There’s a lot of hype in the media about a potential AI bubble. What is an AI bubble and should we be afraid of it? Milford Senior Investment Analyst Andrew Curtayne talks with Ryan Bridge about global investment into AI, key factors that determine a ‘bubble’, and what it all means for us in New Zealand.

Click here to download the MP3 file or listen to the podcast on your favourite platform:

Bridge talks Business: 11 November 2025
Episode Transcript

Ryan Bridge
Kia ora and welcome to Episode 56 of Bridge talks Business with Milford. You’ve heard no doubt the reporting about this AI bubble. This week we’re separating fact from fiction. Andrew Curtayne is in the hot seat from Milford. But first, here’s your top five business bits.

1. Overall, the US employment market looks to be in balance. Reports last week of layoffs going up, offset by the fact it appears that jobs are also being added to the economy. Markets are fuelling up on less reliable data because, well, the proper ones haven’t been available due to the government shutdown.
2. Business surveys in the US showed improving economic conditions, strong services sector, but inflation still a worry with pricing metrics ticking up a bit.
3. The tariff court case continues in the Supreme Court in the US. Aggressive questioning by some of the judges had investors reducing their expectations that tariffs would be ruled legal in the end. That said, if they are ruled illegal, Trump still has plenty of other levers to pull.
4. Signs the shutdown might finally be coming to an end for now, lawmakers have taken the first step towards reopening, but it’s only a stopgap measure until January. Share markets rallied on the news.
5. This week, relatively quiet – company earnings reports have largely concluded, and we’re not due to get any US data even if the government reopens.

Alright. It’s business time. Let’s speak to Senior Investment Analyst at Milford, Andrew Curtayne, about the AI situation. Just a reminder, this segment is informational only and should not be considered financial advice. Andrew, welcome back.

Andrew Curtayne
Thanks Ryan. I’m glad to be here.

Ryan Bridge
Good to see you. Now, the AI bubble situation – that’s potentially a bubble situation. Let’s just reverse a little second here. How did we get to this point? The AI stocks have been leading the stock indices. What’s been driving that demand for AI?

Andrew Curtayne
Well it all comes back to when ChatGPT released what is known as the version three of its model. That came out sort of early 2023, late 2022. And it was sort of the defining moment where AI went from kind of like this machine learning thing that people thought was a bit futuristic and would come at some point, to all of a sudden wow we’re here. There’s this chat bot which has got a lot of flexibility to answer questions about whatever we ask and actually provide pretty good answers.

So that was kind of the start moment of what we might call the recent AI boom. And from then that’s been all about competition between all the companies that we call LLM companies, which are Large Language Model companies. So that’s, OpenAI that makes ChatGPT, that’s Gemini, llama by Meta. And all these companies competing to prove the capability of their models. And it started what is essentially kind of like an arms race. That’s a race between the biggest companies in the world to get AI leadership. So, to have the best models and the best products, that can go serve clients and serve customers like ourselves. And it’s not even just an arms race with companies. It’s an arms race between nations. Both China and the US want to win at AI. It could give them a competitive advantage, both economically but even militarily, if they win in this AI race. So, it’s been a hugely fast paced sort of two – two and half years, since ChatGPT3 came along. It still feels like there’s a very long way to go and it’s going to be very unpredictable exactly where this technology goes.

Ryan Bridge
Andrew, you’re seeing a bit of a pullback just in the last couple of weeks in AI stocks. Is that something we need to be worried about?

Andrew Curtayne
Yeah. I think every time the market has a little bit of pullback, you need to have a look and understand why. And so we have seen that the S&P 500 has come off about two and a half, 3%. It’s not a huge pullback by traditional standards. But we have seen AI stocks come off more like 10 to 15%, which is probably the largest pullback we’ve seen for about six months.
And I don’t think we need to be too worried about it, but it is showing that things are very extended. If I look at the performance in the AI spectrum, since the market bottomed back in 2023, AI semi stocks are up about 330%. Now, this compares to the overall market that’s up about 75%.

So it’s been a huge rally. And so you put that in context 10 or 15% pullback when they’ve rallied so much – it’s not particularly worrying. I don’t think there’s anything fundamentally negative that’s coming out on the AI spectrum. In fact I think it’s all been quite positive. So I think that’s more of a sign that the market’s a little bit exhausted, a little bit long.
It gets a little bit hard for continued positive news to push the market back. Also you had a few other things going on in the market. You’ve had these government shutdowns in the US which is starting to make people a little bit more concerned. Although it looks like just recently, last night, we may get some resolution to that. And we actually saw markets bounce back a little bit this morning. It went up about 2% and AI stocks led the rally back.

Ryan Bridge
Do people want to pay for it? Because that’s the problem you’ve got, isn’t it? And that’s when you start to get worried. You know, people talk about a potential for a bubble is when the earnings aren’t quite keeping up with the market cap that’s been put on them.

Andrew Curtayne
And that’s a very good question, Ryan. And this is what is fuelling some of the scepticism. Some of the AI bubble discussion. Is there enough revenues, and revenues being generated by the AI companies that can support the hundreds of billions of dollars of CapEx that’s going in? So this year alone, we’re talking about $600 billion of CapEx being spent on AI infrastructure. And when I mean infrastructure, I’m talking about AI data centre. So these are the huge buildings that are full of the super advanced computing chips called GPUs, which essentially run the LLM models. At the moment, in my estimate, we’re probably got around about 50 to $70 billion of AI revenues coming. So OpenAI, which owns ChatGPT, is probably making around about 15 to $20 billion of revenue at the moment.
And you compare that to the $600 billion of CapEx alone this year, it’s just not enough revenue. And if you roll forward, where we think CapEx could go on the next few years, we think it could be well over $1 trillion per year by 2028. And if you run the math, how much AI revenue do you need to support a good return on investment on that CapEx? It’s probably about $2.5 trillion. Now, we’re at $50 to $70 billion today. But coming back to your question. Do people want to pay? Well, there are signs of people wanting to pay. ChatGPT has got 800 million users at the moment. And if you compare that to other technology products, that’s the fastest scaling user subscription you’ve ever seen. Only around about 5% or so of those people are paying for it – willing to pay the monthly. That’s probably quite heavy in people that work in information-type jobs, white collar workforces who are looking for information to help in their workplaces. So there’s not quite the conversion rate you’d want to see to justify the CapEx at this point.

Ryan Bridge
The two trillion number that you’re throwing out there, how realistic is that we could get to that one day? That there will be a demand? And maybe it’s not a demand from you or I wanting to search how to make a pancake, but businesses might buy the subscription so that we can use it for work, that kind of thing.

Andrew Curtayne
Yeah, exactly. And that’s certainly what companies are OpenAI targeting, both at the consumer with a subscription, at the businesses. I think in a short period, if we’re looking over the next 4 or 5 years, I’m highly sceptical we’re going to get anywhere close to that number. So, even OpenAI’s own forecasts – they get leaked periodically to the market – they’re talking about doing 200 billion of revenue by the end of this decade. So nowhere near the 2 – 2.5 trillion that you probably require.

Ryan Bridge
So what does that mean for the current market valuations? What does that mean for this bubble?

Andrew Curtayne
So it creates some level of scepticism, right. Ultimately if the revenue is not going to be achieved, or the revenue that we think needs to be achieved to justify investment, then you will have a significant fall back and the prices of stocks that are linked to this. I mean, you can’t go in and spend $1 trillion in CapEx and not get the revenue to support that.
The reason you’re still seeing the market generally move higher, or you’re still seeing strong prices in AI-linked stocks, is people think they’re willing to give time for the technology to develop. If you go back through a lot of technology cycles in history, they tend to under deliver in the short term. The technology’s not quite as good as people hoped it would have been, or as the speculators were hyping it up to be in the short term. But they overdeliver in the long term. We go back and ook at things like when the first computers that were invented in the 1950s or 60s. It wasn’t really until the late sort of 1980s where everyone started getting personal computers, and it took 20 or 30 years of that technology to develop.
Even if you look at some of the technology around, like Microsoft Office and when Excel first came along, it took a long time for this to develop. If AI follows that same pathway, it could be a situation that yeah we’re spending a little bit heavily at the front end, but everyone’s wanting to make sure they are going to be a leading market position. And over the next 10, 15, 20 years, you start to see AI grow into its own.

Ryan Bridge
In the short term. If we have a correction, if there’s to be a correction, some sort of bubble bursting -could that bring the rest of the stock market down? How exposed is New Zealand to it? You know Milford’s investing in AI. What’s the exposure there?

Andrew Curtayne
Yeah, absolutely. If you saw the AI bubble pop, let’s say. Or let’s say there is a bubble – we’re not certain it’s a bubble – hindsight would tell. But let’s say there is a bubble. And it does get to a point where everyone gets sceptical on AI working. And then the big companies go ‘oh look we’ve probably been spending a bit much. Let’s pull back’. I think that’s going to be very negative for the stock market. By some measures, I think JP Morgan came out with a paper about a month ago where they looked over the last couple of years. 80% of the earnings growth in the S&P 500 has been driven by AI-linked stocks.
And yeah, this is a pretty broad term with AI-linked stocks. Basically they got all the big mega-cap tech stocks – Google, Microsoft, Meta, Amazon, Nvidia – all linked to it. But it is showing you there’s a real concentration of both earnings growth and by stock price performance in the AI stocks. So if they all say ‘oh AI doesn’t work’, then you’re going to have a pretty material pullback.
I’m not seeing it being like Dotcom levels. I think the Dotcom period which was about year 2000, when that bubble popped the Nasdaq fell over 75 percent or so. So I don’t think we’re seeing that sort of level of hype in the market yet. And I don’t think we’ll see that pullback. If you compare today, the companies that are spending the most on AI, that’s like your Microsoft, Amazons, Googles, Metas, they’re incredibly profitable companies. They’re trading at valuations that aren’t that unreasonable. Yeah they’re a little bit expensive vs history, but quite reasonable. Whereas in the 2000 period you had companies with no earnings that were spending a lot of money leveraging to spend it all, they didn’t generate much cash flow. So I think the fallout as we sit today would have been a lot worse in 2000 than what we’d likely see today.

Ryan Bridge
So what is the investing strategy then? Because, as you say, you don’t want to miss out on the potential long term benefits, but also you don’t want to get stung by investing in certain stocks – maybe. Or maybe it’s too many eggs in one basket – an AI stock that turns out actually that wasn’t the one.

Andrew Curtayne
Yeah. Good question. Firstly, it’s diversification. Don’t put all your eggs into one or two stocks. Have a spread of the companies you invest in. How we look at AI, we put the companies in different buckets. So you’ve got the companies that benefit from CapEx. So the capital expenditure is what Microsoft, Amazon, Meta, these companies are spending, give or take 100 billion plus on capital expenditure, a year. That money is going into a few places. The first and biggest beneficiary is the companies that are producing the computer chips. So we call them semiconductor companies. The biggest ones are Nvidia, which produces the leading GPU or AI chip at the moment. But there are other companies like Broadcom, AMD that produce competing chips. There’s a company called Taiwan Semiconductor Manufacturing Facility, TSMC. That company actually manufactures all the chips for Nvidia. So Nvidia designs the chip but they manufacture it. So that’s an area at the moment – while you’re seeing AI spending going higher and higher, all the CEOs are saying they want to spend more – we see that area is still relatively attractive because they’re getting the revenue today. We don’t have to predict about AI revenues in five years. We’re seeing it today right behind them. And at the moment we’re still seeing higher expectations of CapEx. So, that area we can play, you know, we don’t want to be all in on that area because that area probably has the most downside risk to earnings if AI spending stops. So we need to be careful.
So then what do we do? We diversify to some companies which we think are a bit more balanced, that could be longer term winners, not as exposed to that short-term risk that like a Nvidia would be. So companies in that bucket include Microsoft, Google, Amazon. So these are the companies that have the AI clouds or the clouds. When people want to use a GPU to train a model, to provide the answers of models – they pay Microsoft or Amazon or Google to do this. And so these companies have sort of a good mix of getting some AI benefit today because their revenues are growing quite fast as people use more GPUs, but they also have the potential to benefit longer term as the AI software comes up, gets developed and improves. So, that’s the second key area.
And then thirdly, we diversify into companies that are getting benefits from AI. So companies that are lowering costs. So these companies are able to save call centre costs, the saving cost for AI coders. And I would say people that are using developers and they can use AI to save costs there, and so they are more of the longer term beneficiaries So the share prices rise as much today, but they could rise further – the biggest beneficiaries over a 5 to 10 year period.

Ryan Bridge
Fascinating. I knew you’d have all the good information. And because there’s kind of this media frenzy. This is what we do in the media, we sort of freak out and we, run a headline about a bubble. And it does well, so we do ten more. And before you know it, it’s like a self-fulfilling prophecy. But what is the last thing you used AI to search for?

Andrew Curtayne
Let me think off the top of my head. This is going to sound really geeky, but I’ve been searching about how custom accelerator chips are designed by Broadcom and how their contract works with Google. I’ll give you a more practical example. Running shoes – I’m training for a running race, and I used ChatGPT to compare the best off-road trail running shoes. It provided fantastic data.

Ryan Bridge
And better advice than you’d get from a human in a store, you reckon?

Andrew Curtayne
Absolutely yeah. I think it’s a way better thing, because it’s going to all the specialist websites. Previously you used Google – you’d go to one specialist website and try to find an article. This will go compare them all. But you know, where I stop short is I then go to Google to go to the actual underlying product website to buy it. What ChatGPT needs to do at some point is keep you there. So do the search and then get you to transact through the website. And they can take like a cut of that transaction fee. So I think that’s something which will develop over the next couple of years.

Ryan Bridge
They need to put you in charge. They’d have no problems with revenue.

Andrew Curtayne
I did miss it – you asked about New Zealand before and how it impacts New Zealand. I forgot to answer that question.

Ryan Bridge
How exposed are we?

Andrew Curtayne
We’re in quite a good position because we’re not actually spending all the money on CapEx. So it’s all being spent by the big companies all over the world like Microsoft, Amazon etc. New Zealand’s not really spending the money, but we benefit from it. So if all these AI tools come out like we’re just using the examples that I’ve just given like me using it a personally, we are able to benefit on that without having to really pay much for it.
Secondly, from a stock market perspective, our stock hasn’t got too many AI stocks in it. So if you have an AI bubble sort of pop and pullback, we’d probably outperform. But an interesting area I think, which isn’t a great sign for young people, is you are starting to see graduate levels of hiring go down because you’re starting to see AI as being a way that it can replace a little bit of what graduates do. You’re sort of your lower-skilled staff, the more repetitive or autonomous work. And so I think over the next few years, it’s going to be a little bit difficult for some grads coming out. Maybe there’s not going to be as many jobs being hired from accounting firms like PwC or legal firms. So that’s a little bit of a concern.

Ryan Bridge
Yeah. Interesting stuff. Andrew, good to see you as always. Thanks for coming on.

Andrew Curtayne
Thanks Ryan, appreciate it, cheers.

Ryan Bridge
That was Andrew Curtayne. He’s a Senior Investment Analyst at Milford talking to us about the potential for an AI bubble and where to here for the stocks. You can like, follow and subscribe this podcast wherever you like to listen. Thanks so much for doing so. Until next week, don’t forget to invest in yourselves.

Missed previous episode? Don’t worry! Click here to catch up now.