AI is flying high. Will it fly too close to the sun and crash like Icarus in Greek myth? 

 

Physical constraints and ROI are the sunrays in this story. The likely downswing in prices and investment could be challenging but need not end in meltdown.

 

Is this really a bubble?

 

The exceptional valuations on AI-adjacent firms are matched only by the extraordinary scale of capital expenditure on data centres, both overseas and in Australia.

 

The use of leverage and sometimes opaque financing structures twitch the antennae of those of us who have been through financial boom-bust cycles before. Naturally, people question the return on investment and whether this is a “bubble”. But is it? 
 

A bubble is usually less about the technology itself than the behaviour around it - investors buying because they believe someone else will pay more later.

 

In the extreme, the market is in “Greater Fool” territory, where investors admit that they do not think the revenue warrants the price, but believe they will find a Greater Fool to buy it later. 
 
Property is one of the asset classes most prone to boom–bust cycles. Data centres are a kind of property, and leverage is clearly being used to finance their construction. 

 

Even so, the speculative behaviour that typically characterises property bubbles is missing: nobody is making these investments in expectation of reselling data centres to make a quick buck. It is all about the expected value of the computing services these data centres will provide. 

 

That ultimately raises a more important question: can the industry generate enough revenue to justify the investment?

 

Hype and FOMO 

 

Currently the “cost of compute” seems to be on the rise as the firms building the models try to shift their customer bases away from free and flat-rate subscriptions onto higher-cost pay-by-use plans. 

 

Yet everyone knows that AI can only achieve the desired returns by expanding usage, which means it needs to be cheaper than the human labour it seeks to replace. Even if investment decisions are based on expected returns on services provided by that capital, the realism of those return expectations can and should be scrutinised. 
 
It is natural to want to look for parallels in history.  
 
The seemingly obvious one is the ‘dot com’ boom of the late 1990s/early 2000s. But while there are some similarities to that episode, there are also important differences. Back then, the tech firms building the new technology were not so financially intertwined with the telco firms buying the spectrum, laying the fibre – and borrowing the money.  
 

Better analogies to the current AI/data centre boom can be found in other price cycles. Many boom–bust cycles start from something real and substantial. A new technology might instigate an equities boom. Opening up a new region might instigate a land boom there. 

 

In other words, a boom can be built on something genuinely valuable and still end with investors paying too much for it.

 

The danger of flying too close to the sun

 

What worries me about the current boom is how far financial considerations seem to sit in the background.

 

Capitalism has its drawbacks, but at least raw commercial incentives limit people’s appetite to overpay for things. When all the focus is instead on geopolitical rivalries or being the first to build god-like intelligence, commercial discipline is lost.

 

A potentially beneficial innovation wave instead risks becoming an Icarus-like mad flap that takes us too close to the sun. That is the perfect environment for a price cycle, including in the tech sector more broadly.

 

Whether any downswing becomes a meltdown depends largely on debt: how much has been borrowed, who borrowed it and whether losses would ripple through the broader financial system. 

 

Regulators will be watching closely to see where the debt sits and whether losses could spread through more highly leveraged parts of the financial system.

 
Humans have managed to transcend Greek mythology and master the art of flying without melting.

 

Yet the message of Icarus’s story endures: there are dangers when hubris goes unchecked.

 

To avoid getting ahead of yourself, it helps to let commercial considerations shape decisions, especially around borrowing. Far better that than basing decisions on imagined scenarios that do not survive contact with real-world economics.  

 

A version of this story was first published on Westpac IQ.