top of page

Don’t underestimate the impact of leverage

2 days ago
3 min read

Charles Kindleberger, an American economic historian and author of Manias, Panics and Crashes, outlined the different phases that lead to asset bubbles and crashes.

 

The first is displacement, when an external “shock”, such as a technological breakthrough, opens up new profit opportunities. Next comes the boom, as credit expands and investment pours into these opportunities, often led by insiders. Then comes euphoria – when outsiders, often with little expertise, start piling in as they see their “friends getting rich”. Speculative herd like behaviour takes hold.

 

At some point, the rosy expectations underpinning the boom, amplified by the euphoria, begin to disappoint. As the narrative supporting the boom starts to unravel, insiders begin selling. Asset prices stall, wobble or fall. This marks the beginning of the financial distress phase.

 

Then comes a trigger – a failure, a fraud or a sharp fall in asset prices – that sets off panic selling as investors scramble for cash. This is the final stage, which Kindleberger called revulsion.

 

(My colleague John Calverley has also put together a useful and in-depth checklist for identifying market bubbles. You can find it in this blog, published a few years ago.)

 

The AI boom seems to be ticking several boxes in the early two stages of Kindleberger’s boom-bust cycle – displacement and boom. Advances in AI models, powered by more powerful chips and greater computing capacity, have provided the displacement. We are now seeing the boom, often led by insiders in the sector, with credit and related financial engineering playing an increasingly important role. For example, some commentators describe Nvidia as “effectively a central bank for AI labs and neocloud companies”.

 

I am not saying the AI sector is heading into euphoria or bubble territory, or will end in crashes. But one of the points Kindleberger highlighted, and which also features on John’s bubbles checklist, is the role of credit in fuelling a boom. It is not just about the amount of credit being extended. It is also about how that credit is distributed and how it creates a vast web of interconnected exposures. This was one of the dynamics at play before the global financial crisis.

 

I can already hear some arguing that the credit boom, or leverage, is confined to the AI sector. This is nothing like what happened in 2007 and 2008. But remember this: AI companies such as Nvidia are using their balance sheets to support loans made to frontier AI labs, neocloud firms and others. Some of these guarantees are also quite circular. For example, Nvidia’s guarantees are underpinned by expectations that borrowers will use its chips to meet growing AI demand. The ECB recently expressed its concerns on this (see here).

 

If the backlash against AI in the US gathers pace – with many political analysts expecting that to intensify as the Democrats take control of Congress in January 2027 – some of the “rosy expectations” we heard from hyperscalers, semiconductor manufacturers, AI labs may fail to materialise.

 

If leverage starts to unwind, the share and bond prices of many AI companies, including start-ups, could come under pressure. Investors who have also leveraged their own investments could suffer significant losses. And what might they do next? They may have to sell shares and bonds in other companies to offset their losses or meet margin calls on their loans.

 

Perhaps the fallout will not be as severe as the global financial crisis. But don’t underestimate the impact of deleveraging in a single sector. It could be enough to trigger the final sell-off in Kindleberger’s boom-bust cycle.

 

 

James Chu CFA

Head of Investment Solutions

 

 

 

Recent Posts

See All

Comments


bottom of page