Welcome to the Merryn Talks Money Market Wrap, where we talk about the biggest

moves in markets this week and what's been driving them.

I'm John Stepek, senior Reporter and author of the Money Still newsletter.

And then join me in the studio today is Bloomberg's private companies managing

editor Neil Callanan. And Neil covers hedge funds, asset

management and real estate in EMEA

So in today's show I basically wanted to talk about the AI industry.

I wanted to give a structure of what it's actually about.

Lay it out so that you can you can understand it better.

They cannot refuse producing the chips and who's producing the models.

And then wanted to look at how that's funded and why there are some concerns

about perhaps the circularity of some of the deals in the industry.

And Neil has written a lot about that. Thanks very much for being with us

today, Neil. Um, it's great to have you on the show.

Thanks for having me, John.

Not at all. Um, so we're recording this on

Wednesday. Um, and probably the biggest story so

far this week has been the row in the South Korean KOSPI index.

Um, and basically, this is driven by the fact that the KOSPI is half of it is

comprised of two stalks that are related to the AI

boom. And that's SK Hynix and Samsung.

Um, and also there is an element of retail investors getting in over their

heads because they've been able to buy all these leveraged ETFs.

And so the KOSPI went up a lot and then down a lot.

I mean, one of the stats that really jumped out at me is that it's actually

still up 35% on the year. But in the last month it's fallen 34%.

And so you okay. That's a proper roller coaster moment.

Um, but does this tell us- what does this suggest about the overarching AI story

because it's not just about Korean leverage.

There's other stuff going on too, isn't there?

Absolutely. Korea, in a way, is a special case

because Korea has almost gamified the finance industry, the stock market.

And you mentioned the leverage ETF, but they've been big on crypto and tether

and things like that. And stablecoins for for a long time.

And and you know those a lot of those investors in retail have grown up very

accepting of those levels of risk. And in many cases it has paid off.

So as you said to KOSPI, is still up this year despite the rout in recent weeks.

But what is driving a lot of the change at the moment is the fear that China is

catching up. And China's emergence in AI, which was

always going to happen but perhaps faster than people thought, is really

concentrating minds at the moment. And we have seen in the past, when DeepSeek

came out of nowhere and released its models at the market disruption that

that caused. And now we're going through a kind of

second bout of that where in the space of just over a week, you've seen massive

progress in terms of memory chips, in terms of language models.

And so China is definitely emerging. And one of the things around all this

spending that we're seeing in AI is the assumption is that people are willing to

pay for it. Yeah, and pay a lot for it going forward.

And then suddenly if you have these cheaper models coming out of China and

people start switching to that, then what happens with the names we're all

familiar like with like, um, OpenAI and anthropic?

Um, and so, you know, there's definitely fears in the market about that at the

moment. And I think people, um, are definitely

taking pause globally as a result of that.

At the same time, a lot of this money is going to be spent.

There is an investment case for it. Um, it's whether it's getting ahead of

itself at the moment. And certainly in terms of valuations, it

did seem to be getting steamy at one point this year.

Yeah. I mean, I thought we'd just kind of lay out

roughly the AI industry and the business model because I think a lot of

the time, you know, we sort of AI and it's quite an amorphous kind of blob

And a lot of time. We're thinking about ChatGPT.

I know Nvidia is in there somewhere and all the rest of these things, but so I

was kind of looking at it, um, and as I was talking to the AI about it, um, and

getting a sense of what the kind of value chain is.

And so you've got, you know, you've got the machines that make the and sorry you've

got the companies that make the machines that make the chips and that's like just

ASML um, that makes the kind of the lithography machines.

And then you get the chip foundries themselves and that's like the Taiwanese

company TSMC. You've got chip designers and that's

Nvidia. So they they're the ones that basically

do the kind of value add element isn't it it's like they're they're making

the brains for these things. And then all of these chips go in a big.

AI hotels the kind of data centers and they're run by the same people that run

what we used to call internet hotels. The kind of data servers like the

Amazons and all the rest of this world? And then you've got the people who

actually make the AI models and the AI models, the brains of the AI live on the data

centers, and that's like your OpenAI and then they're selling them.

to companies, either big enterprise providers who can like, garnish their

existing offerings with AI, or induce users like me and you, uh, of

smaller companies. Does that-

Is that about the size of it? Is that does that kind of sound about.

Right, in terms of what the AI value chain as in who are the companies are

involved in this 100%, but like, this is so massive now that it encompasses

everything else as well. So they need energy.

And so we're seeing a massive boom in energy, and people are talking about

needing up to 300GW of additional energy power by 2030.

That's enough to power 225 million homes for a year.

That's just for data centers now, those things have to be built as well.

You have to get people to build the data centers and build the energy

infrastructure, etc.. So construction firms are taking off as

well. And so, you know, in the US in

particular, it's driving a lot of the economic growth.

And any slowdown in this kind of spending would be negative for growth,

not necessarily turning recessionary or anything, but it would be negative

for growth in the US. But

this goes back to your point about in the end and it comes back to the end user.

This goes back to my point about China in a way, which is that, you know, we

need people at the end to be paying for these services in order to justify these

investments. And when you look at the amount of money

that are being spent by some of the hyperscalers in particular, which are

these internet hotels that originally, as you described them, um, you know, if

you look at Alphabet's recent filing, its, um, forward spending commitments

rose $500 billion essentially in three months.

Uh, revenue in future revenue growth grew, but not by anything close to that.

And that goes back to the fears people have about this entire ecosystem that is

sucking up so much money and so much capital at the moment that people are

kind of going well. I don't know whether I want to have as

much exposure as I could have to this. And you're starting to see that in the

credit markets as well, where people are becoming much more discriminating about

the deals they invest in, underpriced or willing to pay.

And they're also hedging a bit more. So this week CoreWeave’s CDS, which is a form

of hedging against, um, default risk. Yeah, that's risen to almost a record

CoreWeave is one of these companies that is

like a we work of the, uh, GPU world, which is the chips and they basically

rent out to other people what they're doing.

And, you know, people are become much more cynical in the last few weeks of

ideas, partly because they know so much of this stuff is coming to the credit

markets that they don't have to buy everything.

And the question is whether people in the AI industry, as a wider thing, have

become too complacent about the idea that the credit markets will be there

and be supportive of them. And that's not necessarily always going

to be the case. I mean, I think that this is this vastly

from that point of view in the. There really has been a deluge, a bond

issuance hasn't it. I mean, we're talking about, um, what is

by some measures now, the hyperscalers are the biggest issuers of corporate

investment grade debt. And it used to be in the banks.

And the other point about hyperscalers is that before all this they were

they were basically running on their own cash generation, which was deemed as

being essentially impregnable balance sheets.

Um, and while it's not negative, they've started, you know, reason or not

started, they've now got a lot of debt. Um, and for example, Alphabet, its

latest quarter was its first negative free cash flow quarter ever since at

listed, um, it's just sort of sign that

fundamentally being changed by this. And then you throw in China, maybe

turning around and doing it all, oh, much, much cheaper.

And suddenly again, like, oh, wait a minute, your earnings are going to be

crushed. Sorry the earnings that that we're hoping

that you get may get crushed by us. Um, and one thing I thought was really

interesting, going back to the debt point

is you've involved in putting together this very complicated and now famous

and, uh, sort of local way, uh, chart, um, about how this is all being

financed. And one of the big things that's I think

slightly worrying people as well, is that a lot of the money seems to be

coming from the people who, you know, raised the money in the first place.

So, uh, like, almost like a form of vendor financing where you've got companies

that kind of make the chips paying the people who buy the chips, or who are

going to rent the data that use the chips.

Can you talk to us a bit more about about that and how that's, uh, kind of

panning out? Yeah.

Well, what happened like last year was, um, I was sitting there reading story

after story about these deals where companies were doing deals with other

companies in AI the universe. Nvidia has always been at the center of

this, but the likes of Google and Anthropic and OpenAI were doing all

these as well, and I literally couldn't keep up.

And so we just had the idea that maybe we need to do a visual here and kind of

show the levels of, um, circular deals that are happening.

And, you know, this became something of a bad word in the 1990s with fiber

optic. And when there was a big rollout of

that, lots of spending and or vendor financing, and there was capacity

sharing and various other things. And then a lot of those companies went

bust when the demand wasn't there at the end, having invested in all that money.

And of course, the irony, of course, is that, uh, long term the economy

benefited massively from that level of spending.

It's just the companies involved, but no way so many of them ended up going

bankrupt. Well, that's that's the infrastructure

story, isn't it? You know, you the railways are still

here and we still have trains, but a lot of companies that built them went bust

and seen with.com. And that happens with most of

technologies. And again that goes back to why

investors are being somewhat skeptical at the moment about that.

But these are deals that can create misaligned incentives, um, around things

like are you making the decision for in the best interests of the company?

Are you making it in the best interest of the company that invested in you and

is one of your biggest customers, etc.? And does that mean you're not devoting

resources to something else, when perhaps you should be?

It also raises the question of whether there might be misaligned values.

So yeah, the deal has been done out. Let's just say 7 billion valuation.

But if that was with somebody else might have been a 4 billion deal.

Yeah. Yeah.

You're also creating customers for you who are beholden by those, um,

contracts. And then you that that like essentially

creates regulatory risk as well. Where regulators may come along and go

actually we don't like that um, we're not sure whether that's the best deal

for the consumer. Um, and so, you know, all these risks

are emerging, or perhaps we didn't have before, but the biggest one, probably of

all, is that these circular deals can create a false impression of the man.

And you may think that all these companies are generating massive

revenues, but if it's all just moving around and sloshing around and if

something falls out of bed, then there could be wider implications and that

could become systemic. Yeah.

Well, that is that is what I was going to ask is obviously that that's kind of

what we worry about. Because the other thing I thought was

interesting this week is that a low KOSPI is falling bed.

Uh, the Nasdaq took a bit of a bump, but not a big one.

Um, it so far seems to be restricted largely to the AI and the tech sector.

And actually, plenty of other stocks are doing fine.

Um, equal weighted S&P is doing fine. There's an that's the S&P that's not

wholly invested in the tech sector. Um, and also the Footsie 100 is almost

back at a record high. Uh, the other they kind of laggered off

the global stock markets. Um, but it's this issue of how much does

this spread or could it spread beyond the tech sector?

Who else is involved in lending to these companies?

And if something did break down there? So are where are the contagion sort of.

Um. Yeah.

Vectors. Um,

the demand means that the AI industry has had to go to pretty much every

corner of the credit market. Yeah, hand out the ball and kind of say,

give us a few quid to fund what we need going forward.

And so you're seeing everything from direct lending, which is, uh, private

credit usually for infrastructure. So like just to build a data center.

Um, you're seeing investment grade, you're also seeing these neo clients

like CoreWeave, who I mentioned, who are some investment grade, sometimes not,

uh, not all of them, but some of them are some investment grade.

So that's a high yield market. The junk bonds that people might know,

um, they're in the structured credit markets, etc..

Uh, you know, these are basically taking bond payments that are due and slicing

them up by risk and selling them off to people.

And that reminds me, but hey, it's didn't something like that happened in

the mortgage market? Uh, was it 20 years ago at

We haven't got the kind of the CDO level or CDO squared level?

Well that's good. like where I would be concerned around

that lending in particular is around what in real estate is called

speculative lending, which means something else to many people.

But in real estate, it's basically building the stock before you have a

tenant. Yes.

And if you can get, uh, finance for that from a bank or from a private credit

lender that says something about bubble territory because they are taking a

complete risk that you are building this thing over the course of five years and

you will find somebody to occupy and if not or if the industry moves on and

technology moves on, you can end up with a very expensive elephant, uh, at the

end, which may not have much residual value for people.

If you build it, they will come. And it only works in the field of

dreams. Uh, yeah.

I mean, there can be a first mover thing where you can get away with it, but if

you're spending 5 billion credits building a data center, I certainly

would want to be more certain of that. But we are beginning to see elements of

that. And we're also beginning to see, uh,

terms being pushed within those deals for lending, uh, maybe are too generous

to the hyperscalers. That is the point of view of the credit

market. So you're starting to hear things like

after a certain amount of time, if the project is delayed, then the hype is

going to come back up. Uh, if you've invested five years and

spent, as I say, $5 billion in building a datacenter, and it gets a bit delayed

because let's just say somebody got stuck in the Straits of Hormuz at the

moment. That's a very big risk.

And so again, that's part of what's happening with the credit markets and

the pullback. Um, they're kind of reconsidering some

of the levels of risk that they're accepting at the moment.

I mean, from that point of view, this arguably as long as it's not already

going too far, maybe a good thing, as in maybe it gets to reign in the horses a bit

before it does go properly. Pear shaped.

Yeah. And I think it was always going to

happen. Um, and to be fair, like you look at

something like the KOSPI and, it's still up.

You know, many of the companies are set up over 100% here today and in some

cases where AI adjacent companies are up 300% for the year.

So, you know, and there was a natural moment for a pause anyway.

And while we are seeing in terms of the Chinese, um, evolution is that that was

always going to happen as well. Yeah.

And China's big advantage is it has cheap electricity.

So the tokenization is cheaper. So if you're a company that's also been

spending loads of money on AI and your staff, it turns out, are using it to

convert Excel files into PDFs rather than actually, you know, using a much

cheaper technology. For that, you're going to be scurrying

around for cheaper prices. Yeah.

And so, you know, China and the Chinese LLM's become a natural kind of

success story from that. Um,

and, you know, that was all to be expected, I think, within a certain

reason. And it's just all happened very quickly

and all at once. As it does these days in markets.

Um, there's a bit of panic in certain areas, particularly when it comes to

retail money, you know, and I have spent years writing about retail money, being

hot money and how people can panic. And to be fair, if I had leverage

several times, I'm looking at that and kind of gone, oh my God.

So it's understandable in a way. But, you know, if people are sensible

about how they invested some major opportunities there.

Yeah. I mean, yeah, my heart goes out to the

various, uh, kind of rookie retail investors in South Korea who are now

looking at some really nasty the losses. And I'm hoping that they were all young

enough to bounce back from it. Um, the only thanks very much for

listening was really helpful. The one other thing I was asking about

is on the debt side say, to flood the kind of the debt market, if you like.

And we have also seen a bit of equity issuance.

And I suppose the other thing I'm wondering about is how much can the

market take whenever we've had decades of de-equitization

So we see companies buying back their shares or, you know, getting bought off

the market. And now it's going to get quite

unusually kind of net equity issuance quite possibly this year.

I mean what does that say to you about what might happen regarding the tightness of money

overall, um, as in it's not going to be enough to go around for other hungry

mouths. So I think investors will be discerning.

I mean, you can tell you can tell an equity story that people

will follow. Uh, I mean, just think of it in AI how

many of these companies we had heard of three years ago, very few if any.

Um, and yeah, you know, that they may not have very much revenue, but people

are willing to bet on a deal. But, um, there's also the thing of like,

it can go wrong quite quickly, and SpaceX is probably a good example of

that. Not necessarily went wrong, but there

was so much hype. And now it's obviously down since its

IPO price, which was ambitious in the first place.

Um, I looked recently and this short interest on the stock was at near 40%.

So people are being very bearish on the future for it.

And that's probably the big question for investors at the moment.

Space X It's been partially an AI story with

data centres in space etc.. Um, is has uh and what happened to that

stock closed the IPO window for uh, yeah.

And uh, and that since then the focus would shift back to the equity markets,

uh, sorry to the credit markets and the credit markets.

Um, uh, sculpture, I think it was put an investor letter out recently and I said

they talked about how credit capacity is needed at its greatest extent, just as

people are pulling back. Um, and that's something you have to be

a bit fearful of, I think. Um, now, after we have taught these

things and two weeks later, it's the market has said, oh, oversold.

And it's all right back up, uh, and often exceeding the previous size,

obviously. Um, but at the moment it's definitely a

moment. Um, it's just, uh, the length at that

moment takes. Great.

Well, look, thanks a lot, Neil. I think that was really helpful, I hope.

Um, so the main point here was to try and explain to people what's going on.

And I think you've done that excellently.

And obviously, uh, the market can remain excitable for longer than anyone can.

I mean, solvent, especially if you're invested in a leveraged ETF.

Um, so just be careful out there. Thanks again Neil.

Thank you.