AI infrastructure spending is fueled by circular financing

Transcript

From the transcript The Diary Of A CEO publishes with the episode.

1. 1:45:45

   Right. So one could argue that a lot of the crypto industry was the same. It's Wang that is inflated by the media. The difference is, the reason the metaverse and NFTs didn't... escape this was there weren't stocks to speculate on. There weren't big companies that you could invest in. They had record earnings in 2021. There's a bunch of money floating in the system thanks to post-COVID, the PRDC, that basically government, federal money flowed in to the banks.

2. 1:46:09

   There's a bunch of easy money, zero interest for a year. Money was easy to find. Then after that, there was the hangover. Growth started to slow down dramatically. This is actually my rock-com bubble theory, which is they don't have any hyper-growth ideas anymore. So suddenly they started buying GPUs. And when they bought GPUs, people went... They're doing AI. Oh, we better buy the stock. And the stock's won an incredible run.

3. 1:46:30

   Meta's like several hundred percent. In the last few years, their stock has grown by hundreds of percent, despite zero proof. And because the media was just saying, yeah, Meta's revenue is growing because of AI, right? Microsoft's revenue is growing because of AI, right? The fugazi you're talking about was the fact that everyone just gave them credit in advance. And now we're kind of getting to the point where it's like, hey...

4. 1:46:52

   You didn't spend that trillion dollars for no reason, did you, Satya? Amy Hood, just going to take him out back, send him to the glue factory or something? I do think there's overspending. I want to concede that. Dramatic. Yeah, no, I do think there is. And I think the reason why there's overspending, Ed, is I think there is something here. And in terms of like, I think there is practical uses for this technology.

5. 1:47:16

   And I think when people realize that through history, they go crazy because they want to be the person that owns the opportunity. I'm going to be honest. I just, I fundamentally don't agree. You don't agree with which part. I don't agree that the speculation is a result of actual demand. I don't believe it's suspect. I don't think private credit is sinking hundreds of billions of dollars into AI because of actual demand.

6. 1:47:35

   They are doing it because they saw the biggest companies in the world building data centers, making a ton of money from two companies they feed money, and went, I want some of that money. I'm saying that I do think there is value in the underlying technology. Sure. And so I think... I'm not saying how much value. Right. Okay. I get you meaning that. That's fair. I'm not saying it's proportionate to the investment.

7. 1:47:55

   All I'm saying is that, do you know what it's like? It's like, if I take your example, the Rot Economy essay that you wrote. Yeah. Say that you're on a desert island and then someone says they found a banana tree. Right. And there's 10,000 people on the island. Okay. They are going to stam fucking peed towards where they think the banana tree is. They are going to fucking claw each other to pieces.

8. 1:48:17

   And if your essay here is right, that there was desperation because they hadn't found an innovation in a while, maybe that explains it. Maybe there is a bit of value here. Right. And they're stam fucking peeding and killing each other and making irrational decisions like hungry people would. I actually think then we actually agree. That is actually my point, which is these three companies in matter, their main business lines are running out of growth.

9. 1:48:38

   There's only so much they can grow. And indeed, in the next three and a half years, analysts think that these two bastards, these two, OpenAI and Anthropic, are going to spend over $400 billion on these people alone. Microsoft... Google and Amazon. And the crazy thing is, is that's a large part of their future growth. And if this money isn't spent, their growth slows down. Okay. So your point about a bananas, I actually agree.

10. 1:48:58

    That is the rock-on bubble. It's they don't have a new thing and they're desperate. And indeed, they got rewarded for buying the GPUs. They got, when they bought these goddamn GPUs from NVIDIA, All the markets went rock hard overnight. They loved it. There were stories about how they were sending armored cars with the GPUs to Microsoft to make sure Microsoft got the GPUs. And so everyone saw all that money flowing in, even though they never disclosed AI revenues.

11. 1:49:21

    They saw the expenditures and they went... Well, I want to do what these people are doing. I want to get a little of that money, don't I? I think the area where we have a slight disagreement is that I think the underlying technology has a lot more promise over the long term than you do. So the thing I want to push back on there is... To have progress with AI, just taking it in a vacuum, to have progress for these two companies to keep going and to keep progressing, they need to spend tens of billions of dollars a year on training.

12. 1:49:52

    The only way that that can happen is if these companies and venture capitalists and private credit firms and NVIDIA keep circulating money to them. So the progress... that we've got so far is entirely a result of this circular system. So it means that without- Circular, you talked about VCs there. Venture capitalists who are, by the way- Investors. The majority of the funding that OpenAI got in the last six months came from SoftBank, NVIDIA, and Amazon.

13. 1:50:18

    Okay, yeah. So just the point is, you're talking about progress continuing. Progress in LLMs can only continue as long as the money keeps flowing. I completely agree. Once the money stops flowing, the progress stops. But isn't that most like early... Spotify didn't make money for 20 years. Spotify didn't lose $20.9 billion in one year. They didn't need to raise $217 billion in the space of six months.

14. 1:50:40

    Yeah, and Uber's another example. $33 billion since inception before it became a messy kind of profitable. Amazon Web Services between 2003 and 2015, when it became profitable, $29.7 billion. They'd spent. Yeah, that's the total capital expenditures. And that's not just Amazon Web Services. That's the entire logistics operation. Normalized for inflation. So they all lost money for a long period of time, is the TLDR.

15. 1:51:01

    Yes, but the amount of money they lost is... completely just magnitudes different on a level where these three come... Can I argue then that that's because the potential of intelligence permeates everything, whereas Amazon at the time was like selling books. No. That was bringing retail online. When Amazon Web Services grew, it was... Oh, Amazon Web Services. Amazon Web Services. Amazon Cloud. With Amazon Web Services, the reason I bring that up, I'm going to repeat something, but it's really important.

16. 1:51:30

    2003, it was founded. Mm-hmm. And it was founded mostly because Amazon, as a growing online store, needed hardcore infrastructure. 2006, I think, is when they turned it client-facing. I may be wrong on the dates there. But 2015 was the year it became profitable. The total capital expenditures normalized for inflation were $29.7 billion across that 12-year period. And yeah, it lost money. But if we speak cold economics here...

17. 1:51:56

    Amazon didn't have to go into the... They were unprofitable in a way, but their margins actually started improving because AWS was a very margin-heavy business. It was great. These two, Google, cash flow negative. Amazon, cash flow negative. These businesses, the reason you liked software businesses was they are meant to be... Cash-heavy asset light. These companies, along with Meta, have added more than $700 billion of new property, plants, and equipment, so assets, data centers, GPUs, in the last four years.

18. 1:52:28

    They have gone from being these cash machines to these cash... furnaces you said a second ago this can only continue if investors continue to invest yes and i was saying i think that investors are used to pumping money into things that are burning cash your rebuttal to me sounds like well this is burning more cash than ever and then so i would say well is the opportunity bigger than those other case studies you reference like aws and one would say that the opportunity of intelligence is permeates everything so the TAM the total addressable market is enormous maybe the rebuttal back to me is about open source and all these kinds no no no I actually know what you're going at so what you were describing there is the argument that Satya Nadella or Sam Alton would make the theoretical opportunity of large language models and I could have bought that shit into any 24 from them