AI Today: The lenders are watching now
S&P has warned it may cut Meta's credit rating over its AI spending. The question is no longer just whether AI pays off, but who has a way to earn the money back.
On 30 October, S&P Global Ratings changed one word in its view of Meta. The outlook on the company's AA- credit rating went from "stable" to "negative". It is not a downgrade. It is a warning that one may come.
AI Today is a series from Fíor, a natural language classifier that predicts whether a piece of text was more likely written by a person or by AI. Each piece explains one development in AI and shows where the claims come from.
Stock investors have been worrying about the cost of AI all month, in a run of sell-offs. A credit rating agency asks a different question. It is not whether a company will grow, but whether it can comfortably pay back what it owes. That is why this move matters.
According to Reuters, S&P now expects Meta to spend $39 billion more than it brings in this year, and $71 billion more in 2027. It thinks that gap could last beyond 2028, and that Meta may have to borrow to cover it by then. Meta's debt, measured against its earnings, could rise above the level its rating allows as early as next year.
Meta disagrees. A spokesperson told Reuters the company respectfully disagrees with S&P's view of its ability to make money from AI, and sees significant opportunities ahead.
Who has a customer for the computers
The most pointed part of S&P's reasoning is a comparison. It said Meta's ability to earn money from its AI spending is more limited than its rivals'. The difference, as The Next Web noted, is that Amazon, Microsoft and Google all run cloud businesses that rent out computing power to other companies. Meta builds data centres mainly for itself.
Alphabet's latest results show what that means. It spent $51.4 billion on infrastructure in the third quarter, more than double a year earlier, and more than its operations brought in, so its free cash flow was negative too. But its cloud business, which sells that capacity to others, grew 103% to $30.7 billion in revenue. Alphabet is spending heavily as well, but part of what it builds already has paying customers.
Meta's money comes overwhelmingly from advertising. Its AI spending has to pay off through better ads and new products, which is harder to measure and slower to show up.
Borrowing, then selling shares
The credit markets had already moved. The cost of insuring Meta's debt against default reached a record in late October, Reuters reported. Meta has been leaning on share sales rather than new debt, including the plan to raise up to $97.5 billion we covered last week. A lower credit rating would make borrowing more expensive, which helps explain why.
S&P has been here before. Earlier this year it cut Oracle's rating to one notch above junk, over the debt behind its data centres.
For most of the AI boom, the people asking whether it will pay off were investors in shares, who can afford to wait for growth. Now the lenders are asking too, and they want to know something simpler: who has a customer for all these computers, and how soon?
Try the browser extension at Fíor.
