Director's Cut - Veteran money manager James Gubb: Oracle's shock downgrade signals the first crack in the AI bubble

Former hedge fund manager and CFA James Gubb argues the AI boom is real, but many investors are backing the wrong winners. In this interview with Alec Hogg, he explains why Oracle's credit downgrade could signal deeper cracks, why OpenAI and hyperscalers face mounting pressure, how China's AI advances are reshaping competition, and why energy - not AI models - could deliver the biggest long-term investment opportunity. Is this another dot-com bubble, or something even bigger?.Priced for perfection: A retired fund manager's warning on the AI boom.Listen here.Edited transcript of the interview.Alec Hogg (00:08)James Gubb spent his career as a chartered financial analyst, being paid to separate genuine value from a good story. We often hear that from people like Peter Fulloon: don't listen to the narrative; look at the reality.Yesterday, S&P Global downgraded Oracle's credit rating. It cited Oracle's exposure to OpenAI – the company behind ChatGPT – as one of the reasons. That's not a hypothetical in James Gubb's world.I say this because, on BizNews, we've republished an absolutely fabulous insight into what's happening with AI and how the AI value chain stacks up. James kindly allowed us to republish it and has now joined us to bring it all together and explain it in a little more detail.Before we kick off, James, I know the two of us both admire the book Engines That Move Markets, which I've recommended to the BizNews tribe on numerous occasions. It looks at manias, such as the one we're seeing in AI at the moment, over centuries.As I mentioned when I invited you to join us today, it's my go-to book for trying to understand what on earth is going on in the markets.Now, you're an expert in markets, and AI seems to be yet another one of these manias. Yet you're not short of everything – you're long on some things. It's more than just a confused mess, and you're going to have to help us through it.Can we start with that Oracle story? You talk about a first domino to drop. I'm wondering whether this could be the first domino.James Gubb (02:14)Yeah, Alec, it's good to be on your show.It might be.I've learned to be a little more humble over the recent past as this technological wave has come over us, because I was early in saying, "Watch out, guys. This looks a little frothy."So I've been relatively conservative with my positioning in the markets, and I've been wrong.This wave is bigger than I thought. A lot bigger than I thought.It's almost impossible, in an exponential world, to conceive of what things will look like in a year and a half, two years or even three years' time. Our brains find it very difficult to comprehend and understand exponential change.Coming from that confession, I have to be careful not to become too bearish, because what is happening on the technological front is enormously significant.I mean, this is the biggest thing to happen to mankind. It's really that big.So to stand in front of it and say the market is suggesting this wave will go on forever... I might be overstepping the mark.Let me go back a little and explain why I wrote the piece.I live in Plettenberg Bay. I've got beautiful weather, great surf, and the beach is always beckoning. But I suppose I have a restless mind.James Gubb (04:29)I approached this AI phenomenon thinking, "I'm going to try to understand it." The more I tried to understand it, the more rabbit holes I went down.I have a CFA background and I'm a generalist, having run hedge funds in the past. So I put on my analyst's hat and gave it a proper crack.I ended up writing a note that I sent to a couple of friends.The caveat is that I'm not an AI expert. I simply tried to understand it as best I could.The more I looked at it objectively, through my somewhat value-oriented way of viewing things, the more I felt that something just didn't fit.I looked at it through the lens of return on invested capital. The basic question is: how much money are you spending, and how much money do you expect to get back?When I looked at it in those simple terms, it just didn't add up.So I explored it a little further, and that's how this note came about.I suppose that's the introduction to all of this, Alec.Alec Hogg (06:03)It really is a reminder, I guess, for those of us who've been around a while, of what happened during the dot-com boom and then the dot-com crash. Of course, we also had the global financial crisis in 2008.You make an interesting comparison between those two periods.This doesn't seem to be a classic "wham-bam" crash scenario. Could you explain whether you see this more as a global financial crisis-type issue rather than something like the dot-com collapse?James Gubb (06:32)We're simply trying to look at what capital expenditure has been over history.I was actually trying to work out how big this capex spend is relative to the construction of the pyramids, because I genuinely think it goes back that far.In commercial history, the dot-com boom was previously the biggest.Have you seen the size of some of these data centres being built? They are utterly extraordinary.And it's not just one. There are plans for hundreds of them.Yes, there's a lot of red tape to get through, but we're talking about something on the scale of the pyramids.Anyway, coming back to the dot-com comparison...Alec Hogg (07:23)Just explain that, James.For those who haven't seen the scale of these data centres, can you describe just how big they really are?James Gubb (07:34)I can't give an exact comparison – just Google it. The base of these data centres is about the size of the Giza pyramids.Alec Hogg (07:47)So we're talking about something absolutely enormous.Of course, you've got Elon Musk on the other side saying he's going to build data centres in space because it's supposedly much more efficient, and so on. There are all kinds of different angles to this.But when you compare them with the pyramids, I presume you're referring to the fact that, at that stage in mankind's development, they were off the charts – bigger than anything else. And it's a similar situation here.James Gubb (08:11)Yeah.There's been plenty of debate about exactly when the pyramids were built, but let's assume around 6,000 or 7,000 years ago.I'm simply trying to give people a sense of the scale and context of what's happening today.It's massive. Absolutely massive.It dwarfs the capital expenditure of the dot-com era on fibre infrastructure. On a GDP-adjusted basis, this spend is around four times larger.And we all remember what happened during the dot-com boom. They overbuilt, just as they overbuilt the railways.History doesn't suggest that this kind of capital expenditure is rewarded handsomely for those who build it. It's generally the people who buy the assets later – after the original builders have gone bankrupt – who end up making the money.Alec Hogg (09:09)Is it a bit like property cycles?Time and again, people invest heavily right at the peak of the property cycle, and then spend years trying to fill the offices or industrial sites they've built.So we're going through something we've seen before. From your perspective, having done all this work, the winners are likely to be the providers of the picks and shovels, while the losers are everyone else.James Gubb (09:24)One hundred per cent.James Gubb (09:38)The most glaring example of a hyperscaler that hasn't participated in this capex spending spree is Apple.It doesn't have a large language model. It doesn't have an AI model of any significance. Siri has become a bit of a joke.I actually think Apple has been very smart by waiting to see how this all plays out.It seems to have recognised that this could potentially be a bubble.It will also be the company sitting on enormous amounts of cash because it's still generating very strong free cash flow, while all of its competitors are spending virtually everything they earn.Most of them aren't free cash flow positive at the moment.That puts Apple in a wonderful position should any of its competitors stumble over the next few years.Alec Hogg (10:48)At the other end of the spectrum has to be Meta, surely.Meta was nowhere in AI, and now it's spending more aggressively than almost anybody else.James Gubb (10:56)Exactly.If you value these hyperscalers on the basis that they're asset-light businesses generating lots of free cash flow, they deserve a certain valuation multiple.But once they've joined the AI race, they've become asset-heavy businesses.They're building enormous data centres, and all of their free cash flow is being diverted into capital expenditure.That's no longer an asset-light business.An asset-heavy company should trade on a different – and, I'd argue, lower – valuation multiple than it currently does.Alec Hogg (11:47)Going back to the property analogy, which I think many people will understand, Meta is effectively building a huge number of office towers into what could become an oversupplied market.Do you think there are going to be too many data centres?James Gubb (12:05)No, I don't think there'll be too many data centres.I think they'll all eventually be used.The issue is that the returns earned by the companies building them first won't be sufficient.Because they're generating negative free cash flow and poor returns on invested capital, they become vulnerable to being taken over.Alec Hogg (12:32)That's capitalism at work, isn't it?James Gubb (12:34)Exactly.In the end, those data centres will be fully utilised, just as the infrastructure built during the dot-com boom eventually was.The difference is that they'll be owned by different people.The same thing happened with the railways.Alec Hogg (12:48)Perhaps an even better property analogy is golf estates.When there's a boom in building golf estates, the developer who builds them first often loses money. The second owner might break even, but the third owner can make a fortune.The point you made about Apple is very interesting because it'll be sitting on a mountain of cash.Who else do you think is positioning themselves well for this extraordinary period of history?James Gubb (13:24)There are lots of different ways to invest in AI.But, if I'm honest, I'd take a slightly different approach.Where would I put my money?I'd put it into energy.The cost of compute, thanks to the enormous competition we're seeing, is ultimately going to fall towards the cost of energy plus the amortisation – or depreciation – of silicon.That means the real bottleneck eventually becomes energy supply.That's where I'm investing because many energy stocks are still trading on derisory valuations.I wouldn't necessarily chase today's AI hype.You need to be a genuine specialist to pick winners there.Perhaps something like OpenRouter is interesting. CoreWeave is another example.Alec Hogg (14:47)When you say energy...James Gubb (14:51)Energy is where I'm putting my money.Alec Hogg (14:58)Just elaborate on that, James.From a South African perspective, we think of Sasol as an energy company.Would that qualify?James Gubb (15:05)Yeah. I'm thinking more along the lines of nuclear energy.But, more broadly, just traditional energy suppliers.If this exponential trend continues over the next 10 to 15 years, demand from these data centres will eventually become significant.Right now, it's almost immaterial. Over the next three to five years, it remains largely immaterial.But if this growth continues into years seven, eight, 10 and 12, you'll suddenly see a material increase in electricity demand driven purely by compute requirements.Remember, building a new power station – particularly a nuclear one, which still seems to me to be the greenest option – takes 10, 12 or even 13 years.That means there's likely to be a massive gap between demand and available supply.I think that's where this AI wave ultimately leads.One of the smartest ways to invest is through that bottleneck.Alec Hogg (16:37)Coming back to the broader AI picture, when DeepSeek arrived there was a real jolt on Wall Street.Now we've got Moonshot AI, which is apparently even better than DeepSeek.Is this another DeepSeek moment, for want of a better term?James Gubb (16:55)Way better.James Gubb (17:02)Yes.It's the arrival of something called Kimi K3.It's an open-weight model – not open source, but open weight.The developers have promised to release the model weights at the end of this month, which really puts the cat among the pigeons.The level of intelligence it delivers is comparable to the frontier models from Anthropic and OpenAI.The cost advantage, however, is dramatic.That seriously threatens companies like OpenAI and Anthropic.Remember, they're still going to have to come back to investors with their caps in hand and ask for more money.That's when the rubber really meets the road.I don't think, with this level of competition, they'll be able to raise capital at the valuations they've achieved in the past.It's a fairly complex argument, and I explain it in more detail in the note I wrote.Alec Hogg (18:37)Just so I understand...On the US side – Silicon Valley – you've got Claude from Anthropic, which many people regard as superior to ChatGPT, the pioneer that introduced generative AI to the mass market.Then you've got Google's Gemini, along with several others.Meanwhile, China hasn't been standing still.We've seen DeepSeek, and now Moonshot, which is supposedly even better.All of that must make life much more difficult for companies like Anthropic and OpenAI, both of which are expected to raise fresh capital in the near future.James Gubb (19:24)Correct.The Chinese certainly haven't been sleeping.Yes, they've been constrained by limited access to top-end AI chips, but they've simply become smarter.This whole industry gets eroded from the bottom.I explain this in some detail in my note.The latest frontier products are continually undermined by previous generations, which don't disappear.Those earlier models still perform extremely useful work for businesses, often at a tenth – or even a hundredth – of the cost.I don't think the current economics are sustainable, particularly when China also enjoys an energy advantage over the United States. That's more of a long-term issue, but it's still important.The Chinese are pouring money into AI and, rather ironically, have become champions of open-weight models by releasing them to the world, while the Americans seem intent on restricting them.In fact, I heard this morning there are rumours that the US government is considering banning Chinese open-weight models.If that's true, I think it's one of the most misguided ideas I've ever heard.America would be moving in exactly the wrong direction.One other thing, Alec.I saw this only last night.Someone asked: "At what point will frontier AI models be released at a rate of one every day?"The answer was January.In roughly seven months' time, we could be seeing top-end frontier models released at a cadence of one every single day.Alec Hogg (22:13)That's extraordinary.As you said earlier, exponential growth is incredibly difficult for us to grasp because we naturally think in linear terms.But based on what you've just said, if I were a banker lending money to companies like Anthropic – whose Claude model currently sits at the frontier – I'd be thinking a lot more carefully about extending credit, given the wave of competition that's coming.We've already seen Microsoft restructure its relationship with OpenAI.Can you explain what's happening there?It's almost as though Microsoft is saying, "Hang on, we're not quite so sure anymore."James Gubb (22:57)That's exactly right.They're conserving their cash as much as possible.Like Apple, they're trying to sidestep the spending obligations associated with these hyperscalers.They're gradually stepping back.I'm not entirely sure what Microsoft's long-term strategy will be, but they are clearly trying to reduce their potential exposure to funding ever-larger language models.Alec Hogg (23:40)That's a significant signal.If anyone has deep insight into OpenAI, it's Microsoft.They've backed the company from day one.If they're now beginning to step away, despite having access to far more information than the rest of us, that sends a very powerful message.James Gubb (23:56)Agreed.Alec Hogg (23:58)What about the postponement of the OpenAI listing? It was supposed to happen already, but it's apparently been pushed back from this year to next year.What does that tell us?James Gubb (24:14)Well, it's fairly obvious.They're not able to raise capital at a valuation that makes commercial sense to them.It's as simple as that.Alec Hogg (24:32)It's simple for you because you've spent your career analysing these things.But for the average person who thinks, "I'd quite like to buy shares in OpenAI because I use ChatGPT," what does this actually mean?It suggests that, while you may eventually get that opportunity, the people who know the business best are stepping back and saying, "We're not quite so sure anymore."Perhaps that ties in with what you were saying earlier about the pace of development in frontier AI models.James Gubb (25:05)Agreed.It's actually quite difficult to get short exposure to OpenAI because it's not listed.You can gain some indirect exposure through SoftBank, for example, but it's not straightforward.You have to look at private equity firms with stakes in OpenAI, and even then you don't get clean exposure.There are other ways of doing it, but for the average investor I'd simply be standing aside.Alec Hogg (25:46)Explain Jevons paradox.James Gubb (25:50)Jevons paradox simply says that when something becomes cheaper, demand expands to fill the gap.As things become easier and cheaper to use, people use a lot more of them.It's classic economics.James Gubb (26:17)The idea comes from 19th-century coal economics.Making a resource cheaper to use can actually increase total spending on it because demand grows by more than the price falls.Alec Hogg (26:31)A South African example would probably be mobile data.As broadband became more widely available, prices fell dramatically.Those of us who were paying for broadband 20 years ago will remember just how expensive it was.James, what about Elon Musk and SpaceX?What's your view?James Gubb (27:01)I'm a fan of Elon Musk.He's a genius among geniuses.But he also knows how to tell a story.His "vapourware" is legendary. His promises are extraordinary.Ultimately, though, the proof is in the pudding.SpaceX's valuation has roughly halved since its peak. That's a significant decline, and it's still not part of the S&P 500.In my view, it's overpriced.In fact, this was one of the things that prompted me to write my note.I was chatting to a friend of mine, one of the Diggle brothers, who'd written a short note on SpaceX. I agreed with him, and we started drawing parallels with the wider AI industry.SpaceX is an amazing company, but the only consistently profitable part of the business is Starlink.That's a relatively small part of the overall asset base.It's the only division with meaningful earning potential at the moment.The rest is, quite literally, pie in the sky.Alec Hogg (28:49)Elon Musk is, as you say, famous for selling a compelling story.For those outside investment markets, that's what "vapourware" really means.The latest story is that data centres will become too expensive to operate on Earth, while they'll be easier and cheaper to power in space. He's got the rockets, so he can take them there.Is that just more vapourware, or could there be something to it?James Gubb (29:18)It's incredibly difficult to do anything in space.Having said that, Elon has his fingers in an extraordinary number of remarkable projects, so I wouldn't dismiss him.But putting data centres into orbit would require a tremendous amount of effort.Why not simply put them on the ocean?You've got an enormous natural cooling system right there.My point is that you don't necessarily need to go into space. We've got vast amounts of real estate in the world's oceans.So don't discount Elon.But is SpaceX overpriced?Yes.Personally, I'd expect it to fall by at least another 50% before I'd even begin looking at it.I love his ideas.The problem is how much investors are being asked to pay for them.Alec Hogg (30:32)Is there anyone else in history quite like him?Looking back at previous bubbles... Edison, perhaps?James Gubb (30:36)Thomas Edison is probably the closest comparison.If I remember correctly, he also overreached financially at one stage.But there's really nobody quite like Elon Musk.He's an extraordinary individual.I simply think investors are paying far too much for his ideas.There's a lot of fluff in the valuation.Alec Hogg (31:16)You've explained why you're investing in energy rather than AI itself.What are you avoiding?What parts of the market would you steer clear of?James Gubb (31:34)I think this is a stock-picker's market, particularly within AI.The investment cycles are becoming much shorter.As technology moves exponentially, time effectively compresses, and market cycles compress with it.Investors are going to have to be extraordinarily nimble.Technology can help identify outperformers and underperformers, but you still need to move quickly.I hesitate to name specific companies I'd avoid.Instead, I generally protect myself by buying put options on markets that I think are expensive.I buy the things I like and then insure the portfolio with relatively inexpensive out-of-the-money puts while implied volatility remains low.The sectors I do like include biotechnology and resources, particularly where political fragmentation is creating supply bottlenecks.Those supply chains are becoming strategically important to governments.That's where I'd rather have my money.Overall, I prefer protecting myself with broad market index puts.Alec Hogg (33:38)Many members of the BizNews community own stocks like Amazon, Alphabet, Apple and Meta.You've made a bullish case for Apple and a more cautious case for Meta.Would you personally own Tesla, Amazon or Alphabet?James Gubb (34:01)I'm getting into individual stock recommendations there.These days my own investing is becoming increasingly index-oriented because I simply don't have the time or the inclination to analyse individual companies properly.I do have indirect exposure to Elon Musk through the Contrarius fund, run by former Allan Gray manager Milton Hall, who's now based in Australia.He's very bullish on Elon.So that's how I gain exposure.I'd rather not comment on specific shares because there are people far better qualified than I am to analyse individual companies in detail.My focus is much broader than that.Alec Hogg (35:00)Fair enough.I was really asking from a Charlie Munger perspective of inversion — not what you would buy, but what you would avoid.You've already given us some useful guidance.What are you doing these days?Many people remember you from your investment management career — and, of course, for that famous "middle finger" moment.What are you working on now?James Gubb (35:29)I've always had a restless mind.I tend to chase ideas and explore them.With AI-assisted research, it's now possible to develop concepts that would have been incredibly difficult only a short while ago.One project I'm working on came from conversations with younger people in Plettenberg Bay and Cape Town asking me, "What should I do with my money?"Traditionally, they go to an asset manager who puts them into a balanced fund — typically something close to a 60/40 split between equities and bonds.That model has existed for more than a century.I think it's time for a serious rethink.The reason is that equities and bonds are increasingly behaving in the same way because of the world's enormous debt burden.During periods of stress, those asset classes become more highly correlated.So I believe the traditional 60/40 allocation needs to be reimagined.I came up with an idea — literally in the middle of the night.Imagine a character called Bob standing inside a cube.Each face of the cube represents a different asset class.The floor represents the money you can't invest — your essential living expenses: food, clothing, shelter, raising your children and everything else you need simply to live.Once you've deducted that, everything remaining becomes your investable wealth.The other five sides of the cube represent different categories.Behind you is tangible "stuff": property, wine, classic watches, cars, natural resources and other physical assets.One side is equities.Another is debt.Above you sits resilience — anti-fragile assets such as gold and, believe it or not, Bitcoin and Ethereum.The final side represents hybrids: hedge funds, commodity trading advisers, convertible debt and other alternative investments.Each deserves its own allocation.The cube itself sits inside a sphere.That sphere represents the macroeconomic environment, which changes the relative importance of each asset class over time.It's a visual way of helping ordinary investors think differently about portfolio construction.A younger colleague of mine, Adam Westhazen, and I are developing an app around this concept.Hopefully it'll be ready by the end of the year.Coming from an artistic background, I've always found visual models useful for explaining very complex ideas.That's what occupies my time these days — along with painting and making marks on canvases.Alec Hogg (40:35)James Gubb, it's always a pleasure speaking with you.I look forward to catching up again when the app is ready.It sounds like a fascinating new way of thinking about investing.At BizNews, we firmly believe our community should take greater responsibility for their own financial futures rather than simply handing that responsibility to traditional fund managers.What you've described sounds like it could help people do exactly that.James Gubb (41:06)That's exactly the idea.This AI wave should be embraced.Today, you can describe almost any idea to an AI model, and through agentic AI you can build things that would have seemed impossible just six months ago.It's an extraordinary world we live in.Alec Hogg (41:40)It certainly is.Just be careful not to get swept up in the hype when investing.James Gubb is a former top-rated South African money manager.I'm Alec Hogg for BizNews.

Director's Cut - Veteran money manager James Gubb: Oracle's shock downgrade signals the first crack in the AI bubble

Related Stories

BizNews Edge: James Gubb cuts through the AI noise — how top investors are hedging the boom

Director's Cut — Chris Pappas reverses course: the real reason he's running for uMngeni mayor again

BizNews Edge: Chris Pappas — Why the DA's only KZN municipality won't lose its mayor after all

Director's Cut — Dr Mark Burke: What is behind the governance crisis at the R3.5 trillion PIC