Rising US bond yields put AI investment boom under pressure

US Treasury yields have risen to their highest levels in almost two decades, creating a fresh test for an AI investment boom that depends on abundant and relatively affordable financing.

The yield on the 10-year Treasury climbed from 4.95 per cent to 5.16 per cent in a week, with increases also visible across other maturities. The move takes the benchmark rate to its highest point since 2007, before the global financial crisis. smh.com.au reports that investors are now weighing whether higher returns on government debt could eventually undermine expensive technology shares and slow the construction of the infrastructure needed for artificial intelligence.

Why yields matter to shares and AI

US government bonds are generally treated as the safest major investment. When they offer more than 5 per cent, the roughly 1 per cent dividend yield available from the US sharemarket becomes less attractive, particularly while stocks are trading near record valuations and enthusiasm for AI remains intense.

Higher interest rates also affect companies in two ways. They raise the price of borrowing, which can reduce profits and weaken economic activity. At the same time, investors apply higher discount rates when translating expected future earnings into present-day values. That calculation tends to hurt companies whose promised growth lies far into the future.

The effect is especially important for the technology groups leading the AI expansion. They are spending heavily on processors, data centres, electricity networks and water systems, with much of that investment requiring debt as well as internal cash. US AI spending is estimated at more than US$700 billion this year and could exceed US$1 trillion next year.

The major cloud and digital infrastructure companies, known as hyperscalers, represent about 35 per cent of the US market’s value. Their scale makes them central to the sharemarket, but also leaves the wider market exposed if investors begin to question whether future revenue can justify the cost of today’s construction.

A market that has so far looked through the risks

Stocks have not yet responded materially to the latest yield surge. The market has largely absorbed oil prices above US$100 a barrel, persistent inflation and the Federal Reserve’s first policy-rate increase in more than three years. A short-lived fall followed the US and Israeli attack on Iran in February, but the broader advance has continued.

One explanation is the strength of the belief that AI will transform the economy and deliver enormous profits to the companies at the forefront. Another is that the US economy remains robust. AI-related spending, government support, a federal deficit approaching 6 per cent of gross domestic product, strong consumption among higher-income households and powerful corporate earnings have all helped sustain demand.

Yet even the largest technology companies are turning to debt. Amazon and Alphabet are among the hyperscalers using borrowing to help finance their expanding programmes, after AI spending absorbed much of the cash generated by their established businesses. The increased use of debt suggests that internal funds and the sharemarket are no longer sufficient to carry the entire burden.

Reported earnings growth has been unusually strong, with third-quarter estimates showing an increase of about 25 per cent from a year earlier. But the figures may overstate the underlying strength of the AI economy. Some profits arise from changes in the value of companies’ stakes in one another, creating revenue and gains that circulate within the sector rather than coming entirely from outside customers.

Debt, volatility and the wider economy

The rising cost of finance is a problem beyond technology. The US has more than US$40 trillion in government debt, while its deficit is close to 6 per cent of GDP. Refinancing older, cheaper borrowing at today’s rates would place increasing pressure on public finances. The Federal Reserve has only begun its tightening cycle, and markets are pricing in at least three further quarter-point increases over the next year.

The bond market itself may also be more fragile than in the past. Foreign central banks and institutions that once bought large quantities of US debt have become more cautious amid trade disputes and sanctions, shifting some holdings towards gold and other assets. Hedge funds and other traders now play a larger role, often using borrowed money in trades involving bonds and futures.

That structure can amplify a sudden market move. The MOVE index, which tracks volatility in government bonds, rose from 78.56 to 96 in a week. The earlier figure was close to its average for the past decade, suggesting that some leveraged investors were unsettled by the speed of the yield increase.

Credit markets are also signalling greater concern. The cost of insuring US investment-grade corporate debt against default has risen from 51 basis points at the start of the year to 58. Protection on US$100 million of loans now costs about US$580,000, compared with US$510,000 previously. For some hyperscalers, the cost is above 90 basis points.

The gap between two-year and 10-year Treasury yields has narrowed from about 85 basis points to roughly 30. That flattening indicates expectations of slower growth as policymakers respond to persistent inflation. With share valuations at levels last approached in 1999, before the dot-com collapse, a prolonged period of higher rates would pose a significant challenge.

For now, the market continues to place faith in AI-driven productivity and profits. The scale of investment means that much depends on those promised gains becoming real, not only for technology investors but also for the broader US and global economy.