professor. The three of them had just formed the Google Brain Team within Google. To turbocharge all of this AI work that has been unleashed by Alex D. And of course.

15:29 To turn it into Huge amounts of profit. Four. Google. Turns out.

15:34 Individually serving advertising that's perfectly targeted on the internet through Facebook or Google or YouTube. Is an enormously profitable business. And one that consumes a whole lot of NVIDIA GPUs. Yes. So about a year later, Google also acquires DeepMind famously. And then right around the same time, Facebook scoops up Computer science professor Jan Lacun, who also is a legend in the field. And the two of them basically establish a

16:00 Duopoly on leading AI researchers. Now, at this point, nobody is mistaking what these companies and these people are doing. For true human level intelligence or anything close to it.

16:14 This is AI that is very good. at narrow tasks, like we talked about social media feed recommendations. So the Google Brain team and Jeff and Alex and Ilya One of the big projects they work on is redoing the YouTube algorithm. And this is when YouTube goes from like money losing, you know, crazy thing that Google acquired to the just absolute juggernaut that it is today. In like

16:42 twenty thirteen, twenty fourteen. We did our YouTube episode not that long after. The majority of views of YouTube Videos were embeds on other web pages. This is when they build it into a social media site. They start the feed, they start autoplay. All this stuff is coming out of AI research.

17:00 Some of the other stuff that happens at Google Famously after they acquired Deep Mind. Deep mind. Built a bunch of algorithms to save on cooling costs. And

17:09 Facebook, of course. They probably had the last laugh in this generation because they're using all this work and Yan Lacoon is doing his thing and hiring lots of researchers there. This is just a couple of years after they acquired Instagram. Man, we need to like go back and redo that episode because

17:26 Instagram would have been a great acquisition anyway, but it was AI powered. recommendations in the feed that made that into a hundred, two hundred, five hundred billion dollar asset. For Facebook.

17:40 And I don't think you're exaggerating. I think that is literally what Instagram is worth to meta now. By the way, I have bought a lot of things on Instagram ad, so that the targeting works. It absolutely does. There's this amazing quote. from Astro Teller who ran Google X at the time and still does.

17:56 In a New York Times piece. Where he says that the gains From Google Brain. during this period. I don't think this even includes Deep Mind, just the gains from the Google Brain team alone. In terms of profits to Google.

18:08 More than funded. everything they were doing. In Google X. Which has there ever been anything profitable out of Google X? Google Brain.

18:18 Yeah, I mean, yeah. We'll leave it at that. So This takes us to twenty fifteen. When a few people in Silicon Valley.

18:27 start to realize that this Google Facebook AI duopoly is actually a really, really big problem. And Most people had no idea. About this. This is

18:40 really visionary of these two people. And not just a problem for like the other big tech companies,'cause you could make the argument it's a problem'cause like Siri's terrible All the other companies that have lots of consumer touch points, have pretty bad AI at the time. But the concern is for a much greater reason.

18:57 I think there are three levels of concern here. One. Obviously is the other tech companies. Then there's The problem of startups.

19:05 This is terrible for startups. How are you gonna Compete with Google and Facebook when this is the primary value driver of this generation of technology. I mean

19:16 There really is another lens to view what happened. With Snap. what happened with musically and having to sell themselves to bite dance and becoming TikTok and going to the Chinese. Maybe it was business decisions, maybe it was execution or whatever that prevented those platforms from getting to independent scale.

19:35 Snap's a public company now, but like it's no Facebook. Maybe it was that they didn't have access to the same AI researches that Facebook and Google had. Hm. That feels like an interesting question. It's probably w a couple steps too far in the conclusion, but still sort of a fun straw man to think about. A fun straw man, nonetheless. This is definitely a problem.

19:54 The third layer of the problem is just like This sucks for the world that all these people are locked up in Google and Facebook. This is probably a good time to mention this founding of OpenAI was motivated by the desire to find AGI or artificial general intelligence. first before the big tech companies did. And Deep Mind was the same thing. It was gonna be this winding and circuitous path at the time, since really nobody knew then or knows now the best path to get to AGI. But the big idea at OpenAI's founding was whoever figures out and finds AGI first

20:27 will be so big and so powerful so quickly, they'll have an immense amount of control. And that is best in the uh open. So these two people. We're quite concerned about this. Convene a very fateful dinner. In twenty fifteen.

20:43 At of all places. Is it the Rosewood? The Rosewood Hotel on Sand Hill Road. Naturally. It would have been way better if it were a Denny's or Bucks and Woodside or something like that. But it does actually just show like where the seeds of open AI come from. It is very different than this sort of organic scrappy way that the NVIDIA of the world got started. You know, this is Powers on high and existing money saying no, we need to will something into existence. Yep. So of course those two shadowy figures are Elon Musk.

21:15 And Sam Altman. Who at the time. was president of White Combinator. So they get this dinner together and they invite Basically all of the top AI researchers at Google and Facebook and they're like

21:28 Yep. What is it gonna take? for you to leave and to break this duopoly. And the answer is From almost all of them is

21:38 Nothing. You can't. Why would we ever leave? We're happy as clams here. We've gotten to hire the people that we want. We've built these great teams. There's a money spigot pointed at our face. Right.

21:49 Not only. Are we getting paid? Just Ungodly amounts of money. But

21:56 We get to work. directly with the best AI researchers in the field. If we were still at academic institutions, you know, say you're at University of Washington, amazing academic institution for computer science, one of the top in the world. Or the University of Toronto, where these guys came from. You're still at a fragmented market.

22:14 If you go to Google or you go to Facebook, you're with Everybody. Yep. So the answer is no from basically everybody. Except.

22:23 There's one person Who's intrigued by Elon and Sam's pitch. And to quote. An amazing wired article from the time by Cade Metz that we will link to in our sources. Quote.

22:36 The trouble was So many of the people most qualified to solve all these AI problems were already working for Google and Facebook, and no one at the dinner was quite sure that these thinkers could be lured to a new startup, even if Musk and Altman were behind it. But one key player. was at least open to the idea of jumping ship, and then they have a quote from that key player. I felt

22:57 There were risks involved. But I also felt it would be a very interesting thing to try. And that key player. Was. Yeah.

23:04 So let's give her. Yeah. So after the dinner, Ilya leaves Google. and signs up to become, as we said, co founder and chief scientist of a new independent AI nonprofit research lab backed by Elon and Sam.

23:18 Open AI. All right listeners. Now is a great time to talk about a new partner of ours here on Acquired, Lagora. The agentic operating system that is redefining how the world's best legal teams work. Yep. It's sort of obvious that AI is gonna completely change the legal industry. I bet most of you listening have dropped a contract into some sort of AI chat bot out there. Lagora took that insight and asked the question, what if you really built something with that power from the ground up for the legal industry?

23:50 So the founders did exactly what great founders do. operate with obsessive customer focus. They embedded inside a massive law firm for months. They sat with the lawyers just watching how the work really gets done. And that's how you get features that customers love, like tabular review, where you drop in a folder of hundreds of contracts and it pulls every key term into a grid a lawyer can actually work with. Legor's Bet Here is interesting. Since it lets each lawyer handle more complexity, any given person can increase the quality of their work.

24:23 and do higher value work. And this means that the pie can grow even as each individual task takes less time. And they recently launched Lagora Agent, offering greater intelligence and performance. The agent lets lawyers set an objective. Then it can handle the planning and the execution and delivery of the final product. Legal teams get to maintain full control and transparency since they're still involved where judgment is required. And Lagora works where you already work. You can use it within Microsoft Word while redlining or drafting. The early Lagora numbers essentially speak for themselves. When they have a head-to-head pilot with their top competitor, they win 70% of the time. Legora now has over a hundred thousand lawyers on the platform from twelve hundred legal teams in fifty countries. And crazily, they went from one million

25:13 To a hundred million in AR. In about Eighteen months. truly insane numbers. And that is the real test.

25:21 Plenty of things demo well, but the question is whether a busy associate actually reaches for it during crunch time, or whether a partner trusts it before going into a conversation with a major client. If your legal team wants to check it out, whether you're a law firm or you're in house at a company. You can learn more at Lagora.com slash acquired And just tell'em that Ben and David sent you. Okay, so David.

25:44 OpenAI has formed. It's twenty fifteen. Here we are eight years later and we have chat GPT. Super linear path from there to here, right? Turns out uh no. So As we were talking about a little bit. AI at this point in time.

26:01 Super good for narrow use cases. Looks nothing like GPT four. Today. The capabilities that it had

26:11 We're Pretty limited. And One of the big reasons Was that

26:17 The amount of data that you could practically train these models on. With Pretty limited. So the Alexnet example, you're talking about fourteen million images.

26:29 In the grand scheme of the internet. Fourteen million images is A drop in the bucket. And this was both a hardware and a software constraint. On the software side, We

26:40 just didn't actually have the algorithms to sort of suppose that we could be so bold to train one single foundational model on the whole internet. Like it wasn't a thing. Yeah, that was a crazy idea. Right.

26:53 People were excited about the concept. of language models, but we actually didn't know how we could algorithmically get it done. So in twenty fifteen Andre Carpathy.

27:05 who was then at OpenAI And went on to lead AI for Tesla and is actually now back at OpenAI. writes this seminal blog post called The Unreasonable Effectiveness of Neural Networks. And David, I don't think we're gonna go into it on this episode, but note that recurrent neural networks are a little bit of a different thing than convolutional neural networks, which was the

27:26 twenty twelve paper. The state of the art had evolved. Yes. And right around that same time There is also a video that hits YouTube right a little bit later in twenty sixteen.

27:36 that is actually on NVIDIA's channel. And it has two people in this very short one minute and forty five second video. One is a young Ilya Sitskiva, and two is Andre Carpathy. And here is a quote from Andrew. from that YouTube video.

27:53 One algorithm I'm excited about is a language model. The idea that you can take a large amount of data and you feed it into the network and it figures out the pattern in how words follow each other in sentences. So for example You could take a large amount of data on how people talk to each other on the internet. You can train basically a chat bot.

28:12 But You can do it in a way that the computer learns how language works and how people interact. Eventually we'll use that to talk to computers just like We talk to each other. Wow, this is twenty fifteen.

28:26 This is two years before the transformer, while Carpathy is at OpenAI. He both comes up with the idea or espouses the idea of a chat bot. So that sort of had already been discussed. But even before we had the transformer, the method to actually pull this off. he sort of had the idea that

28:44 There's an important part here. It figures out the pattern in how words follow each other in sentences. So there's this idea that. the very structure of language and the way to interpret knowledge is actually embedded in the training data itself rather than requiring Labeling.

29:03 This is so cool. So at Spring GTC this year. Jensen did a fireside chat with Ilya. And

29:12 It's amazing. You should go watch the whole thing. But In it this Question comes up. Jensen kinda poses as a straw man. Like

29:20 Hey, some people say that GPT three, four, chat GPT, everything going on, all these LLMs. They're just probabilistically predicting The next word in a sentence. They don't actually have knowledge. And Ilya has this amazing response to that.

29:37 He says, Okay. Well consider a detective novel. Yes. At the end of the novel. The detective gathers everyone together in a room. And says I am now going to tell you all

29:50 The name of the person who committed the crime. And that person's name is Blink. Ha ha The more accurately An LLM predicts that next word.

30:02 I. e. the name of the criminal. Ipso facto, the greater its understanding not only of the novel But of all general human level. Knowledge and intelligence.

30:15 Because You need all of your experience in the world, and as a human. To be able to guess who the criminal is. And the LMs that are out there today GPT three, GPT four.

30:28 Lama They can guess who the criminal is. Ooh yeah, put a pin in that. Understanding versus predicting.

30:37 uh hot topic du jour. So David, is now a good time to fast forward two years to twenty seventeen to the transformer paper. Absolutely. Ben tell us about the transformer. Okay, so

30:50 Google. twenty seventeen transformer paper. Paper comes out. called attention is all you need. And it's from the Google Brain team, right?

31:00 Yes. That Ilya just left. Just left, two years before to start open AI. So machine learning on natural language, just to set the table here. had long been used for things like autocorrect or foreign language translation. But in twenty seventeen

31:16 Google came out with this paper and discovered a new model that would change everything for these fields and unlock another one. So here is the scenario. You're translating a sentence from English to French. You could imagine that a way to do this would be one word at a time. In order.

31:32 But for anyone who's ever traveled abroad and uh tried to do this. you know that words are sometimes rearranged in different languages. So that's a terrible way to do it. you know, United States in Spanish is Estados Unidos, so failure on the very first word in that example. So enter this concept of attention, which is a key part of this research paper. So this attention, this fairly magical component of the transformer paper.

31:56 It literally is what it sounds like. It is a way for the model to attend to different areas of the input text at different times. You can look at a large amount of context. while considering what word to pick next in your translation. So for every single word that you're about to output in French You can look over the entire set of inputted words to figure out

32:20 What words you should wait heavily In your decision. for what to do next. This is why AI and machine learning was so narrowly applicable before.

32:30 If you anthropomorphise it and you think of it like a human. It was like a human with a very, very short attention span. Yes. Now here's the magical part.

32:40 While it does look at the whole input text to consider what the next word should be, it doesn't mean that it throws away the notion of position entirely. It uses a technique called positional encoding. So it doesn't forget the position of the words altogether. So it's got this cool thing where it weights the important part relevant to your particular word. And it still understands. Position. So remember I said the attention mechanism looks over the entire input every time it's picking what word to output.

33:10 That sounds very computationally hard. Yes. In computer science terms, this means that the attention mechanism is O of N squared. Oh. That's given me the heebie jeebies back to my intro C S classes in college. Oh, just wait till we get through this episode. It gets deeper. So obviously, yes. Traditionally you'd say this is very, very inefficient, and it actually means that the larger your context window, aka token limit, aka prompt length, gets. The more computationally expensive it gets on a quadratic basis. So doubling your input means quadrupling the cost to compute an output or

33:44 Tripling your input means Nine times the cost. It gets real gnarly. Yeah, it gets real expensive real fast. But GPUs to the rescue. The amazing news for us here. Is that these transformer comparisons can be done in parallel. So even though there are lots of them to do, if you have big GPU chips with tons of cores, you can do them all at exactly the same time. And previous technologies to accomplish this, like recurrent neural networks or LSTMs,

34:13 long short term memory networks. which is a type of recurrent neural network. Excera. Those required knowing the output of each step before beginning the next one, before you picked the next word. So in other words, they were sequential since they depended on the previous word.

34:30 Now with transformers, even if your string of text that you're inputting is a thousand words long. It can happen just as quickly in humid measurable time. As if it were 10 words long, supposing that there were enough cores in that big GPU. So the big innovation here is you could now train sequence based models. In a parallel way.

34:50 You couldn't train models of this size at all before, let alone cost effectively. Yeah. This is huge and probably for All listeners out there starting to sound very familiar to The world that we live in.

35:02 Today. Yeah, I sort of did a sleight of hand there morphing. translation to using words like context window and token length, you can kind of see where this is going. Yep. So this transformer paper comes out in twenty seventeen.

35:15 The significance is huge. But For whatever reason. There's a window of time where the rest of the world doesn't quite realize it. So Google obviously knows how important this is.

35:28 And There's like a year. Where Google's AI work Even though Ilya has left and open AI is a thing now. accelerates again beyond anybody else in the field. So this is when

35:41 Google comes out with smart compose in Gmail. And they do that thing where they have an AI bot that'll call local businesses for you. Remember that uh demo from IO that they did? Did that ever ship? I don't know. Maybe it did, maybe I mean it's just Google here. Like the capabilities are there. The product sense not as much. This is when they really start investing in Waymo. But again, where it really manifests is just

36:05 back to serving ads and search and recommending YouTube videos. Like they're just crushing it. In this period of time. OpenAI and everyone else, though, they haven't adopted Transformers yet. They're kinda Stuck in the past. And they're still doing

36:20 These really researchy computer vision projects. So like this is when they build a bot to play Dota Two, Defense of the Agents Two, the video game. And super impressive stuff. Like they beat the best Dota players in the world at Dota by literally just consuming computer vision, like consuming screenshots. and inferring from there. And that's a really hard problem because Dota two is not a game where you get to see the whole board at once. So it has to do a lot of like really intelligent construction of the rest of the game based on. just a single player's worth of input. So it's unbelievably cutting edge research. For the past generation. It's a faster horse, basically. Maybe, yeah. I mean they were also doing stuff like uh Universe, which was the

37:00 three D modeled world to train self driving cars. You don't really hear anything about that anymore, but they built this whole thing. I think it was using Grand Theft Auto as the environment, and then it was doing computer vision training for cars using the GTA world. I mean, it was crazy stuff, but it was kind of scatter shot. Yeah. It was scattered. And I guess what I'm saying is It was still in this narrow use case world. They weren't doing anything approaching

37:27 GPT at this point in time. Meanwhile, Google had kinda moved on. Yeah. No. One thing I do wanna say in defense of open AI and

37:37 Everybody else in the field at the time. They didn't just have their heads in the sand. To do what Transformers enabled you to do. Ben you're gonna talk about in a sec. Cost.

37:48 A lot in computing power. DPUs and NVIDIA and the Transformer made it possible. But To work with the size of models you're talking about. You're talking about spending

38:01 an amount of money that's certainly for a nonprofit. And Anybody really except Google. was untenable. Right.

38:10 It's funny, David, you made this leap to expensive and large models. All we were doing before was merely talking about translating one sentence to another. The application of a transformer does not necessarily require you to go and consume the whole internet and create a foundational model. But Let's talk about this. Transformers lend themselves

38:29 quite well, as we now know, to a different type of task. So for a given input sentence, Instead of translating to a target language, They can also be used as next word predictors. to figure out what word should come next in a sequence. you could even do this idea of pre-training with some corpus of text to help the model understand.

38:49 how it should go about predict that next word. So Backing up a little bit, let's go back to the recurrent neural networks, the state of the art. before transformers. Well they had this

38:59 Problem in addition to the fact that they were sequential rather than parallel. They also had a very short context window. So you could do a next word predictor. But it wasn't that useful because it didn't know what you were saying more than a few words ago. By the time you'd get to the end of the paragraph, it would forget what was happening at the beginning. It couldn't sort of hold on to all that information at the same time.

39:22 So this idea of a next word predictor that was pre-trained with a transformer could really start to do something pretty powerful, which is consume large amounts of text. And then complete the next word. Based on a huge amount of context. Yeah.

39:37 We're starting to come up to this idea of a large language model. And We're gonna flash forward here just for a moment to do some illustration and then we'll come back to the story. In GPT one. The first

39:49 Open AI model, this generative pre-trained transformer model, GPT. It used unsupervised pre training, which basically meant that as it was consuming this corpus of language, It was unlabeled data. The model was inferring the structure and meaning of language Merely by reading it.

40:09 Which is a very new concept in machine learning. The canonical wisdom is that you needed extremely structured data to train your smallish model on because how else are you going to learn what the data actually means? This was a new thing. You can learn what the data means from the data itself. It's like how a child consumes the world.

40:28 Where Only occasionally does their parent say, No, no, no, you have that wrong. That's actually the color red. But most of the time they're just self-teaching by observing the world. As a parent of a two year old. Can confirm. And then a second thing happens.

40:42 After this unsupervised pre-training step. Where you then have supervised fine tuning. The unsupervised pre-training used a large corpus of text to learn the sort of general language, and then it was fine-tuned on labeled data sets for specific tasks that you sort of really want the model to be actually useful for. So to give people a sense of Why we're saying that.

41:04 The idea of Training on Very, very, very large amounts of data here. is crazy expensive. T P T one.

41:14 had roughly a hundred and twenty million. parameters that it was trained on. GPT two had one point five billion.

41:23 GPT three had a hundred and seventy five billion. And GPT four, OpenAI hasn't announced, but it's rumored that it has about one point seven Trillion. parameters that it was trained on. This is a long way.

41:38 From Alex Net here. It's scaling like NVIDIA's market cap. Yeah. There is this interesting discovery, basically, that the more parameters you have, the more correctly you can predict the next word.

41:51 These models were basically bad. sub 10 billion parameters. I mean, maybe even sub a hundred billion parameters. They would just hallucinate or they would be nonsensical. It's funny when you look at some of the like one billion parameter models, you're like There is no chance that turns into anything useful ever. But by merely adding

42:08 more training data and more parameters. It just gets way, way better. There's this weirdly emergent Property. where transformer based models scale really well due to the parallelism. So as you throw huge amounts of data at training them. You can also throw huge amounts of NVIDIA GPUs at processing that. Exactly. And the output sort of unexpectedly gets magically better. I mean I know I keep saying that, but it is like Wait, so we don't change anything about the structure. We just give it way more data and let it run these models for a long time and make the parameters of the model way bigger. And like

42:45 No researchers expected them to reason about the world as well as they do. But It just kinda happened as they were exploring larger and larger models. So in defense of open AI They knew all this.

42:58 Bye. The amount of money that you would have to spend to buy GPUs or to rent GPUs in the cloud. To train these models. is prohibitively expensive. And you know, even

43:10 Google at this point in time, this is when they start building Their own ships. T P Us. 'Cause you know, they're still buying tons of hardware from NVIDIA. But they're also starting to source their own here.

43:21 Yeah. And importantly, they've at this point or getting ready to release TensorFlow to the public. So they have a framework where people can develop for stuff and they're like, look, if people are developing using our software, then maybe it should run on our hardware that's optimized to work with that software. So they actually do have this very plausible story around Why their hardware. Why their software framework. It was kind of a surprising move when they open sourced it because people were like, Gasp, you know, why is Google giving away the farm for free here? But

43:48 This was three, four years early and a very present move to really get a lot of people using Google Architecture Compute at scale. Yep. All within. Google Cloud.

43:59 Yeah. So with this it starts to look like maybe this whole open AI boondoggle didn't actually accomplish anything. And the world's AI resources are more than ever Just locked back into Google.

44:12 So in twenty eighteen. Elon. Get super frustrated by all this. Basically throws a hissy fit and quits. And pieces out of open AI.

44:22 There's a lot of drama around this that We're not gonna cover now. he may or may not have given an ultimatum to the rest of the team that he would either take over and run things or leave. Who knows? Did you want.

44:34 But Whatever happened. This turns out to be a Major catalyst. For the rest of the open AI team and truly a

44:42 History turning on a knife point. Moment. It was also a Probably super bad decision by Elon. But again, story for another day. So there's this great explanation of what happened in the semaphore piece that uh we'll link to in our sources.

44:55 The author says. That fall, it became even more apparent to some people at OpenAI that the costs of becoming a cutting edge AI company were going to go up. Google Brain's transformer had blown open a new frontier where AI could improve endlessly. But that meant feeding endless data to train it. a costly endeavor.

45:13 OpenAI made a big decision to pivot toward these transformer models. On March 11th, 2019, OpenAI announced it was creating a for-profit entity so it could raise enough money to pay for all the compute power necessary to pursue the most ambitious AI models. We want to increase our ability to raise capital while still serving our mission, and no pre-existing legal structure that we know of strikes the right balance. the company wrote at the time. OpenAI said it was cappits for investors with any excess going back to the original nonprofit.

45:42 Less than six months later, OpenAI took a one billion dollar investment from Microsoft. Yeah. And I believe this is mostly, if not all. Do to Sam Altman's. Influence and

45:55 Taking over here. So You know, on the one hand, you can look at this sort of skeptically and say, Okay, Sam, you took your nonprofit and you converted it into An entity worth. thirty billion dollars today.

46:08 On the other hand Knowing this history now. This was kinda the only Path. They had. They had to raise money.

46:16 to get the computing resources to compete with Google. And Sam goes out and does these landmark deals with Microsoft. Yeah, truly amazing. And their opinion at the time of why they're doing this is Basically, this is gonna be super expensive. We still have the same mission to ensure that artificial general intelligence benefits all of humanity, but it's gonna be ludicrously expensive to get there. And so we need to basically be a for-profit enterprise and a going concern and have a business that funds our research eventually to pursue that mission. Yep. So twenty nineteen they do the conversion to a for profit company.

46:49 Microsoft invests a billion dollars, as you say. And becomes the exclusive Cloud provider. For open AI. Which is going to become highly relevant here for NVIDIA.

47:01 More on that in a minute. June of twenty twenty, GPT three comes out. In September of twenty twenty, Microsoft licenses exclusive commercial use of the underlying model for Microsoft products. twenty twenty one GitHub Copilot comes out. Microsoft invests another two billion dollars in open AI.

47:19 And then of course this all leads to November thirtieth. Twenty twenty two. In Jensen's words. The AI heard around the world.

47:29 OpenAI comes out with chat. DPT. As you said, Ben, the fastest product in history to reach a hundred million users. In January twenty twenty three, this year, Microsoft invests another ten billion dollars. In open AI.

47:42 announces they're integrating GPT into all of their products. And then in May of this year, GPT four. Comes out. And that basically catches us up. Two.

47:52 Today. We eventually need to go do A whole nother episode about all the Details here of the OpenAI and Microsoft.

48:00 But for today The salient points are One. Thanks to all this. Generative AI as a user facing product.

48:08 Emerges as this. Enormous opportunity. Two To facilitate that happening. You needed

48:17 enormous amounts of GPU. Compute. Obviously benefiting. NVIDIA. But just as important, three

48:25 It becomes obvious now. That the predominant way That companies are gonna access and provide that compute. Is through the cloud.

48:35 Um The combination of those three things. Turns out to be basically. The single greatest moment. That could ever happen.

48:44 For NVIDIA. Yes. So You're teeing all of this up. And

48:49 So far I'm thinking, so this is like the open AI and Microsoft episode. Like what does this have to do with NVIDIA? And God, there's a great envidia story here to be told. So Let's get to the NVIDIA side of it.

49:01 All right, listeners, now is a great time to tell you about a longtime friend of the show, Vanta. AI has scrambled the whole security picture. It used to be that you proved that you were secure once a year on audit or a static PDF, then everyone would nod and you're done. But in an AI first world, that doesn't hold up anymore. Yep, your risk surface changes every week now.

49:23 A vendor turns on an AI feature or someone writes in a new model without telling IT And your posture is different than it was last week, let alone at your last audit. Fanta's own research found that around seventy percent of companies have this quote unquote shadow AI running with no security review at all. Right. And that's where Vanta comes in.

49:43 They're the leading agentic trust platform, meaning they've built the thing that closes the gap. And the way that they close that gap is Vanta Agent. Think of it as a GRSC engineer, that's governance, risk, and compliance, except that it's software and it doesn't sleep. It finds the issues, drafts the fixes, and cuts the time that you'd spend on vendor assessments in half. In half.

50:04 Which is exactly why more than sixteen thousand companies today run on Vanta. Companies like Ramp, Cursor, and Snowflake. All stay audit ready and catch the risks that crop up between audits across every vendor and Every AI tool. The whole environment. And that's the real value. Trust has to be continuous now, which is why Vanta automates your security, your compliance, and the work to earn and prove trust.

50:30 We're huge fans of Vanta over here, and literally hundreds of acquired listeners have become Vanta customers at their companies over the years. So you can get$1000 off Vanta at vanta.com slash acquired. That's V-A-N-T-A.com slash acquired for a thousand dollars off. And just tell them. That Ben and David sent you. Okay.

50:49 So NVIDIA. Okay, so we just said These three things that we've painted the picture of on the first part of the episode here that A generative AI is like possible a thing and it's now getting traction.

51:02 B. It requires an unbelievably massive amount of GPU compute to train. And three. It looks like the predominant way that companies are going to use that compute.

51:14 Is gonna be in the cloud. The combination of these three things is Is I think the most perfect example we've ever covered on this show. Of the old saying about luck being what happens when preparation meets opportunity for NVIDIA here. So obviously the opportunity is generative AI. But the preparation front.

51:33 NVIDIA has Literally. Just spent the past five years. Working insanely hard. To build a new

51:42 Computing platform. For The data center. A GPU Accelerated

51:50 Computing platform. Two In their minds Replaced The old CPU led

51:58 Intel dominated X eighty six architecture. in the data center. And for many years, I mean, they were getting some traction, right? And the data center segment was growing for NVIDIA, but people were like Okay, you want this to happen, but like why is it gonna happen? Right. There's these little workloads here and there that will toss you, Jensen, that we think can be accelerated by your cool GPUs.

52:20 And then you know, crazy things like crypto happened, and there was like AI researchers in academic labs that are using it as, you know, supercomputers. But for the longest time the data center segment of NVIDIA It just wasn't Clear that organizations had enormous

52:37 parts of their software stack that they were gonna shift to GPUs. Like why? What's driving this? And now we know what could be driving it, and that is AI. Uh not only could be, but if you look at their most recent quarter. Absolutely freaking is. Okay. So

52:54 Now it begs the question. Why is it driving it? And David Are you open to me giving a little computer science lecture on computer architecture.

53:04 Ooh. Please do Do my best professor impression here. Dude, I loved computer science in college. They were my favorite classes. I will say doing these episodes, this TSM C

53:17 It really does bring back the thrill of being in a C S lecture and being like Oh, that's how that works. Like it's just really fun. So Let's take a step back. and consider the classic computer architecture.

53:31 The von Neumann architecture. Now the Von Neumann architecture is what most computers most CPUs Where they can store A program

53:42 in the computer's memory and run that program. You can imagine why this is the dominant architecture. Otherwise we'd need a computer that is specialized for every single task. The key thing to know is that the memory of the computer can store two different things. The data that the program

53:59 uses. And the instructions of the program itself. The literal Lines of code. And in this example we're about to paint All of this is wildly simplified because I don't want to get into caching and speeds of memory and you know, where memory's located and not located. So let's just keep it simple.

54:18 So the processor in the von Neumann architecture. executes this program written in assembly language, which is the language that compiles down to the bytecode that the processor itself can speak. So it's written in an instruction set architecture, an ISA. From ARM, for example.

54:36 Or Intel before that. Yes. And each line of the program is very simplistic. So we're gonna consider this example where I'm gonna use some assembly language pseudocode. To add

54:49 The numbers two and three to equal Five Ben, are you about to program live on acquired? Well, it's pseudo assembly language code. So the first line is We're gonna load

55:02 The number two. from memory. We're gonna fetch it out of memory and we're gonna load it. Into A register on the processor. So now we've got the number two. actually sitting right there on our CPU ready to do something with.

55:17 That's line of code number one. Two. We're gonna load the number three in exactly the same fashion into a second register. So we've got two CPU registers with two different numbers. The third line. We're gonna perform an add operation, which performs the arithmetic

55:33 to add the two registers together on the CPU and store the value in some either third register or into one of those registers. So that's A more complex. Construction since it's

55:45 arithmetic that we actually have to perform, but these are the things that CPUs are very good at, doing math operations on data fetched from memory. And then the fourth and final line of code in our example. is we are going to take that five that has just been computed and is currently held temporarily in a register on the CPU is And we're gonna write that.

56:04 back to an address in memory. So the four lines of code are load, load, add, store. This all sounds familiar to me. So you can see each of those four steps is capable of performing one and only one operation at a time. And each of these happens with one cycle of the CPU. So if you've heard of gigahertz.

56:22 That's the number of cycles per second. So a one gigahertz computer could handle the simple program that we just wrote. Two hundred and fifty million times in a single second. But you can see something going on here. Three of our four clock cycles are taken up by loading and storing Data to memory.

56:41 Now this is known as the von Neumann bottleneck. And it is one of the central constraints. of AI, or at least it has been historically. Each step. Must happen.

56:51 In order. And only one at a time. So in this simple example, it actually would not be helpful for us to add a bunch more memory to this computer. I can't do anything with it. It's also only incrementally helpful to increase the clock speed.

57:06 If I double the clock speed, I can only execute the program twice as fast. If I need like a million X speed up for some AI work that I'm doing. I'm not gonna get it there with just a faster clock speed. That's not gonna do it. And

57:19 It would of course be helpful to increase the speed at which I can read and write to memory. But I'm kinda bound by the laws of physics there. There's only so fast that I can transmit data over a wire. Now, the great irony of all of this is that the bottleneck actually gets worse over time, not better, because the CPUs get faster and the memory size increases. But the architecture is still limited, so this one pesky single channel known as a bus.

57:44 I don't actually get to enjoy the performance gains nearly as much as I should because I'm jamming everything through that one channel and it only gets to sort of be used one time. Per every clock cycle. So the magical unlock, of course, is to make a computer that is not a von Neumann architecture. To make programs executable in parallel

58:06 and massively increase the number of processors. Or cores. And that is exactly what NVIDIA did on the hardware side, and all these AI researchers figured out. How to leverage On

58:18 The software side. But interestingly. Now that we've done that, David. The constraint is not the clock speed or the number of cores anymore.

58:28 For these absolutely enormous language models, it's actually the amount of on chip memory that concerns us. I thought you were going. And this is why the data center and what NVIDIA's been doing is so important. Yes. There's this amazing video that we'll link to on the Asianometry YouTube channel that we link to also on the TSMC episode. But the constraint today is actually in how much high performance memory is available on the chip. These models need to be

58:55 in memory all at the same time. And they take up hundreds of gigabytes. So while memory has scaled up, I mean we're gonna get flashing all the way forward, the H one hundreds on ship RAM is like eighty gigabytes.

59:08 The memory hasn't scaled up nearly as fast. as the models have actually scaled in size. The memory requirements for training AI are just obscene. So it becomes imperative to network multiple chips. And multiple servers of chips. and multiple racks of servers of chips together.

59:26 into one single computer, and I'm putting computer and air quotes there. in order to actually train these models. It's also worth noting. We can't make the memory chips any bigger due to a quirk of the extreme ultraviolet photolithography that we talked about, the EUV on the TSM C episode. Chips are already the full size of the reticle.

59:47 It's a physics and wavelength constraint. You really can't etch chips larger without some new invention that we don't have commercially viable yet. So what it ends up meaning is You need huge amounts of memory very close to the processors. All running in parallel. with the fastest possible data transfer. And again, this is a vast oversimplification. But you kinda get the idea of

1:00:10 why all of this becomes so important. Okay. So back to the data center. And here's what NVIDIA is doing that I don't think anybody else out there is doing. And why it's so important for them.

1:00:24 that all of this new generative AI world, this new computing era, as Jensen Dubs it. Runs in the data center. So NVIDIA has done three things.

1:00:35 Over the last five years. One and probably most importantly. Related to what you're talking about, Ben. They made One of the best acquisitions of all time.

1:00:45 Back in twenty twenty and nobody had Annie. Idea. They bought a quirky little networking company. out of Israel.

1:00:53 Called Melanox. Well, it wasn't little. They paid seven billion dollars for it. Yeah. And it was already a public company, right? It was, yep. Yeah. But it was definitely quirky.

1:01:04 Now what was Melanox? Melanox's primary product. was something called InfiniBand. Which we talked about a lot. with Chase Lockmiller on our A C Q two episode with him from Crusoe.

1:01:15 And actually, InfiniBand was an open source standard or managed by a consortium. There were a bunch of players in it. But the Traditional wisdom was While InfiniBand is way faster, way higher bandwidth, a much more efficient way to transfer data around a data center. At the end of the day, Ethernet is the lowest common denominator.

1:01:36 And so everyone had to implement Ethernet anyway. And so most companies actually exited the market, and Mellanox was kind of the only InfiniBand spec. Provider left. Yeah. So you said wait, what is InfiniBand? It is a competing standard to Ethernet. It is a way to move data.

1:01:54 Between Racks in a data center. And Back in twenty twenty. Everybody was like Ethernet's fine.

1:02:02 Why do you need more bandwidth than Ethernet between racks and a data center. What could ever require thirty two hundred gigabits a second of bandwidth down a wire. Yeah. A data center.

1:02:15 Well, it turns out if you're trying to address Hundreds. Maybe more than hundreds of GPUs as one single compute cluster to train a massive AI model. Yeah, you want really fast data interconnects between them. Right. People thought, oh sure, for supercomputers for these academic purposes, but

1:02:33 What the enterprise market needs in my shared cloud computing data center is Ethernet. And that's fine. And most workloads are going to happen right there on one rack. And maybe, maybe, maybe things will expand to multiple computers on that rack, but certainly they won't need to network multiple racks together. And NVIDIA steps in. And you got Jensen saying Hey dummies. The data center is the computer. Listen to me when I tell you.

1:03:00 The whole data center needs to be one. computer. And when you start thinking that way. You start thinking Geez, we're really gonna be cramming huge amounts of data through

1:03:12 wires that are going between these like how can we sort of think about them as if it's all sort of on ship memory or as close as we can make it to onship memory, even though that's in a box located three feet away. Yep. So

1:03:27 That's Piece number one. Of NVIDIA's. Grand data center plan over the last five years. Piece number two.

1:03:35 is in September twenty twenty two. NVIDIA makes a Quite surprising. Announcement. Of a new

1:03:44 Chip. Not just a new chip. An entirely new class of chips. That they are. Making.

1:03:50 Called the Grace CPU processor. NVIDIA is making a CPU. This is like heretical. But Jensen, I thought all

1:04:00 Computing was gonna be accelerated. What are we doing here on these armed CPUs? Yeah. These grace CPUs are not for Putting in your laptop. They are for

1:04:12 Being The CPU component. of your entire data center solution. That is. specifically from the ground up design to orchestrate with

1:04:22 These massive GPU clusters. This is the end game of a ballet that has been in motion for thirty years. Remember when The graphics card was subservient to the PCIe slot. In Intel's motherboard. And then eventually, you know, we fast forward to the future, NVIDIA makes these GPUs that are these beautiful standalone boxes in your data center, or perhaps these little workstations that sit next to you.

1:04:45 while you're doing graphic programming, while you're directly programming your GPU. And then of course they need some CPU to put in that. So they're using AMD or Intel or they're licensing some CPU. And now they're saying, you know what? Or actually just gonna do the CPU two. So now we make a box.

1:05:01 And it's a fulgrated NVIDIA solution. With R GPUs, R CPUs. R NV link between them, our InfiniBand to network it to other boxes. And you know, welcome to the show.

1:05:14 One more. piece to talk about the third leg of the stool there strategy before we Get to what it all means. That I think you're about to go to spoiler alert. You say solution. I hear gross margin. The third part of it is The GPUs.

1:05:29 Up until NVIDIA's current GPU generation, the hopper. generation of GPUs for the data center. There was only one GPU architecture. At NVIDIA.

1:05:40 And that same architecture And those same Chips from the same wafers made at TSM C. Some of them went to consumer gaming graphics cards. And some of those dyes went to

1:05:54 A one hundred GPUs in the data center. It was all The same. Architecture. Starting in September of twenty twenty two.

1:06:03 They broke out the two business lines into different architectures. So there's the Hopper architecture named after Great computer scientist Grace Hopper, I think rear admiral in the US Navy, Grace Hopper. Got it. Grace CPU. Hopper G P U Grace Hopper.

1:06:19 The H one hundreds. That was for the data centers. And then on the consumer side, they start a whole new architecture called Lovelace, after Ada Lovelace. And that is the RTX forty XX. So you buy uh, you know Top of the line RTX forty what have you gaming card right now.

1:06:37 That is no longer the same architecture as the H one hundreds that are powering chat GPT. It's got its own architecture. This is a really big deal. Because what they do with the hopper architecture is they start using what's called chip on wafer on substrate. C O W O S.

1:06:53 Chaos. When you start talking to the real semi nerds, that's when they start busting out the coas conversation. This is when a certain segment of our listeners are gonna get really excited. So essentially what this is Back to this whole concept of memory being so important for GPUs and for AI workloads. This is a way.

1:07:13 To stack. More memory. On The GPU chips themselves, essentially by going vertical in how you build the chips. This is the absolute bleeding edge technology. That.

1:07:26 is coming out of T SMC. And by NVIDIA bifurcating. They're chip architectures into a Gaming segment.

1:07:35 That does not have this latest Cowas technology. This allows them To monopolize like a huge amount of TSMC's capacity. To make the Kowaz chips.

1:07:47 specifically for these H one hundreds, which allows them to have way more memory. than other GPUs on the market. Yes. So this gets to the point of why can't they seem to make enough chips right now? Well, it's literally a TSM C capacity problem.

1:08:01 So there's these two components that are extremely related that you're talking about. The coas chip on waiver on substrate and the high bandwidth memory. So There's this great post from Semi Analysis where the author points out A two point five D chip, which is basically how you assemble this coas stuff. to get the memory really close to the processor.

1:08:19 And of course Two point five D. It is literally three D, but three D means something else. It's even more three D, so they came up with this two point five D denominator. Anyway. D chip packaging technol from TSMC is where you take multiple active silicon dies, like the logic chips and the stack of high bandwidth memory.

1:08:40 And they stack'em on one piece of silicon. And there's more complexity here, but the important thing is coas is the most popular technology for GPUs and AI accelerators for packaging these chips. And it's the primary method. to co-package high bandwidth memory.

1:08:57 Again, remember think back to the thing that's most important right now is get as much high bandwidth memory as you can closest to the CPU, next to the logic to get the most performance for trading and inference. So coas represents right now. About ten to fifteen percent of TSMC's capacities. And many of the facilities are custom built for exactly these types of chips that they're producing. So when NVIDIA needs to reserve more capacity. There's a pretty good chance that they've already

1:09:23 reserved some large part of the 10 to 15% of TSMC's total footprint. And TSMC needs to like Go make more fabs. In order for NVIDIA to have access to more coas capable capacity. Yeah. Which

1:09:38 As we know. It takes years for T SMC to do this. Yep. There are more experimental things that are happening, like I would be uh remiss not to mention. There are actually

1:09:48 experiments of doing compute in memory, like as we shift away from von Neumann and sort of all bets are off now that We're open to new computing architectures. There are people exploring, well, what if we just process the data where it is in memory instead of doing the very loss, expensive, energy intensive thing of moving data over the copper wire to get it to the CPU.

1:10:10 All sorts of trade offs in there, but It is very fun to sort of dive into the academic computer science world right now where they really are rethinking like What is a computer? So These three things that NVIDIA has been building.

1:10:23 The Dedicated hopper data center GPU architecture. The grey CPU platform. The Melanox powered networking stack. They now have

1:10:34 A full Sweet. Solution. For generative AI. Data centers.

1:10:41 And then When I say solution I hear margins. But let's be clear. You don't need to offer some sort of solution to get high margins if you're NVIDIA.

1:10:51 Price is set where supply meets demand. And They're adding as much supply as they possibly can right now. Like, believe me, for all sorts of reasons, NVIDIA wants everyone who wants H one hundreds to have H one hundreds. But for now. The price is kind of like a

1:11:06 I'll write you a blank check and Nvidia, you write whatever you want on the check. So their margins are. crazy right now. Just literally because there's way more demand than supply for these things. Yes. Okay, so let's break down what they're actually selling. So like you were saying, but

1:11:22 Of course. You can and lots of people do. Just go buy H one hundreds. I don't care about the gray CPU. I'm care about this Melanax stuff. I'm running my own data center. I'm really good at it. And the people who are most likely to do this are the hyperscalers, or as NVIDIA refers to them, the CSPs, the cloud service providers. This is AWS, this is Azure, this is Google, this is Facebook for their internal use. Like NVIDIA, don't give me one of these DGX servers that you assemble. Just give me the chip and I will integrate it the way that I want to integrate it. I am a

1:11:55 World class. data center architect and operator. I don't want your solution. I just want your tips. So they sell a lot of those. Now

1:12:04 NVIDIA of course has also been seeding new cloud providers out there in the ecosystem like our friends at Crusoe, also Core Weave and Lambda Labs, if you've heard of them. These are all new GPU dedicated clouds. that NVIDIA is working closely with. So they're selling H one hundreds and A one hundreds before that to all these cloud providers. But let's say you are an arbitrary company in the Fortune five hundred that is not a technology company.

1:12:30 And my God, do you not want to miss the boat on generative AI and you've got a data center of your own. Well, NVIDIA has a DGX for you. Yes. They do. full GPU based supercomputer solution in a box.

1:12:44 that you can just plug right into your data center and it just Works. There's nothing else on the market like this. And it all runs Cuda. It is all speaking the exact language of the entire ecosystem of developers that know exactly how to write software for this thing.

1:12:59 Which means that Whatever developers you already had who were working on AI or anything else. Everything they were working on. It's just gonna come right over and run within your brand new shiny AI supercomputer,'cause it all runs CUDA.

1:13:14 Amazing. More on Cuda in a minute. But as we said, you say solution, I hear gross margin. NVIDIA sells these DGX systems. For like A hundred and fifty to three hundred thousand dollars a box.

1:13:26 That's Wild. And Now With

1:13:30 All these three new legs of the stool, Hopper, Grace, and Melanox. These systems are just getting way more integrated, way more proprietary, and Way better. So if you want to buy A new top of the line DGX.

1:13:44 H one hundred system. The price starts. At five hundred thousand dollars. For one box. And if you want to buy

1:13:53 The DGX G H two hundred superpod. This is the AI wall that Jensen recently unveiled, the huge like Room full of AI. And it's like twenty racks wide. Imagine an entire row in a data center. Yes, this is two hundred and fifty six.

1:14:09 Grace Hopper DGX racks all connected together in one Yeah. They're building this as the first turnkey AI data center that you can just buy. And can train a trillion parameter GPT four class model. The pricing on that is

1:14:27 Call us. But I'm imagining like Hundreds of millions of dollars. Like I doubt it's a billion, but Hundreds of millions, easily. Wild.

1:14:38 Well, let's talk about the H one hundred. I've the baseball card right here on this uh insane thing that they've built. So they launched it in September 2022. It's the successor to the A one hundred. One GPU, one H one hundred cost forty thousand dollars. So that's how you get to that price point you're talking about. That's what they're selling to Amazon and Google and Facebook. Right. And you mentioned that five hundred thousand dollar price point. The five hundred thousand dollars is the eight forty thousand dollar H one hundreds in a box.

1:15:06 with the grey CPU and, you know, the nice bow around it. Yep. Which Do the math on that. So Eight times forty thousand.

1:15:15 That's Three hundred and twenty thousand dollars. So that's essentially an extra hundred and eighty thousand dollars of margin that NVIDIA is getting out of selling the solution. It's an ARM CPU. It doesn't cost them anything to make that. And these forty thousand dollar H one hundreds have margin of their own. So like every time they bundle more.

1:15:33 There's more margin in the fulfilled. I mean that's literally bundle economics. you are entitled to margin when you bundle more things together and provide more value for customers. But Just to like illustrate the way that this pricing works. So

1:15:46 The reason you want an H one hundred is they're thirty times faster than an A one hundred, which mind you is only like two and a half years older. It is Nine times faster for AI trading. The H one hundred is literally purpose built for training LMs.

1:16:01 Like the full self driving video stuff. It's super easy to scale up. It's got Eighteen and a half thousand CUDA cores. Remember when we were talking about the von Neumann example earlier, like

1:16:13 That is one computing core that is able to handle you know, those four assembly language instructions. This One each one hundred, which they're calling A GPU. Has eighteen

1:16:25 and a half thousand cores. That are capable of running. CUDA software. It's got six hundred and forty tensor cores, which are highly specialized for matrix multiplication. They have 80 streaming multiprocessors. So what are we up to here? Close to 20,000 unique cores on this thing.

1:16:44 It's got meaningfully higher energy usage than the A one hundred. I mean a big takeaway here is that NVIDIA is massively increasing the power requirement every time they come out with the next generation. They're both figuring out how to push the edge of physics, but they're also Constrained by physics. Some of this stuff is only possible with way more energy. This thing weighs seventy pounds.

1:17:06 This is one H one hundred. Jetson makes a big deal about this. Every keynote that he gives them like ah I can't lift it. It's got a quarter trillion transistors across Thirty five thousand parts. It requires robots to assemble it. Not only does it require physical robots to assemble it, it requires AI to design it.

1:17:24 They're actually using AI to design the ships themselves now. I mean, they have completely reinvented the notion of what a computer is. Totally. And this is all part of Jensen's Pitch.

1:17:37 Here, two. Customers. Yes. Our solutions Are very expensive.

1:17:43 However. He uses The line that he loves. The more you buy The more you save.

1:17:48 If you could get your hands on some. Right. But what he means by that is like Okay. Say you're McDonald's. And You're trying to build A generative AI.

1:17:57 So that I don't know. Customers can order something. You're using it in your business. If you were gonna try and

1:18:04 build and run that in your existing data center infrastructure. It would take so much time. And cost you so much more over the long run in compute. Then if you just went and bought my superpod here, you can plug and play and have it up and running in a month. Yep. And by the fact that this is all accelerated computing, the things you're doing on it, you literally wouldn't be able to do otherwise or might take you a lot more energy, a lot more time, a lot more cost.

1:18:28 There is a very valid story too. buying and running your workloads here or renting from any of the cloud service providers and running your workloads here is more performant. Because the results just happen. Much faster, much cheaper, or at all. Yeah.

1:18:42 You mentioned energy here, like this is Also. Jensen's argument. He's like Yes, these things take a ton of energy. But

1:18:49 The alternative takes even more energy. So we are actually saving energy if you assume this stuff is going to happen. Now there's a bit of caveat here in that It can't happen except on These types of machines. So He enabled this whole thing, but

1:19:05 He has a point. Oh, I totally buy it though. I mean, I think there's a very real case around look, you only have to train a model once. And then you can do inference on it over and over and over again. I mean the analogy I think makes a lot of sense for model training. Is

1:19:21 to think about it as a form of compression. LLMs are turning the entire internet of text into a much smaller set of model weights. This has the benefit of storing a huge amount of usefulness in a small footprint. But also enabling a very inexpensive amount of compute, again, relatively speaking.

1:19:40 in the inference step for every time that you need to prompt that model for an answer. Of course, the trade-off you're making there is once you encode all of the training data into the model, it is very expensive to redo it. So you better do it right the first time or figure out little ways to modify it later. Which a lot of ML researchers are working on. But I always think a reasonable comparison here is to compress a zillion layer Photoshop file. For anybody that's ever

1:20:03 dealt with, oh, I've got a three gigabyte Photoshop file. Well, that's not a thing you're gonna send to a client. You're gonna compress it into a JPEG and you're gonna send that. And the JPEG is in many ways more useful as a compressed facsimile. of the original layers comprising the Photoshop file. But The trade off is you can never get from that compressed little JPEG back to the original thing.

1:20:24 So I think the analogy here is like You're saving everyone from needing to make the full PSD every time because you can just use the JPEG the vast, vast majority of the time. So Hopefully we've now.

1:20:36 painted a relatively coherent picture of Both the Advances that made the Generative AI opportunity. Possible.

1:20:45 Got it. has truly become a real opportunity. And why NVIDIA. Even above the obvious. reasons was just so well positioned here, particularly because of the

1:20:57 data center centric nature of these workloads. And that they had been working so hard for the past five years. to fundamentally re architect the data center. Yeah. So on top of all this

1:21:10 NVIDIA recently announced. Yet another. Pretty incredible piece of their cloud strategy here. So Today.

1:21:19 Like we've been saying. If you want to use H one hundreds and A one hundreds, say you're an AI startup. The way you're probably gonna do that is you're gonna go to a cloud, either a hyperscaler or a dedicated GPU cloud like Crusoe or Core Weave or Lambda Labs or the like. And you're gonna rent. Your

1:21:37 GPUs. And Ben, you did some research on this. So like what does that cost? Oh, I just looked at the pricing pages on public clouds today. I think Azure and AWS were where I looked. You can get access to a DGX server that's eight A one hundreds for about thirty bucks an hour. Or you can go over to AWS and get a P5.48X large instance.

1:21:58 Which is eight h one hundreds, which I believe is an HGX server. for about a hundred dollars an hour. So about three times as much. And again. When I say you can get access, I don't actually mean you can get access. I mean that's the price. Right.

1:22:11 If you could get access, that's what you would pay for it. Correct. Okay. That's just getting the GPUs. But if you buy everything we were talking about a minute ago, say your McDonald's or UPS or

1:22:23 Whoever. And you're like, you know, I really like Jensen, I buy what you're selling. I want this whole integrated package. I want an AI supercomputer. in a box that I can plug into my wall. And have it run.

1:22:37 But I'm all in on the cloud. I don't run my own data centers anymore. NVIDIA. Has now introduced

1:22:45 DGX Cloud. Yeah, and of course you could rent these instances from Amazon, Microsoft, Google, Oracle. But like you're not getting that full integrated solution. Right. And you're getting some integration the way that the cloud service provider wants to create the integration using their proprietary services. And to be honest.

1:23:06 You might not have the right people on staff to be able to deal with this stuff in a pseudo bare metal way. Even if it's not in your data center and you're renting it from the cloud. You might actually need based on your workforce to just use a web browser and just use a real nice, easy web interface.

1:23:24 to load some models in from a trusted source. that you can easily pair with your data. And just click run. and not have to worry about any of the complexity of managing a cloud application. That's in

1:23:38 Amazon or Microsoft or something a little bit scarier and closer to the metal. Yep. So NVIDIA has introduced DGX Cloud. Which is a virtualized

1:23:47 DGX system. That is provided to you right now. Via other clouds. So Azure and Oracle and Google. Right, the boxes are sitting in the data centers of these other CSPs. Right. They're sitting in The other cloud service providers. But as a customer

1:24:05 It looks like you have your own Box. that you're renting. You log into the DGX Cloud website. Through NVIDIA. And it's all nice WYSIWYG stuff. There's an integration with Hugging Face where you can easily deploy models right off of Hugging Face. You can upload your data. Like everything is just really

1:24:26 WYSIWYG is probably the way to describe it. This is Unbelievable. NVIDIA launched their own Cloud.

1:24:33 Service. Through other clouds. And NVIDIA does have, I think, six data centers, but that I don't believe is what they're actually using to back DGX Cloud. No. So starting price.

1:24:46 For DGX Cloud. Is thirty seven thousand dollars a month. While A one hundred based system.

1:24:54 Not An eight one hundred base system. So The margins on this are insane for NVIDIA and their partners. A listener helped us out and estimated.

1:25:03 That the cost to actually build An equivalent A one hundred. DGX system. would be today something like a hundred and twenty K. Remember, this is the previous generation. This is not H one hundreds.

1:25:16 And you can rent it for thirty seven K a month. So that's three month payback. On the CapEx. For this stuff for NVIDIA and their cloud partners together. And even more for NVIDIA.

1:25:28 More important longer term. For enterprises that buy this. NVIDIA now has a direct sales relationship. with those companies, not necessarily intermediated by sales through Azure or Google or AWS, even though

1:25:45 The compute is sitting. in their clouds. Which is crucially important because at this point, the CFO Colette Crest said on their last earnings call that about half of the revenue from the data center business unit. is CSPs. And then I believe after that is the consumer internet companies and after that is enterprises.

1:26:06 So There's a few interesting things in there. One of which is Oh my god. There

1:26:11 revenue for this is concentrated among like five to eight companies with these CSPs. Two, they don't necessarily own the customer relationship. They own in the developer relationship through CUDA. You know, they've got this unbelievable ecosystem right now of NVIDIA developers that's stronger than ever. But in terms of the actual customer. half of their revenue is intermediated by

1:26:31 Cloud providers. The second interesting thing about this is Even today. in this AI explosion, the second biggest segment of data centers is still the consumer internet companies.

1:26:43 It's still all that stuff we were talking about before of the uh uses of machine learning to figure out what should show up in your social media algorithms and match ads to you, that's actually bigger than all of the direct enterprises who are buying from NVIDIA. So the DGX Cloud Play to sort of shift Some of that CSP revenue.

1:27:03 into direct relationship revenue. So All of this brings us to twenty twenty three. In

1:27:12 May of this year. Nvidia reported there. Fiscal twenty four earnings. Nidias on this weird January fiscal year end thing. So

1:27:21 Q one twenty four is essentially Q one twenty three, but Anyway. In which Revenue was up nineteen percent. quarter over quarter to seven point two billion, which is

1:27:31 Great,'cause remember they had a terrible end of twenty twenty two. With the write offs, crypto falling off a cliff and all that. Yes, it's amazing that in that strategy interview, when was that? And uh March of twenty twenty three, Jensen said last year was unquestionably a disappointing year. This is the year ChatGPT was released.

1:27:52 It is wild the roller coaster this company has been on. The time frame is so Compressed here. And part of that, of course, is Ethereum moving to proof of stake, the end of the crypto for NVIDIA, which I'm sure they're actually thrilled about. But part of it was they also put in A ton of pre orders for capacity.

1:28:11 With T S M C That then. they thought they weren't gonna need so they had to write down. So from an accounting perspective, it looks like a big loss, like a really big blemish on their finances last year. But now, oh my God, are they glad that they reserved all that capacity. Yep, it's actually going to be quite valuable.

1:28:29 So speaking of, you know, this Q one earnings is like Great. Up nineteen percent quarter over quarter. But then they drop the bombshell. Due to

1:28:40 Unprecedented. Demand. For generative AI compute. in data centers. NVIDIA forecasts Q two revenue of eleven billion dollars.

1:28:52 Which would be up another fifty three percent quarter over quarter over Q one. And sixty five percent. Year over year. The stock goes Nuts.

1:29:02 twenty five percent in after hours training. Yeah. This is a trillion dollar company, or at least this made them a trillion dollar company, but like a company that was previously valued at around eight hundred billion dollars popped twenty five percent After earnings.

1:29:18 Well, and it's even crazier than that. Back when we did our episodes last April. NVIDIA was the eighth largest company in the world by market cap, had about a six hundred and sixty billion. dollar market cap. That was down slightly off the highs, but that was kind of the Order of magnitude back then.

1:29:35 It crashed down below three hundred billion. And then within a matter of months. It's now back up over a trillion. Just wild. And then

1:29:44 All of this culminates. Last week at the time of this recording. When NVIDIA reports Q2 Fiscal twenty four. Earnings.

1:29:53 And this earnings release, we usually don't talk about like individual earnings releases unacquired,'cause like In the long arc of time, who cares? This was a historic event. I think this was one of if not the most incredible

1:30:08 Earnings release. by any scaled public company. Ever. Seriously, no matter what happens going forward Last week was a

1:30:16 Historic moment. The thing that blows my mind the most is that their data center segment alone did ten billion dollars in the quarter. That's more than doubling. off of the previous quarter.

1:30:31 In three months. They grew from four ish billion to ten billion of revenue in that segment. And revenue Only happens. when they deliver products to customers. This isn't pre-orders.

1:30:43 This isn't clicks. This isn't Wave your hands around stuff. This is we delivered stuff to customers and they paid us an additional six billion dollars this quarter than they did last quarter. So here are the full numbers. For the quarter. Total company revenue of thirteen point five billion.

1:31:00 Up. eighty eight percent from the previous quarter and over a hundred percent from a year ago. And then Ben, like you said, in the data center segment. Revenue of 10.3 billion. So 10.3 out of 13.5 for a segment that basically didn't exist five years ago for the company. That's up a hundred and forty one percent from Q one and a hundred and seventy one percent from a year ago.

1:31:24 This is ten billion dollars. That kind of growth at This scale? I've never seen anything like it. No. Neither has the market. That's right.

1:31:32 And so this This is the first time I noticed it. Jensen had talked about this in Q one earnings, so it wasn't. The first time. But he brings back

1:31:41 The trillion dollar dam. Not in a slide, I think this time he just talks about it. No, but in a new way that I think is a better way to slice it. This time it's different. You know, look, we'll spend a while here now talking about what we think about this, but

1:31:54 This is very different. This time he frames NVIDIA's trillion dollar opportunity. As the data center. And this is what he says. There is One trillion dollars.

1:32:06 worth of hard assets sitting in data centers around the world right now. growing at two hundred and fifty billion a year. Annual spend on data centers to update and add to that capex is two hundred and fifty billion dollars a year.

1:32:22 And NVIDIA has Certainly the most Cohesive Fulsom.

1:32:30 And coherent. Platform. To be the future of what those data centers are gonna look like for a large amount of compute workloads. This is a very different story than like, Oh, we're gonna get one percent of this hundred trillion dollars of industry out there.

1:32:48 And the thing you have to believe now, because whenever someone paints a picture, you say, Okay, what do I have to believe? The thing you have to believe is There is real user value being created. by these AI workloads and the applications that they are creating. And there's pretty good evidence. I mean, ChatGPT made it so OpenAI is rumored to be doing over a billion dollar run rate now.

1:33:11 maybe multiple single digit billions. And still growing, meaningfully. And so that is like the shining example. Again, that's the Netscape navigator here of this whole boom. But The bet, especially with all these Fortune five hundreds, is that

1:33:26 There are going to be GPT like experiences. In Everyone's private applications. in a zillion other public interfaces. I mean Jensen frames it as in the future every application Will have

1:33:42 a GPT front end. It will be a way that you decide that you want to interact with computers that is more natural. And I don't think he means like Versus clicking buttons. I think he means everyone can kinda become a programmer, but the programming language is English. And so when you're sort of like

1:33:59 Well, why is everyone spending all of this money? It is that the world's executives with the purchasing power to go right a ten billion dollar check last quarter to NVIDIA. For all this stuff. wholeheartedly believes from the data they've seen so far.

1:34:14 That this technology is gonna change the world enough. for them to make these huge bets. And the thing that we don't know yet. Is is that true? Is the GPT like experiences?

1:34:26 going to be an enduring thing for the far future. Or not. There's pretty good evidence so far that people like this stuff and that it's quite useful and transforming the way that you know, everyone lives their lives and goes about day to day and does their jobs and goes through school and

1:34:40 Yeah. On and on and on. But that is the thing you have to believe. All right listeners. Now is a great time to thank our longtime friend of the show, ServiceNow.

1:34:51 If you are running a large enterprise, AI agents are likely spread across every team, and deploying them is uh no longer the hard part. Yeah. The hard part is knowing what permissions they have, what employees are using them for, or what decisions AI is making. AI security for an enterprise at scale is not a small concern. Like the risks Are real. Exactly. And the challenge with AI is governing it, securing it, measuring it, and making sure that it actually delivers value. That is why ServiceNow built the AI control tower.

1:35:23 Yep, AI control tower gives enterprises a single place to see, manage, govern, and optimize AI across the entire business. And it works with Any AI, not just theirs. Every device on your network, every permission across every system. Every AI agent visible and secure in one place. And ServiceNow can do this because they've spent more than twenty years building the operational backbone of the enterprise, the workflows, governance, approvals, security controls, and institutional knowledge that power how work actually gets done across IT, HR, customer service, finance, and security. ServiceNow

1:35:57 already runs more than a hundred billion workflows annually and trillions of transactions for more than eighty five percent of the Fortune five hundred. So when companies need a place to govern AI at enterprise scale, they're building on a platform at the center of how their business already operates. And in a future, that isn't going to be one AI, it's going to be thousands of AI agents working across every function of the company. But the question is. Who's managing them all? So if you're trying to turn AI ambition into real business outcomes and make it work safely, securely, at scale. Go check out service now.com slash acquired and tell'em that Ben and David sent you. Okay.

1:36:34 So David. Analysis. We gotta talk about Kuda. Before we start analyzing anything else here. Talked about a lot of hardware so far on this episode.

1:36:44 But There's this huge piece of the NVIDIA puzzle that we haven't talked about since Part two. And CUDA, as folks know, was the initiative started in two thousand and six. By

1:36:56 Jensen and Ian Buck and a bunch of other folks on the NVIDIA team. to really make a bet on scientific computing that people could use graphics cards for more than just graphics and they would need great software tools to help them do that. It also was the glimmer in Jensen's eye of Ooh, maybe I can build my own relationship with developers. And you know, there can be this notion not of a Microsoft or an Intel developer who happens to be able to

1:37:19 you know, have a standard interface to my chip, but I can have my own developer ecosystem, which has been huge for the company. So Kuda. Has become the foundation

1:37:30 That Everything that we've talked about, all the AI applications. are written on top of today. So you know, you hear Jensen in these keynotes reference coup to the platform, coup to the language. And I spent some time trying to figure out like, when I was watching developer sessions and like literally learning some CUDA programs.

1:37:48 What is the right way to characterize it? And what is the right way to characterize it today? Because it has evolved a lot. Yes. So today. Cuda is starting from the bottom and going up.

1:37:58 A compiler. A runtime. a set of development tools like a debugger and a profiler. It is its own programming language, CUDA C plus plus. It has industry specific libraries.

1:38:10 It works on every card that they ship and have shipped since two thousand and six. Which is a really important thing to know. And if you're a CUDA developer, your stuff works. On everything. Anything NVIDIA, all this unified interface.

1:38:24 It has many layers of abstractions and existing libraries. That are optimized. So these libraries of code that you can call to keep your development work short and simple instead of reinventing the wheel. So you know, there are things that you can decide that you want to write. in C plus plus and just rely on their compiler to make it run well on NVIDIA hardware for you, or you can write stuff in their native language and try to implement things yourself in CUDA C.

1:38:50 The answer is It's incredibly flexible. It is. Very well supported. And there's this huge community of people.

1:38:59 That are developing with you and building stuff for you to build on top of. If you look at the number of CUDA developers over time. It was released in two thousand and six. It took

1:39:10 four years to get the first hundred thousand people. Then By twenty sixteen, thirteen years in They got to a million developers. Then just two years later, they got to two million.

1:39:22 So thirteen years to add their first thirteen million. Then two years to add their second. Twenty twenty two they had three million developers, and then just one year later. In May of twenty twenty three.

1:39:33 CUDA has four million registered developers. So at this point. There's a huge moat. For NVIDIA.

1:39:41 And I think when you talk to folks there, and frankly, when we did talk to folks there. They don't describe it this way. They don't think about it like, well, CUDA is our moat versus competitors. It's more like Well, look, we envisioned A world of accelerated computing in the future, and we thought there are way more workloads that should be parallelized and made more efficient. that we want people to run on our hardware and we need to make it as easy as possible for them to do that.

1:40:05 And we're going to go to great lengths. And have one two thousand people that work at our company, they're gonna be Full time software engineers building this. programming language and compiler and foundation and framework and everything on top of it.

1:40:18 to let the maximum number of people build on our stuff. That is how you build a developer ecosystem. It's different language, but The bottom line is they have a huge reverence for the power that it gives them at the company. This is something we touched on in our last episode, but has

1:40:34 really crystallized for me in doing this one. NVIDIA Thinks of themselves as and I believe is A platform company. Especially

1:40:46 This week. After the blowout earnings and everything that happened this quarter and the stock and whatnot. Sort of a popular take out there that you've been seeing a lot is Oh, we've seen this movie before.

1:40:58 This happened with Cisco. You could say over a longer timescale this happened with Intel. Yeah, these hardware providers, these semi conductor companies. They're hot when they're hot and people wanna, you know, spend CapEx and then when they're not hot, they're not hot. But I don't think that's quite the right way to characterize NVIDIA.

1:41:19 They Do make semiconductors. And they Do make data center gear. But really they are a

1:41:26 platform company. The right analogy for NVIDIA. Also is Microsoft. They make it. The operating system, they make the programming environment, they make many of the applications.

1:41:38 Right. Cisco doesn't really have developers. Intel never had developers. Microsoft had developers. And Intel had Microsoft. But Intel didn't have developers.

1:41:48 NVIDIA has developers. I mean, they've built a new architecture that is not a von Neumann computer. They've bucked fifty years of progress, and instead every GPU has a stream processor unit. And as you'd imagine You need a whole new type of programming language and compiler and everything to deal with this new computing model and that's CUDA and it freaking works.

1:42:09 And there's all these people That develop their livelihood in it. You talk to Jensen and you talk to other people at the company. And they will tell you we are a

1:42:17 Foundational computer science company. We're not just slinging hardware here. Yeah, I mean it's interesting. They're a platform company for sure. They're also a systems company. They're effectively selling memeframes. It's not that different than IBM way back when

1:42:31 they're trying to sell you a you know a hundred million dollar wall that goes in your data center. And it's all fully integrated and it all just works. Yeah. And maybe IBM actually is a really good analogy, like old school IBM here. They make the underlying technology, they make the hardware.

1:42:47 They make the silicon, they make the operating system for the silicon, they make the solutions for customers. They Everything. And they sell it. As a solution.

1:42:57 Yeah. Okay, so a couple other things to catch us up here as we're starting analysis. One big point I want to make is Let's look at a timeline,'cause I didn't discover this until like two hours before we started recording. In March of twenty nineteen, NVIDIA announced they were acquiring Melanox for seven billion dollars in cash. And I think Intel was

1:43:18 considering the purchase and then NVIDIA came in and kind of blew them out of the water. And It is fair to say nobody really understood what NVIDIA was going to do there and why it was so important, but the question is why. Well, NVIDIA knew that these new models coming out would need to run across multiple servers. Multiple racks.

1:43:36 And they put a huge level of importance on the bandwidth the machines. And Of course, how did they know that? Well In August of twenty nineteen.

1:43:46 NVIDIA released what was at the time the largest transformer based language model called Megatron. Eight point three billion parameters trained on five hundred and twelve GPUs for nine days. Which at the time at retail would have cost something like half a million dollars to train, which at the time was a huge amount of money to spend on model training, which is what, only four years ago, but Today that's quaint. Nvidia did that. Because they do a huge amount of research at the company and they work with every other company doing AI research and they were like, Oh. Yes.

1:44:18 This stuff is gonna work. And this stuff is gonna require the fastest networking available. And I think that has to do with why no one else saw how valuable the Mellinox technology could be. Yeah.

1:44:31 Another thing that I want to talk about for NVIDIA's business today. is this notion of the data center is the computer. And Jensen did a great interview with Ben Thompson last year. Where he talks about the idea that they build their systems full stack, like their dream.

1:44:47 Is that you own and operate A DGX superpod. And he says, we build our systems full stack, but we go to market in a disaggregated way, integrating into the compu fabric of the industry.

1:45:02 So I think that's his sort of way of saying Look. Customers need to use us in a bunch of different ways. So we need to be flexible on that. But we want to build each of our components such that If you do assemble them all together, it's this unbelievable experience and we'll figure out how to provide the right experience to you if you

1:45:19 only want to use them in piecemeal ways, or you want to use us in the cloud, or the cloud providers want to use us. Again, it's Build the product as a system, build the system full stack, but go to market in a disaggregated way. And I think if I remember right in that interview, Ben picked up on this and was like Wait, are you building your own cloud? And Tencent was like, Well maybe, we'll see.

1:45:40 And of course then they launched DJX Cloud in a Well, maybe we'll see sort of way. Yeah, you could imagine there are more NVIDIA data centers likely on the way that are uh fully owned and operated. Speaking of all of this, We gotta talk some numbers on margin.

1:45:55 This last quarter they had a gross margin. Of Seventy percent. And they forecast it for next quarter to have a gross margin. of seventy two percent.

1:46:06 I mean If you go back pre CUDA when they were a commoditized graphics card manufacturer. It was twenty four percent. So they've gone twenty four to seventy on gross margin.

1:46:17 And with the exception of a few quarters along the way for these strange one time events, It's basically been a linear climb quarter over quarter as they've deepened their moat and as they've deepened their differentiation in the industry. We're definitely at a place right now that I think is temporary due to the supply shortage of the world's enterprises and in some cases even governments. You look at the UK or some of the Middle Eastern countries. Like blank check I just need

1:46:45 access to NVIDIA hardware. That's gonna go away, but I don't think this very high, you know, sixty five percent plus margin is gonna erode too much. Yes, I mean I think two things here.

1:46:57 One. I really do believe what we were talking about a minute ago that NVIDIA is not just a hardware company. They're not just a chips company. They are a platform company And there is a lot of differentiation baked into what they do. If you want to train GPT or a G P T class model.

1:47:15 There's one option. You're doing it on NVIDIA. There's one option. And yes, we should. Talk about it's lots of less than GPT class stuff out there that you can do. And especially inference is more of a wide open market versus training that you can do on other platforms. But they're the best. And they're not just the best because of their hardware. They're not just the best because of their data center solutions. They're not just the best because of CUDA. They're the best because of all of those.

1:47:39 So The other thing. Sort of. Illustrative. thing for me that shows how wide their lead is.

1:47:45 We haven't talked about China yet. The land of eight hundreds. Yes. So what's going on? Last year China was twenty five percent or sales to mainland China. was twenty five percent of

1:47:58 NVIDIA's revenue. And a lot of that is they were selling to the hyper scalers, to the cloud providers in China. Baidu, Alibaba, Tencent, others. And by the way, Baidu has potentially the largest model of anyone.

1:48:10 their GPT competitor is over a trillion parameters and may actually be larger than GPT four. Wow, I didn't know that. Yep. Oh, it's wild. So Then

1:48:21 I believe also in September of twenty twenty two. Last year the Biden administration announced Pretty sweeping. regulations and bans on sales of Advanced computing infrastructure.

1:48:34 David, they're export controls. Don't say bans. I mean Yes. That's a fine line. Uh this is pretty close. To bands what the administration introduced. As part of that, NVIDIA can no longer sell their top of the line H one hundreds or A one hundreds to anybody in China. So

1:48:54 They created a Nerfed Skew, essentially. that meets the regulations, the performance regulations. The A eight hundred and H eight hundreds. Which I think they basically just crank down the N V Link's data transfer speeds. So it's like

1:49:09 Buying a Top of the line A one hundred, but not with as fast of data connections as you need, which basically makes it so you can't train large models. Right, or you can't train them as well or as fast as you could with the latest stuff. The incredibly telling thing.

1:49:24 Is that Those Chips and those machines are still selling like hotcakes in China. They're still The best hardware and platform that you can get in China.

1:49:35 Even a crippled version. And I think that's true anywhere in the world. And there's been a even a more recent spike of them because A lot of Chinese companies are reading the tea leaves and saying, Ooh, export controls might get even more severe. So I should get them while I still can, these eight hundreds. Yep. So I mean

1:49:52 I can't think of a better illustration of just how. Wide their lead is. Yeah, that's a great point. Talking about the rest of NVIDIA, just for a moment, I mean this episode is about the data center segment, but Oh, you mean they still make gaming cards too? It is worth talking about this idea that

1:50:08 Omniverse is starting to look really interesting. As of their conference six months ago, they had 700 enterprises who had signed up as customers. And the reason this is interesting is it could be Where their two different worlds collide. 3D graphics with ray tracing, which is new and amazing, and the demos are mind blowing. And AI. They have been playing in both of these markets. Since the workloads are both massively parallelizable.

1:50:34 That is the sort of original reason for them to be in the AI market. If you recall back to Way back our part one episode, the original mission of NVIDIA was to make graphics a storytelling medium. And then their mission has expanded as they've realized my God, our hardware's really good at other stuff that needs to be parallelized too. But fascinatingly with Omniverse. The future could actually look like applications where you need both amazing graphical capability and AI capability for the same application.

1:51:05 And I mean for all the other amazing uniqueness about NVI that we've been talking about and how well positioned they are. Adding this on top. where they're the number one provider for

1:51:17 Graphics hardware and software and AI hardware and software. Oh, and by the way, there's this Huge. application emerging where you actually do need both. They're just gonna knock it out of the park if that comes true. There was a super cool demo at a recent keynote. It might have been at Siggraph.

1:51:34 Where NVIDIA created a Game environment. You know, fully ray trace game environment. Looks like a Triple A game. You know, looks amazing. You know, basically. distinguishable from reality, but like you really gotta look hard to tell that this isn't real and this isn't a real human you're talking to. So there's a non playable character that you're talking to in N P C who's giving you like a mission.

1:51:57 They show this demo, it looks amazing. Then they're like The script. that that character was saying to you were not scripted. That was all generated.

1:52:07 With AI. dynamically. Oh. So you're like, holy crap. You know, you think about you play a Dirk mm The characters are scripted. But

1:52:18 In this world that you're talking about, you can have generative AI controlled avatars that are unscripted that have their own intelligences and that drives the story. Totally. Or, you know, an airplane that's in a simulation of

1:52:33 Not just a wind tunnel. But simulating millions of hours of flying time. using real time weather that's actually going on in the world and using AI to project the weather in the future. So you can sort of know the Real world potential things that your aircraft could encounter.

1:52:50 all in a generated graphical AI simulation. I mean, there's gonna be a lot more of this stuff to come. Yep. Totally. Another thing to know about NVIDIA. That we really didn't talk about on the last episode.

1:53:03 They're pretty employee efficient. They have twenty six thousand employees. And that sounds like a big number, but for comparison. Microsoft, whose market cap is only twice as big has two hundred and twenty thousand

1:53:18 So that is five X the number of employees per dollar of market cap going on over at Microsoft. And this is a little bit Farcicle since you know NVIDIA only recently has had such a massive market cap. But the scale of the platform that NVIDIA is building. Is On the order of magnitude of Microsoft scale. Right. They have forty six million dollars of market cap per employee.

1:53:42 Wild. Crazy. Which I think translates into the culture there as we've gotten to know Some folks there. It really is a very unique kind of culture. Like it is a big tech scale company. But you never hear about

1:53:56 The same kinda. Silly big tech stuff that you hear at other companies at NVIDIA. As far as I know, I could be wrong on this. There is no like, you know, oh, work from home or return to the office policy at NVIDIA. It's like No, just like

1:54:10 You do the job and you know Nobody's forcing anybody to come into the office here and like they've accelerated their ship cycles. Well, I also get the sense that it's a little bit of a Do your life's work or don't be here situation. Like Jensen is rumored to have forty direct reports and his office is basically just an empty conference room because he's just bouncing around so much and he's on his phone and he's talking to this person and that person. And like you can't manage 40 people directly if you're Worrying about someone's career ambitions.

1:54:40 Yeah. He's talked about this. He's like, I have forty direct reports. They are the best in the world at what they do. This is their life's work. I don't talk to them about their career ambitions. Like I don't need to. Like you know, yeah, for like recent college grads, we do mentoring it, but if you're a senior employee, you've been here for twenty years. You're the best in the world of what you do.

1:54:58 And we're hyper efficient and I start my day at five AM seven days a week and you do too. Crazy. Yeah, there's actually this amazing quote from Jensen that I heard on an interview with him that I was listening to where towards the end of the conversation the interviewer asked him. Jensen. You and NVIDIA do these

1:55:17 Just amazing things. What do you do to relax? And Jensen's answer is I'm reading this is a quote, direct quote. I relax all the time. I enjoy relaxing at work because work is relaxing for me. Solving problems is relaxing for me. Achieving something is relaxing for me. And he's a hundred percent serious. Like a thousand percent serious. How old is Jensen? The dude is sixty years old.

1:55:45 It kinda feels like all of his peers have either decided to retire and relax Or our uh you know, relaxing while running their companies. I think there's another crop of people that are doing that. And that is just not at all interesting to him or what he's doing. And I kinda get the sense like He's got another thirty years in him.

1:56:03 And he's architected the company in such a way that That's the plan. I don't think there's anyone else there where they're like getting ready for that person to take over. I think the company is a Extension.

1:56:15 of Jensen's thoughts and will and drive and belief about the future, and that's kind of what happens. I don't know if there is or isn't a Jensen and Lori Huang Foundation, but if there is, he's not spending his time on it. He's not Buying sports franchises. buying mega yachts or if he is, he isn't talking about them and he's working from them. Yeah, he's not buying social media platforms and newspapers. Yeah, totally.

1:56:40 I mean it is quite telling that when you watch one of their keynotes, it's Jensen on stage. And it's some customer demos. But it's not like The Apple keynotes where Tim Cook's calling up another Apple employee it's the Jensen show. Right.

1:56:53 Nobody would accuse Tim Cook of not working hard, I don't think, but you go to those keynotes and it's like Tim does the welcome. And then the handoff and you know a parade of other executives talk about stuff. Good morning. Tim Apple. I love it. Love Tim Apple. We gotta have Tim on the show sometime. That would be amazing. Yeah, text him. Text.

1:57:13 All right listeners. Now is a great time to talk about one of our favorite companies, Statsig. Yes, there is a reason why the best product teams rely on Statsig, whether they are iterating on their core product features or shipping AI powered experiences at scale. Yeah. In the

1:57:32 crazy speed of today's AI world. Shipping fast is just table stakes now. It's basically trivial to build and deploy your app constantly. The real advantage is how quickly you learn what changes actually created value for customers and how fast you can use that signal to guide what you ship next. This is where StatsIG comes in. It brings experimentation, feature flags, and product analytics into one unified system so teams can ship safely, test rigorously, and directly link what they changed. to how users actually behaved. So if you want to make learning your competitive advantage, whether you're building new AI experiences or just evolving your existing core product, go to statsig.com slash acquired to get started.

1:58:14 All right, power. Let's talk power. All right. So for listeners who are new to the show, this is the section where we talk about what it is about the company that enables them to Achieve persistent differential returns. Or in other words, to be more profitable than their closest competitor.

1:58:31 And do so sustainably. And NVIDIA is fascinating because They sort of have a direct competitor, but that's not the most interesting form of competition for them. Disintermediation is.

1:58:44 Sure, ostensibly there's NVIDIA versus AMD, but like AMD doesn't have all this capacity reserved from TSMC, at least not for the 2.5 D packaging process for the high-end GPUs. AMD doesn't have the developer ecosystem from CUDA. They're the closest direct comp. But It's

1:59:03 Amazon building training and inferentia. It's if Microsoft decides to go and build their own ship as they're rumored to with AMD. It's Google and the TPU. Facebook developing PyTorch and then leveraging their foothold with PyTorch with the developer community to figure out how to extend underneath of PyTorch. There's a lot of competitive vectors coming at Nvidia, but not directly. Not to mention

1:59:26 All the Data center hardware providers that are their direct competitors now too. Yep. Intel, et cetera. On down the line.

1:59:34 Yep. Now all that said They've got a lot of powers. So As we move through these one by one, I think let's just say them all and we can decide if there's something to talk about here.

1:59:45 Counter positioning is the one where I actually don't think there's anything here. I don't think there's anything that NVIDIA does. where there's another company that's actively choosing Not to do that. Because

1:59:57 Any company would want to be NVIDIA right now. I would have agreed with you, but I actually think There is strong counter positioning in the data center world right now. Nvidia and Jensen.

2:00:08 put a flag in the ground several years ago where they said, We are going to re architect the data center. And all the existing data center hardware and compute providers. had strong incentives not to do that. But like right now, what do you think other data center hardware providers What are they not doing?

2:00:26 Yeah, fair point. They're all Trying to put GPUs in the data center too. Everyone's just gonna chase exactly what NVIDIA is doing. Years behind them. That's the market right now.

2:00:36 Yep. Okay, fair enough. And the question is Will NVIDIA be able to stay? ahead in ways that matter. That I think is the entire analysis on the company right now is

2:00:47 in what ways that matter to customers at large scale and large markets. Will they be able to sustainably be ahead of people that are just chasing them and trying to copy what they're doing because the margin profile is so fat and juicy that people don't want to pay it.

2:01:02 Yep. So the second one, scale economies. This has Cuda written all over it. You can make massive fixed cost investments.

2:01:12 when you have the scale to amortize that cost across. And when you have four million developers who want to develop on your platform. You can justify Whatever it is, sixteen hundred people who actively on LinkedIn at NVIDIA today have the word CUDA. in their job title.

2:01:28 I mean, I'm sure it's actually even more than that. who just aren't, you know, they're saying software or something like that, but thousands of people of an investment. that they don't make any money on software. They may they make a de minimis amount on software. But That is amortized across

2:01:43 the entire developer base. I think it's worth saying a bit more. Here on this too, which we also talked about in our last episode. To me the dynamics here. Are a lot like Apple and iOS.

2:01:56 Yes. Versus Android. Apple has Thousands and thousands and thousands of developers working on iOS. Android also has thousands and thousands of developers.

2:02:08 working on it across a wide spread. Ecosystem. But at Apple it's all tightly controlled and it's coupled with hardware. At Android it's not and like as a user Maybe you'll get the latest operating system update. Maybe you won't.

2:02:23 I think this is exactly the right framing here, that NVIDIA is the apple of AI and PyTorch is sort of Android because it's open source and it's got a bunch of different companies that care about it. Open CL is the Android as it pertains to graphics, but it's pretty bad and pretty far behind. Rock M is the CUDA competitor made by AMD for their hardware, but again new not a lot of adoption. They're working on it, but

2:02:49 They've open sourced that because they realize they can't go directly head to head with NVIDIA. They need some different strategy. But yes, they are. Absolutely running the Apple playbook here. Yep. And

2:03:01 I think in the current state of things it's even more favorable to NVIDIA. Then iOS versus Android. Because NVIDIA has had First.

2:03:10 Dozens and then hundreds and now thousands of engineers working on CUDA. Fourth. Sixteen years. Meanwhile, the Android equivalent out there in the open source ecosystem

2:03:22 just been getting going. You know, if you think about the Delta of the timeline between iOS and Android. It was A year and a half, two years. There's a Probably at least ten, probably closer to fifteen year lead. That NVIDIA has.

2:03:38 And so we talked to a few people about this and we're like, Oh, what's going on in the open source ecosystem? Is there an Android equivalent? And even the most bullish people we talked to were like, Oh yeah, you know, now that Facebook has really moved. PyTorch into a foundation and outside of Facebook. That means that other companies can now contribute, you know. couple dozen engineers to work on it.

2:03:58 And you're like Cool? So AMD's gonna contribute a couple dozen, maybe a hundred engineers to work on PyTorch. And so will Google and so will Facebook and so will everybody else. NVIDIA has

2:04:09 Thousands of engineers working on CUDA. Ten years ahead. I sent you this graph, David, of my estimated number of employees working on CUDA per year since inception in 2006. And then If you look at the area under the curve and just take the integral, it's approximately 10,000 person years.

2:04:27 That have gone into CUDA. Like good luck. No, again. Open source is a very powerful thing. The market incentives are absolutely there for this to happen.

2:04:37 Right. That is the interesting point is every moat Only works if the castle is sufficiently small. If the prize at the end of the finish line becomes sufficiently large You're gonna need a bigger moat.

2:04:51 And you need to figure out a uh you know, how to defend the castle harder. I'm mixing so many metaphors here, but you get the idea. Yeah. I love it. This was a perfectly fine moat when the addressable market was a hundred billion dollars. Is it at

2:05:04 A trillion dollar market opportunity? Probably not. Basically it means margins come down and competition gets more fierce over time. And I think NVIDIA totally gets this because Part of this, as I was alluding to, is COVID related, but We talked way back in part one about how NVIDIA

2:05:23 ended up to save the company moving to a six month shipping cycle for their graphics cards when their competitors were on a one to two year shipping. That persisted for several years and then they relaxed back to a annual shipping cycle. There were annual GTCs. Since Covid

2:05:41 Nidia has reaccelerated to a six month shipping cycle. They've been doing two GTCs a year. most years since Covid. Which is insane. For the level of technology complexity that they're doing.

2:05:55 Yeah. Imagine Apple doing Two WW D Cs a year. Yeah. That's what's happening in NVIDIA.

2:06:02 It's crazy. So On the one hand, that's a culture thing. On the other hand, that is an acknowledgement of like we need to be pedal to the floor right now to outrun competition. We've built some structural ways to defend the business, but we need to continue running as fast as we've ever run to stay ahead. because it's such an attractive race that we're in.

2:06:20 Yeah. All right, so that's scale economies. Let's move to switching costs now. So far Everything of consequence, especially model training, especially on LLMs.

2:06:31 has been built on NVIDIA. And That alone is just a big pile of code and a big amount of organizational momentum. So

2:06:42 switching away from that, even from the software perspective, is gonna be hard. But there are companies Today. In twenty twenty three. both at the hyperscalers and Fortune five hundred companies that own their own data centers.

2:06:55 Making data center purchase and roll out decisions. that will last at least the next five years. because these data center re architectures don't happen very often. And so you better believe that

2:07:08 NVIDIA. is trying as hard as they can to ship as much product as they can while they have the lead. In order to lock in that data center architecture for the next N years. Yeah, we talked to many people in preparation for this episode, but one of the most interesting conversations was with some of our favorite public market investors out there, the NCS Capital guys. Who I stole many insights from for this episode. Oh, they're just so great.

2:07:35 And obviously I've been following NVIDIA in the space for a long time. They made the point that Data center revenue and data center capex. Is some of the stickyest

2:07:45 Revenue. That is known to humankind. Just the organizational switching costs involved in data center procurement and data center architecture standardization decisions. God, that's a mouthful even to say. At Fortune five hundred companies and the like is Like they're not changing that more than once a decade.

2:08:06 At most. So even if we're sort of in this bubbly moment around the excitement of generative AI before we necessarily know the full set of applications, NVIDIA is leveraging this excitement to go get to Block in. I've seen some people on the internet being like, they love how supply constrained they are. I don't think so. I think they're looking for capacity in every way they can get it. to exploit this opportunity while it exists.

2:08:30 I completely agree with that. Yeah. I think Yeah, again. We didn't talk to Collette, NVIDIA CFO, about this. I strongly suspect if I were them, I would be happy to trade some of this curse margin right now for increased throughput on sales. Yep.

2:08:45 But there's only one T SMC and There's only so many fabs that they have that can do the what do they call it, the Two point five D. Architecture. So Should we talk cornered resource?

2:08:56 Yeah. This is probably the textbook cornered resource. Nvidia has access to a huge amount of capacity at TSM C that none of their competitors can get their hands on. I mean, they did luck into this quartered resource a little bit. They reserved all that wafer supply for a different purpose, partially cryptomining. But AMD doesn't have it.

2:09:16 AMD does have a ton of capacity, it's worth saying at TSMC for their other products. data center CPUs, which they've actually been doing very well in But NVIDIA did end up with this wide open lane all to themselves on coas capacity at TSMC. And they gotta make the most of that for as long as they have it. Yep.

2:09:34 And I guess to say a little more though. It's not like This is not a commodity, as we talked about on our TSM C episode. Although TSM C is a contract manufacturer, it is the opposite of a commodity, especially at the highest end leading edge. It's like a invention delivered by aliens that very few humans know how to actually do. Yes.

2:09:57 It is worth acknowledging. It's kind of a two horse race for LLM training. I know we've been harping on NVIDIA. But Google TPUs are also manufactured at volume. You can just only get them through Google Cloud. And I think I don't know if you have to use the TensorFlow framework. Which has been waning in popularity relative to PyTorch, but it's certainly not an industry standard to use TPUs the way that it is to use.

2:10:24 NVIDIA's hardware. I suspect a lot of the volume of the uh TPUs is being used internally by Google for Bard. for doing stuff in Google search. Like I know they've added a lot of the generative AI capability to search. Yep, totally. Two points on this. Just sticking to the scope of

2:10:41 This This is a major casualty of a strategy conflict at Google obviously the way you wanna do this is the way NVIDIA is doing this of like Your customers.

2:10:53 Wanna buy through the cloud, you wanna be in every cloud. But obviously Google is not gonna be in A W S and Azure and Oracle and all the new cloud providers. They're only gonna be in G C P

2:11:04 Maybe, David. But I was gonna say though, through the expanded lens though, I think this makes sense for Google'cause Their primary business is their own products. Right. And they run among the most profitable businesses the world has ever seen.

2:11:19 So anything they can do to further advantage and extend that runway, they probably should do. Nothing has changed all of this with respect to the fact that what the previous generation of AI enabled with machine learning with regard to social media and internet applications being the most profitable Cash flow geysers known to man.

2:11:40 None of that has changed. That is still true in this current world and still true for Google. Yep. The last one that I had highlighted is network economies. They have a large number of developers out there and a large number of customers. that they can amortize these technology investments across and who all benefit from each other. I mean, remember

2:12:00 There are people building libraries on top of Cuda. And you can use the building blocks that other people built to build your code. You can write Amazing CUDA programs that just don't have that many lines of code because it's calling other pre-existing stuff. And NVIDIA made a decision in 2006 that at the time was very costly, like big investment decision, but it looks genius in hindsight to make sure that every GPU that went out the door was fully CUDA capable. And today there are 500 million CUDA capable GPUs for developers to target. It's just very attractive.

2:12:32 I'm putting this in network economies. I think it's probably more a scale economy than a network economy, but You could imagine a lot of people ho humming around NVIDIA in two thousand six to twenty twelve saying Why do I have to make it so that my software fits on this tiny little footprint and we can include CUDA taking up a huge amount of space on this thing and make all these trade offs and our hardware so that we can write why are people gonna use CUDA? And today it just looks so genius. Yeah, I mean we've talked about this many times on the show, including with Hamilton Helmer and Chen Yi themselves.

2:13:02 But four. platform companies like NVIDIA clearly is. There is this special brand of power that is a combination of scale economies and network economies, and this is What you're getting at. Yeah.

2:13:13 They do have branding power. For sure. Yeah, I actually think it's worth talking about this a little bit. This is the nobody gets fired for buying IBM. I mean, NVIDIA is the modern IBM in the AI era. Yeah.

2:13:25 Look, I don't feel confident enough to like Pound the table on this, but Given the nature of how the company started. And the fact that they Also have

2:13:37 the market leading product in a totally different business in graphics. Yeah, which is both consumers but also professional graphics. I Think that probably does lend some brand power to them, especially when

2:13:52 the CIO and the C suite at McDonald's is making a buying decision here, like everybody knows NVIDIA. Hm. You're saying that they carried their consumer brand into their enterprise posture. This is way, way, way down the stack in power, but I don't think it's hurt them. They've always been known as a technology leader and And

2:14:11 the whole world has known for decades at this point that the stuff that they can enable is magical. Yeah. There's a big strength leads to strength thing here, too, where I bet the Revenue results from last quarter. massively dwarf any brand benefit that they ever got from the consumer side. I think it's just the fact that like, hey, look, everyone else is buying NVIDIA.

2:14:34 I'd be an idiot not to nobody is getting fired for buying NVIDIA any time soon. Yep. Right, or taking a big dependency on them or targeting that development platform. It's just the like If you're innovating in your business, you don't want to take risk on the platform you're building on top of. You want to be the only risk in the value chain. All right, then the last one, right, is process power. Yeah, and this is probably the weakest one, even though I'm sure you could make some argument that they have process power. It's just that all the other powers are so much more valuable.

2:15:02 It's always so tricky to tease out. Yeah. You know, I think the argument here would just be like NVIDIA's Culture in their six months

2:15:11 shipping cycle that Clearly they had in the past, then they didn't have for a while, and now they have again. I don't know, I think you can make an argument here. Is it feasible? Let's do a thought exercise. Could any of their competitors really in any domain.

2:15:26 move to a six month ship cycle. That'd be really hard. Yeah. You know, could uh Apple sized company.

2:15:34 Do two WW D Cs a year, like No. The question is, does that actually matter? There are so many people that are using A one hundreds right now. And in fact, most workloads can be run on A one hundreds unless you're doing model training of GPT four.

2:15:49 I just don't know that it actually matters that much or as much as other factors. And I'll give you an example. AMD does have 3D packaging on one of their latest GPUs. It's a more sophisticated way of doing real copper to real copper direct connection.

2:16:07 without a silicon interposer. I'm getting into a little bit of the details, but basically it's more sophisticated than the process that the H one hundred two point five D is using to make sure that memory is extremely close to compute. And Does that matter? Not really.

2:16:23 What matters is everything else that we've been talking about, and nobody's gonna make a purchase decision on this thing because it's, you know, a little bit of a better mouse trap. Yeah, thinking about this more, I think actually brand is a really important power for NVIDIA right now. Yeah. And in a strength leads to strength way. So you can see why they're trying to sort of seize this moment. Yeah.

2:16:42 Playbook? All right. Let's move on to Playbook. So one thing that I want to point out is Jensen keeps referring to this as the iPhone moment for AI. And when he says it, The common understanding is that He means a new mainstream method for interacting with computers.

2:16:57 But there's another way to interpret it. Does this sound familiar, David, when I say A hardware company differentiated by software that then expanded into services. Yes, yes it does.

2:17:10 It's quite tongue in cheek to be referring to the iPhone moment of AI. When referring to oneself Nvidia as the apple. Because I really think that the parallels are uncanny that they have this vertically integrated hardware and software stack provided by NVIDIA. You use their tools to develop for it. They've shipped the most units. So developers have a big incentive to target that market. It's the best individual buy to target because they're the least cost sensitive and they appreciate you building the best experiences for them.

2:17:40 I mean it's the iPhone, but in many ways it's better because the target is a B to B target instead of consumers. Yeah. The only way in which it's different is Apple has always had a market cap that sort of lagged its proven value to users, whereas NVIDIA right now is uh Exactly over their skis.

2:17:58 Well Let's save that for Bull and Bear at the end. Great. The second one is that they've moved on from becoming a hardware company to truly being a systems company. Well, NVIDIA's chips are typically ahead. It really doesn't matter on a chip to chip comparison. That is not the playing field. It is all about how well multiple GPUs and multiple racks of GPUs work together as one system with all the hardware and networking and software that enables that. They have just

2:18:22 entirely change the vector of competition, which I think lots of companies can learn from. And my third one here is this quote that Jensen had again from the same strategy interview, which is You build a great company by doing things that other people can't do. You don't build a company. By fighting other people to do things that everyone can do.

2:18:41 And I think it's so salient. It comes out in all these interesting ways, one of which is NVIDIA never dedicated resources to building a CPU. Until there was a differentiated way and a real reason for them to build their own CPU, which is now. And the way that they're doing it, by the way.

2:18:58 It's not terribly differentiated. It's an off the shelf ARM architecture that they're putting some of their own secret sauce on, but it's not like They're doing Apple style M3.

2:19:09 creation of a chip from scratch. It's not the hero product. Right. There are many ways that NVIDIA sort of applies this where I think we talked about it in the last episode. If they think it's going to be a low margin opportunity, they don't go after it. But the nicer way to say that is

2:19:24 Well, we don't want to compete for things that anybody can do. We want to do things that only we can do. Oh, and by the way, we will fully realize the value of those things when we do them. Yeah. I think there's maybe a related Playbook theme here for NVIDIA of Strike when the timing is right.

2:19:39 I suspect that a lot of the inner competitive drive and motivation for Jensen and the company. Over the past. Ten. Fifteen years here.

2:19:49 Has been To really fight against Intel. Intel. Tried to kill them, as we talked about. many times in the previous episodes.

2:20:00 We talked to somebody who framed it as Intel was the country club and NVIDIA is the fight club. And back in the days the Intel country club didn't want to let NVIDIA in. Intel controlled the motherboard. Intel controlled the most important chip was the CPU. Intel would Integrate and commoditize all other chips into the motherboard eventually. And if they couldn't do that well, then they'd try and make the chips themselves. And they tried to run all these playbooks on NVIDIA. And NVIDIA just Barely survived. And then in the data center.

2:20:29 Intel controlled the data center for so long. PCI Express, you know, that was the interconnect in the data center for so long and NVIDIA had to live in there. And I'm sure they hated every single minute of it. But they didn't turn around ten years ago and just be like, Guess what? We're making a CPU too. They waited until the time was right. It is crazy. They used to have to plug into other people's servers.

2:20:51 And then they started making servers that plugged into other people's racks and rows and architectures. And then they started making their own entire rows and walls. And at some point here, they're gonna start running their own buildings full of servers too. And they're gonna say, We don't have to plug into anything. Yeah.

2:21:06 But I think for a lot of other leaders. It would have been hard to have the patience that they've had. Totally. You only get to do the stuff they're doing. If you Invested ten years ahead of the industry.

2:21:19 were wildly inventive and innovative in creating these like true breakthrough innovations. And we're really, really right about huge markets. Yeah. None of this stuff applies unless you're doing those three things. Yeah. Fortune five hundred CIOs aren't making buying decisions if

2:21:36 None of what you just said isn't true. Right. So there's this interesting conversation I wanted to have with you ahead of winding it up with the bull and bear case. So Think back to our AWS episode. We talked a lot about

2:21:50 How AWS is just Locked in. The Databases are a ridiculously durable advantage. Once your data has been shipped to a particular cloud.

2:22:01 Often literally in semi trucks full of hard drives. Snowball, yeah. It's hard to move off of it. There's this sort of interesting question of Will Winning. Cloud one point oh for all these Google, Microsoft, Amazon.

2:22:16 Will that Toe hold actually enable them to win in the cloud AI era. On the one hand, you'd think, yes, absolutely, because I want to train my AI models right next to where my data is. It's really expensive to move my data somewhere else to do that.

2:22:31 Case in point. Microsoft is the exclusive cloast provider for open AI. Which runs. As far as we know, solely on NVIDIA infrastructure, but they buy it all through Microsoft. Right.

2:22:44 On the other hand The experience that customers are demanding is the full stack NVIDIA experience, not this Oh, you found the cheapest possible cost of goods sold way to offer me. something that's like the experience that I want. And sometimes the cloud providers have to offer me an A one hundred or an H one hundred because

2:23:04 my code is way too complicated to ever re architect for whatever accelerated computing devices they're offering me that's first party and cheaper for them. I don't know. I just think for the first time in the last five years or so, I've sort of cocked my head a little bit at The moat. of these existing cloud providers and said, Huh.

2:23:23 maybe there really is a vector to compete with them and cloud is not a settled frontier. Yeah. Well, This is pejorative here. Cloud.

2:23:33 Is a euphemism for data centers, right? There's so much more too. The hyperscalers in public clouds than just Data centers, right? But physically they're data centers. Yeah. There is a Mile of distance. Metaphorically between like an equinix

2:23:47 And uh AWS. Yep. But They're data centers. And there is a fundamental shift, at least according to Jensen. A fundamental shift that is happening in data centers. So I think that probably does create some

2:24:01 Shifting sands that the clock is gonna have to navigate. Yeah. I bet the way it plays out is that Where you landed in cloud one point oh

2:24:10 strongly dictates where you will land in this AI cloud era,'cause at the end of the day, if customers are demanding NVIDIA stuff, then the cloud providers have every incentive in the world to make it so that you can run your applications great in their cloud. But also like there's more to this too. Crusoe exists. Core Weave exists, Lambda Labs exist. These are well funded startups with billions of dollars that

2:24:31 A lot of smart people think there's a major cloud sized opportunity for. Yep. That would not have happened a few years ago. Super true. All right. Let's do

2:24:40 The bull case and bear case and bring this one home. Oh boy. We've been trying to delay this as long as possible. This is the crux of the question right now. Yeah. I mean part of it is Is their existing moat big enough if GPUs

2:24:55 actually become a hundred billion dollar a year market. I mean right now GPUs in the data center are like a thirty billion dollar a year market going to like a fifty billion dollar next year. And like if this actually goes the way that everyone seems to think it's gonna go. There's just too many margin dollars out there for these big companies to not invest heavily.

2:25:17 Meta through Tens of billions of dollars making the metaverse. I mean, Apple's put fifteen billion dollars into rumored into their headset. Amazon's put tens of billions of dollars into devices which By all means was a terrible investment. How is Echo paying anything back?

2:25:35 Oh man, total sidebar. I'm so disappointed. I have standardized my house on the echo ecosystem and it keeps getting dumber. How in this world of incredibly accelerating AI capabilities Are my echoes getting dumber? Well, they need to train and inferenti a little bit harder. Oh Jesus.

2:25:54 Okay. Rant over. Yeah. I mean Never doubt big tech's ability to throw tens of billions of dollars into something if the payoff could be big enough. These are ludicrously profitable monopolies, except for Amazon's not that profitable. AWS is. Yeah. But Google, Facebook, Apple, at some point here.

2:26:12 There's a game of chicken that ends. And Some of these companies go all in and say, Yeah, we have smart engineers too. Like We're gonna figure this out. Yeah.

2:26:21 But also never underestimate. The inability of big tech to execute on stuff that it thinks can, especially with major strategy shifts. Yeah. Yeah. All right, so let's actually do this.

2:26:33 Bear case. Let's start with the bear case. So you just illustrated, I think, bear case number one, which is literally everybody else in the technology ecosystem. is now aligned and incentivized to say, I wanna take a piece of NVIDIA's Pi. And these companies have

2:26:50 Untold resources. Yep, and to put a finer point on that. Let's look at PyTorch for a minute. Now that all the developers or lots of developers are using PyTorch, it does Enable PyTorch to aggregate customers.

2:27:04 Which gives them the opportunity To disintermediate. Maybe you've got to write a lot of new stuff underneath and ship a lot of hardware. I mean, the cloud service providers have taken some steps here. It was originally developed by Meta, and while it's open source.

2:27:19 It's still hard for all these companies to invest in it if it's really sort of owned and controlled by Meta. So now PyTorch has been moved out into a foundation that a lot of companies are contributing to. And again, It is a absolute false equivalence to be like PyTorch versus NVIDIA. But in real Ben Thompson aggregation theory parlance, if you aggregate the customers, you have the opportunity then

2:27:43 to take more margin to disintermediate to direct where that attention is going. And PyTorch has that opportunity. That feels like the vector that a lot of these CSPs will try and compete on and say, look, if you're building for PyTorch, it runs really well on our thing too. Yeah. For sure.

2:27:59 No doubt that that's gonna happen. Alright, so that's bear case number two, kinda as part of bear case number one. The next one is like literally the market isn't as big as the market cap reflects. I think there's a pretty reasonable chance that there's some falter.

2:28:16 in the next twelve to eighteen months. where there's a crisis of confidence among investors. Where At some point something will come out. Where we all observe, oh, maybe GPTs aren't as useful as we thought.

2:28:29 Maybe people don't want chat interfaces. And that crisis of confidence and that mini bubble burst will trickle out to America's CIOs and CEOs, make it harder to advocate in the boardroom to make this big fundamental purchase and re-architecture of our whole budget from this year that we agreed on that I'm trying to propose us changing. There's a crypto like element to a excitement bubbsting. That will for some companies, slow their spend.

2:28:56 And The question is sort of like when that happens,'cause it's not an if, it's a when. I have a hard time believing that Given all the hype around everything right now. AI will be even more useful than everyone.

2:29:10 And it will continue in a linear fashion where without any drawdowns, everyone's excitement only gets bigger from here. It may end up being way more useful than anyone thought, but there at some point will be some valley or trough. And it's sort of about how Does NVIDIA fare? During that.

2:29:29 crisis of confidence. It's funny. You know, again, we talked to a lot of people for this episode, including A set of some of the foremost A I researchers and practitioners out there and Founders and C suites of

2:29:43 Companies that are Doing all this. And pretty much to a T, they all said the same thing when we asked them about this question. They all said Yeah. This is

2:29:52 Overhyped right now, of course, obviously. But On a ten year time scale, you haven't seen anything yet. The transformative change that we believe is coming. You can't even imagine.

2:30:03 The most interesting thing about the overhype is that it's actually showing up in revenue. It's everyone who is buying access to all this compute. believes something and for NVIDIA, because it's showing up in the form of revenue, the belief is real. And so they just need to make sure that they smooth the gap to customers actually realizing as much value as the CIOs of the world are currently investing ahead of. Yeah.

2:30:27 So I think the sub point to that that's worth the discussion right now is like Okay. Generative AI. Yeah, is it all it's cracked up to be. Well, David, I haven't asked you about this in like a month or so, but a month ago you were pounding the table insisting to me like I have no need for I've never used chat GPT. I can't find it to be useful. It's hallucinating all the time.

2:30:47 I never think to use it, it's not a part of my workflow. Like, where are you at? Still basically there, including forcing myself to try to use it a bunch in preparation for this episode. But Also, as we talk to more people.

2:31:00 I think I've realized that like David Rosenthal's use case doesn't really matter here at all. Right. A because As a business we are such a Hyper specialized Unique little unicorn thing where

2:31:14 Accuracy and the depth of the work and thought that we ourselves put into episodes is the paramount thing. Well and we have no co workers. There's so many things about our business that is weird. Like We never have to prepare a brief for a meeting. Right. All this stuff.

2:31:31 Anything external that we prepare is a labor of love for us. And there is nothing we prepare internal. I know people who use chat GPT to set their OKRs and I'm like, Okay, what's an OKR? And they're like, I wish my life were like that too. That's why I have Chat GPT do it. Right. Honestly, like I think through doing this and talking to some folks and reading

2:31:50 I think there's a very compelling use case for it for writing code right now. No matter what level of software developer you are, from zero all the way up through Elite software developer. You can get a lot more leverage out of this thing in GitHub Copilot. So Is that valuable? Yeah, for sure that's valuable. Yeah, the LLMs are unbelievably good at writing and helping you write code.

2:32:12 I'm a huge believer in that use case. Yep. And then I think, you know, there's the slightly more speculative stuff, but you can actually sort of see it now of like That gaming demo that I mentioned recently from NVIDIA of like Oh.

2:32:26 You're talking to a non playable character. That wasn't scripted. We Did an A C Q two episode recently with Chris Valenzuela from the CEO of Runway. That was used in

2:32:36 Everything everywhere all at once. And he said that's just the tip of the iceberg. Like the stuff that you can Do you think that's the same. that is happening, that's out there today. with generative AI in these domains is Astounding.

2:32:50 Yeah, I think what you're saying is One could be a bear. on your own experience. Every time you try to use a generative AI application, it doesn't fit into your workflow. You don't find it useful, you're not sticky. But on the other hand

2:33:03 Actually what AI will be is a sum of a whole bunch of niches. There's a video game market. There's a writing market. There's a creative writing market. There's a software developer market. There's a marketing copy market. You know, there's a million of these things and you just may happen to not fall into one of the first few niches of it. Yeah.

2:33:22 I think for me, at least again, just speaking personally too. I had a very strong element of skepticism initially because The timing was just too perfect. You know, it was like all UV C's out there. You just told everybody about how crypto's the future and whatever you're talking about and then Interest rates went to, you know, five percent and your world fell off a cliff.

2:33:44 Oh the number of people Who were like outraising a fund and they're like, The future is AI. Yeah, right. This the best time ever to be investing. And so there was a large part of me that I was just like, Come on, guys. Yeah. It's too perfect. You're right. It's too perfect. But This most recent couple months in this quarter for NVIDIA.

2:34:07 Put all that aside. Fortune five hundreds are adopting this stuff. CIOs are adopting this stuff. NVIDIA is selling real dollars. And learning also about What it takes to train

2:34:17 These models. And the step scale function of knowledge and utility. Going from

2:34:24 A billion to ten billion parameters to two hundred to a trillion parameter models. Yeah, like something's going on there for sure. So this leads me to my next bear case, which is The models will get good enough.

2:34:38 And then they'll all be trained and then we'll shift to inference. And most of the compute load will be on inference, where NVIDIA is less differentiated. There's a bunch of reasons I don't believe that. That is a popular narrative, though. One of the big reasons I don't believe that is The transformer is not the end of the road.

2:34:53 In a bunch of the research that we did, David, it's very clear that there are things beyond the transformer that are in the research phase right now. And The experiences are only gonna get more magical. and only gonna get more efficient. So there's sort of a second bear case there, which is

2:35:09 Right now we threw a brute force kitchen sink at training these things and all of that revenue accrued to NVIDIA because they're the ones that make the kitchen sinks. And over time, like you look at uh Google's chinchilla or lama two, they actually use less parameters. than GPT four.

2:35:28 and have equivalent quality, or you know, many other people can be the judge of that, but we're high quality models with less parameters. So there is this potential bear case around future models will be more clever and not require as much compute. It's worth saying that even today, the vast majority of AI workloads don't look like LLMs, at least until very recently. LLMs are like the current maxima in human history of jobs to be done. that require a ton of compute. And I guess the question is, will that continue? I mean, many other magical recent AI experiences.

2:36:04 have happened with far less expensive model training, like diffusion models and the entire genre of generative AI on images, which we really haven't talked about a lot on this episode because they're less compute intensive. But many tasks don't require an entire internet of training data and a trillion parameters to pull off. Yeah. That makes sense to me. And I think there also is some merit to

2:36:26 workloads are shifting to inference. That is happening. I agree with you. I don't think training is going anywhere. But Until recently, you know, thinking back to the Google days. Training was what everybody was spending money on. That's what everybody was focused on as usage scales with this stuff. than inference. And inference of course being

2:36:45 the compute that has to happen to get outputs out of the models after they're already trained. that becomes a bigger part of the pie. And as you say The infrastructure and ecosystems around doing that is less differentiated than training. Yeah. Okay, those are the bear cases. There's probably also a bear case around China.

2:37:02 Which is a legitimate one,'cause That's Gonna be a problem for lots of people. a large market that they won't be able to address for the foreseeable future in a meaningful way.

2:37:13 And just what's gonna happen generally, like obviously. China is racing to develop their own home grown ecosystems and competitors. And like that's gonna be a closed off market. So what's gonna come out of there? What's gonna happen? Yeah.

2:37:26 That's definitely one too. My last one is a bear case, but it ends up not being a bear case. For most companies I would say That If they were trading at this very high multiple and they just experienced this tremendous real growth in revenue and operating profit.

2:37:44 But that sort of spike to the system, when it goes away, will irreparably harm the company. When things slow down. Stock compensation's an issue, employee morale is an issue, customer perception's an issue. But this is NVIDIA. Yeah, this is nothing new. The number of times that

2:38:00 They've risen from the ashes after you know, years long terrible sentiment with something mind blowingly innovative. They're probably the best positioned company or the company with the best disposition. Two handle that when it happens. Oh, I thought that's a great turn of phrase there.

2:38:18 You uh upped your training model on uh language there. You should see the number of parameters. I love it. All right. Just to list the ball cases, one, Jensen is right about accelerated computing.

2:38:30 The majority of workloads right now are not accelerated. They're bound to CPUs. They could be accelerated, and that shifts from some crazy low number like five or ten percent. of workloads being accelerated today to fifty plus percent in the future. And there's way more compute happening in parallel, and that mostly accrues to NVIDIA. Oh, I have one nuance I want to add to that. On the surface I think a lot of people

2:38:52 look at that and they're like Yeah, come on. But I think there actually is a lot of merit to that argument in the generative AI world and everything we've talked about in this episode. I don't think Jensen and NVIDIA are saying

2:39:07 That traditional compute is going away or getting gets smaller. I think what he's saying is that AI compute. will be added onto everything and the amount of compute required for

2:39:22 Doing that. Will dwarf. what's happening in general purpose compute. So like it's not that people are gonna stop running SharePoint servers or that whatever products you use are gonna Stop.

2:39:33 using their whatever interfaces that they use. It's that Generative AI will be added to all of those things, and new use cases will pop up, which will also use traditional general purpose CPU based computing. But the amount of workloads that go into making those things magical. Is just gonna be so much bigger.

2:39:50 Yeah. Also, just a general statement on software development. Writing parallelizable code is really hard unless you have a framework to do it for you. even writing code with multiple threads, like if anybody remembers a C S college in class where they had a race condition or they needed to write a semaphore.

2:40:08 These are the hardest things to debug. And I would argue that a lot of things that could happen in an accelerated way aren't just because it's harder to develop for. And so if we live in some future where NVIDIA has reinvented the notion of a computer to shift away from von Neumann architecture into this stream processor architecture that they've developed, and they have the full stack to make it just as easy to

2:40:31 write applications and move existing applications. Especially once all the hardware's been bought and paid for and sitting in data centers, there's probably a lot of workloads that actually do make sense to accelerate if it's easy enough to do so. Yeah. That's great, but so your point is that There's a lot of latent

2:40:48 accelerated addressable computing out there that just hasn't been accelerated yet. Right. It's like uh this workload's not that expensive and I'm not gonna pay an engineer to go re-architect the system, so it's fine how it is. Ah bah là. I think there's a lot of that.

2:41:03 So Bullcase one, Jensen is right about accelerated computing. Bull case two. Jensen is right about generative AI. I mean, combined with accelerated computing, this will massively shift spend in the data center to Nvidia's hardware.

2:41:16 And as we've mentioned, OpenAI is rumored to be doing over a billion dollars in recurring revenue on chat GPT. So I think there's let's call it three billion because that's the most sort of credible estimate that I've heard and maybe that was a forecast for next year. But like They're not the only one. I mean Google with Bard, which I've found tremendously useful actually preparing for this episode. is not directly monetizing that, but they're sort of retaining me as a Google search customer by doing it. There is a lot of real economic value even today. Not nearly the amount that's sort of baked into the valuation, but

2:41:49 I suppose the bare case of this is that everything has to go right for NVIDIA, but the bull case is Indications are things are going right for NVIDIA. Yep. Third. Bull cases, NVIDIA just moves so fast.

2:42:01 Whatever the developments are, it's hard to believe that they're not gonna find a way to be really well positioned to capture it. That's just a cultural thing. Four is the point that you brought up earlier that there's a trillion dollars installed in data centers, 250 billion more being spent every year to refresh and expand capacity, and that NVIDIA could take a meaningful share of that. I think today what's their annual revenue at, like thirty billion or something. Well, if you run rate this current quarter, then it's uh like fifty. Fifty.

2:42:29 Plus, yeah. So right now that puts them at like twenty percent. of the current data center spend.

2:42:37 You could imagine that. Being much higher. Okay, wait, that includes the gaming revenue. It's about forty because the data center revenue is forty is ten. So forty annualized. All right, so fifteen, eighteen percent. Yeah. But you could imagine that creeping up. Again, if the accelerated computing and generative AI belief comes true, like they'll expand that 250 number and they'll take a greater percent of it.

2:42:59 Yep. An interesting way to Do a sort of a check on this math is to look at what other people in the ecosystem are reporting in their numbers.

2:43:08 TSMC in their last earnings said that AI hardware currently only represents six percent of their revenue. But all indications over there is that they expect AI revenue to grow. 50% per year for the next five years. Wow. So

2:43:24 We're trying to come at it from the customer workload side and say, is it useful there? But if you come at it from this other side of what do NVIDIA suppliers forecasting. And they have to put their money where their mouth is building these new wafer fabs to be able to facilitate that. And packaging and all the other things that go into chip. So

2:43:42 It's expensive for TSM C to be wrong. Yeah. That's another ball case. The last one that I have before leaving you with one final thought. Are you saying you have one more thing?

2:43:53 is that NVIDIA isn't Intel. And I think that's the biggest realization that you helped me have. And it's not Cisco. Yeah. The comparison we were making in the last episode was wrong.

2:44:04 They are Microsoft. They control the whole software stack and they simultaneously can have relationships with the developer and customer ecosystems. And I mean, it may even be better than Microsoft because they make all the hardware too. Yeah, it may be old school IBM. Right. Imagine if IBM operated in a computing market. of today's magnitude. Computing was tiny little market back then. Right. I mean it was like I mean it took P C wave to disrupt IBM.

2:44:29 Which was a personal computer in today's part edge computing, you know, device based computing. IBM dominated the B2B mainframe. Cycle of computing. And again, if you believe everything Jensen is saying and how he's steered the company for the last five years, we are going back into a

2:44:48 centralized data center modern version of a mainframe dominated computing cycle. Yeah. I suspect a lot of inference will get done on the edge. You think about the insane amount of compute that's walking around in our pockets that is not fully leveraged right now. There's gonna be a lot of machine learning done on phones that are gonna like call up to cloud based models for the hard stuff. No doubt.

2:45:10 I don't think training is happening at the edge any time soon, though. No. I certainly agree with that. All right, well, just like our T S M C episode, I wanted to end Of what it would take.

2:45:22 to compete with NVIDIA because my big takeaway from the TSMC episode was like, wow, that's a lot of things you have to believe. about a government putting billions of dollars in and hiring all this talent. And I was like, what's the equivalent for NVIDIA? So here's what you would need to do to compete. Let's say you could design GPU chips that are just as good. Which arguably AMD, Google, and Amazon are doing.

2:45:45 You'd of course then need to build up the chip to chip networking capabilities like N V Link that very few have. And you'd of course need to build relationships with hardware assemblers like Foxconn to actually build these chips into servers like the DGX. And even if you did all that, you'd need to create server to server and rack to rack networking capabilities as good as Mellanox, who was the best on the market with InfiniBand that NVIDIA now fully owns and controls. Which basically nobody has. And even if you did all that, you'd need to go convince all the customers to buy your thing.

2:46:16 Which means it would need to be either better or cheaper or both, not just equal. T NVIDIA. And by a wide margin too, to this brand, you're not gonna get fired for buying NVIDIA any time soon. Like This is the canonical, you gotta be ten X better than NVIDIA on this stuff if you're gonna convince a CIO. Yep.

2:46:34 And even if you got the customer demand, you'd need to contract with TSM C to get the manufacturing capability of their newest cutting edge fabs to do this 2.5 D Coas, lithography, and packaging. Which there of course isn't any more of. So you know, good luck getting that. And even if you figured out how to do that, you'd need to build software that is as good or better than CUDA.

2:46:58 And of course, that's gonna take ten thousand person years. Which would of course cost you not only billions and billions of dollars, but all that actual time. And even if you made all these investments and lined all of this up, you'd of course need to go and convince the developers to actually start using your thing instead of CUDA. Well, NVIDIA also wouldn't be standing still, so you'd have to do all of this in record time. to catch up to them and surpass whatever

2:47:26 additional capabilities they developed. since you started this effort. So I think the bottom line here is it nearly impossible to compete with them head on. And if anybody's gonna unseed NVIDIA in the future of AI and accelerated computing, it's either gonna be from some unknown flank attack that they don't see. Or the future will turn out to just not be accelerated computing in AI, which seems very unlikely. Yeah.

2:47:51 Well When you put it that way. I think the conclusion that we can come to. is that Mark Andrewson was right in

2:48:01 What years was this that we were talking about on uh it was like twenty fifteen or something. Yeah, like twenty fifteen, twenty sixteen. They should have put every dollar of every fund that A sixteen Z raised into NVIDIA's market price of the stock every single day. Yeah. 'Cause they were seeing all of these startups. doing deep learning, machine learning at the time, early AI And they were all building on NVIDIA and they should have just said

2:48:26 No, thank you, Tall of them, and put it all in video. Mark is right once again. Strength leads to strength. There you go. There it is. Well, listeners, I acknowledge that this episode generalized a lot of the details, especially for technical out there, but also for the finance folks who are listening. Our goal was to make this more of a lasting NVIDIA part three big picture episode than sort of a

2:48:49 How did they do last quarter and what are the implications on that of the next three quarters? So hopefully this holds up a little bit longer than Just some current NVIDIA commentary. But thank you so much for going on the journey with us. Yeah. We also, as we've alluded to throughout the show, we owe a bunch of thank yous to lots of people who were so kind to help us out, including

2:49:09 people who have way better things to do at their time. So we're very, very grateful. I mean one, Ian Buck from NVIDIA, who leads the data center effort and is one of the original team members that invented CUDA way back when. Really grateful to him for speaking with us to prep for this. Absolutely. Also, big shout out to

2:49:27 friend and listener of the show, Jeremy from ABC Data, who prepared four PDFs for us. Completely unprompted, like an insane write up for us about a lot of the technical detail behind this. Private blog posts. Yeah, private blog post, so Our acquired community. Is just the best. Like you guys

2:49:46 Continue to blow us away. So Thank you. Julian, the CTO of Hugging Face. Ornit's Yone from AI2, Luis from Octo ML. And of course our friends at NZS Capital. Thank you all for Helping us research this.

2:50:02 Indeed. All right. Carvats. Let's shift gears. Carvalho.

2:50:07 What'd you get? My wife and I have been on an alias binge. Oh wow. Yeah, Jennifer Garner? Yes. I never saw it when it came out. It is like the perfect early two thousands junk food when you have one more hour at the end of the day and you're just laying on the couch.

2:50:22 Then I never have one more hour at the end of the day. I have a two year old. But I really appreciate it. For uh sixteen years from now when she goes to college. I'll keep that on my list. Oh, you play games. Oh, that's true. But that's research. I'm just checking out the latest graphics technology. So my review of alias is it's a little bit campy. They repeat themselves pretty often. I mean, it's weird to observe how much TV has changed between now and then because they make very similar shows today.

2:50:47 But They're just much more subtle. They're much darker. they leave much more sort of to the imagination. And in the early two thousands everything was just so like explicit and on the nose and restated three times.

2:50:59 I'm just glad the show doesn't have a laugh track. But it's well worth the watch. Sometimes you have to imagine it has a different soundtrack. 'Cause every episode has like a Matrix type song to it. Bomba da bump ba da bump ba dump dump. Yes, that's right. This is like the T V version of the Matrix, right? Yes. But it's great.

2:51:16 I don't know, we're having a lot of fun watching it. My car out. Related for my stage of life. Also something I missed and discovered recently. We just watched Our first.

2:51:26 Full Movie. Fall Disney movie. with her. Who'd you major milestone. Um

2:51:34 She freaking loved it. I think we picked a great one. Moana. Which neither Jenny or I had seen before. And in reading just a little bit about it afterwards, You know how?

2:51:46 Super sadly. Pixar kind of fell off in recent years, like Stuch up. Bummer. I mean they're still Pixar, but like they're not Pixar. It's not the uh guaranteed hit every time that it used to be.

2:51:58 Yeah. So Moana came out in this kind of generation with Tangled and some of the other stuff out of actual. Disney animation after the Pixar. acquisition that are just like these are return to form, Eisner era, Disney animated, just like

2:52:13 Fires on all cylinders and We loved it. We watched with our brother and sister in law. Who Don't have kids and are

2:52:21 thirty something's living in San Francisco. They love to Our daughter loved it. highly recommend Moana, no matter what live phase you're in. All right. Great.

2:52:30 Adding it to my list. And it's got the rock. How can you complain? There you go. Well, listeners. If you want to be notified every time we drop a new episode and you want to make sure you don't miss it and

2:52:40 You want Little hints to play a guessing game at our next episode or you want follow ups from our previous episode. In case we learn from listeners, hey, here's a little piece of information that we wanted to pass along, we will exclusively be dropping those in the email, acquire.fm slash email. It was so fun. I think you're about to talk about our Slack. It was so fun watching people in Slack talk about the hints for this episode.

2:53:04 We wrote the little teaser and I was like, Oh, everybody's gonna know exactly what this is. No one got it. I was shocked. Yeah. Eventually somebody did, but it took a couple of days. Yeah. Uh we have a hat. You should buy it. Um

2:53:17 This is not a thing that we make a lot of margin on. We just are excited about. More people sporting the ACQ around. So Participate in the movement. Show it to your friends. It's not our superpod, but you know. Yeah.

2:53:29 Mm-hmm. The pod is the superpod. Uh, you become an acquired LP. You can uh come closer to the kitchen and help us pick an episode once a season, and we'll do a Zoom call every other month or so. Acquire.fm slash LP. Check out ACQ2 for more acquired content in any podcast player and come talk about this in the Slack, acquired.fm slash slack.

2:53:51 Listeners. We'll see you next time. See you next time. Who got the truth? Is it you, is it you, is it you who got the truth now

What you see above is a preview of the first minutes. One unlock costs 10 credits and covers this episode forever: full segment and word-level timestamps on this page, plus .txt, .srt, .vtt and word-level JSON downloads, as many times as you like.