NVIDIA's Pivot from Chipmaker to Financier

0:07Aaron: Good morning, good evening, wherever you are, and welcome back to the Enterprise AI Show. This is your host, Aaron, and I'm recording this from my massive remote studio down at the beach here in North Carolina. It's August. It's hot. There's not much more to say about that. We hope everyone out there is surviving as we go into fall. One quick note, for those of you wondering, we have a bunch of interviews coming up soon and we'll be getting back to normal programming format shortly. But moving on to today's show, you get all of us, Brian, Brandon, and me. And today we discuss NVIDIA's recent role in financing more and more data centers and data center buildouts.

0:44Aaron: And all of that is coming up right after this quick break. Today's show is sponsored by Nasuni. There's a growing gap in AI right now between what's possible in theory and what successfully works at scale inside an enterprise. The difference comes down to unstructured file data. Many AI initiatives struggle because the file data they depend on is scattered, unstructured, and disconnected from where and how work actually happens. Nasuni changes that. It brings your unstructured file data into a single, secure foundation so AI, both generative and agentic, can access it with the context, governance, and performance it needs in production.

1:21Bring AI to where your unstructured data lives. See what it takes to activate your data for AI and request a demo at nasuni.com slash AI.

1:31Aaron: And we are back and we actually have everyone. We've got Brian, we've got Brandon, and we've got Aaron this week. And we actually want to do another quick talking or quick topic. And actually, this one kind of came out in the news this week and I thought was super interesting. And Brian and Brandon and I were kind of kicking this around. But we want to talk a little bit about, I guess you could say, GPUs as an asset class, what's going on with NVIDIA. And there's all this news about going to kind of being a financier, if you will, going forward. And so we wanted to really break that down a little bit.

2:09Aaron: What does that mean for enterprise AI? What does that mean also just at a technology level as well as a business level? And so with that introduction out of the way, I'm actually going to kick it over to Brian to kind of start this off here. Brian, you had some kind of interesting things and some thoughts on this. So what's your initial thoughts when you read all of this about this? And by the way, too, it's the NVIDIA$105 billion backing of OpenAI's Ohio Data Center is what we're talking about. Link will be in the show notes. Go ahead, Mike. I think the basics of this is NVIDIA has one, and we've talked about this a million times, right?

2:44They have one sort of focus and purpose in life. Forget about all of their five-tier stuff and whatever it is. If we do not have the biggest share price in the world, which will drive people buying our chips for forever, what's the purpose of what we're doing? And what they're kind of realizing is at some point, people have burned through their VC money. They've burned through their free cash flow money. There's only so much debt that they can take on for these companies, especially the startup ones. And so they're looking, they are the open AIs, the anthropics of the world, the neoclouds of the world are like, hey, we would love to keep buying your chips because people seem to enjoy using AI for whatever purpose they do.

3:21We just can't afford it anymore. And so NVIDIA goes, okay, let's get creative here. And this is where it gets really interesting because two things have happened. One is NVIDIA has gone from the, I shouldn't even say NVIDIA. Jensen Wong has gone from this idea that if you are not buying the latest and greatest chips, if you are not within one day of buying the next Grace Hopper, Blackwell, whatever's coming next, if you're not buying that, you are close to going out of business. You are close to not having the biggest model on the biggest chips on the baddest data center in the world. This is what we heard from him for two and a half years, right?

3:58If you are a day behind, your business is completely screwed. And so the implication of that was these chips only have a certain amount of useful life cycle, even though they might continue to run for a long period of time. They only have a realistic, useful life cycle because the people that buy the next generation, if they're a day ahead of you with a slightly better model ahead of you, you're completely out of, you're useless. And so we heard this drumbeat for two and a half, three years or whatever it was. And now Jensen has flipped it over because the people that were buying those things are saying, I don't have enough money to do that anymore.

4:31I can't figure this out. And so what Jensen has done is he's gone, I'm going to switch from being the Pied Piper of technology to being the Pied Piper of financing. And he's convinced a whole bunch of financing companies, private equity, banks, whoever they might be, that GPUs don't really have this super short lifecycle of which they are going to be incredibly valuable. They have a almost infinite lifecycle, 10 plus years of how long they can continue to make money for me. And if they have a very long lifecycle of how long they'll make money for you, then we no longer should really think of them as a technology like leading edge thing.

5:09It's all about innovation. This should be about something that is long running and stable, like a mortgage or something that generates cashflow for long periods of time, like a bond or something. And therefore the risk of taking on the technology should no longer be something that NVIDIA is burdened with or their customers are burdened with. We should push this back onto these financial organizations and the financial organizations can then treat GPUs as a commodity, like I said, like a bond or an annuity that will forever throw off cashflow and they can go ahead and do all of their financing magic.

5:43And Jensen has essentially, he's done a 180 from you have to be one day ahead or you're useless to 10, 12 year horizons for AI and GPUs is perfectly fine. As long as it creates a mechanism in which folks can keep buying my GPUs, regardless of whatever technology story I have. So it's been interesting to watch him do that. I guess it's a survival technique, but it is an interesting shift from AI is at the center of innovation. And if you're not at the leading edge of it, you're essentially dead to folks. This is just stable stuff. This is just annuities and pension funds, and you folks should be buying all this stuff up.

6:20It is exactly the same stuff that we heard in the big short in terms of let's take a bunch of stuff that may be volatile, make it seem really stable, and then let's package it up as a financial instrument. It's a very interesting kind of flip that he's made over the last six months for what's going on. Brandon, what's your thoughts? I got some thoughts, but go ahead. You next. Yeah, I think Brian's made the case there pretty well. I think I kind of look at it maybe a little bit different. It's like it's a point in time thing. The one thing I think is true right now, most of the companies out there can't get enough GPUs, right?

6:52That's what's made, if you will, the older GPUs, giving them the longer life cycle that Brian was alluding to. Maybe it's gone from three to four years to 10 years, which makes sense. I get that. And I think he's also in a place where he's already sold everything. He's sold everything he can sell for the next probably two years. So I think that has maybe potentially given us a weird cycle, whereas at this very moment, there probably needs to be a way that people can buy more of this, use more of this, especially as the demand increases. As long as you believe that Anthropic and OpenAI and others are continuing to have this demand crunch, there's going to be a need to buy more.

7:29Right. So I get that. And that part makes sense to me. I think where Brian's getting at, though, is and I think Ben Thompson's made this point as well. It's like smart people and markets, if you will, market behavior is such that like this this hole will get filled. Right. It may take us three years, but at some point everyone's going to show up with what Brian was talking about, with like the latest GPU. right and when all the latest gpus are available and they run at less power they give you better performance then you're out of potential i think to exactly what brian is saying then you're going to be at a point where it's like all this debt on this old these old things that are worthless right it's almost like a car that's broken is potentially going to be a problem so i don't know this is one of these things it's like ultimately i guess the market and jensen's like you're trying to thread this needle of can we use this financing for the next three to five years to manage this demand and And then, if you will, not magically, and then transition perfectly to the new GPUs that maybe even NVIDIA doesn't make.

8:27Maybe Google's making them or Amazon at that point. And the financing is needed. Does it all go well? History would show it doesn't normally go well. So I think there's probably plenty of warning signs out there. So I don't know if that's what I think is going on. What do you think, Aaron?

8:42Aaron: So I think two big things. Number one, okay, GPUs as a product. the model had to change because okay you go back to nvidia origins when they were making 3d graphics cards for gamers right basically the top end was something you could fit in a pc the life cycle was a couple years you bought another one you moved on we're to the point now where i like the analogy of thinking like gpus like airplanes now they're so expensive that you have to keep them running all the time they have to be depreciated off over a number of years and you can't think of especially the high end, you can't think of it the way we traditionally thought of, hey, it's just a chip or it's just something else.

9:22Aaron: It is something that is a major asset. It's probably one of the most expensive assets. I like the comparable to like it's an airline in the airline industry. You got to keep those planes flying and you got to keep those planes full. And if the product model doesn't fit, you need to come up with a new product model. And the biggest thing we have right now is yeah okay great there's the airplanes but there's where this is where the analogy breaks down the data centers that you have to build on new data centers at this point because the existing data centers you can't get enough power you can't get enough liquid cooling into them all these other things so if you got to build an entire new data center because the plane's so big and so expensive you've just taken something super big and expensive and made it even bigger and more expensive because you got to build the data center to go around it now.

10:12Aaron: So this almost seems like a natural progression. And it also is the single biggest limiting factor in our industry. Or profits. Exactly. Exactly. And so this is the difference between this and the plane analogy. Like yes, plane analogy, you know, while we've seen, we've seen airlines go bankrupt because they do dumb things or whatever they do, or they don't hedge oil prices changing or whatever. They do find ways to become a profitable business. It's probably not the best analogy because airlines tend to run, they almost look like a grocery store at the level of profits they make. But I think what we're seeing here and what we've done to ourselves over the last, I don't know, several years is we've continued to base everything on when more chips show up, things will get better.

10:55When more data centers show up, things will get better. And people just ignore the, but will it be profitable, right? We look at, oh, okay, Anthropic has this much revenue and now they're starting to get into sort of the bullshit phase I'm not really going to tell you what my revenues are because maybe I'm an IPO. Maybe I'll tell you, maybe I won't. But none of these guys are profitable. And we've gone from, okay, this is being funded. For example, this is being funded. Oh, this is funded out of the hyperscalers, just their cashflow. This is not a big deal. They're just funding it out of their own money.

11:24They're not even going to VCs or the debt market. And then it became like, oh, we're spent all of our free cashflow. And now it's becoming, oh, we're taking on debt ourselves. It's nowhere in any of these equations is anybody highlighting profitability. And we're now, we're now touching like trillions of dollars of investment. And again, you go back to the companies that everybody kind of models themselves after, whether it was Microsoft with software or Google with ads or Apple with smartphones and stuff like they became profitable in a very reasonable amount of time before any of this stuff happened.

11:57And so we're very much seeing the shenanigans that people do when they can't find profitability and they're just going to keep throwing money at this thing until I think what they're ultimately doing is none of them are doing these things because it's a smart thing to do as an individual basis. I think they're all doing it based on who's going to be the last one around when a whole bunch of these start collapsing. And then I can start to buy the next sort of tranche of this stuff on the cheap. Because right now there is tons of money going into this thing. Nobody's profitable on it. And they keep throwing it at it in the same way.

12:32And we haven't seen anything break yet. And I think at the point when something breaks, then we'll figure out if any of these approaches are any good. But right now, all these things are looking very zitron-y in that if you look at them rationally in sort of a straight line, you're like, that doesn't make sense. You're throwing money at it. You're not making it. You're not becoming profitable. And then they hide it behind this idea of, these are smart people. These are bankers. These are computers. These are smart people. They would make smart, rational decisions. And you're like, no, it doesn't.

12:59That's not how it works in any of these bubbles. There is no rationality that happens at some point. They're chasing, I'm going to be the only one left at the end of it. And we just haven't seen anybody crash yet. And I think it's just a matter of time before some big piece of this falls out, whether it's debt that doesn't come due or one of these frontier companies goes belly up or something like that. Then we figure out how bad of these decisions they are or how smart they were to get ahead of this new unique financing. But I think until that happened, either somebody shows a profit or somebody crashes out, we're just chasing funny numbers.

13:33Yeah, no, I'm just reminded as Brian was talking there is that we can't go a show without a sports metaphor. I say winning is a great cure for culture. It's like when everyone's winning, the culture of your team is it's fine. Oh, they don't get along or they're just playing hard, right? And things like that. And when you start losing, though, that's when you start to see like real conflict. And I think that's where we are today. It's like as long as demand is this high, right? As long as it continues to grow, like this, this is going to keep going, right? No one's going to stop, right? Because it's just like the machine needs to be fed.

14:00If you're really on the bear case, if you feel like at some point, the people that are buying Anthropic and OpenAI and paying the big subscriptions, like the nine figure subscriptions, if they suddenly decide this is not worth it anymore, right? Then suddenly another famous quote, tides come in and tides come out. I think it's like a Buffett quote. It's like when the tides go out, like everybody, you see who's naked, right? You're going to know like who actually has money. So I don't know. I get it though. I am more sympathetic maybe to if I was running these companies and then you feel like Jensen, Wong, because I think he's all in mentally.

14:33He's like, this is life choice. I think it is true, but I think he's even, I think it isn't even just a, I think for him, it's more like, it's not even so much. He's even intellectually questioning it. He's just bought everything and he's just like, this is going to change the world. I've got to do it. And when you're at that point, it's just, we're all going for the ride. And I guess all of our 401ks are going for the ride as well.

14:53Aaron: And it actually goes back to Brian, the kind of the origins of this podcast. One of the unofficial mottos has always been following the money, right? And trying to figure out how and where things are going to get profitable, all the going back, all the technology waves we've covered. And I think one of the biggest things here is, and we've said it once or twice now, the maths don't math and they're still not math and, and we're still trying to figure that out. Go ahead. Yeah. This is, this is the stuff that's interesting. So when we all watch the big short or margin call or any of those things, and you look back on it afterwards, you were like, oh, okay, this is, that's how they built stuff.

15:30That's how they levered stuff. Okay. I see why that sort of happened and so on and so forth. But like nobody was really talking about it. And maybe it was because again, it was the earliest days of the internet and social media. So people didn't cover that beat the way they do today. But you look at it now and you just walk, if we're following the money, it was like, it started with VC money, which is essentially free money because if you lose it all, it doesn't cost you anything. Then it became, I'm going to spend my free cashflow money. And then it became, I'm going to do a bunch of debt stuff and I'm going to do it off the book on these sort of, what do they call them?

16:01SRV special, basically the, I created a shell company to put my debt over there. And now they're like, okay, I've loaded that up enough. And now I'm doing my own debt on this stuff. We've been through all the money. We've been through at least four or five tranches of the money. But again, we're not profitable necessarily in this stuff. So it's going to be very interesting to see where this stuff goes. I see why they're doing it. I understand Brandon's argument is essentially like, Jensen's all in this because he's a true believer. He's also a true believer because if he's not a true believer, he's going like they're They're in trouble.

16:37He doesn't have any choice, right? He doesn't have an ability to walk away from the table because he is not only the player at the table with the biggest stack, he's also the table, right? Become both the casino and the biggest player. And so he can't walk away from this thing because he's got to feed himself as well as feeding the industry on this thing. So it might work out. This might be, he's playing the ultimate game of, like we talked about during COVID, like Zoom was worth infinitely more than they should have been and they spent none of their chips. He's at least spending all of his chips because he's, if this doesn't work, I have no chips.

17:10So, you know, he's playing the only game he can play. It's just, it's telling that he's now had to get to this point that like literally nobody can buy his chips. They can't afford to buy his chips anymore. They don't have any money to buy his chips. And so he's okay. I'm going to, I'm going to create a different way for you guys to have the thing that will call money that you can therefore buy my chips, which I'm buying myself. It's an interesting follow the money, but it does look and smell like other things we've seen where we go, oh, this didn't end well. Again, it's like you can do whatever analysis you want of it, but it looks very much like other stuff.

17:44And what's really interesting is people like to make the analogy where they're like, at least with the internet, you got fiber. At the end of the day, Google got to buy up all the old fiber that nobody used. And people were like, when they laid the railroads, yes, the railroads all went bankrupt, but the tracks were laid. And he's now trying to make the argument because people have said forever, that doesn't work for GPUs because they've only got a three, four year life cycle. He's now trying to create the perception that not only can you trade on these things, but if this stuff goes belly up for some number of these things, these are like the fiber, right?

18:16Like they're going to have a very long life cycle and maybe they really will, but it's a very different perspective than what he had two years ago. and what are we really interesting maybe this is entirely is if that's the world that we live in which it's like hey we shouldn't be chasing the latest and greatest because the the financial model for gpus is this thing should last 10 12 15 years in terms of like getting profits out of them that sort of has a ripple effects to like why do we keep building frontier models right because the frontier models are chasing biggest and jensen is essentially saying that's not going to be profitable for you guys.

18:53I'm showing you what profit, where, how you're going to get the profitability. And I don't know what business model you're in anymore. So it's an interesting sort of second ripple that goes along with him trying to figure out how to get himself paid. He's essentially told the frontier folks, like your business model doesn't work anymore. It's not worth it to keep training the biggest models all the time. I'm kind of wondering too, like the, the, I'm thinking of like the, what's a, what's leftover analogy like the train tracks or the fiber is the idea behind it too okay we're going

19:26Aaron: to build all these data centers and the data center is what's left behind because again there's you're going to need somewhere to host all of this stuff and i've never seen the power needs go down and so there is almost like this data center build out that's going to have to happen and even if stuff does crash there's a data center build out that maybe is the i think what people say is we'll have the power will be lost. The power generation is always used. So that could be it. But I guess I'll just take, I'll take the bold case here. I'll just be more optimistic and say, if you were to be super optimistic would be something along the lines of, we just need to fund these models, the frontier models.

20:03So they get to some point where we actually start to see these incredible societal benefits that everyone keeps talking about, right cure for cancer the list of things go on and on right that could be done so if you wanted to say like a bet maybe that's a different bet but a similar bet that jen is making is like listen we i'm just i'm trying to get us to the point where we're going to have the great models that can do all of the things like and provide all the things that we talk about right that the potential of this and if we did reach that point that the businesses and things that we create will create so much value that it will pay for all this debt 10 times over, right?

20:42I think there is a, and I think that's a very optimistic view of it. But I think you can make that case, right? You can say it's like, we're close. And this is the hard thing about all of this because we're living through it. Like right now, I think it's incredible. Like it's amazing. It can write code. It can do all these things. And I use it every day and I'm totally blown away with it. But at the same time, I always say if you were like in charge of the AICMO, like we need a big win. We need a win that all of society points to and says, this is incredible, right? That they're just like the thing that we wanted to happen finally happened.

21:12And everybody in society can easily see the benefits of it. Because today I think people just see like data center and they're like, that's bad. I don't want a data center in my neighborhood. I don't want this in there. And so that's like the underlying bet. And I guess you could be, that's maybe the Dario, the singularity guys that are out there. They're making that case. They're trying to take a page out of the Elon playbook, which is I'm going to set a goal that's so big that you can't really fathom it. But I'm going to tell you it's right around the corner. And then once the corner, we start getting towards the corner, they go, it is.

21:41But there's a couple more corners to get to. But we've made a lot of progress. We're close to flying cars and we're close to data centers on Mars and data centers floating. That's the other game that they're playing is they're like, we have to be right on the edge of also being a meme stock because that's the only way that you can value the expectations is something that you just can't measure. which is it's fine no i get it i think that game has proven itself to be wildly successful at least for one person and yeah it can i think you're right i think sometimes it works out and people sometimes say valuation is what today's earnings plus a story so if you can tell a great story and people believe you for a long period of time like bitcoin's another one it's just like it just sits out there forever and forever and people like that story so we'll see i don't know are we living through is it just a story are we actually on the verge of some breakthrough i don't know listen to the enterprise ai show we'll just tell you we'll tell you when it's all going to happen one way or the other.

22:32That's right.

22:33Aaron: All right, man, we talked about a lot of stuff. You want to wrap it up? Yeah, absolutely. So everyone out there, thank you very much for listening. We certainly appreciate it. And also, by the way, wherever you get your podcasts, if you have the ability to leave a review, we'd certainly would love a review. And also, as always, reach out to us if you have any ideas for guests or any feedback as well. And on behalf of Brandon, Brian, and myself, thank you everyone for listening. And we'll talk to everyone next week. Thanks for listening. Check us out at the enterprise AI show.com for past shows, newsletters, and all things enterprise AI.