If a financial bubble bursts, the initial losses will typically be borne by those who own the assets or have financed them. This is not necessarily where the largest investments have been made. What happens next is more uncertain. Losses can spread through credit, collateral and refinancing, and from there to companies and the real economy.
The previous three blog posts identified the indicators that are reasonable to monitor when assessing whether a bubble may be approaching its breaking point. The next three outline what could happen. The answer is uncertain and depends on the size of the losses, who owns the assets, how the investments have been financed and how much leverage sits behind them. It also depends on when refinancing is required and how easily the assets can be sold.
Since the financial crisis, business risk has changed form and location
Since the 2008 financial crisis, legislators have introduced extensive regulation of banks to contain systemic risks. Capital and liquidity requirements have been strengthened, for example, and risk definitions have been harmonised globally through, among other things, the Basel framework. The financial system has also been structured so that shareholders and certain creditors are, in principle, required to absorb losses if a bank experiences severe difficulties (“bail-in” rather than “bail-out”). This has increased banks' resilience to a range of well-known types of financial risk.
This is an important reason why banks have significantly reduced their exposure to corporate risk since then. The risk has not necessarily become smaller, however. It has instead partly changed form and location and is now increasingly provided through Private Credit. In the United States alone, Private Credit has grown thirteenfold since the financial crisis to USD 2–3 trillion.
This means that a larger share of financing now sits in structures where authorities have limited visibility. This applies, for example, to leverage, ownership, liquidity and cross-exposures. The current AI boom is financed to a significant extent through a combination of equity, bonds, bank financing and Private Credit. At the same time, large parts of the value are concentrated in a relatively small number of companies and a limited number of assets, as well as around specific expectations about the technological foundation and direction of development.
The initial effects hit the owners of the assets
The initial effects of a bubble burst are normally the easiest to understand. When expectations of future earnings decline, the value of the assets based on those expectations also falls.
For listed assets, this happens directly through market prices. Shareholders therefore lose wealth. Bond prices can also fall if credit spreads widen or market interest rates rise. New investors will then demand a higher required return to take on the same level of risk.
If the company has also financed its investments with debt, leverage makes the lender's risk closely linked to the company's ability to service interest payments and refinance its debt. When asset values and expected earnings decline, the debt burden becomes larger relative to the underlying assets and earnings.
This can create a self-reinforcing negative feedback loop. Falling asset values increase financing costs, which in turn put further pressure on asset values.
Credit can become the most important transmission channel
In an AI scenario, credit is particularly interesting because the investment boom is highly capital-intensive. Data centres, chips, electricity infrastructure and other physical capacity, such as cooling and water infrastructure, require large and often irreversible investments before they can generate returns. If expectations for future demand suddenly decline, the basis for the financing also changes.
This can quickly shift credit risk. Investors will typically demand higher risk premiums. The lower the credit rating, the higher the premium and the wider the credit spread. This can make bond issuance more expensive and refinancing more difficult. The same applies to Private Credit, where investments can be more difficult to realise because the loans and underlying assets are not continuously traded in liquid markets.
It is also important to distinguish between a solvency problem and a liquidity problem. A company can be profitable and have sound, valuable assets, yet still default if it is unable to refinance because of a lack of liquidity. Conversely, liquidity cannot solve a permanent solvency problem where the economic value of the assets is no longer sufficient to support the debt.
An equity market correction can therefore quickly develop into a credit problem. Unlisted assets may simultaneously exhibit relatively low visible price volatility because they are valued or traded only infrequently. This can conceal part of the underlying economic risk until many investors require liquidity at the same time. At that point, the difference between the book value and the realistic sale price of an asset can quickly become substantial.
Leverage amplifies the transmission mechanism
Leverage can increase the losses for asset owners. Consider, for example, an asset worth 100, financed with 60 in equity and 40 in debt. If the asset falls by 20%, its value is now 80. But the debt is still 40, so equity has fallen from 60 to 40. Although the asset fell by 20%, equity therefore fell by 33%.
The same applies to the lender. If the collateral securing a loan loses value, the lender may require larger safety margins or additional collateral to reduce its own risk. If the company also needs to refinance its debt, the lower equity position may make it more difficult to maintain financing on the same terms.
If many investors attempt to sell assets at the same time, this can amplify price declines while also reducing market liquidity. We saw this mechanism play out particularly clearly during the financial crisis and the dot-com bust.
Who bears the risk in the AI boom?
It is therefore relevant to view the AI boom as a chain of financial exposures.
- At the top are the companies expected to generate future earnings.
- Beneath them are companies supplying chips, software, data centres, energy, networks and other infrastructure.
- Next come the financial investors that own shares and bonds or provide credit.
- Behind these are pension funds, insurers, investment funds, banks and other investors.
This makes it difficult to see the overall risk through any single market indicator. A decline in AI equities does not necessarily affect the same actors as a decline in AI-related credit. Nor does a loss at a Private Credit fund necessarily affect the same investor who has bought a listed AI stock.
This is an important reason why bubble bursts are often very difficult to understand while they are unfolding. Where the investment is visible is not necessarily where the financial risk is ultimately borne.
Corporate investment is affected quickly
If the correction becomes sufficiently broad, companies will typically change their behaviour. Higher interest rates or tighter credit conditions can suddenly lead to investments in, for example, new data centres being postponed. Suppliers may then lose orders, while contractors may find themselves with spare capacity. Their response will typically be to invest less and reduce costs, including by reducing their workforce. The effect therefore moves from financial markets to corporate decisions and ultimately into the real economy.
This is where the transmission chain becomes significantly more difficult to predict. How large will the losses at Private Credit funds be? Who owns the underlying loans? Which funds hold the same assets? How much leverage sits behind them? Which banks have indirect exposures? How much capital is tied up in data centres that are no longer economically viable? How large are the capital reserves of investors and lenders?
The answers depend on what triggers the correction, how large the losses become and how the financing around the AI boom is structured.
This is why the initial effects are the easiest to understand
The first stage of a bubble burst is therefore relatively straightforward. Asset prices fall, affecting the wealth of their owners. This in turn increases broader credit risk and generally makes financing more expensive.
The picture then becomes more complex. If credit conditions tighten, companies may reduce investment. If investment falls, suppliers are affected. If suppliers are affected, employment and earnings may decline. And if asset prices continue to fall at the same time, collateral values may decline further.
A financial correction can therefore develop into a broader economic downturn.
The next stage raises more difficult questions: How large will the indirect effects be, where will they occur, and how will the physical economy be affected? That will be the subject of the next post.