Episode Transcript
The Finance Ghost: Welcome to episode 261-and-a-half of Magic Markets. The reason it's a half is that we've had to re-record this entire show because of a technical issue. So, that’s our commitment to you: to make sure we get one of these out every week, even when we have to do it twice!
Luckily, we do enjoy doing it, Moe. I'm not sure we enjoy it so much that we'd like to make this a regular occurrence in terms of the redo, but nonetheless, we now get a second bite at the cherry to talk about what we wanted to talk about this week.
And that is the world of Big Tech – how they’re throwing tons of capex at AI and how all of this capex is going into a world where AI is supposedly taking everyone's jobs.
Even though our audio recording sometimes doesn't work, it's now kind of funny to actually compare where tech works and where it doesn't. But anyway.
The other thing we're going to talk about is free cash flow (FCF) margins as well. This is something very topical because we covered Shopify in Magic Markets Premium this week. If you’re interested in that, go and check out Magic Markets Premium. It's only R99 a month.
I think one of the really good things about it is we’re not trying to screen for stocks that are a great buy this week, or next week, or the month thereafter. We're actually just looking at it and saying, “Well, there are a whole lot of stocks in the world. Let's learn about them,” so the learnings are somewhat more evergreen.
You can go back and enjoy the library of stocks. If there are names that you've been looking at that you want to get some research on, it's pretty much all in there.
It's unlikely that you would have a big stock top of mind that isn’t somewhere in that library, because we’ve done many reports now. So, go and check it out. Join us in Magic Markets Premium and learn about why Shopify's valuation is so enormous.
And Moe, let's dig into this Big Tech story, and capex, and then some FCF margins – again, might I add.
Mohammed Nalla: indeed Ghost. We love doing this. So yeah, I guess it's an episode-and-a-half.
The Finance Ghost: Don't let the technology hear you. We don't love it as much as that. Once is fine.
Mohammed Nalla: [laughing] We love doing what we do. We don't love redoing the exact same show again. But I mean, tech is your friend until it's not your friend.
Is AI eating the world? Well, we're going to answer (or try to answer) that.
One of the key themes that we've certainly been seeing – not just on the macro front but certainly in terms of some of the stocks that we've been covering – has been this massive (and I literally mean ‘massive’) build-out of capex spend, in the US particularly.
We saw that most recently when we covered Alphabet, which is the parent company of Google. When we looked at Alphabet, their guidance for capex for 2026 (I said this on the Premium show, but I'm going to say it here as well) was between $175 billion and $185 billion for 2026.
Now, if we round that off to, let's call it, $200 billion, that's roughly half of South Africa's GDP. And that's a loose stat that you can actually use. But just for context, that's one mega-cap tech company's capex spend for a single year. So, that tells you what's happening behind the scenes.
Why is this so important? The absolute numbers become almost hard to imagine, but when you actually look at these numbers in terms of a ratio, if you look at it relative to the company's FCF generation, that's when you actually see just how much of the kitchen sink some of these tech companies are throwing at the problem.
I've mentioned Alphabet, but let's look at Meta. I know that's a stock that you kind of like. We've covered it a couple of times on Magic Markets Premium. Meta's capex guide for 2026 is in the region of around $115 billion to $135 billion.
If you compare that to their FCF, I stand to be corrected on the number, but I think they're using 100% (if not more) of their FCF on capex. And that's Meta. We know what kind of roads they've walked down in the past, in terms of the metaverse, so you can determine whether it's money well spent or not.
That's Meta. We've covered Alphabet.
If you look at another player, if you look at Amazon, I think their capex plan for 2026 is around $200 billion. These are really, really large numbers that are out there.
So, the question is twofold. One, how much runway is there actually? Because these companies are throwing all of their FCF into capex. That means not a lot of that actually accrues to investors or to shareholders at the end of the day.
The other thing is, does it actually generate returns? Because these are effectively depreciating assets. There's an obsolescence curve – lots of talk around that. And we haven't even gone into whether this is starting to create some stressors on these companies’ balance sheets.
We know a lot of them have remarkably strong balance sheets, but just after we covered Alphabet in Magic Markets Premium a couple of weeks ago, they actually went out there and announced that they were considering a 100-year bond issuance.
I think it's in sterling (some technicalities around that), but I mean that's a 100-year bond issuance. Again, I stand to be corrected on the size (Ghost, if you have the size on that one), but these are really large numbers that are out there.
The Finance Ghost: It was a couple of billion pounds, I think. It was a number. It's small in the context of Google/Alphabet, but it was a number, and it is 100 years.
Mohammed Nalla: [laughing]
The Finance Ghost: Imagine lending someone money for 100 years. That’s amazing. That's ridiculous, actually. Talk about creating a problem for your kids.
Mohammed Nalla: Ghost, before I even hand over to you, it almost reminds me… A while ago, when someone was discussing global financial crises and all of these things, and they were saying, “Well, technically, you could issue a perpetual zero-coupon bond,” right?
And what is that? Effectively, it's money printing. So, I'm waiting for these tech companies to offer zero-coupon perpetual bonds. That's when we know this is fully cooked! [laughing]
The Finance Ghost: Yeah, that’s great. No, look, that will be a dark day. Don't give them any ideas, Moe. The AI is probably listening. It's probably trying to figure out why the first recording went wrong.
Anyway, shall I jump into some FCF margins then, after we've discussed this? But actually, before I do that, I do just want to set the scene with a couple of share price moves.
Because what’s interesting is, a year ago, you might have thought, “Well, I'm very clever and I'm going to avoid all of these crazy, capex AI what-what's and I'm going to go and buy Apple because everyone loves Apple.”
If you had done that, you would be up 4.6% in the past year, which is not a lot. In fact, if you consider the rand-dollar, you would have been better off just putting your money under the mattress. So, not great.
The winner here is very much Alphabet, up 66% in the past year (something we highlighted in our recent research on Alphabet). But that share price has come off its highs quite sharply and is very much on a downward trend, so be careful of that.
Still, all of that is much better than if you had bought Meta, down 10%. Microsoft, down 2%. Amazon, down 12%. So, the market clearly has jitters.
Basically, what's happening is all this capex is just a flow of value out of these names – not necessarily Apple so much, but out of the other ones I've mentioned – and into the providers like Nvidia, for example. So, you definitely want it to be as far up the value chain as humanly possible, in this particular case.
But let's deal with this FCF margin because I think it will help with explaining why this is the case and where this money is going.
Let's start with the software as a service (SaaS) businesses that no one seems to want anymore, beginning with Salesforce. Highly seasonal business, but FCF margin generally above 35%. That's really good for a business that no one seems to want anymore.
Adobe, much the same – high 30s, great business. Everyone is kind of assuming that people are just going to vibe code replacements and everything's going to go great. Again, I will remind you, we've just had to re-record this episode because our tech broke, so I think the over-reliance on tech will only get you so far.
Shopify, which we just covered in Magic Markets Premium, that’s a high-teens FCF margin. Not so much SaaS, that business. They've actually become a little bit of a weird hybrid. I'm not sure they really know what they are – elements of SaaS, but also lots of fintech and that kind of thing. Bit of a drag on margins there, so go check out that report.
Now, let's look at the more capex-heavy plays, and let's really bookend this thing by looking at Taiwan Semiconductor Co. (or TSMC). This is a really, really interesting situation. Their foundry business is an absolute capex pig – they need to keep putting capex in this thing all the time. Literally, they have no choice.
So, what's interesting is it's incredibly cyclical. They have to go and invest a fortune, then they wait for the demand to come through for the chips, then they make a lot of money, and then they go and reinvest, and the FCFs basically disappear. Then they come back again.
It's quite incredible. Go and chart it over several years, and what you'll see is their FCF margin at the moment is around 17.5% on a last 12 months (LTM) basis. Ironically, kind of similar to Shopify, which shows you the extent to which the Shopify model just doesn't quite make sense.
What’s interesting is that if you go back to TSMC in the heat of the pandemic, when cloud was absolutely everything, their FCF margin was in the low single digits. They were just ploughing almost everything back into the business.
Now, they’re enjoying life in the high teens – although it actually went up to the mid 20s in 2024 and much of 2025, so that's the cyclicality that I referenced.
The SaaS names are not cyclical, and that's why they tend to get a premium valuation. But the market doesn't really seem to be applying that rule of thumb anymore, because SaaS has been sold down really heavily.
AI spend, maybe it's not being celebrated right now, but it's not being punished nearly as much as these other companies that seem to be getting punished just for existing right now. Let's get into some of the names that are being punished.
Microsoft, start of the pandemic, FCF margin: mid- to high-20s. Very much a software company. Yes, they had Azure, they were investing in infrastructure, but at that stage, Microsoft was still seen as enterprise software.
That has now changed quite a lot. It's now seen more as an infrastructure group, and they’re struggling to keep it above 20% in terms of FCF margin. But, unlike TSMC, the capex is not going into something that is proven at scale. It's not going into chip manufacturing.
It's going into data centres. It's going into stuff that is really risky, and we've seen margins deteriorate quite sharply and the risk go up along the way as well. So, that's why the valuation has come under pressure here.
I just want to pause for a moment and explain what this FCF story really means. Imagine you own a small business and it makes R1 million profit a year – what do you do with those profits?
Well, if your business is nice and capex-light, you can pay them out to yourself as a dividend, for example. But if your business is capex-heavy, yes, you made a profit, but that money is going to get reinvested in capex, and then it's about, “Well, what is the useful life of that thing?”
And that's the problem with these data centres. No one's really sure – is it three years, is it five years, is it ten years? No one quite knows because the technology is evolving so rapidly.
So, if all this money is flowing back into tech that becomes obsolete quicker than everyone is expecting, then those profits may as well never have happened, right? They're just being reinvested in stuff that may or may not work. And that is why FCF is so important.
If we look at Alphabet, that's even more severe. They've never been shy to throw money at their Other Bets segment and the marginal parts of the business. So their FCF margins have been in the mid- to high-teens for most of the past five years. Peaked at 22% but currently at around 12.5%, reflecting their investment in AI.
And the worst part, Moe, which is what you flagged, is that they are talking about a higher spend in future, not a lower spend. They are taking a leaf out of TSMC's book here.
It’s worth talking about Amazon before I sign off on that point. Another absolute beast in terms of capex. Very, very capex hungry.
And if you have a look at their LTM margins on a quarter-by-quarter basis (so basically, rolling 12-month basis), then what you’ll find is that their FCF margin has touched 10% once in the past few years, and on every other rolling 12-month basis, it's in the mid-single digits.
In fact, it even went negative at one point in 2022 because they were actually reinvesting profits that they weren't even making. They were investing more than they were making.
So, revenue – yes, it may be the main driver of a valuation, but the market is paying a lot of attention to FCF margin, and that means something.
Mohammed Nalla: Yeah, Ghost, those fracture lines are certainly showing up here. Some of these companies are spending close to 100% of their FCF, if not more, on effectively the next year's capex, and I think you raise a very important point.
You had the analogy there in terms of if your business generates a million bucks and you put that million bucks back into the business, what have you actually gotten out of the business? And that means that not a lot is accruing to shareholders, at the end of the day.
I think there's an additional point, though. Because investing in these data centres, with a very steep obsolescence curve… It’s almost as though the business goes out and buys this massively depreciating asset, like a car, that doesn't necessarily generate the same kind of revenue in order to justify any sort of return on investment.
So, I think those are the fracture points on the FCF story that we need to pay attention to.
The other thing I want to land on as well is that we've mentioned how Alphabet was considering a 100-year bond – and these are chunky numbers. They're in the billions and billions of dollars. Why would they be doing that?
Because a lot of these tech companies have fairly strong balance sheets – I say a lot of them, not all of them. We'll touch on one that is arguably the canary in the coal mine.
But the fact of the matter is that if they're seeing opportunities in terms of raising long-term funding at preferential rates, maybe it's cheaper for them to actually do that on the back of the debt, rather than on the back of equity?
So, you've got to look at the weighted average cost of capital there. I think that's important. But that's why you've got to pay attention to what's happening in the credit markets, as well.
And here's where we get to that canary in the coal mine – Oracle. That's a company we covered (I think it was the latter half of last year) in Magic Markets Premium. We were quite bearish at the time, and from the time we covered that stock to where we are today, that stock has almost halved in value.
That's on the equity side. But if you look at the credit side of the market, lots of news flow around Oracle CDS (or credit default swaps). Those, almost doubling – and that’s the cost of insuring against a default by Oracle, almost more than doubling over the course of the last quarter, there and thereabouts.
That's showing you that there are stressors. There are strains coming through. Is it a capex hole at the end of the day or not? We don't know. We don't know when the music stops.
They say, “You only know who's swimming naked when the tide goes out.” Same with musical chairs.
Now I want to move on from that, Ghost, into a look at another point that we maybe haven't touched on. That is if you look over the course of the last week or so (in fact, a couple of weeks), you’d have seen a bit of a rotation within the tech sector. So, you've seen a rotation out of some of those heavy capex players into the beneficiaries of that money.
For example, Micron. They did really well. That's because they are beneficiaries of this ‘memory shortage’ that we're seeing in the market. Again, the money's flowing through the ecosystem. You're going to have to actually track that if you're playing this game.
When the music stops, though, I think it actually filters throughout the entire value chain. You've got to be very sensitive. Interestingly, last week you actually saw an underperformance amongst the Big Tech names, but then you also saw banks underperforming.
Now, I know there's a lot of macro news out there, so it could have been more than that. I'm not saying it's definitely related to some of this, but my question is, when do we get to a point where companies are too big to fail?
If, for example, Alphabet starts to hit the wall, are they too big to fail? Does that actually then filter through to their banks? And if their bankers are JP Morgan, whoever that might be, are they too big to fail?
And so you end up with this contagion-type of risk where the capex spend that's happening in this AI space becomes a risk flag in terms of a much wider contagion effect that could filter through the financial ecosystem.
And then another point that you've got to note is, where are you on that value chain? Because you mentioned TSMC, for example, but then you've got a stock like ASML that sits even further up that value chain. They effectively sell the equipment that helps manufacture the chips in the first place.
So, how far do you go up that value chain? Where does the pressure actually materialise, and where is that pressure going to be felt most acutely?
Ghost, I don't know if you have any views on that, specifically?
The Finance Ghost: What I do have a view on is what might happen to the poor ASML share price.
I don't know how many of their shareholders are actually Dutch, but you've got to love the fact that on one side of the pond, you've got all these American tech giants spending an absolute fortune, and on the other, you have European regulators who have gone and now passed a tax on unrealised capital gains in the Netherlands, which is an absolute disaster. Extremely stupid. Anyone who understands anything about finance will know just how dumb that is.
And it's almost like the European regulators looked at the entire continent and were like, “Wow, we have one really cool tech stock. It's in Holland. There must be some unrealised capital gains there, let's go!”
I mean, who knows? Honestly, it is a bit of a shambles, really, Moe.
But on the ‘too big to fail’? Look, we're probably there already. I would imagine that, even if things go wrong with the amount of money that has gone into capex on AI and there's too much supply, there's so much cash in the system – if they basically just all stood still and said, “No. That's it. No more,” their core business still makes a fortune, right?
They're spending money that they make every year at the moment. We're not in negative FCF margin territory yet.
Nasty for valuation. Lots of money will have been wasted. A catastrophe for anyone who's left holding NVIDIA, if that happens. But I don't think they would fail. I don't think we're there yet.
It's interesting. Remember when Zuck was pushing the metaverse so hard and he had that ridiculous video with “legs”, and the market absolutely panicked because they thought he was basically going to run the thing into the ground?
And then all he did was he just slowed down the spending. Just said, “Okay, fine,” pulled back a bit, and the machine that is Facebook and Instagram made a whole lot more cash, and the market was like, “Okay, cool. Now that Zuck's calmed down, we can buy this thing again.”
I think that's the big difference here. They all have absolute cash machines at the centre. They just have to be careful not to overextend themselves.
Mohammed Nalla: Yeah, Ghost, I think that's a fair point. Right now, the pressure point is how fast they are letting the cash out the door.
If these Big Tech companies eventually say, “Enough, we're going to stand still.” Yes, their core businesses will still generate lots of cash, effectively. Certainly, the stronger ones. Then, in that instance, it means that the pressure flows down into the equipment manufacturers and further downstream.
So, it's always going to be a question of who's left holding the can. Because if we actually continue down this path, how much runway do we have before even the Big Tech giants run out of runway?
And then it's a question of who's the weakest player in that entire value chain. I think that's really where it's got to land.
Are they too big to fail? Can they rescue it? I mean, we're talking orders of magnitude in terms of where Zuck was spending on the Metaverse. You know, “Hey, look – legs!” vs. the amount of money across the entire ecosystem that's being spent here.
Because we've kind of contextualised that, but if you wrapped up Meta, if you wrapped up Alphabet, all of the big players, Amazon, you're pretty easily sitting on close to $1 trillion. And that's really large, even if you look at that in terms of just a proportion of US GDP.
A great analogy that has kind of done the rounds (and I've seen it and I think it's valid) is that some people who study history have gone and they've had a look at the extent of investment in US Railroads at its peak.
Again, that was a bubble that then famously popped. We're not quite at railroad percentages of GDP, if memory serves, in terms of peak railroad investment as a percentage of US GDP.
At the moment, on this capex-led AI spend, we're probably close on around 4%. That's a number that I saw out there. But to contextualise that, that's 4% of what is still arguably the world's largest economy, and if that bubble pops, it's going to be pretty painful further downstream.
Ghost, we haven't even discussed other places in the value chain (perhaps something for another day), but there are the companies that are maybe REITs that give you exposure to data centres. They might not have the obsolescence issue that some of these large tech names have.
That would be companies like Equinix or Prologis. That's something we've covered in Magic Markets Premium before.
All I'm saying is there are a number of pressure points throughout the ecosystem. The choke points here are more than just memory availability. There are choke points on water, there are choke points on power.
I saw Meta, for example, signed a deal with Oklo. They're going to be effectively generating their own power on location on some of these data centre investments. Some of it will probably be useful infrastructure.
If energy infrastructure goes in, that could arguably be repurposed even if we get through some of the exuberance on AI investment in data centres. But the fact of the matter is that at the end of the day, a lot of this tech that's going in, may well end up being obsolete.
It may also not generate the required return on investment. And going all the way back to your ASML point in the Netherlands, if they just wait, depending on where we are in the cycle, if the music stops here… Heck, they may not actually have any unrealised capital gains to realise if, actually, you see that pressure filter all the way through to some of those players higher up the value chain.
Ghost, I think we've tried to cover a lot of ground here. I don't know if there's anything further you'd like to add before we wrap up?
The Finance Ghost: Yeah, so the benefit of you being able to talk there and me being able to Google is I could go and check the railway point – worth making sure listeners get the perfect number here. So, according to Google AI, which is hopefully right, it peaked at around 6% to 7% of GDP.
And (again, according to Google AI, with all the health warnings that comes with), data centre investment was around 4% of US GDP in the first half of 2025. So, not that far off…
Mohammed Nalla: You're scaring me!
The Finance Ghost: …is kind of my summary. Not that far off.
Mohammed Nalla: I thought we were a lot higher.
The Finance Ghost: Yeah, no, it's not that far off.
Mohammed Nalla: Maybe the fact that you used AI to get that answer might mean that we've got a little bit more runway there, but that means we're a lot closer.
The Finance Ghost: Google doesn't give me a choice. Do a normal search, and you don't get much choice; you just get the AI one. But yeah, it's pretty interesting to do that. Anyway, maybe something we can unpack in more detail at some point, but certainly worth keeping in mind for investors at the moment.
And I think especially those SaaS names are almost worth a look. I think we should do it soon in Magic Markets Premium, because they do feel like people are just assuming that these things just go away. I'm not sure life is that simple.
I think Adobe's maybe in trouble because the creative industry is right in the crosshairs of AI. But something like Salesforce, Intuit, some of the other things…I mean, I use QuickBooks in my business. I don't think AI is about to change that. Something for us to go look at, I reckon.
Mohammed Nalla: Yeah. I think there's so much. It's such an interesting space. Definitely lots for us to unpack. Your point around Google not giving you a choice – they've got to try and embed the use case for AI in their core business, which is still search.
I would argue you could see similar kinds of developments in some of those SaaS offerings out there.
The risk here is one of disruption. It's the risk of people, as they call it, ‘vibe coding’ and just massive disruption there where AI can effectively generate apps much faster, that then negate the need for some of these SaaS companies. Again, lots of health warnings come with that. These companies weren't built overnight. They have lots of checks and balances, governance issues that come into play.
I don't know if I would trust an AI with my tax submissions right now. I think Intuit has a pretty nice niche on the market, but this is moving at such a frantic pace.
Again, maybe just to wrap that up, you keep on seeing these memes around early AI with Will Smith eating the pasta. I don't know if you saw that? It was horrendous, back in the day, and now it's actually pretty decent.
The Finance Ghost: I have seen that. He looks like Sid the Sloth in the first one, actually.
Mohammed Nalla: [laughing]
The Finance Ghost: He looks exactly like the sloth from Ice Age. And then, yeah, go to number four, whatever we're on now, and he looks real, which is becoming a very interesting IP fight for Hollywood, by the way.
Mohammed Nalla: It is indeed. I think lots of pressures in the creative space coming through from AI. I have no doubt this is a topic we're going to keep on unpacking, maybe until the music stops, maybe even after the music stops.
Maybe the use case gets built beyond Internet memes and all the funny stuff that we've been discussing on the show.
Lots to unpack and again, lots that we're learning, and we'd love to learn from you, our listeners. So that's where we've got to leave it this week. But let us know what you thought of the show.
Hit us up on social media. It’s @MagicMarketsPod, @FinanceGhost and @MohammedNalla, all on X. Or you can find us on LinkedIn. Pop us a note on there. We hope you've enjoyed this. Until next week – same time, same place. Thanks, and cheers.
The Finance Ghost: Ciao.
This podcast is for informational purposes only and is not financial or investment advice. Please speak to your personal financial advisor.
