Is AI debt making US$1 trillion of bonds behave strangely? The warning signs ASX investors should watch

Here’s the US$1 trillion question: Is a huge chunk of the global corporate bond market simply cheap – or is it trying to tell investors something is amiss?

It matters because developments in the usually opaque world of credit can eventually flow directly into company profits, dividends, and share prices – as we so spectacularly witnessed during the Global Financial Crisis. When businesses must pay more to refinance their debts – or suddenly find lenders less willing to provide the money at all – shareholders may ultimately pick up the bill.

In July 2007, problems were already emerging in US subprime mortgages and lower-quality corporate credit. Yet the US Federal Reserve said financial conditions remained generally supportive, corporate credit spreads were still narrow, and equity markets had recorded sizeable gains.

Nobody is suggesting another GFC is inevitable. But almost 20 years later, research by Bloomberg has uncovered another curious disconnect. Government bond yields are surging around the world, while one of the most closely watched measures of corporate financial stress remains extraordinarily calm. And yet, around US$1 trillion of highly rated company bonds are telling a very different story.

So, what is the credit market really telling us – and when should share investors start worrying? This article examines why the cost of money can rise even when credit spreads stay calm, where private credit is already showing strain, and the warning signs that could signal whether cracks are starting to spread.

Fine on the surface, but...

First, two bond-market concepts you really need to understand. A government bond yield represents the base cost of borrowing in an economy, while a credit spread is the extra interest investors demand to lend to a company instead of the government. The greater the perceived risk, the wider that spread should generally become.

Right now, credit spreads are remarkably relaxed. Investors are happy to receive a historically low premium to endure the extra risk of owning riskier corporate versus risk-free government bonds. In the US, investment-grade corporate bonds pay an average spread of just 0.78 percentage points, close to their lowest level in 25 years, while European spreads are similarly subdued at 0.79 percentage points.

Yet Bloomberg’s research also identified about US$580 billion of US bonds and almost US$400 billion in Europe trading at unusually wide spreads compared with bonds carrying similar credit ratings. Twenty-five A-rated US borrowers were even trading at wider spreads than the benchmark for lower-quality BBB debt.

In plain English: Investors are demanding unusually high returns to lend to some supposedly safer companies – a sign the market may see more risk than their credit ratings suggest.

There are perfectly innocent explanations. Alphabet, Meta, Amazon, and Oracle have issued more than US$300 billion of bonds since the beginning of 2025 to help fund the AI investment boom. That enormous new supply has forced some investors to sell existing corporate bonds to make room for debt from these and other so-called “hyperscalers” – the handful of tech giants building the world’s AI data centre capacity. As investors sell, the prices of these bonds fall, pushing their yields – and therefore their credit spreads – higher. Even if the companies themselves remain financially strong.

That’s important. This isn’t a story about credit markets freezing or investors refusing to lend. Quite the opposite – money remains plentiful. The question is why, despite that abundance, so many individual bonds are being priced as exceptions.

In share market terms, imagine two companies considered roughly equal in quality. One suddenly trades at a much cheaper valuation. Either investors have overreacted and created a bargain – or the market knows something has turned sour in the company’s performance. That’s essentially the US$1 trillion conundrum in credit markets today.

Why the echoes of 2007 deserve attention

The useful comparison with 2007 isn’t that today’s corporate bonds are another version of subprime mortgages. They’re not. Nor are today’s banks in anything like the same position they were heading into the GFC. The similarity is what markets looked like before everybody realised there was a crisis.

In the first half of 2007, the Fed acknowledged that subprime mortgage conditions were deteriorating and lower-quality corporate credit was showing strain. Yet business debt was still expanding strongly, bond issuance remained solid, defaults were near zero, and corporate credit spreads remained generally narrow. The Fed even noted that financial markets remained supportive despite the worsening subprime problem.

The danger only became obvious when seemingly isolated problems began feeding into each other. Borrowers found it harder to refinance, defaults rose, lenders became more cautious, credit became harder to obtain, and falling asset prices made refinancing harder again.

Former New York Fed president Bill Dudley drew attention to a similar feedback loop in August. Vast spending on AI has lifted demand, profits and share prices, he noted – and more and more of that spending is financed by debt. If the boom disappoints, Dudley warned, that loop “can run powerfully in reverse” – weaker demand hits cash flows, lenders reassess the risk, and borrowing gets dearer just as growth slows.

Credit markets are already showing some unease around AI. Almost 80% of hyperscaler bonds issued since early 2025 were recently trading at wider spreads than when they first changed hands, although much of that deterioration can be explained by the sheer volume of debt being issued.

Oracle (NYSE: ORCL) is an example of a company where those concerns have become most visible, as its huge AI infrastructure spending and rapidly rising debt load have made investors increasingly nervous about its finances. Société Générale estimates the price of insuring against an Oracle default now implies a default risk above 16% – far higher than for its hyperscaler peers and a striking example of how quickly perceptions of credit quality can shift.

Set against these pressure points, the Fed says that US household and business debt relative to GDP has fallen to levels not seen since the early 2000s, publicly listed investment-grade companies generally retain solid debt-servicing capacity, and US banks remain well capitalised.

But they also told us that things were benign back in 2007 – just before the greatest financial calamity since the Great Depression! In the Fed’s defence (then and now), financial stress doesn’t necessarily arrive with a flashing warning light all at once. Sometimes the first clue that something is wrong is that individual parts of the market stop behaving the way they normally should.

The rising price of money – the last straw?

The potential common denominator behind today’s apparently unrelated cracks may be much simpler than subprime mortgages or AI. It may simply be the price of money.

Government bond yields are rising across much of the world. Bloomberg recently found two-thirds of 32 major global interest-rate markets are pricing further official rate increases, including around 4 percentage points of combined tightening across seven major economies over the following year.

US long-term borrowing costs have also risen sharply, with the 30-year Treasury yield reaching its highest level since 2007 earlier this year. This is where calm credit spreads can become misleading for share investors.

Imagine a company that borrowed five years ago when the government bond yield was 2%. Add a 1% corporate credit spread, and its borrowing cost was 3%. Now the government bond yield is 5%. Even if investors remain just as confident in the company, its new borrowing cost is now 6%.

For a company refinancing $10 billion, the difference between paying 3% and 6% is $300 million a year. That’s $300 million no longer available for profits, dividends, capital expenditure, acquisitions, or share buybacks.

And then comes the more dangerous version. If economic growth slows while interest costs rise, earnings weaken. Lenders become nervous and demand an even wider credit spread. Refinancing becomes more expensive again, squeezing earnings further.

That’s how an interest-rate problem can become a credit problem – and eventually a share market problem.

None of this stops at the US border. Australian companies borrow in the same global market, and the ASX’s most debt-reliant sectors – listed property, infrastructure, and utilities – are precisely the ones whose valuations were built on cheap long-term money. When the risk-free rate resets higher, their refinancing costs reset with it, however calm credit spreads look today.

One place where that pressure is already visible is private credit, lending undertaken by investment funds rather than traditional banks or public bond markets. Many of its borrowers are lower-quality companies using floating-rate debt, meaning higher rates hit them much faster than a company sitting on a long-term fixed-rate bond.

The Federal Reserve says debt-servicing capacity is already weaker among some riskier private companies relying on private credit and other floating-rate loans. Its May financial stability report judged the redemption requests themselves to have “remained manageable”. At the same time, Bloomberg reported that several of the largest funds had blocked investors from taking their money out.

Indeed, as early as March, J.P. Morgan Private Bank was noting “cracks emerging at the margins” in private credit, including elevated redemption requests and growing concerns around underwriting standards. But it argued those problems remained concentrated rather than systemic, with private-credit defaults still broadly in line with historical averages.

More recently, Pimco president Christian Stracke said investors were still waiting to get their money back from some funds, with withdrawal queues equivalent to around 15% of assets at many business development companies – the listed funds that lend to private businesses. Bloomberg estimates, using data from Robert A. Stanger & Co, that more than US$14.5 billion of investor capital remains stuck in more than a dozen funds, while Stracke warned of a pipeline of troubled software loans extending into 2027 and 2028.

That’s still a long way from a systemic credit crisis. To put the present exposure into perspective, J.P. Morgan estimates private credit accounts for only around 9% of total corporate borrowing, while bank exposure to the sector remains far smaller than banks’ exposure to real estate lending before the GFC.

And there is an entirely benign interpretation of Bloomberg’s US$1 trillion credit-market dislocation: Disruption is temporary but enduring – each phase lasts only as long as it takes the new to usurp the old, and then another begins. AI companies are flooding bond markets with new debt; carmakers’ bonds are being marked down on Chinese competition; software companies’ bonds on AI disruption; and so too are the bonds of insurers carrying private-credit exposure.

Different industries have different problems, and active bond investors are simply pricing those risks more precisely than a broad index can. Indeed, continued strong demand for both public and private debt argues strongly against claims that global credit markets are already seizing up – no smoking gun here, at least not yet.

Conclusion: What share investors should watch next

US$580 billion of American bonds and almost US$400 billion of European ones are priced as exceptions to their own ratings. Yet headline investment-grade credit spreads remain close to historic lows.

The risk-free cost of borrowing is rising around the world, and companies are gradually having to refinance cheap debt at much higher rates. Some weaker private-credit borrowers are already struggling, and investors are queuing to withdraw money from some funds.

That disconnect doesn’t make 2026 another 2007. But it does make 2007 worth remembering – and it should be a catalyst to staying on one’s toes.

For share investors, the warning signs from here are relatively simple. Watch:

• whether today’s company-specific credit problems begin appearing across more industries;
• whether corporate credit spreads start widening broadly;
• whether rating downgrades accelerate;
• whether private-credit defaults and withdrawal restrictions increase; and
• whether more companies – A-REITs and infrastructure owners especially, as their interest-rate hedges expire – begin warning that rising interest bills are eating into earnings and cash flow.

Most importantly, watch whether credit remains readily available. As long as lenders remain eager to provide money, today’s cracks can probably remain isolated. However, if lenders begin demanding substantially higher returns – or stop lending altogether – the equation changes quickly.

That was one of the crucial lessons of 2007. Best we don’t forget it.

This article was first published on Livewire's sister site Market Index on Friday 4 September 2026. For the full references list, please refer to the bottom of that article.