The perverse effects of the AI investment boom - The Business Times

Direct Source Verification: This story is aggregated from The Business Times (businesstimes.com.sg). Full reporting rights and copyright belong to the primary publisher.
SOVEREIGN and corporate debt levels have been flashing red for some time.

SOVEREIGN and corporate debt levels have been flashing red for some time.

In the US, federal debt now exceeds 100 per cent of gross domestic product, and the cost of servicing it – US$1 trillion this year – has overtaken defence spending.

And with French public debt reaching roughly 119 per cent of GDP, yields have come under pressure as investors question the sustainability of the country’s fiscal path.

Meanwhile, big technology companies have been borrowing as if there were no tomorrow, both on public markets and through private credit.

The so-called hyperscalers – those investing tens of billions of dollars in AI data centres – had hitherto funded their investments mainly from existing cashflows and equity. But now they are turning increasingly to debt financing.

Amazon, Alphabet, Meta and Oracle raised US$194 billion through bonds in the first half of 2026 alone – compared with roughly US$108 billion for all of 2025 – and their combined 2026 issuance is on track to reach US$250 billion.

Moreover, in September, Amazon raised £4.25 billion (US$5.8 billion) through its first-ever sterling-denominated bond, having already tapped the euro and other major bond markets.

The maturities on these recently issued bonds span from three to 19 years, with yields between 5.2 and 6.7 per cent. And separately, Alphabet stretched maturities and issued a 100-year “century” bond earlier in 2026.

This massive AI-driven fundraising has had three immediate consequences.

First, it demonstrates that large US tech companies can easily tap debt markets globally, and in currencies other than the dollar.

Second, the scale of the issuances means that hyperscalers can attract substantial institutional capital.

And third, by issuing long-dated bonds, they are now competing with governments for capital from pension funds and insurers.

The hyperscalers’ combined capital expenditure up to 2028 has been estimated at US$1 trillion or more, the equivalent of 3 per cent of US GDP.

That means their demand for capital will continue to grow. But such demand cannot be met from cashflows and equity alone, nor from the US corporate bond market.

So, as the AI data-centre market matures, its reliance on debt is likely to increase quickly.

Most hyperscalers have balance sheets strong enough to earn higher credit ratings than would normally be the case for corporate debt, meaning AA- and higher.

With little additional perceived risk, their bonds are therefore attractive even compared to risk-free assets such as US ten-year Treasuries or German ten-year Bunds.

But what will this competition for capital mean for other sectors and markets?

One risk is that it could have a crowding-out effect, not only among corporate borrowers but also in sovereign-debt markets, where yields could be pushed up, raising governments’ borrowing costs.

When combined with already-high interest rates, this dynamic could generate adverse spillovers for countries that need capital to finance economic development.

With inflation proving persistent, the US Federal Reserve returned to hiking interest rates in September – the first increase since 2023 – and the European Central Bank recently raised its deposit rate for the second time this year, to 2.5 per cent.

But if rates are poised to stay higher for longer, advanced economies and developing countries alike will have less fiscal space, owing to higher debt-service costs.

And this tightening will come just as political pressures for more public spending are mounting, and as the need for greater investment in environmental sustainability is becoming increasingly evident.

While G7 governments already carry historically high debt, the outlook for developing countries is even bleaker.

They may face a double squeeze, because as external capital becomes scarcer and more expensive, domestic financing costs rise in step, crowding out local investment.

For countries whose access to funding is already constrained, a severe fiscal squeeze or even a debt crisis cannot be ruled out. After all, many developing economies are starting from a weak position.

According to the World Bank, they paid US$741 billion more in principal and interest on their external debt than they received in new financing between 2022 and 2024 – the largest such gap in at least 50 years.

Now, the world’s richest companies could be outbidding the world’s poorest countries for the same pool of savings.

It is too soon to know how far this competition will go, but the scale of capital that AI infrastructure clearly demands warrants greater scrutiny.

Given how little evidence there is that AI is boosting revenue growth or lifting margins outside the tech sector, we could well be witnessing an extraordinarily costly misallocation of capital.

Therefore, AI’s promised gains need to be weighed against those investments (such as in climate change mitigation and adaptation) that may never happen now.

And even if the promised gains are borne out, the costs and benefits are unlikely to be distributed equally, both between and within countries.

These adverse spillovers are not receiving the attention they deserve.

What should be done? The G20, which the United Kingdom will chair from Dec 1, is the ideal forum for any coordinated discussion of AI-related capital flows and their spillovers.

This topic should be high on the agenda, with the aim of understanding these dynamics better and agreeing on a coordinated policy response based on shared data and joint analyses.

At the same time, the international community should scale up financial instruments that provide affordable funding to developing countries.

Guarantees from multilateral development banks can crowd private investors in rather than out, and more lending in local currencies can shield borrowers from exchange-rate risks.

The entire debate about AI has revolved around whether it will boost economic growth. But it would be far more practical to ask: Whose growth, and at whose expense? PROJECT SYNDICATE

Taylor Swift has over 300 trademark filings. Why do most SMEs have none?

Original Source
https://www.businesstimes.com.sg/opinion-features/perverse-effects-ai-investment-boom
Visit The Business Times ↗
SHARE STORY:
𝕏 f in

Related Coverage in Economy