Hyperscalers are significantly increasing their AI capital expenditure — and are more frequently relying on credit markets for financing.
However, investors assert this transition is challenging the ‘fortress balance sheet’ label attributed to mega-cap tech firms and disrupts what they refer to as the “unwritten agreement” that kept speculative AI investments mostly disengaged from debt markets.
After Amazon, Meta and parent company of Google Alphabet all revealed substantial increases in their yearly capital expenditure plans during earnings reports, with UBS data suggesting that the total capex among AI hyperscalers could surpass $770 billion in 2026 — approximately 23% above prior forecasts.
In a note dated February 18, UBS credit strategists indicated that these hikes imply a $40 billion to $50 billion increase in borrowing from hyperscalers, driving public market debt issuance to between $230 billion and $240 billion this year.
Al Cattermole, a fixed income portfolio manager at Mirabaud Asset Management, stated that this movement towards the bond market is significantly altering the relationship between hyperscalers and investors.
“For years, we have been informed that this AI expenditure would be financed through generated cash flow — that it represents equity risk, it is speculative, and that there’s no need for credit concerns,” Cattermole told CNBC during an interview.
“There now appears to be a shift in the unspoken agreement that while we would continue lending to these firms, AI capex would still be funded through equity or cash….By incorporating capex expenditure into debt markets, the question of creditworthiness arises.”
‘Break point’
Last September, Oracle accessed the bond market for around $18 billion in one of the largest debt issuances on record. Other companies have quickly followed suit, with Google-parent Alphabet recently issuing approximately $20 billion in debt, including an uncommon 100-year sterling-denominated bond.
This is intensifying scrutiny on the sector’s debt levels.
“Previously, people viewed these AA- or A-rated tech firms as effectively cash-like. Suddenly, however, adding this volume of debt to the balance sheets has represented a significant change,” Cattermole remarked. “It’s only been a few months since this shift occurred — everyone is adjusting to it in a new manner.”
Investors are now anticipating challenges ahead. BlackRock noted that mega-cap tech companies are leveraging the current credit issuance “bonanza” to bridge the distance between present investments and future income.
“The challenge: increasing corporate borrowing adds supply to bond markets that are already struggling to absorb substantial public deficits,” BlackRock stated in its weekly market commentary.
“The market’s focus has shifted: it now wants to know how AI adoption will convert into revenues and profits. This sorting of winners and losers makes this an ideal time for active investment,” BlackRock added.
The world’s largest asset manager highlighted that AI developers have primarily tapped the U.S. investment grade market, “thus we prefer high-yield and European bonds.”
As Oracle’s stock price has declined over the past half-year, credit default swaps on its bonds — which provide protection in case a borrower defaults — have experienced significant volatility.
Cattermole, on the other hand, pointed to Alphabet’s projected capex of nearly 50% of its revenue for the upcoming year, which he described as nearing an “unprecedented level.”
“You wouldn’t encounter such figures for a typical company at any time,” he noted. “We’re clearly at a turning point in natural cycles.”
‘Hidden risks’
Emphasizing worries over a potential debt-driven AI overspend, investors are concerned that the massive data centers crucial for expansion could become obsolete due to swift technological advancements that enhance chip efficiency and lessen the need for capacity.
This has serious repercussions for debt holders, according to Cattermole.
“What if, in three years, these Nvidia chips are outperformed by a Chinese rival, and I’m lending for five or eight years, only to find my data center is outdated by year three?” he inquired.
Shaan Raithatha, senior economist at Vanguard, stated that AI hyperscalers are beginning from a robust standpoint, featuring “strong balance sheets, significant free cash flow, and formidable moats” — but he recognized that they are accumulating more leverage now.
“Are there concealed dangers emerging in the system, whether through special purpose entities, increased asset leasing, or enhanced off-balance sheet activities? The concealed risks are indeed accumulating, and it remains uncertain if they will ever surface,” Raithatha told CNBC’s “Squawk Box Europe” on Wednesday.
“However, investors need to be aware of that as they start factoring in stock market returns into the future.”
—CNBC’s Michael Bloom contributed to this report.
