There are always early indicators of when booms are about to end. Sadly, they rarely get noticed before it’s too late.

Whether it is the 19th century rail boom in the UK or the dot com boom around the world they have always been heralded by statements about how it’s different this time.

In the case of AI and data centres there might, however, be another early indicator – for Australia at least – the Albanese Government’s pledge to start building data centres.

If you haven’t noticed the latest AI and data centre boom is suffering from the traditional problem which arises – too much debt and not enough income. It is difficult to imagine how Australia can ensure it doesn’t get itself entwined in the messy situation the industry is facing.

The Weekend Financial Times (1/11) reported that Calpers, California’s giant $US640 billion retirement fund was pivoting away from the big US tech companies and boosting its exposure to private equity. Calpers is the US’ largest retirement fund for government workers.

Anton Orilich, Calpers deputy chief investment officer said: “We are diversifying our equity risk. We are less concentrated.”

Calpers has reduced the combined number of shares it directly held in the so-called Magnificent 7 stocks – Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta and Tesla – by 26% according to Financial Times research.

The Economist (24/10) also reported that the AI industries gigantic debts are making investors nervous. It cites as a prime example the company CoreWeave, which rents out data centre capacity for AI models. The price of its bonds has fallen by 16% since July and it’s bond yields are at a level associated with companies in financial distress.

The Economist said Alphabet, Amazon, Microsoft and Meta have borrowed $US214 billion between them in 2026 – up from US83 billion last year.

On top of that there is $US79 billion issued by special purpose vehicles (SPVs). Providing data-centre gear and electricity demand the total borrowed in 2026 is $US850 billion.

Taken together the various companies carry more the $US3 trillion in debt.

And all this is complicated by the incestuous relationships between the various companies. For instance, The Economist reports Nvidia had covered as much as $US300 billion of debt to guarantee debt used to purchase chips if its customers are unable repay their creditors.

The Economist concludes: “the bigger the tower of AI debt grows, the more vulnerable it will become – and the graver the consequences should it come crashing down.”