Is Big Tech spending too much money?

0:05Pushkin. Big tech companies are big spenders. Like, seriously, these guys are not mucking around. One of the biggest is Meta, the company behind Facebook, Instagram and WhatsApp. Its shares took a massive knock the other day, wiping a cool$200 billion worth of its value off after it said it would spend, get this,$72 billion on AI stuff this year and a, quote, notably larger amount in 2026. A few dozen billion here, a few there, and soon enough, you're talking about some serious money. Now, the big reason markets have not been freaking out about all this big tech spending is that the companies have been doing this out of their own cash.

0:53Well, guess what? Now, Meta has issued a$30 billion bond. So today on the show, we're asking, literally, what the hell? Are these just like huge spending machines? How come this is all OK? This is Unhedged, the markets and finance podcast from the Financial Times and Pushkin. I'm Katie Martin, a markets columnist at the FT in the allegedly broke, destitute, crime-infested hellhole that is London. It's actually fine. Don't listen to that Elon Musk guy. And I'm joined down the line by a little twofer for you here, listeners. Firstly, the big fella, Rob Armstrong, who you all know and largely tolerate from the Unhedged newsletter.

1:34and secondly by Hull's finest John Foley from the FT's Lex Column, both in New York. John has promised to serve full finance bro here today, listeners. Together, we are joined by a love of cheap plastic watches, us three, and by a fascination into big tech and big markets. So John, FT's superfans are well aware of the Lex Column, but what is it? What are you doing over there, John Foley? Well, Katie, the Lex column is we're an opinion column and we write short, punchy opinion articles on companies and finance and deals and investment for people who have not much time and, like me, very short attention spans.

2:17Lex, ladies and gentlemen, is the true brains behind the FT operation. So with that in mind, let's all assume you know what you're talking about. John, please mansplain to me all of this craziness. You get extra points for very large numbers. What are big tech companies spending all these big bucks on? Well, there are two kinds of big number that are slightly freaking some investors out at the moment after the earnings that were released over the last couple of weeks. One is the amount that they're spending their capital expenditure sums, which are now running around, according to Morgan Stanley, about$620 billion next year of cloud computing.

2:56I mean, who's counting at this point? That's a lot of money, isn't it? That's between 11 companies. And most of it is actually the big four. It's Meta Platforms, which owns Facebook. It's Amazon. It's Google. Google's parent Alphabet. And it's Microsoft. I have numbers here. Warning, numbers incoming. Those four, Alphabet, Meta, Microsoft, Amazon. Their expected capital expenditure next year, respectively, 107 billion, 104 billion, 103 billion, 122 billion. So that's 400, almost 500, 450 billion dollars worth of capex just from those four. It sure is. That's number-wing people. And it's being used to build these big data centers, right?

3:38Big warehouses with servers in that are going to be used to train and run AI models. So that's the first big number that are freaking people out. And the number's getting bigger and bigger every single time. And last week, we had all of those companies, Microsoft, Meta, and Google explicitly saying, and Amazon implying, that they're going to quite significantly increase their spending next year. Meta said it was going to notably increase. Google said significantly increase. So investors are like wondering how big these numbers can get. And particularly Meta's case, Mark Zuckerberg, who runs Meta and founded Facebook, is saying that he's going to aggressively front load spending.

4:16And you never want to hear those words. And by the way, it wasn't restricted to capital investment. They also, these four companies, or at least three of them, suggested there would be pretty significant increases in operating expense. So they'd be spending more on people, on depreciation, which is what capital expenditure becomes when it gets old. And so that's a part of it too. Well, also for Meta, if you think Meta, all the data centers that Meta builds, and this is a difference between Meta and its peers are for its own use. It doesn't like rent server space out to other people like Google does.

4:51So Meta, everything is like for it to make its own AI models super amazing. But it's also on top of that, to your point, Rob, about operating expenses, as well as building this stuff, which comes out as CapEx, it's also paying other companies to use their space, which comes out as OPEX. So they're spending increasingly large amounts on data center capacity that isn't theirs. But it just suggests that their appetite is limitless. It's just spiraling outwards. I have two points to make about this, and I'd be interested to hear, John or Katie, what you think about the two and what you think is most important.

5:23So two differences between Meta and Alphabet, Amazon, and Microsoft. One difference you just pointed to that those other three have a business model, which is we'll add this AI capacity and we will rent it to other people or sell people AI, other companies, AI services. Meta is just going to do something fantastic later, I think was the basic line, whatever that may be. Yeah. Point number. And do we have to sort of, yeah, guess what that is? Because I'm old enough to remember their whole like boondoggle into the metaverse. Well, it's super intelligence, right? It's super intelligence. A artificial general intelligence, you can call it lots of things.

6:06But it's the omniscient, super clever, strangle us in our sleep, super AI. That's what he called it. Okay, cool. I make it sound so good, John. But there is another salient fact about meta, which is that in percentage terms and in dollar terms, its spending is expected to go up more next year by Wall Street analysts. And it's expected to go up so much that it's free cash flow. The actual dollars the business spits off with which it can pay dividends, buy back stock, do acquisitions, whatever. At Meta, that number, free cash flow, is expected to fall in dollar terms. At the other three, they're still expected to grow free cash flow next year.

6:53Now, analysts could be all wrong about this, but there's a difference in business model here, and there's a difference in financial outlook, which explains why. Some balance of those two explains why. Alphabet says, we're going to spend a load of money. Its stock goes up. Meta says, we're going to spend a load of money. Its stock goes down. Right? So investors seem to be saying one of these things is not like the others. Yeah. Like, it's impossible to put, like, because we don't know exactly how much they're going to spend. And this is particularly true in Meta's case. Everyone is just reading the route.

7:26It's like Vibes based. So when Zuckerberg says aggressively front load, and he said, he was asked last week, what happens if it's all too much? What happens if you don't need all of this capacity, these data centers? And he sort of said, well, the first step is that we'll give it, instead of doing super intelligence, we'll give it to our advertising business because that also needs lots of service space and AI technology, which is true. But then if that's not enough, the analyst said, what then? And he just was like, well, no, we'll kind of grow into it. We'll just see what happens. Whereas the other company is like, well, we'll rent it to everyone.

7:56Like we are cloud computing companies. So if we don't use it, then you will use it. But Meta is like, we'll see. So that's a bit concerning. John, I have questions about this. Katie, you can answer them too. One, if you have built a very large, expensive data center specifically for artificial intelligence and crammed it full of NVIDIA chips, can that be used for non-AI things? Does it make sense to say, we'll just use this compute for any old damn thing? That seemed to be the impression Meta was giving. My second question is, what is the useful life of an NVIDIA chip? And I've heard different numbers on this thing.

8:42People are depreciating at different rates. So, like, if AI is as big as these guys say it is and they're going to build the super-duper intelligence, are they going to have to basically replace all these GPUs in, like, three years? And so they're going to be spending$100 billion again every X number of years until the super intelligence liquidates us all and flies off to another galaxy. That's the implication, isn't it? You know, these things aren't designed to last forever. It's not like the railroads. Do we have other facts here, John, that we need to be aware of? There's not time for facts.

9:23This is not a podcast for facts. It's 2025, man. What are you talking about? So not all of the spending is chips, right? So the first thing is that about, depending on who you ask, like some of the numbers that have been coming from NVIDIA suggest that about$3 out of five that they spend will go on chips and related servery kind of stuff. And then the rest of it is like, you know, the warehouses and the cooling systems and all that kind of stuff. So this all depreciates at different speeds. Like I'm not a sort of chip expert, but I've been told that most of them are depreciating over roughly six years.

9:55and we haven't really been using GPUs for AI for very long. The discovery that GPUs, these kind of chips are good for AI, only came about kind of accidentally a few years ago. But for CPUs, which are the less fancy chips that you use for general kind of like cloud computing, those often can still be used. They can be used a bit longer than you might think. So you depreciate them, but then they still have a kind of useful life for a while afterwards. I don't know if that's the case with GPUs. I don't know if the GPUs just like melt after six years or three. But it won't be that they have to replace all of the CapEx.

10:28Yeah. But do they have the equivalent of like my four-year-old laptop, which everyone laughs at because it's like so thick and runs so slowly. And the chip cycle is moving really quickly. Like what if next year someone comes – Do you remember the DeepSeek thing earlier this year? Which like no one's really talking about as much anymore. But it was the idea that a thing we haven't thought about suddenly changes the whole trajectory of all of this because someone's worked out a way to do something much faster and cheaper. What if a chip comes out that makes what we currently have more or less obsolete or a cooling system comes out that makes what we have obsolete?

11:00Like that is something you have to think about. If they've like planned to spend$100 billion on NVIDIA chips, but actually you don't need NVIDIA, you only need one chip by some new supplier. Unlikely, admittedly, but the point is like when you're dealing with really long numbers and multiple years out into the future, you can get it wrong. John promised us initially finance, bro. I think he's serving tech bro here. Oh, no, don't say that. That's worse. But moving it back to, I don't know, markets bro. So maybe this question is for Rob. Like, why is it that certainly up to this point, and maybe this is running out of road a little bit with Meta, but up to this point, the markets have been like totally cool with these gigantic amounts of spending.

11:43That feels like a little bit of, you know, not quite the script of how markets normally work. You can't normally just throw money at the promise of some sort of magic beans without the market, you know, taking you to task for it. I mean, this is a weird, slightly weird thing to say, or I should say argument to make. But you could make the case for any of these four companies, Alphabet, Meta, Microsoft and Amazon, that they are the best businesses the world has ever seen. Amazon, it's slightly harder, but certainly Alphabet, Meta, and Microsoft, you can make that case. Amazon, actually Amazon too.

12:21The point being, these guys generate so much cash and have such incredibly high returns on capital that you kind of like, if they want to take a$100 billion flyer at something for a couple of years, you're like, okay, guys, whatever you like, because that business will still be there. right you know what i mean so if you're going to have any four companies do something a little bit risky these are the guys who can afford to do it i think also some of the reason why mark has been very tolerant of this is that ai has actually already done quite amazing things to these companies business models so meta people forget this but like in 2022 meta's revenue was shrinking a huge company has managed like a company with like trillions of dollars of market cap and hundreds of billions of dollars of revenue, has managed to go from shrinking revenue to revenue that's growing at 20%.

13:13That's quite amazing. And a large part of that is to do, there are various things in that, the advertising business recovering and so on. But also AI is a big part of it. It has actually generated returns already. So there is a limit to how much spending makes sense, but you can see why people would like more of it rather than less. Now, I want to just pivot a tiny bit because we mentioned at the top of this conversation that Meta has issued something like$30 billion worth of bonds. Now then, there's a lot of these big tech companies issuing absolutely shared loads of bonds, like to the extent where, so reading here from a piece of ours the other day, Goldman Sachs had previously estimated there's going to be about$180 billion worth of these bond sales by these big tech firms this year.

13:59It's looking like it's going to be more than that. But this is already like a quarter of all net new supply of U.S. corporate debt this year. That is something, right? This is becoming a really dominant factor in U.S. corporate bond markets as well as in stock markets. Rob, what potential issues does that throw up? Well, it's a lot of issuance into a market where everything is already really expensive. What I'm referring to here is like the investment grade bond market where the spread, which is the difference between what you pay for a treasury, which is issued by the United States, a country with the printing press, so it literally can't default.

14:48And like a high grade, high quality corporate bond, that difference is historically skinny. And so if you dump a huge amount of supply into that market, that spread can widen and a little widening of spreads really hurts bond investors. So it's not that anybody thinks that Meta is going to default on these bonds or Google or Microsoft or Amazon. I think you could argue correctly that these are actually better credits than the United States government. You know what I mean? It is actually more likely that the U.S. will default than that Microsoft will. But it's just the sheer supply is my point.

15:33Just jumping the supply on the market is a big deal. And when you get lots of what we call supply, right, so when lots of new bonds hit the market, If you are a corporate bond investor, you sit there, you've got your portfolio full of corporate bonds. Some new corporate bonds come along, like a lot of them from Meta or Microsoft or whoever it is. And you think, oh, I want to buy some of that. So I have to sell something else to make room in my portfolio to put that there. So sometimes when you get a lot of supply like this, it means that fairly random, unconnected, you know, other corporate bonds weaken in response.

16:07And this doesn't seem to be what's happening here, but you can get little accidents that happen around there where you end up with little spirals of selling in random bonds that get a little bit ugly. So, you know, the bankers that, you know, look after these bonds and launch them onto public markets know what they're doing here. But you do have to be careful when you're splurging that much new debt onto the market. I think that's right. so the 15 % drop in Meta's shares after it put out these spending plans last week what do we think do we think this is a one-off or do we think this is a little bit of a sign that investors are starting to take a slightly more discerning view of big tech spending writ large I would hesitate to say that it shows a serious change of mood in investors I think that also included a reaction to meta issuing a big chunk of bonds, as you say, right?

17:03And that made people say, oh, what if this AI capex boom also features leverage, features debt, which makes people a bit concerned? I actually don't think they need to be because I think, although, as you've said, these amounts of debt are very large in the context of the market and normal issuance. They're very small in the context of meta. Google's bond that it's issuing is very small. These are companies that don't really have debt. And you actually want companies to have some debt because it's not an efficient way to run your balance sheet to do it all with expensive equity. So there isn't really any kind of like Meta is not going to go bust over this.

17:40And if Meta's share price falls, even if it falls quite a lot, the problem, if you could argue, is that Mark Zuckerberg doesn't really care. Like he doesn't care. He doesn't need the money. He doesn't change his plans. He's not looking to raise equity in the market to fund this CapEx. He has voting control of the company. He has control of the company. And an activist investor cannot make his life difficult. Nope. So he can just say, well, investors don't get it. And that's the story of Facebook is often about Zuckerberg making a decision. Other people saying this is terrible. You shouldn't do this.

18:11And him saying, I know what's best and being right a lot of the time. So I think – But sometimes being wrong. There was a dramatic fall. The metaverse was a terrible call that crushed the share price. Yeah. And it took years to recover. I think he sees that, though, still. I think he still sees that that was all money well spent on the long-term Facebook project and VR, for example, headsets and things. So I didn't think he cared. The risk is that he's not listening to what equity markets are telling him. I would take a slightly different angle on your question, Katie. I would say that this latest earnings season, which came at a time when everyone was saying these companies are just spending like crazy and it's an AI bubble and, you know, it's tulip bulbs or whatever.

18:59This earnings season did show that investors care about earnings and cash flow. And I think that I really believe that. And I think they're making a distinction between Meta and its big peers because Meta's earnings and cash flow outlook is both not as good and not as certain as the others. And they care about that. And you can see that, for example, in the price earnings valuation of the various companies. So I think this all sounds healthy to me. The market actually cares one way or the other about the economic fundamentals of these businesses. So that's a good thing that I feel good about. That is reassuring.

19:46I hope you're right and there's a first time for everything. Chaps, we're going to have to come back in just one sec with Longshot.

20:03Okie dokie, it's time for Long Short, that part of the show where we go long, a thing we love, or short, a thing we hate. John, what you got? I'm going long M &A. Nice. We're seeing a big resurgence in companies buying each other, and not just like big deals happening, which we are seeing again this week already. We've seen this big deal, Kimberly Clark buying Kenview, which kind of sells paracetamol in the US, like tens of billions of dollars. But also we're now getting like deranged shoot-em-up style M &A where companies just wildly outbid one another like this bid for a weight loss drug owner called Metzera in the US, which is sort of ratcheting up the tension.

20:40So M &A exuberance. I'm long. Yeah, merger mania. Love it. Merger mania. Rob, what about you? I am short free buses. We have an election here in New York City today. the likely winner although not the certain winner is the democratic socialist zoran mamdani he promises to make buses free i think services from the city should cost things i think making them free regardless of the economics sends the wrong message because i am a tiresome middle-aged man this is your most boomer opinion i've ever heard you utter i am is that like how if you give If people free mosquito nets, they don't use the mosquito nets.

21:22But if you charge them a pound, they use them. It's important. Symbolically, it's important that you pay so you value the thing that you're getting, even if the price is nominal. Who doesn't like free buses, though? Katie, if you're going to go all pinko on me, too, I'm going to quit this whole show. I would love free buses. I am long, clean fluences. so earlier today i was on i was doing an ft event with a uk cleaning products company and apparently the big thing for selling cleaning products now the big kind of marketing strategy is tiktok and there were like people on tiktok who go around cleaning things and then people buy the cleaning products and i'm like i had no idea that there were clean fluences that they were a thing that people are cleaning things on tiktok certainly not encouraging the teenagers in my life to clean in the house, but who knew?

22:19Who knew? The internet is terrible. We should shut it down. On that bombshell. Listeners, we will be back in your ears on Thursday, so listen up then. Unhedged is produced by Jake Harper and edited by Brian Erstad. Our executive producer is Jacob Goldstein. Topher Forhears is the FT's acting co-head of audio. Special thanks to Laura Clark, Alistair Mackey, Greta Cohn and Natalie Sadler. FT Premium subscribers can get the Unhedged newsletter for free. A 30-day free trial is available to everyone else. Just go to ft.com slash unhedged offer. I'm Katie Martin. Thanks for listening.

23:13We'll be right back.