Big Technology Podcast - Big Tech’s Insane Hidden AI Spending, Ranking Anthropic vs. OpenAI, AI For Travel Debate

Episode Date: August 21, 2026

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) Big Tech is spending trillions more than it tells us on AI infrastructure 2) The mechanisms of the off-balance-sheet AI buildout 3) What would happen if these projects were on the balance sheet? 4) Can Wall St. actually not figure this out? 5) Will the tech giants pay the money back? 6) Is a soft landing in AI possible at this point if things go poorly? 7) Anthropic's revenue numbers are soaring 8) OpenAI, meanwhile, is in more tumult 9) Why OpenAI is dealing with so many executive departures 10) Startups vs. established companies, and what are the AI labs exactly? 11) Why travel is a good eval for AI --- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. Want a discount for Big Technology on Substack + Discord? Here’s 25% off for the first year: https://www.bigtechnology.com/subscribe?coupon=0843016b Learn more about your ad choices. Visit megaphone.fm/adchoices

Big Tech is spending trillions more in AI than it appears on the surface. Is it a glaring red flag? New open AI and anthropic numbers show one company surging and another sputtering? And should we actually be talking about AI for travel use cases more? That's coming up on a Big Technology Podcast Friday edition featuring Ron John Roy's glorious return right after this. Welcome to Big Technology Podcast Friday edition where we break down the news in our traditional cool, added, and nuanced format. We have a great show for you today. we're going to talk all about the off balance sheet borrowing and spending that big tech is doing to enable the AI buildout and a movement that is potentially putting $3 trillion more towards the AI buildout than it even looks like on the surface. So we're going to get into that. We also have new Anthropic and Open AI revenue numbers. Are the companies healthy? What does it look like as they move towards their IPOs? And then we will close with the debate on whether we should actually talk about travel use cases when it comes.

comes to assessing the performance of AI products. Ronan and I have been at odds on that one. And now I feel strong enough after a couple weeks of travel to be firmly in the pro camp. And we'll see if Ron John sticks to his guns. So Ron John, great to see you again. Welcome back to the show. It's good to see you, man. What do we do here again? It's been so long. I don't know. We look at a bunch of numbers and say, hmm, I think we'll do plenty of that today. Well, I'm excited to be back. Excited to be back. Me too.

I've almost made it back to New York, flown from Indonesia to Dubai and then back to New York on a flight tomorrow. Excited about it. When this IPO season heats up, I think a lot of, and I'm talking specifically about Open AI and Anthropic, I think there's going to be a great amount of scrutiny to whether this AI boom is sustainable. and whether those backing a lot of the growth that we're seeing today are going to be in a position to sustain it. And we're going to talk a lot about that today. And I think the first place that we should really start is this great story by the Wall Street Journal, which talked about something that I think we all knew was going on behind the surface, but actually put the numbers together, and it's

excellent. The story says why big tech's AI spending is $3 trillion higher than it seems. I'll read a little bit of it. Each quarter, big tech companies disclose their massive capital expenditures and artificial intelligence infrastructure from data center to chips, but those figures don't come close to expressing the full extent of future spending to which Google parent alphabet, meta platforms, Oracle, and many others have committed. This is because a huge swath of their coming financial obligations aren't reflected on their balance sheets. Nine, top tech companies had some $3 trillion of off-balance sheet commitments, mostly related to AI.

Those obligations are growing faster than traditional. CapEx, which totaled about 600 billion over the past year, past year they reported, and were triple what the companies owe under the outstanding leases and long-term borrowings. So, Ranjan, we're going to get a little bit into the mechanics of this, but I'd love to hear you just sort of break down what an off-balance sheet commitment is and why this has become something that's so in favor among big tech to fund their AI. infrastructure investments. Well, I think in terms of why it's so in favor, it's clear because it's just, it does not

draw down your cash. Remember, Google, meta, these are cash generating operations. I mean, for their advertising businesses just spit out cash. So the fact that normally if they want to invest, just simply, you know, reaching into that pool of money that you've already accumulated and using that to build a data center seems like a pretty standard thing that you would want to do. But it would obviously be a lot better for you to not have to actually have any of that live on your balance sheet and create a completely external financing vehicle to actually make these investments.

Where it's going to get really interesting. And I mean, there's been... Wait, wait, wait, wait, before you go on, just explain what this off-balance sheet situation is. Because I think it's important to talk about the mechanisms that are involved here. So instead of spending the cash, what do they do? They, you know, you work with Blue Owl is kind of one of the famous ones and the external investor. And they're going to actually set up this financing vehicle. And then they're going to actually raise the money in potentially conjunction with you,

but it's not going to live on your balance sheet, this new asset that you're creating. You're instead going to invest some amount of cash or invidia or all these others. There's all these other very creative ways that they're going to be. using chips as collateral to actually raise this money. But the main thing is, when you're reporting earnings, when you're actually showing investors the state of your business, these data centers do not live there directly. So it's a massive investment that you've managed to creatively push off your balance sheet and move and spread that risk to other people or other pools of capital. And I think, you know, we've, there's been endless talk about what, how this mirror.

a lot of 2007-2008. And I think that central point of not reflecting the risk of this investment you're making or going to be depending on on your kind of traditional financials is how this entire spending spree is taking place. Yeah, so here's one example that the journal points out, I think is worth reading just to flesh out exactly how this works. So it says, Meta's gigantic Hyperion Data Center project in Louisiana, which is the size of about 1,700 football. fields helps explain how obligations wind up off tech companies' balance sheets. Though Meta is the builder, neither Hyperion nor the $27 billion in debt that's financing its construction show up on Meta's balance sheet. Funds managed by the Wall Street firm

Blue Owl Capital, there you go, owns a majority of a joint venture that in turn owns the campus. Ben-Yea investor, which I would say is the perfect named investor for something happening in Louisiana, a holding company that owns the Blue Owl steak, raise the construction financing in a bond sale. Meanwhile, Meta is Hyperion's minority partner and tenant. Its lease payments will provide the cash flows to help make the payments to bondholders. Sorry, Roger, this just kind of, it looks bad. It kind of smells, I have to say.

It's like you were saying, we're starting to see some companies, including Google, for the first time, go out to negative free cash flow, right? which is like very rare for these cash printing businesses. And the others, you have a company like Meta, basically what they're doing is they're building a data center, but the risk or the numbers are being held by these other companies. And meanwhile, Meta is making the promise here. This is the financial commitment. Meta initially agreed to lease Hyperion for a four-year term starting in 2029 with options to renew for up to 20 years.

It guaranteed that it would make bondholders whole if it doesn't stay the entire two decades. The company doesn't think payments under the guarantee are probable, so it hasn't recorded any liability on its balance sheet. I mean, it's crazy. Basically, what the big tech companies are doing is they're making these huge investments. They're setting up these like weird financial, you know, sort of daisy chains to run, you know, to sort of create these things. And they're guaranteeing that they're going to sort of make the, you know, these, it almost feels like cut. out entities, even though they're not really cut out entities, but something close to them. They're making guarantees to make them whole if something goes wrong. So effectively, this is,

this is their spending. And they're just like not listening to get on the balance sheet. So go ahead. Go ahead. It's not their spending. I mean, the money is going to be coming from, again, it's not cash outlays from these companies for the most part. It's some, it's possible. I mean, this is pension fund money that's going into these. It's, private equity money that's being raised from traditional kind of institutional investment pools of capital. So it's not their money. It's everyone else's money along with some kind of guarantee. And typically it's not even cash again in the Nvidia situation. So I think to me the biggest issue of this is it's like the tangled web, the daisy chain element of it is why you have to have

these reports and investment bank firms trying to actually total up what this potential spending is, is that it's not clearly laid out in any way. And again, open AI and anthropic, private companies, we're going to talk about their financials in a bit. We have no real idea. But Google and meta, we're supposed to understand if you're a public market investor and you're buying Google stock or meta stock, you're supposed to know what are the risks and what's the actual state of their business. And this does not give you an accurate reflection of that. And to me, that's the most worrying part of this. And especially in the backdrop of the overall AI story, how's it all going to play out. I mean, this has been bubbling for a little while. I think while

we've been off over the last couple of weeks, it's definitely coming more front and center. I mean, this has lived in Ed Zittron territory for a long time. And to his credit, he's been on this for a while. but it's interesting to now see the investment banks, Morgan Stanley, the Wall Street Journal, everyone finally catching up to the story. Okay, a quick technicality. Then I want to get on a little bit more to the bigger picture aspects of this. So how is this not the big tech spending? Because, yes, it's okay, the pension funds are outlaying the money to build these data centers,

all right, the investors in these capital groups, the private equity. But if meta guarantees that it's going to make the bondholders, whole if it doesn't stay the entire two decades. You know, isn't that guaranteed good, I guess, or is it still risky for, you know, these funds that are putting the money out to build the data centers? Well, they're not guarantee. Again, so this meta-high period one is like the perfect example. So meta is guaranteeing to make bondholders whole if it doesn't stay for the entire two

decades, but everyone is betting if AI demand explodes and if these data centers start printing cash and then they're able to pay bondholders and then some and generate additional cash, then meta's not on the hook for anything. Then meta does not have to actually kind of provide that backstop. So it's binary in the sense of if data centers, if the whole story works, they're not spending, meta's not spending cash. like that so again as ridiculous as that might sound when I say it out loud it's that that's the argument that can be used in a very technicality legal financial structuring creative sort of way

is that if the data center story works then it is not a cash outly because these data centers will actually be so in demand and generating cash and then meta is not on the hook for anything I see. So basically, meta is giving the guarantee that this company will have a big customer and therefore that gives the company the confidence to go ahead. But there's no commitment from the company to be the only supplier to meta. And if demand for data centers goes up, you know, significantly in the next decade, then they could be a big business and metal will just have a stake in it. No, where it gets even more ridiculous is in many of these cases like meta is the customer and the guarantee. For now or forever? Well, no, no, not forever. Not forever. But still, I mean, they're building and investing, they're building Hyperion so they can have access to Hyperion.

Like, they're not building this for the good of humanity, as much as Mark Zuckerberg might try to spin that story sometimes. Like, they're doing this. This is creative financial structuring. It is. Like, I think the question is, is it, is it legal? Is it okay? Yes, I mean, I'm guessing so. Like, there's a lot, I'm sure a lot of lawyers have redlined a lot of documents for this.

Is it okay for the kind of like overall economy as a whole? I don't know. We're going to find out. We're going to see very soon. Actually, not even very soon. The problem with all this stuff is there's a lot of this that should play out over a long period of time as demand increases, as these data centers kind of achieve some kind of maturity in terms of their economics, but also, you know, any kind of fear, and we're going to get into overall government

yields and just the cost of borrowing, I think there's like certainly things in the short term that could cause some issues within this whole structure. Right, and the numbers are huge. I mean, this is, again, from the journal story, meta-syperian lease obligations will remain off balance sheet until it starts paying rent. It said its aggregate initial lease commitment is about 12.3 billion. It disclosed $347 billion in total obligations for leases that haven't kicked in yet as of June. Right.

$347 billion in spending that, well, the company's market cap is in the one and a half range. Let me just make sure that's right. Yeah, $1.39 trillion. And so it's got $3.44. 47 billion total leases that haven't kicked in yet, many of which are not on the, not like reflected in its financial statement. That's insane. I mean, when you say it out loud again, when we're saying these things out loud, it does all sound insane. I think what's happened is this is good that this story is really bubbling up right now because this has just happened

in the background. And also the numbers are so massive that it is hard to comprehend or actually like, you know, you picture some junior investment banking analyst trying to kind of put together a financial model for meta and pulling up Claude for Excel and trying to put these numbers in and like, they don't make sense and they don't reflect any kind of past way that meta has ever operated or any of these companies have ever operated at this kind of scale at all. So this is new. This is definitely new in terms of four big tech, they operate. In the past, remember, these companies were like didn't know what to do with these piles of cash they had and were just buying back stock. And we were all like that was not a great thing for a lot of people. And now they have the cash. But they're also being creative and not even spending it down and actually just coming up with very creative ways to kind of forecast $347 billion in obligations. which is yeah it's absurd okay so i have two dumb questions that i'm going to throw out there

to you and i think this is sort of going to help us illustrate what's going on here um first one is why would they do this like why would the companies um not you know not just take what what you might characterize as a more honest approach to this and say okay well we're just going to spend our cat i mean they don't do they have you know i don't think meta has 347 billion and cash on hand, right? So is it that they don't have the money? Maybe, but like, why go through the extra effort to obscure and sort of play these games where, you know, where, you know, to the point that they are at this point? Well, so looking, meta has $91 billion of cash and liquid investments and marketable securities as well. What is? Google 187 billion. So,

they have large pockets, not even pockets, my God. They have large piles of cash. But again, like, why would you, if, what's the margin call, like, famous line from Jeremy Irons, but it's like, you know, like, if there's a buyer, why would you draw down your own cash? If you can find someone else's money, obviously for your business, you would rather do that than actually spend your own cash. I mean, from a purely tactical way standpoint, to me, there's no reason to not do this from like an individual problem. It's again, it's like local optimization versus global optimization. It's like for that company, of course it makes sense. Okay, but I think, all right, so let me go to this.

This is going into my second question. Isn't part of the reason to do this to obscure it from Wall Street, to obscure that spending from Wall Street so you don't take a hit? So let me throw this scenario at you. They could do this scenario where like the numbers aren't there in public and it takes a Wall Street Journal article to sort of expose some of them and even then it's still opaque. Or they could have just gone and spent their cash and shown Wall Street how much. much they're actually spending versus the typical KAPEX numbers. Like, don't you think the stocks would have taken a significant hit if they would have shown, like, the true amount of spending that they're

doing as opposed to this like obfuscated runaround that we seem to be getting? Of course, but that's why you don't do it. I mean, like, I think like, again, I'm not not trying to be overly cynical here, but yes, that is exactly why you would do things in exactly this way. And again, like, you know, like in a more aggressive regulatory environment, imagine like, I don't know, Lena Kahn still around or whoever else, maybe you will say this is a systemic issue and we are going to require that these companies actually provide full disclosure for any of these kind of obligations. Like maybe we know that's not going to be happening in today's environment. from every purely tactical level, this is logical, unfortunately, but it is. I'm not, again, I'm not trying to say this is right or wrong. I'm just saying what it is.

If you're, yeah, like if you're saying, we are taking on this massive amount of potential, like business risk, of course the stock's going to take a hit. But if you don't have to do that, then why would you? So here's my question to you. These just built. You think Wall Street doesn't know this? I mean, if a Wall Street Journal reporter can like look at the footnotes of SEC filings and make these determinations, clearly like anyone who's running money, you know, for any like respectable outlet would know or wouldn't they? I think you'd be able to give us an answer to that.

I think one of my friends who runs like a technology hedge fund. We've been going back and forth because, like, he's always asking me, you know, like, what's going on in AI? And for the last, like, year and he's more traditional software focused, you know, he's like, he's still desperately clinging to the assumption that markets are rational. These are smart, rational people, especially managing large pools of capital. I have argued the other side. I have said, like, you would think Wall Street, whatever, that. might mean exactly. But like, yes, that there would be smart analysts sitting there saying, how does this affect things? But the mania has been so strong, the phomo of not being in the next

whatever, anthropic or what, NVIDIA on the public side. So, no, I do not think that Wall Street obviously should be on top of this. I think they will be. This is the fact that these deals have been going on for the past 18 months maybe, maybe even longer. And now you're finally getting these research reports from investment banks trying to put together what this looks like. So no, I do not think it's a given that Wall Street is smart and should be ahead of this. Well, I think, you know, maybe it's not a given that Wall Street is functional, but they should be ahead of this. Because if Ed Zitron can do it in the Wall Street Journal can do it, surely Morgan Stanley should be able to do it.

And here's the quote. I always love looking at the quote at the end of these stories because it tells you what the author really thinks, you know, after they make through their like carefully hedged language, like laying out the facts. Here's the quote. As these off balance sheets, as these off balance sheet commitments become frequent, larger and more complex,

it's becoming increasingly difficult for investors to assess companies' total potential leverage. And that comes from none other. than a Morgan Stanley accounting analysis. They're starting to get on it. I mean, even the way you outline that order, Ed Zitron first, then the Wall Street Journal starts to get in on it, like the story.

Like, it's, again, the mania is so strong that it shows you that you have the kind of Ed Zitron persona, the kind of like gadfly truth teller blogging in the corner, like, trying to scream about this story. And again, like, I disagree with him a lot on the idea that, like, generative AI and agentic AI is all, like, you know, not a useful technology. But he's gotten, he's been ahead on a lot of this stuff. And I think, like, to be the one saying that the data center financing doesn't make sense

when the company's investing in it are just their stocks are. skyrocketing, again, if you missed NVIDIA, you do not look good. Like, if you missed Anthropic on the private side and Open AI, like, if you're the one trying to say this 12 to 18 months ago and managing money, you'd be looked at as a failure by your investors. So I think, like, that's, those are the mechanics why it takes so long for these things to come out. Right. And so I think, like, another reason you do this, of course, you don't want the street to, you know, maybe see it and hit your stock. But you may not be so sure you're going to pay it all back.

I think that could be part of it. And ultimately, this is where it becomes a problem is when you pay it all back. Like, if you're sure you can pay it all back, you like, go to your friend and you're like, hey, you know, let me get five bucks. If you're not sure you can pay it back. That's what it is. You're like, can you, like, go to this person and get money? And I'm sure, like, my other friend will pay them back. And you're sure to say, hey, am I going to get my five back or am I not going to get it back?

So this is like one more piece of the story. This from the Wall Street Journal. Again, there are reasons to believe tech companies will make good on all their obligations. Optimity, skyrocketing demand for AI tools. As a proof point, that demand is going to be strong for years and the money will pay all those bills that will be rolling in. For the more anxious set on Wall Street, it's a warring sign that some tech companies that once seem to have a fortress balance sheet have needed to tap the capital markets frequently. I don't know. What side do you want?

What are the sides? How are we defining them? Side one is they're going to clearly pay it back because the AI boom is going to like bring in so much money that they won't know what to do with it. And side two is like, well, these are some of the strongest companies in the world. They had these huge balance sheets of cash. And now they're going to be debt ridden. I don't know if I. So the danger is very clear that these are the largest companies in the world. and they're taking a levered bet.

And they're all taking the same levered bet. So that's not good. Like you never like to hear that, again, where everyone's money is just sitting is all betting on the same thing. Again, and I firmly believe that AI demand will explode and the economy is going to get agentified and all of that's going to happen in the years to come.

but I guess I would lean on the anxious side in terms of like when everyone is taking the same bet, it's always a bit concerning. What about you? No, not just, I'm going to agree with you and not just the same bet. Like you mentioned, a lever bet. And we've seen in the past couple weeks that leverged bets on AI don't always work out. So talk a little bit about the risk here. I mean, it's one thing if it's Leopold. It's another thing if it's like freaking like.

We were on vacation for Leopold. Yes. And Demis, by the way, which we have covered on the show, but you and I haven't had a chance to speak about it yet. But, but yeah, the idea that you're going to, these are big bets. I mean, $347 billion that meta has and obligations for leases that haven't kicked in yet. Like this is, you know, the term bet the company is used way too often, but this is bet the company type activity here. It is. But that's what, so that's again, if we're going to lean on the anxious side. And man, I can't believe we never got to talk about Leopold. But maybe we can start working that in. I think,

no, we did for listeners. Ron John and I did happen like just as Ron John went on break. And we did like exchange a bunch of, bunch of texts being like, we knew there was going to need to be an emergency podcast. So anyway, we'll definitely get a chance to give our Leopold analysis. There's going to be a round two. There's going to be a round two and a three. I'm not worried about that. I think, but when you say that, remember, Mark Zuckerberg has said, I'm not going to remember the exact quote, but it was around like, oh, well, if I were spending this much, we should be spending more. Like every, this is, I honestly wonder what the group chats are like for Big TechC. Are there group chats where it's like Sundar, Mark and everyone? But like, they're, they're all feeding on each other and they're competing on. each other against each other and thinking the same thing. And yes, they are all saying effectively

bet the company. This is the single biggest like moment in their history, the text history, humanities history potentially. So of course you're going to bet the company. I mean, it all goes back to what Paul Kodroski said on the show a week and a half ago, which is effectively that what we're looking at is a call, basically the way that it's basically the way that it works is it's a what all the spending is a call option on AGI. That if you get something that what people are envisioning to be AGI, like a powerful AI that I don't know, maybe does people's jobs and cures cancer, money will spend. You would think if you could hoard it or, you know, maybe. But if not, there's going to be a reckoning here. It seems, it seems like as you read deeper into this

activity, and I agree with you, by the way, that I'm not a skeptic on, you know, let's say the capabilities of AI, at least the way they stand today, they're exceptional. And we're going to talk about this in the second half. How far AI has come in the past few years is insane. But you can say, you can at once say that and also say, by golly. I'm afraid of the numbers that I'm looking at. And it seems, you know, quite concerning when you put it all together. I love that your phrase of exclamation is by golly. that's if there's one thing that i've changed on on break can you can you say that on cnb can you say that on cnbc soon i will i will i will i will attempt it absolutely by golly these numbers are

astounding i think uh they are is there any other way to refer to it no no no by golly is the only way that we can ever refer to this again and and and and as leopold round two and round two and three or whatever else comes next. That's our exclamation. That's what we're both yelling. By golly moment. Okay. Reckin with the substance, Ron John.

Don't lose sight of the substance. I think I forget the substance at this point. The substance is that this is not going in a good direction. Well, okay. So, no, it is. So to me the question, and I don't even say this is highly. I say this as, I don't think this plays out completely smoothly. And in that call option framework, and again, a call option, the right to buy, basically

having unlimited upside for kind of like smaller premium, taking that kind of bet, it's, I don't think it's, the question to me is what does an unwind look like? And I will say like it's it when we say $347 billion bet it's not $347 billion in the sense of if things go wrong. This is that's like a guarantee of lease payments over a large number of years. It's no way that like metas has to pay $347 billion. There's going to be it's going to play out slowly potentially. It's over time. There's lawyers involved.

There's people losing money left. You know, like maybe it's in a very distributed fashion and people notice their pension fund investments are down 10%, which is not good, but like it's not like META's $91 billion of cash evaporates next week. Unlike the Leopold situation of highly levered bets for a hedge fund where you can go from $45 billion to $6 to $10 billion or whatever it is and lose a massive amount of money in minutes or hours. So I think, so that's, to me still, let's say six to 12 months, open AI and Anthropics

IPOs are duds. People stop kind of like having this absolutely manic drive towards the story, the AGI call option. What does a drawdown look like? I think to me that's the more interesting question. Because again, meta is still going to be raking in. advertising dollars, that's not going anywhere. Google between their cloud and consumer businesses, that's not going anywhere either. So to me, I think, like, how does this actually play out? It's going to be, I don't know, and it's very hard to say, but I think that's going to be the

interesting part of this. Oh, definitely. I mean, I think we talked about it on the show numerous times that you need perfect execution, really, by many of the players here for this all to work out. if the execution is not perfect, then that's when things get somewhat interesting and probably very rough for some of the players involved. And we just, we don't exactly know what that looks like. But with the size of the numbers that we're seeing, some extraordinary things have to happen in order for this, you know, it almost reminds me of like the soft landing scenario or the no landing scenario, right? That's the Fed is trying to get inflation down. It's like you need so many things to go right.

That's a good. Or you're going to have a problem. That's a good kind of like corollary. The softly, is there a soft landing? Is there no landing? I think that's like a good way to think about it. And what do those different scenarios look like is, I think, I mean, I think we're going to be talking a lot about that, especially as August 21st and

Anthropic is supposedly coming out to the market next month. I think we're going to start seeing a lot more concrete numbers assigned to this whole story. Yeah. For what it's worth, I just want to state my perspective on this, because I have been thinking about over the past couple weeks. And then we're going to get to the Anthropic and IPO and revenue numbers and Open AI revenue numbers. I'm starting to think there's really only three true outcomes for AI. Like, one is AI works and it's good, right? So it leads to economic prosperity and like a rollback of disease. The other is AI works and it's bad.

So like we start to see more of like the rogue hacking activity that we've seen and it gets in the hands of bad actors and it leads to like bio weapons and things like that. Or the third possible outcome is that AI doesn't work. And I'm I've really struggled like the idea of this soft landing, I struggle to see it. Like if AI doesn't work, there's been so much money tied up into this system. that there's going to be some form of crisis. So I really see it as, and I think I'll write about this on big technology when I get the newsletter back up and running, probably like post-labor day. But I really see it as three true outcomes here.

What do you think about that? I'm going to be your editor and say that I vehemently disagree with your framing of three possible outcomes. Okay, good, good, yeah. Because everything we have been talking about is completely independent of the three scenarios that outlined. It's if it works and it's good or it works and it's bad, actually can be totally independent of will data centers work over the next few years and be able to actually make bondholders whole on their, because it could work over a longer time period. It could work, but then the actual compute requirements because of innovation dramatically decrease and then

actually the data center story fails, but AI is either curing diseases or hacking into all of our bank accounts. So I think there are two very, very separate things, the story that we're talking about and those possible outcomes. I think like independent or separate from the whole financing story, I think, yes, those are three outcomes that kind of cover the entirety of anything. And like, but I also, the curing diseases thing I'm kind of tired of. And I say this only because that feels like such a crutch that Dario and everyone just keeps falling back to just to be like, guys, please like us.

We'll cure disease. We haven't outlined any possible way of like any concrete things that actually are going to get us there. But you should like AI because it's going to cure disease. I strive. Okay. So first of all, we'll add a fourth potential outcome is that AI works, but it works too slowly and it causes financial collapse. There we go. There we go. Now I'm good. That's the fourth outcome. But I'm very, I very much disagree with you about this disease thing. I mean, alpha folds. I know it's been kind of talked about forever. But like some of these personalized health care examples that you're getting, you know, maybe it's not pure. It could be managed. They're very, very difficult to argue. with and I'm personally like, I think that it could be very big in a way that, and the amount of focus

that's going into it. No, no, but. Within the labs, to me, I think it's not worth, you know, it's sort of waving away. No, no, but I firmly believe that AI will bring tremendous progress and innovation around rare disease, disease as a whole. I firmly believe that. I don't think that tells a story about why Anthropics IPO should succeed or you should like Anthropics specifically. Like them, those CEOs using it as kind of a crutch, that's what I have the problem with.

I think like if you're a university professor doing research on how large language models can kind of like parse through mass amounts of unstructured data around rare disease, like that is the future and that's great. is the most beautiful story to tell. If you're Dario and you're worried about like where, or even Sam remember, I feel Open AI has been kind of like talking about this more and more now because like their enterprise story, which we'll get into is not going great. So it's like another one where it's just,

again, it's like moonshottie type stuff when for these very mature businesses. So that's the rant there. Yeah, but you don't think that's going to be economically valuable? I don't think they have shown anything specific to their business today that makes them any more, I don't know, like, they're going to be the ones to do this. I think like startups, some Chinese company, whatever, it's like everyone, it is open season. It's like, which is good. And that's exciting to me.

But that's my rant, which I'll continue for weeks on end, as we had in time. IPO season. Okay, I love it. I love it. All right, we've teased it enough. When we come back from the break, we're going to talk about Anthropic and Open AI's new numbers where it seems like Anthropic is surging and Open AI is, well, we'll get to it. Back right after this. There are a lot of things AI can replace. Teamwork isn't one of them. And covering technology every day I'm seeing firsthand just how quickly the relationship between people and AI is evolving. For me, there's constantly new information to research, interviews to prepare for, and ideas I'm working through with my team. What I like about Notion is that AI becomes part of that process, where we work

from the same context instead of sitting off on our own. Notion is the AI workspace where your team's knowledge, projects, and agents all come together in one place. Teams using Notion move faster, cut friction and cost by consolidating tools, and stay aligned. That shared context means AI can help answer questions, surface information, and handle some of the busy work while the people on the team can focus on the thinking and the collaboration that actually matters. That's the future of AI at work that makes sense to me. Technology strengthening teamwork rather than replacing it. Learn more about how Notion can support your business at notion.com slash big tech. That's all lowercase letters, notion.com slash big tech to try Notion today. And when you use our link,

you're supporting our show. Summer always changes how I get dressed. I want pieces that feel lighter and more comfortable, but still put together. That's where Kintz comes in. They focus on well-made essentials in breathable linen and soft organic cotton, the kind of basics you will keep coming back to. Everything is priced 50 to 80% less than similar brands because they work directly with ethical factories so you're paying for quality, not brand markup. Personally, moving between AC and summer heat, a lightweight, zip-down sweatshirt from Kintz has been a life changer. upgrade your everyday download the kins app for app exclusive offers or go to kins.com slash big tech get free shipping on your order and 365 day returns now available in

canada and the UK too that's q you i nc e dot com slash big tech this episode is brought to you by aft point everyone's racing to roll out AI right now co-pilots chatbotes chatbop agents doing real work. But here's the part nobody loves talking about. All that AI runs on your data, and most teams have no single way to see it, secure it, and prove it's under control. That's exactly what AvePoint does. For 25 years, they've been the trusted layer beneath the world's most demanding data, now extended across your entire AI estate, your data, your cloud, and the agents acting on your behalf. It's how more than 28,000 organizations deploy AI with confidence. So innovation scales without scaling risk.

It's a single platform instead of a pile of tools, bringing security, governance, and resilience altogether. AvPoint, the unifying trust layer for AI. Learn more at AVPT.com slash big technology podcast. That's AVPT.com slash big technology podcast. And we're back here on Big Technology Podcast Friday edition with Ron John Roy of margins. All right. Let's talk a little bit about the numbers that we teased before the break looking at Anthropic and Open AIs revenue that have started to leak. This is from Bloomberg. Anthropics annualized revenue top $65 billion before IPO.

Anthropic is on track to generalize annual revenue with more than $65 billion based on its current performance up more than sevenfold from its pace at the end of last year. The company's run rate hit $65 billion by the end of July. the dramatic acceleration and revenue bolsters. Anthropics plan for a public listing, Anthropic and Open AI filed confidential paperwork to go public with Anthropic expected to make its Wall Street debut as soon as this fall ahead of Open AI. So on the quarter, $11.5 billion

compared with $787 million in the corresponding period in 2025. All right, that's those numbers, by the way. They say a lot. This company is growing, like, exceptionally fast, and we have numbers that look more like real financial statements, quarterly revenue, not just run rate, which I know you hate. Your reaction here to Anthropics growth. I am, this made me even more excited about the day we get to see. You said more in line with real financial statements, but we're not quite there yet.

Because, because, again, okay, two parts of this. One, my God, still, it's like trying to make us backward calculate your like July and August revenue when you're like, well, actually, quarter ending June 30th, we made $11.5 billion, but it's a $65 billion run rate. So what was the revenue acceleration, like trying to backwards calculate that versus just saying here's what we made in April and May and July and here's a graph that shows it's going up into the right? that's what should be happening because again 11.5 billion in a quarter obviously if it's

accelerating but that's still not 65 billion over a one year period like I don't know we're I am excited for real numbers to come out I hope they come out I like there's part of me that's like is this IPO so hyped that they could just be like nope we're only going to give you vague run rate numbers even as a public company somehow and they get away with it but I'm not comfortable with these numbers yet. Yeah, I don't see how they could, how could they ever do that? I'm always curious how the numbers come out in stories like this. We know how the numbers come out.

As part of its regular update with investors, but there's also the prospect that they're going public and the sourcing in the Bloomberg story is, according to people familiar with the matter, which sounds like it might have been delivered in an envelope from Anthropics. I'm getting trouble for speculating, but it's like, Philly inanthropics interest to get the numbers out there to scare opening eye away from trying to go public before them. I don't know.

What do you think? No, no. This is, like, whenever this happens, I, when I was in my past direct-to-consumer e-commerce experience, was part of a potential IPO proceeding. The level of secure, like, if we were to ever even say figures to our wives and families and friends, we would get annihilated by the bankers. Everyone did everything they could to prevent any kind of leaking of any figure to the press. And meanwhile, like, every conveniently at any moment, I mean, we're all just seeing these numbers. Like, this is a very, and again, anthropic sandwich in the park, there are comms masters. Like, all this stuff is orchestrated. I can't, like, it has to be.

And whether it's an investor providing that envelope or whoever else, like, because almost by definition, as this, if this number were truly leaked, leaked, like, there should be held to pay within Anthropic, within the bankers that are actually working on this IPO. But this has happened for a while now. We're like, again, in the past, remember, valuations didn't, weren't attached publicly to a funding round. That was supposed to be actually leverage for investors in a competitive space. And now it's a given that that's going to come out. So, yeah, I actually think, I don't think that's just speculation. I think that's actually an important part of how this story does get reported. Yeah, yeah.

I think it's, you know, it does feel inside baseball a little bit, but it's, I think it's our duty year to sort of talk through sort of how the sausage is made. in situations like this. So I'm glad that we covered it. And now, if basically Anthropic is seemingly on track to IPO this fall, which is not so far away, even though it's warm, you know, in the northern hemisphere now.

It's, we're not too far away for the crispy days of fall. And we could be seeing an anthropic IPO with potentially Dario ringing the bell. And how crazy would that be? Does Dario ring the bell? Does that? feel too pedestrian to ring the bell? I mean, maybe does you, do you sort of, I don't know,

is Claude do it? That seems cliche. He'll ring it. He'll ring it. He'll ring it. You only get to do it once, or twice, three times, but. I don't know. Maybe Claude ringing the bell. I don't know how that, what that's going to look like, robotic arm with an agent attached. That's what I want to see. That's what I want to see. Yeah. That would be seriously a visual for the ages. All right, Let's move over to OpenAI because we also have their second quarter numbers, which just so happened to make it out into the public as well. And OPE is from the Walshry Journal. Open AI told investors its revenue grew by 18 percent from the first quarter to the second quarter while its losses deepened. Results had disappointed some shareholders who had hoped that the startup would show more progress catching up to rival Anthropic.

Sorry, that sounds like the investors leaked it. I don't know. It doesn't seem like Open AI decided that that. If Open AI did put those numbers out, they're probably not happy without was framed. No, exactly. And again, there's reporting that Sarah Fryer in an all hand said that they're not IPOing until 2027. Like, Open AI, okay, we were both off. Denise Dresser leaving as chief revenue officer.

And again, I know this one hits closer to home for me working in Enterprise AI. And I can say that this writer, like everyone was like, Jaws. hitting the floor because eight months into the job, the chief revenue officer who is like as a stellar background in the, she was, I think, Slack CEO under. Yes. Like, I mean, what do you think's going on over there? What do you think is going on? I've been waiting to ask you this for a month now.

So it's not just Dresser, right? So this is from this journal story. Last week, the company replaced its chief revenue officer, Denise Dresser, after she spent last than the year on the job. Her departure followed a string of other exits, including former chief operating officer Brad Lightcap and Fiji Simo, once seen as the heir parent to chief executive Sam Altman. So it is a lot of departures. Fiji, of course, health-related reasons, but Brad Lightcap is a big one. Look, I think this is just the consequences of them, you know, really getting their butt kicked by Anthropic on the way to coding. And, and,

And they've taken a few months to sort of turn their ship and focus it on Codex, which is still in the process of being released. So, you know, they got their first draft out of it before, you know, sometime in July. And now they're going to refine. But this is obviously very powerful and very lucrative form of artificial intelligence. And they are behind the eight ball right now. So when that happens, yeah, revenue tends to grow more slowly than the arrival. I mean, they made $7 billion in the quarter. It's not bad.

Yeah, but that's not. But that sort of, that leads to tumult. But there's tumult and then there's by golly because that's like. We're going to get second by golly in one episode, okay? I think the by golly, three senior executives leaving also not just senior executives, each the one that is supposed to be like, especially Fiji and then Denise, like enterprise is the future. We're going to focus, et cetera, et cetera. These are the people leading those efforts.

And you're not, your business isn't collapsing. I don't know. To me like, and also like if you're to be executing on some giant vision to pivot one of the biggest consumer stories of our lives into enterprise, you would get more than eight months. Like you would, that does not happen in eight months anyways. And again, for her to come in and then even the leaving process, you know it's not, it doesn't happen in a moment. Like it's going to happen over weeks and months.

So like, I don't know, like that, to me, that's the part that is just still kind of crazy. So the timing of all of it. Yeah. Well, I have a theory. on this and I kind of want to run it by you. It's not like anything too juicy, but it is organizational, right? So like most companies, they go through, and I've been a part of fast-growing startups. I know you, yeah, I have at least two of them. I know you have in the past, um, are now. Uh, don't you like in the way that startups work is there's a certain group of

leaders that are there from like, you know, sort of no money to 10 million. Then there's like a certain group that's there from 10 million to a home. hundred in a certain group from 100 to like 500, right? So part of the chaos of working for a startup is constant leadership turnover because there's the people with the idea. Then there's the people that notice sort of take you from like series A to series B. So there's like they start to standardize some of the processes and they start to like really structure your hypergrowth.

And then you bring in like more seasoned executives from bigger companies who like turn your startup into a big company. Right. And so like, you know, it's not uncommon to like, work. work in a startup. One day, the person that hired you, you've been like working underneath for months is gone. And there's like some new person that comes in because they're used to working with big pre-IPO companies. And then you IPO and there's another person. So, but here's no, I, I just want to talk through it. I'm not, not excusing. I just want to explain something. Open AI is, you know, as with

everything in this cycle, open AI is compacting that thing into the shortest amount of time that it ever has. Like even hypergrowth startups when they do that, they take a moment, they grow, they breathe. They take a moment. They grow. They breathe. I mean, just think about the numbers we read from Anthropic. You know, under a billion dollars to 11.5 billion in a quarter and maybe more in a quarter we're currently in. Right. It's crazy. So that's sort of you get, you get this kind of madness. No, no, but I 1,000 percent agree with you having been and currently in a very fast-growing, scaling startup, like scaling startup and that keyword startup, you described it perfectly. These are companies that are seeking $2 trillion market capitalization valuations that are bigger than the vast majority of companies in the world. Like, vast, these are not startups. And like, if they want those valuations, you have to at least, to me, like, the operating

model should be more in the line. least pre-IPO startup, as you said, which they are, rather than that series B to C, C, C to D, kind of like testing the waters. If you want that valuation, you got to show a little bit of maturity. They're asking for $2 trillion. Yes. Okay, here's the thing. They are startups in one really important way.

And that is, like, yes, the numbers are as big as any enterprise company, most, right? but one of the things that defines a startup is you're not fully set on your product. You know, you become an established, like people look at the old question, well, what's a startup and what's not? I mean, part of it is you sort of, when you're not a startup anymore when you have a thing that you sell, right? And it's just kind of the thing. You sort of find that product market fit and you settle into it and that's what you are. Now, of course, you adapt. But the thing with Open AI and Anthropic, by the way, that gives them the nature of a startup is

the technology that they're developing is not mature. It's still growing, the capabilities are growing fast. The offering is growing fast. So you don't have the stability of a traditional big company, even if the numbers look like you do. So this is the kind of point I was going to make. When you bring in somebody like Denise Dresser from Slack, and I'm just this is speculation.

When you bring in somebody like the former CEO of Slack, that person is a good fit for the numbers. But are they a good fit for the K? and the figuring it out part. That is what's going to make these roles so difficult to hire for and so difficult to keep people in because it doesn't really line up the two things that you're looking for. All right. No, no, no, no.

I'll give you that. I'll give you that. That, like, is Enterprise the future? Is a cloud business the future? Is personal devices or wearables or household? Like, especially in the open AI case, wide open. So in that sense. And even if it is, by the way, what is Enterprise?

What are you selling? You know, these are the sort of open questions. And again, that's in my world. Like, is it codex to developers within enterprises, which has been the current story? Is it knowledge work, which they're kind of like chat GPT work and trying to get into? But, okay, that's fair. Like, they are a startup like anyone else in that sense. Yeah.

Anyway, we will see. But I think bottom line from these reports is anthropic. juggernaut to like it is it seems like anthropic is putting some distance between itself and open AI and I don't know I you know it's sort of you know it was clear that Anthropic had a small lead over the past you know basically since the beginning of the year this year but that that gap seems to seems to be growing it's not to say open AI isn't capable of closing it or surpassing Anthropic again but you know at least like when you look at state of play right now Yeah, Anthropic is putting some distance between itself and open.

I think in terms of AI use cases and markets that are mature, I'd like to hear about travel because we're going to be talking about anthropic and open AI for a few months and weeks on end. So I want to hear you're sitting in a Dubai hotel room. I want you to pitch that actually travel is the right story for Agentic. And that's okay. So, okay, so yeah, before we go, we have to talk about this. Ranjan, you and I think we've had like this back and forth relationship on talking about travel as like a worthwhile key use case that tells us something about

AI's capabilities. And, you know, at our summit, for instance, like I wrote up a travel use case and you're like, you know, leave that at the door, man, or something of that nature. And I've been traveling over the past couple weeks. And, you know, and I'm probably going to. write about this again in the newsletter. But I think that like travel when it comes to assessing AI's capabilities is actually an excellent proving ground. All right, I'm going to make the case with three quick points. The first one is when you're traveling, you're you need to use something

to sort through a lot of information, a lot of information, what to do, where to stay, what to eat, how to get there, how to get home, what days, things are open on, what they're, not. The internet, the web, is terrible for these things. And there are terrible incentives to provide the information for you, like all the affiliate links online and all the fake review sites. So I think that this is like something that's kind of made for generative AI, which takes in all the information, adds context, adds little intelligence, and then provides it to you in a succinct way. The second thing is that with travel, what you're dealing with basically is a cascading series of small and big tasks that are always changing in some way. So I think as a, you know,

an eval, so to speak, for AI, it's actually really good because it resembles, you know, some more high stakes type of cases where, uh, you're like maybe at work and you're, you're, you're working a project. Well, what does you have on a project? You have a cascading series of small and big tasks that are always changing in some way. Um, but the best thing about, about the, you know, the travel example, um, is that you can experiment. more at lower cost on a travel on a trip than you can, let's say, at work. And that's a great way to, and if AI aces, you know, your tasks there, you can, you can, you know, looking at it again, as an evaluation, you can be somewhat clear that it's going to, you know, maybe not ace your work

test, but you can feel more comfortable putting it towards more economically valuable tasks. Because travel is in what I'm going to call like the Goldilocks zone of important but not so important tasks. So there's real stakes when you're traveling and you fail at something. Like if you show up and like the hotel that you booked was actually hallucinated by Chachipu T, you know, there's real stakes. But it's not a disastrous stake. Like you don't lose your company. You just kind of have to walk down the block and book a different hotel maybe for a little bit more than you were anticipating to pay. But you can sort of survive that.

That's why I think travel is such a great e-val for AI. And I certainly like, you know, traveling this year compared to. to last summer, was really able to experience, like, the capability increase of, you know, let's just say, like, the latest versions of chat GPT, where I did like almost all the planning and logistics work. What do you think? Are you going to, are you on my side now? Are you with me? I, okay, just coming back from my own extended travels over the past month, I actually am not going to disagree with you. I'm not going to, I like the framing. I like. I like. like the framing travel as the e-vall.

I actually kind of love that, in fact. And honestly, like, I believe there needs to be more actual normy evils as opposed to whatever latest benchmark there is. So the Cantorwitz, TripAdvisor, travel booking, combing, Google flights eval, we should construct. But it's funny, I do, the one thing I'll disagree with is you said if chat GPT hallucinates your hotel, it's not disastrous. Traveling with a wife and child, I would actually disagree because if I did that. But I had my own kind of, so I went to Spain and I was like joking.

It was kind of like an AI generated vacation that basically my wife wanted to go to Barcelona. I've been there a bunch. So like, I was like, we'll do a few days there. But I basically gave AI a bunch of requirements, like really specific around like I want best value for Airbnb in a specific price range with a pool at the Airbnb. I want like snorkeling, but shallow water as my son is only still learning to swim. I want sandy beach.

I gave all this stuff and it came up with Denia, Spain. And we went for four days and not touristy. And a lot of Spain in August is just like Barcelona, we didn't see a single Spanish person other than people working. Like it was, Denya was amazing. It was like one of the best, it had been so long since I'd like actually felt kind of like traveling that, you know, like you discover something and it really is kind of magical. And it was AI got me there. It was literally set of requirements.

And then also during the process as well, creating a markdown file with the context of all your rental car booking numbers, even now having pay. passport information, everything, hopefully secure. But like, I mean, no, no, like, and then being able to quickly ask, oh, wait, what's my rental car number? What's required for a rental at budget and, like, in Spain, all of that. So, so AI got me to Denia, Spain. I don't want to uncover this hidden gem, but to our listeners, you deserve certainly to make it there and would highly recommend it.

Yeah. This was supposed to be a debate, but I'm glad you sort of see it. see it my way on this one. And I had such a similar experience. I mean, for me, you know, this trip, it was a very logistically challenging trip. Like, I think I was on maybe four or five islands in Indonesia over like the span of like 15 days. And then, yeah, coming back through Dubai. And it was just logistically so, so seamless. And I had what I had to do. I planned it in one chat. And before I boarded the plan, I said every day, six a.m. I want to be a lot. I want you to search the chat, search

Gmail, search my calendar, give me a brief of everything that's going on today, all the numbers I should be aware of, all these things. And it did perfectly until, I think I had like canceled because my account because I wanted to switch credit cards and I was defaulted back to the free version of chat chitp-t

like two days ago. And I'm going to have to premium when I get back. But the free version of chat chitp-t sucks so bad. And I cannot wait to be back on the good stuff again. Go ahead. So actually one thing you said, I kind of love is like the ambition of complex logistics and travel increase is increasing significantly. Because I remember like, again, whether travel with family, even before, I remember

the being old, the days of like printing out map quests into a folder and all your information and itinerary. Like, you would actually include less ambitious logistics because you just didn't want it to fuck up. You didn't want to like end up in a travel disaster. Again, when I was traveling like alone throughout Asia and stuff, maybe you took on ambition because they're going back to like whatever happens happens. But like now this trip, there was a lot of, a lot of, a lot of changes and transitions and cars rented and hotels booked and the Airbnb like in a remote place where no host is checking you in and there's a lockbox and like all of that and it was still perfect like because and I'm not like a huge planner planner in the past and this actually like

kind of makes me feel as empowered as my kind of like obsessive friends who would create spreadsheets for trips back in the 2000s. So yeah, I like that, ambition to travel. Yeah, definitely. It's not my strong suit. So it's cool to see it. All right, we'll leave it with that. Ranjan, great to have you back. Your glorious return.

Not a minute too soon or too late. I don't know. We're glad to have you and looking forward to doing this. We're back. It's going to be a fall. It's going to be a fall. Hold on to you. No, wait, sorry.

By golly. By golly. Here we go. Thanks, Rajah. All right, thanks everybody for listening and watching, and we'll see you next time on Big Technology Podcast.