When will the AI boom turn to bust? – Fat Tail Daily

Hi, Greg Canavan here, Investment Director at Fat Tail Investment Research.

Productivity crisis be damned!

As a result, I’m filling in for Bill Bonner today.

I won’t be as eloquent or as philosophical, but hopefully you’ll get some insights…

Last week, our team of analysts spent the day together discussing the longer-term implications of AI on the economy and financial markets.

If you’re not thinking about this stuff, it will hurt your portfolio in the years to come.

I’m not just talking about the obvious stuff like computer chips and data centres. It’s more about the second-order effects.

Will mass job losses result from widespread adoption of AI tools?

What will the effect of that be on house prices (banks) and household consumption (consumer discretionary stocks)?

We don’t know. We just know it’s going to be a big deal. And it’s something we’re thinking very hard about.

Right now, though, the market isn’t showing any signs of thinking about it. AI presents no risks, only opportunities.

That’s because there is a huge amount of money going into the sector.

AI investment boom continues

The hyperscalers — Amazon (AWS), Microsoft, Google, and Meta are expected to spend over US$345–350 billion in 2025 on data centres, accounting for a 50%-plus year-on-year increase compared to 2024

Their combined capital investment is expected to exceed US$1 trillion over 2025–2027.

McKinsey forecasts suggest that by 2030, nearly US$7 trillion will be spent on data centre infrastructure, including major investments in advanced AI hardware, facilities, and supporting grid infrastructure.

No wonder the market doesn’t want to go down!

Now, US$7 trillion in investment over five years sounds impressive. But one of the things we discussed last week was the fact that energy is a major constraint around these global data centre rollout plans.

Back in April, the International Energy Agency released a report titled ‘Energy and AI’. It projects that data centre consumption globally will grow to 945TWh per year by 2030, from 415TWh in 2024.

That’s a massive increase in a short amount of time.

The US will account for the largest increase, followed closely by China. Together, the countries represent nearly 80 per cent of projected global growth.

That makes sense given the two countries are locked in a race for AI supremacy. That’s why they’re both so focused on keeping energy costs low.

In Australia, meanwhile, crazy Chris Bowen is doubling down on our renewables commitment. With high energy costs, Australia will be an AI backwater.

The bottom line is that the age of AI will require all forms of energy. Renewables, batteries, coal, gas and nuclear will all grow significantly in the next five years to support the rollout.

According to US real estate company JLL, approximately 10 GW of new data centre capacity is under construction worldwide.

This represents enough electricity demand to power several medium-sized countries!

My guess is that in the next few years this AI race will bump up against electricity grid constraints. There is a limit to how much and how quickly you can get new capacity through the grid.

What’s the return on investment?

It will also likely encounter its own capacity constraints. At what point will the market question the return on each incremental dollar of investment?

Initially, the returns on this investment seem attractive. That’s because it has allowed the hyperscalers to improve efficiency and reduce their workforce.

The result, according to this report, will be a loss of 235,000 tech jobs globally in 2025.

But the tech sector can only sustain so many job losses. Soon, we’ll be looking for something else to provide the return on incremental investment.

It is at this point that the market might start asking questions.

The thing is, we’re in an AI race. Companies are all building capacity because not building it means you’re out of the race. And no one wants to be out of the race.

So everyone builds capacity, whether it’s actually needed or not.

This leads to overcapacity and declining returns on investment.

When will this start to show up? As always, I have no idea. If I had to give you a guess, I’d probably say by mid-2026. But it’s a guess.

The other thing to think about is what happens when the efficiency gains that tech is enjoying right now spread to every other sector in the world?

The AI boom will no doubt create new jobs. But will the net effect be positive? Initially, I doubt it.

Given the productivity crisis that Australia is in, and the fact that our high energy costs mean we’ll only get the bare minimum in terms of data centres built here, it’s a solid bet that you’ll see a negative net effect on the local jobs market.

That will be good for some companies but bad for others.

A major positive for us is that AI can’t manufacture iron ore, gold, coal, or LNG out of thin air. The world will still need our goods. The ‘old economy’ might bail us out again.

So, enjoy this AI-fuelled boom while it’s going. There’s more gas in the tank. But keep in mind that as more companies adopt AI, job losses will start to creep higher.

That’s not going to be good for a debt and house price-dependent economy.

It might be a 2026 problem, but we’re starting to think about it now.