The 2008 Playbook Nobody Is Reading: Could the AI Boom Trigger the Next Credit Squeeze?
The 2008 Playbook Nobody Is Reading: Could the AI Boom Trigger the Next Credit Squeeze?
Every investment boom begins with a convincing story.
In the run-up to the 2008 financial crisis, the story was that property prices would continue rising, credit could be extended ever further, and financial engineering had made risk easier to manage.
Today, the story is different, but some of the underlying questions are familiar.
Artificial intelligence promises to transform entire industries. Technology companies are committing enormous sums to computing infrastructure, data centres and the energy required to run them. Investors are racing to secure exposure to what may prove to be one of the defining technologies of our time.
The technology may be revolutionary. That does not mean every investment being made in its name will prove profitable.
When investment runs ahead of returns
A recent warning from investment bank Jefferies, reported by Seeking Alpha, has drawn attention to potential parallels between today’s AI infrastructure spending and the vulnerabilities that preceded the 2008 real estate crisis.
The concern is not simply that technology valuations might fall. It is that an enormous buildout of infrastructure, financed through increasingly complex arrangements and justified by expectations of future demand, could leave investors exposed if those expectations are not met.
Three risks deserve particular attention.
First, the scale of capital expenditure. Building and operating advanced computing infrastructure requires substantial upfront investment. Facilities, power supply, networking equipment and specialist chips all demand capital before the ultimate commercial return is known.
Second, the speed of technological change. Today’s leading models and hardware may not retain their economic advantages indefinitely. Competition, efficiency improvements and cheaper alternatives could change the economics of infrastructure that appeared essential only a few years earlier.
Third, the financing structure. When large projects depend on continued borrowing, refinancing or consistently rising demand, a change in expectations can have consequences far beyond the original investment. If revenues disappoint while debt obligations remain fixed, the pressure can spread from individual projects into lenders and wider credit markets.
These risks do not prove that an AI crash is coming. Nor do they mean artificial intelligence will fail to create enormous economic value. They do, however, raise a question investors should ask during any investment boom: what happens if the future turns out to be less profitable than the present valuation assumes?
The uncomfortable lesson of 2008
The lesson of the global financial crisis was not that property was inherently bad or that every tangible asset was safe.
It was that the relationship between an asset’s apparent value, its underlying economics and the debt used to finance it matters enormously.
Property could be valuable and still be a poor investment when purchased at the wrong price, financed with excessive leverage or dependent on refinancing that was no longer available.
The same principle applies to AI infrastructure, listed technology companies and any other sector attracting large amounts of capital.
A compelling narrative can explain why investors are enthusiastic. It cannot, by itself, guarantee that the eventual returns will justify the price paid.
The danger is not innovation. It is the assumption that innovation automatically makes the financing behind it sound.
What if the next two years look different from the last two?
The next 24 months could bring continued expansion in AI investment. They could also bring a reassessment of valuations, project economics and the debt supporting the infrastructure buildout.
If expected returns disappoint, some projects may be delayed, financing may become more expensive and investors may become more selective. In a more severe scenario, a contraction in investment could affect suppliers, lenders and other businesses exposed to the same spending cycle.
Whether that develops into a broader credit squeeze remains uncertain. Forecasting the timing and severity of such an event is difficult, and there are important differences between today’s technology sector and the mortgage-driven financial system of 2008.
But investors do not need to predict the precise timing of a downturn to take its possibility seriously.
They can examine how much of their portfolio depends on a single investment theme, how much debt sits behind their investments, how easily those assets can be sold, and whether the underlying income can support the valuation without continued enthusiasm from the market.
A different way to think about capital
At Willow Rivers Wealth, we believe investment decisions should begin with the economics of the asset rather than the excitement surrounding it.
That means looking at what supports demand, where income comes from, how an investment is financed and what happens if circumstances are less favourable than expected.
Tangible assets and essential services can form part of that approach, but neither is automatically safe. Physical assets can be overpriced, income can disappoint, debt can become difficult to refinance and supposedly defensive investments can still lose capital.
The objective is not to find an asset class that is immune to every downturn. It is to understand the risks being taken, avoid relying on a single optimistic outcome and assess whether the investment still makes sense when assumptions are tested.
The AI revolution may ultimately justify much of the capital being committed to it. It may also leave some investors paying too much for assets whose economics change faster than expected.
The next financial crisis, if one emerges, may not look exactly like 2008. The underlying lesson is unlikely to change.
When capital is abundant, almost any investment can look compelling. When credit tightens, the quality of the underlying economics becomes much harder to ignore.
