Welcome everyone to the Information TI. My name is Akos Pasicha. It is Thursday, September 17th. Today on the show, we're going to be talking about what the Fed raising rates means for the AI sector. We're then going to hear about Bridgewater co-CIO Greg Jensen's view on AI regulation. My colleague Dacon Campbell spoke with him and we're going to unpack that conversation. We're also bringing on Menlo Ventures to discuss some new data they've collected on consumer AI. We're going to close out the show with a conversation about the Information's newly released list of the 50 most influential people in the tech sector. It's going to be a great show. So, let's get right on into it. The Fed raised interest rates yesterday. It has been 3 years since that has happened. Looks like there will be another hike later on this year that will certainly impact the AI sector as debt has become so crucial to the buildout. I want to bring on Paul Meeks, head of technology research at Freedom Capital Markets for a conversation. Paul, welcome back to the show. It's great to have you here.
Okay, so first in three years, more to come. We've got this AI buildout going on in the background. How do you think that this is going to affect the AI story?
It's a clear negative, but dampened somewhat by the fact that everybody sees it coming. Because you can take a look at Fed funds futures contracts and know with what percentage probability, which is awfully high, when rate hikes are coming or when rates fall on the other end of the cycle. But it's going to really hurt things in two ways. The first way is the obvious problem which is all these AI infrastructure builders are aggressively borrowing money because they have to fund these huge builds. You know, cumulatively by the end of the day when we get through all this AI infrastructure building, we're going to cumulatively have spent several trillion dollars. And so most of these companies, the Neo clouds, the AI collocation companies, they are subprime credit rated. And so this is not good. But there's going to be other ways around that. Now, the offset is...
What are the ways around that?
Well, the key is this. The offset is that recently because demand is so voracious and there's still not enough supply and actually the gap is getting worse and worse, they are getting much better pricing on their deals. So much higher, two or 3x what they were getting a year or two ago revenue per megawatt. So their P&L should improve so they can pay off higher interest expense. So that's the offset. Nothing really mechanical. The other way that we get dinged is, you know, all these companies have cash flows that are estimated so far out in the future that just on a valuation basis for an equity analyst like me, rising interest rates equals rising discount rate for these cash flows really hammers the stock valuations. So one through higher borrowing costs, two just damps the valuations with a higher discount rate.
So how do you think the AI companies change their strategy in an environment like this? I mean pricing is one thing. I hear you. Do they temper their ambitions at all? I'm thinking about the CFOs of all... I mean and by the way we should say it's not like they haven't modeled this stuff in. I mean, like you said, this is stuff they've been expecting, but I wonder how you might see the dialogue from these companies change on earnings calls, guidance. Talking about the public companies, right? We'll talk about the private companies in a second here. What are you expecting there?
So, one of the things that they've even started to emphasize in this last round of quarterly calls is the inflection in the adjusted EBITDA margin or even EBIT margin for some companies where they're starting to see as you turn on the capacity. Usually there's a couple weeks between what they call active power and revenue generation that the inflection in the margins are so strong and actually shows you a business that's attractive to finance. And so they really have to show hey it's not just bringing on contracted capacity and over time converting that to revenues. Obviously, that's really important, but the fact that as they scale these businesses, they get to adjusted EBITDA margins 60, 70, 80% in some cases, which is people turn around and say, 'Wow, I'm cool with financing this business because it looks like it's going to be a good one.'
So, you're talking about like what, Oracle?
We haven't seen that uptick yet, but we're starting to see it. The inflection in the margins and...
So, are you talking about sort of, you know, Oracle's earnings? I mean they talked a lot about the ways that they were moderating their costs and keeping margins somewhat afloat with the prepayments and the...
Yeah, I'm talking about the Neoclouds and the AI colos, of which for example Oracle and the rest of the hyperscalers would be customers. Out of the five major hyperscalers including Oracle, Oracle seems to be the only one that's, you know, too far out on its skis with, you know, burning through its free cash flow. I mean, the others, you think about it, and this is different than those who financed the internet bubble back in the late 90s, early 2000s. Those were a lot of really crappy companies that should not have gone public. But the 90% to date of the AI infrastructure building has been with four or five of these hyperscalers that have very strong balance sheets, all investment grade. Even if they bleed some free cash flow over time people should be very comfortable with their balance sheets and their financial wherewithal.
And what are you expecting for capex then going looking ahead? The growth rates that we've seen, do you expect them to continue at that or do you expect them to temper?
I expect them to temper a little bit but I expect capex from the hyperscalers and now we're getting into enterprises also doing these deals directly not through the hyperscalers, I expect it to go fast and furious at least through '28, maybe through the end of the decade. And you can't expect, you know, 60, 70, 80% increases per annum but I don't think we're going to be shrinking either. I think we have robust growth for several years and I used to say '28, now I'm comfortable saying end of decade.
Right, and the rate environment won't affect that?
Well, it depends. If we have two rate increases here in this cycle and then done, if we end up having a slew of them, of course part of it has to do with the price of oil which drives inflation and of course it has to do with, you know, what's happening with the US and Iran. But as of now I am a little bit worried but not too worried about two increases in rates with one already behind us this week.
Right, let me ask a question about the private companies, the labs. I mean one thought that I was having this morning was if you're a public company and you're investment grade, you know, you can show to investors that hey here's what our financials are looking like. You know that's the premise on which they can choose to take your debt or not. Do they have an advantage at all over the private companies right now in so far as how they can fundraise in this new rate environment or is it still the same story? I'm basically asking are the private companies at a disadvantage here in an environment of rates going up.
I would say that the smaller incognito private companies maybe, but when you think about some of the Goliaths, you know, these $2 trillion market cap aspiring companies, you know, SpaceX, Anthropic, OpenAI, they are so large and they are so carefully scrutinized and if they like it or not, maybe it's not a press release, maybe it's a leak, we know their financials because they're essentially in the public domain. So I think for those guys and those are the ones that really move the needle. Don't think it's necessarily a disadvantage and in some aspects of their business might be attributes compared to public companies because when you're a public company every 90 days you have to come to the altar, you know, kind of suck.
Yeah. What about like the venture capital sector? What do you think this means for, you know, and you're in the business of investing in bigger companies but what do you think happens to that landscape?
Well, I think the AI infrastructure build and hopefully followed by AI monetization will be a strong and durable theme and that should really trump maybe no pun or pun intended the push back that you might get from higher rates. Now remember, my view is rates will be up another 25 basis points before Christmas and then hopefully we're steady state. All bets would be off if we continue this cycle pretty far into 2027.
Paul, can I ask you a question about a different headline? We saw last night that OpenAI disclosed six new concerning incidents with their models. This is, you know, the whole question of rogue agents continuing to reverberate through the sector. What's your reaction to this? I mean, certainly as an investor, how are you viewing this?
I think a lot of it is marketing spin, positioning ahead of IPOs. You know, for a while, Dario over at Anthropic wanted to be seen as the AI good guy. Remember months ago they pushed back on the Department of Defense saying hey you're not going to be able to use our technology for surveilling Americans or anybody else for that matter and then that was juxtaposed with Sam Altman over at OpenAI and he was the bad guy, you know, he went to court and got really messy with Elon Musk. And so I will tell you that I watch it, of course I'm concerned about it, of course I think we need to have at least some sort of guardrails on AI development not just domestically but globally. But part of it is the race between Anthropic and OpenAI to go public. Anthropic positioning itself as the good guy. OpenAI not happy that they're being perceived as the bad guy. And so they make some of these mea culpas and try to get ahead of the bad PR. So I think part of it is smoke and mirrors and IPO positioning.
Right. Well, Paul, I want to thank you for coming on. That is Paul Meeks, head of technology research at Freedom Capital Markets here on TI TV. Many people are calling out for and against regulation in AI. Our colleague Dacon Campbell spoke with Greg Jensen, co-chief investment officer at Bridgewater Associates, one of the world's largest hedge funds, about what he thought. I want to bring on Dacon to talk about that conversation. Dacon, welcome to the show. It's great to have you back.
Thanks. I love being here.
It was a great interview. I encourage everyone to read the full transcript on our website. But I do want to talk to you a little bit about your takeaways from it. Set the stage for us a little bit. I mean, Greg Jensen, for those people who don't follow the finance world as closely, who is he and why did you ultimately want to talk to him about AI?
Sure. So, as you said in your lead-in, Greg Jensen is the co-chief investment officer of Bridgewater Associates. It's one of the biggest and most famous hedge funds in the world. Greg has been there for 30 years now. So he knows, lives and breathes Bridgewater. He's also been thinking about AI since at least 2012, he told me. And you know, he was an early investor in OpenAI and I believe Anthropic paid its payroll the first week it was a company with a check that he wrote to them. So he's not just a lion in the investing landscape but also somebody who's been following AI for a very long time.
And we'll go through the interview step by step here but what was the most surprising thing that he told you in the conversation?
So Greg has written an op-ed in the New York Times, he has given other interviews but I guess I was not prepared and talking about this but I was not prepared for the degree to which he wants the big AI labs and the AI industry to be regulated. You know, Bridgewater as one of the world's most successful hedge funds. The people who live there, who work there are capitalists, I would say, with a capital C. And so it's not often that you sit down with somebody who has benefited so much from capitalism and sort of lived and breathed it to be calling for heavy regulation of an industry. That's just not a common message.
What was he suggesting specifically?
Yeah. So he really wants the AI labs to be regulated. He wants anybody that is controlling a significant amount of compute, which he threw out a number 5%, to be regulated. He wants that to be a cap too. 5% or something like that to be a cap. He made the point that OpenAI and Anthropic in the next couple years are going to control between 35 and 50% of the world's compute resources. He would also like any models that are deployed in the US ecosystem as he put it to be regulated, to be tested and overseen. So he's sort of laid out a pretty extensive regulatory regime that he would like to see.
And was he, so is he drawing a parallel here to, I mean, banks and this idea of systemically important financial institutions? I mean is that sort of the model that he was trying to apply?
Yes, that is another thing I should have mentioned that as well. He does want the big compute owners to be regulated like banks. You know, honestly, we didn't get into it too much, but you know, as viewers may know, the big banks are regulated quite heavily by the Federal Reserve and other regulators because they have so much control of largely of deposits, of consumer deposits. And so he was laying out or suggesting that maybe a similarly large and extensive regulatory mechanism or regime should be directed at the large compute owners. So not just the AI labs but others that have large compute resources.
I'm trying to think about this critically here, how the AI labs themselves would feel about this type of regulation specifically because they've obviously said, you know, we agree that we should pace the frontier, things should go slower, there should be an independent review body to keep things safe. I mean, they certainly have not said anything to the effect of cap our growth or, you know, put a cap on how much compute we have. So my assumption here is just that they would be very against this unless I'm missing any angles here that you thought about after.
No, that's my assumption as well. It did occur to me as I was talking to him what the big labs would think about one of their first investors coming out and publicly calling for them to be regulated in this way. I mean over the years I've talked to lots of bankers, they don't like being regulated as much as they are almost to a single banker. So I can't imagine if you're not regulated like that that you would have any enthusiasm for stepping into that kind of a regime.
So why do you think then he is calling for this? Someone who as you said has benefited tremendously from being a capitalist last 30 years. I mean why is this an inflection point do you think in his mind?
Yeah that's a good question. We didn't get into it a ton, but you know, from all that I can tell and all that I've read of other interviews he's given, he feels strongly about this. He feels AI is at a place where if it is not regulated, it can do real damage. You know, one of the things he told me that I found surprising, one of Bridgewater's models cheated on a test that they gave it. So he has seen his models go or at least a single model go rogue in his own lab environment. So and because he's such a, because he was such an early investor, he's having conversations with folks at the big labs and hearing about their concerns and hearing about sort of the fear that certainly some of them have. So it feels like we are, like he believes that we are at an inflection point. You know, there may be some other alternative motive. I haven't thought about that or sort of played that out on the chessboard.
Well, no, but I mean, look, it's always interesting to see how people's perspectives not necessarily change because we haven't been in this AI moment before, but again, people surprise you in all sorts of ways. So, you know, it's kind of interesting to think about why they think what they do. You know, you talked about one of Bridgewater's models cheating which was surprising. You also talked to him about enterprise data security and Bridgewater using these models and whether or not he has any concerns about the models training on their proprietary data. What was his response to that?
Yeah, I was surprised by that. I asked him, you know, what their infrastructure setup looked like and how they were thinking about their own compute resources and he told me that they rent all of it honestly and that a lot of it is held in the cloud and that they actually do some training on other people's compute and on other people's servers. So, you know, he made the point to me that 15, 20 years ago when AWS was first becoming, first establishing the cloud business, Bridgewater had a long conversation with them about getting around intellectual property and getting comfort with putting their intellectual property into the cloud. So, he didn't seem all that, you know, flexed about it to be honest. You know, on the other hand, we've seen other hedge funds, certainly Jane Street and others, you know, securing their own compute, building their own or renting out their own data centers. So, you know, his willingness to use the clouds and to simply rent his compute I thought was pretty interesting.
Right. Well, it's a great interview. Like I said, I encourage everyone to check it out. He also has some comments on the AI trade itself which I will let listeners read for themselves. That is Dacon Campbell, our AI finance reporter here at the Information. Menlo Ventures released a new report on the state of consumer AI. There are some interesting findings in that study. It was a canvas of about 5,000 people. I want to bring on Amy Wu Martin, a partner at Menlo. She was one of the authors of the report. Amy, welcome to the show. It's great to have you here.
Thanks so much for having me on.
