Hyperscalers are significantly increasing capital spending on AI and are increasingly using credit markets to finance it.
But investors say the changes challenge the so-called “fortress balance sheet” status of the mega-tech giants and tear apart what they call the “tacit contract” that has kept speculative AI spending largely insulated from bond markets.
rear Amazon, meta and the owner of google alphabet While the companies revealed large increases in their full-year capital spending plans during the fiscal year, UBS data shows that AI hyperscalers’ total capital spending could exceed $770 billion in 2026, about 23% higher than previously expected.
UBS credit strategists said in a Feb. 18 note that the increase would mean a $40 billion to $50 billion increase in borrowing from hyperscalers, bringing total open market debt issuance to $230 billion to $240 billion this year.
oracle.
Al Cattermole, fixed income portfolio manager at Mirabeau Asset Management, said the tilt towards fixed income markets is dramatically changing the dynamics between hyperscalers and investors.
“We’ve been told for years that this AI spending will be covered by the cash flow generated. We’ve been told this is equity risk, it’s speculative, and from a credit standpoint we don’t need to worry about it,” Cattermole said in an interview with CNBC.
“Now it appears that the implicit contract has changed. While we would continue to lend to these companies, the reality was that the AI capex was still going to be financed with equity or cash….By taking the capex into the bond market, credit quality issues started to arise.”
“Breakpoint”
Last September, Oracle tapped the bond market to raise about $18 billion, making it the largest debt offering in history. Other companies are quickly following suit, with Google’s owner Alphabet recently issuing about $20 billion in bonds, including a rare 100-year pound-denominated bond.
This will put the sector’s debt burden under greater scrutiny.
“Everyone was treating AA or A-rated tech companies as essentially cash-positive companies, and to suddenly have this much debt on their balance sheets is quite a turning point,” Cattermole said. “It’s only been three or four months since we’ve had our turn, so everyone is getting used to the new way of doing things.”
Investors currently see a difficult road ahead. BlackRock said giant tech companies are taking advantage of the current credit issuance “bonanza” to bridge the gap between current investments and future returns.
“The problem is that increased corporate borrowing is increasing supply to a bond market that is struggling to meet large public deficits,” BlackRock said in its weekly market commentary.
“What has changed is the focus of the market. People are now asking how the adoption of AI is reflected in revenue and profits. This classification of winners and losers means that this is the prime time for active investing,” BlackRock added.
The world’s largest asset manager noted that AI builders are primarily tapping into the U.S. investment-grade market, “so we prefer high-yield bonds and European bonds.”
As Oracle’s stock price has trended lower over the past six months, volatility in the credit default swaps on the company’s bonds, which provide protection if a borrower defaults on its debt payments, has increased sharply.
Meanwhile, Cattermole noted that Alphabet’s capital spending next year is scheduled to reach nearly 50% of sales, which he said is approaching “unprecedented levels.”
“You wouldn’t see something like this at any point in a normal company,” he added. “We are clearly at the breaking point of the natural cycle.”
“Hidden risks”
Investors have highlighted concerns about the potential for debt-driven AI overspending, worrying that rapid technological improvements that make chips more efficient and reduce capacity demand could render the huge data centers key to the buildup obsolete.
Mr Cattermole said this had far-reaching implications for debtors.
“What will happen in three years?” Nvidia Will chips be overtaken by Chinese competitors and my data center will be obsolete in the third year, even though I have a 5 or 8 year loan? ” he said.
Shaan Raisata, senior economist at Vanguard, acknowledged that AI hyperscalers are starting from a very strong position, boasting “strong balance sheets, strong free cash flow generation, and strong moats,” but are now even more influential.
“Are there hidden risks building up in the system, like special purpose vehicles, more asset leasing, more off-balance sheet activity? Hidden risks are growing, and we don’t know if they’ll ever come to the surface,” Raisata told CNBC’s “Squawk Box Europe” on Wednesday.
“But it’s clear that investors need to keep that in mind when they start discounting stock market returns into the future.”
—CNBC’s Michael Bloom contributed to this report.
