AI-related firms have raised about $500 billion this year, with major hyperscalers alone accounting for roughly $200 billion; new debt from AI big tech is projected to exceed $1 trillion over the coming years, forcing governments and ordinary companies to rethink when, where and at what cost they can borrow
The AI investment frenzy has spread beyond equity markets and is now shaking global bond markets. Tech giants including Meta, Amazon, Alphabet, Microsoft and Oracle are flooding debt markets worldwide to raise the enormous capital needed to build out data centers, semiconductors and power infrastructure.
As these AI hyperscalers emerge as the new dominant borrowers in global bond markets, corporations and governments are being forced to recalculate when, where and at what cost they can raise money, the Financial Times reported Monday (local time). AI hyperscalers are big tech companies that operate massive data centers and provide cloud and AI computing services worldwide.
According to Goldman Sachs, AI-related companies have raised about $500 billion this year. Of that, major hyperscalers — Amazon, Alphabet, Meta, Microsoft and Oracle — accounted for roughly $200 billion. The group is expected to issue more than $1 trillion in new debt over the coming years to fund AI infrastructure investment.
"The big theme is borrowing wherever, in every possible place," said Greg Peters, co-chief investment officer for credit at PGIM. "The volume of debt hitting the market is historic. The numbers are staggering and completely changing the market."
Meta plans to enter European bond markets for the first time this fall to fund its AI expansion. Big tech companies are tapping debt markets not only in the United States but also in Europe, Canada, Australia and the United Kingdom.
Amazon issued 14 billion Canadian dollars ($10.3 billion) in bonds in Canada in June, breaking the country's all-time record and far surpassing the 8.5 billion Canadian dollar record Alphabet had set just weeks earlier. Alphabet also set local records in Australia, raising A$5.5 billion ($3.6 billion), and in Switzerland, where it raised 3.1 billion Swiss francs ($3.5 billion).
An even more unusual deal emerged in the United Kingdom. Alphabet issued 5.5 billion pounds ($7.2 billion) in sterling bonds in February, including 1 billion pounds with a maturity of 100 years. Given how difficult it is to predict whether AI technology and current business models will remain competitive a century from now, observers say investors are taking on extraordinary long-term risk.
"Investors are buying unsecured bonds backed by transformative technology with maturities of up to 100 years," Peters said. "A hundred years is a very long time. The uncertainty makes it hard to judge whether the risk-reward is favorable."
The ripple effects are reaching the government bond market. Federal Reserve Chair Kevin Warsh and Treasury Secretary Scott Bessent have said hyperscalers are competing with the $31 trillion US Treasury market for capital. With the 10-year Treasury yield recently climbing to its highest level since 2007, massive government bond issuance and surging AI financing demand are simultaneously swelling the pool of borrowers vying for the same funds.
The European Central Bank has raised similar concerns. ECB staff recently said the rapid rise in big tech borrowing "could make it harder for other companies and industries to raise funds," questioning whether eurozone financial markets can smoothly absorb such large and concentrated debt inflows.
There are already signs that AI big tech bond issuance is crowding out other private-sector borrowers. Ordinary companies are adjusting their issuance timing to avoid overlapping with offerings from AI giants such as Oracle and Amazon, since coming to market at the same time risks investors setting aside cash for AI bonds and demanding higher yields in return. Some companies are also shortening maturities as the supply of long-dated debt swells.
Investors are adapting accordingly. "When a large AI bond deal is on the calendar, we definitely prepare," said David Brown, co-head of global investment-grade bonds at Neuberger Berman. "We either delay other purchases or make sure we have room to participate in that deal."
Even so, the borrowing appetite among AI big tech shows little sign of cooling. The combined capital expenditure guidance that Google, Amazon, Microsoft and Meta provided at their second-quarter earnings calls this year totals $745 billion. The fear of falling behind competitors if they slow investment in GPUs, data centers and power grids means these companies will keep raising funds even if interest rates rise somewhat.
The deeper concern is that these capital flows are weaving the entire AI ecosystem into a tightly interconnected web. Semiconductor companies, data center operators and power utilities are borrowing simultaneously in public bond markets and private credit markets, and companies are also lending to one another directly. Meta alone issues corporate bonds at the parent level while also raising funds through special purpose vehicles tied to individual data center projects.
As a result, pension funds and insurance companies are finding it increasingly difficult to gauge how much of their capital is actually tied up in any given AI company or the sector as a whole.
Torsten Slok, chief economist at Apollo, said the hyperscaler credit market is "in effect resting on one big assumption" — that their operating cash flows will more than triple, from roughly $600 billion today to $2 trillion. "If that assumption does not materialize, AI investment enthusiasm could cool, credit spreads could widen, capital expenditure could fall, and US economic growth could slow," he warned.
Debriefing: The Korea Herald's international desk breaks down the hidden stories behind the hottest global issues. Leave your questions in the comments.
mokiya@heraldcorp.com
