S&P keeps Meta at AA- but turns its outlook negative over AI spending

Picture a bank manager who has lent to the same customer for years without a worry. The customer still pays every bill on time. But the manager notices the spending plans growing faster than the income, and writes a quiet note in the file: keep an eye on this one.

That, in plain words, is what S&P Global Ratings did to Meta on Friday. It kept the Meta credit rating at AA-, one of the strongest grades a company can hold. But it changed the outlook, its view of where the rating is heading, from stable to negative, because of how much Meta is spending on AI.

What a negative outlook means, and what it does not

A credit rating is a grade that tells lenders how safe it is to lend to a company. AA- sits just below the top tiers. An outlook is a signal about the next year or two. Negative means a downgrade is possible if things go the way the agency fears. It is a warning, not a downgrade.

The warning matters because Meta now borrows and leases heavily to build AI data centres. A lower grade would usually make that borrowing cost more.

S&P’s numbers behind the Meta credit rating warning

S&P’s own page could not be read directly, so these points come from the reports by Reuters, Bloomberg and 24/7 Wall St.

• The core worry: “Meta’s ability to monetize its AI investments is more limited than that of its peers,” S&P said, as quoted by Bloomberg and 24/7 Wall St. It added that it is unclear when AI revenue will outgrow the rising costs.
• Spending: S&P now expects Meta’s capex, or capital spending (money for buildings, chips and equipment), to pass US$225 billion in 2027, up from its earlier forecast of US$163 billion.
• Cash: S&P expects Meta’s free operating cash flow, the cash left after running the business and paying for those investments, to stay negative beyond 2028. It sees the gap peaking at about US$71 billion in 2027 and the shortfall being covered with debt.
• The trigger line: S&P could lower the grade if Meta’s debt stays above 1.0 times its yearly operating earnings for a sustained period. Reuters reports that could happen as early as 2027.

Meta’s own figures show the pressure has started. In its third-quarter results, Meta reported capital spending of US$34.9 billion and free cash flow of about minus US$2.1 billion. It expects to spend US$135 billion to US$145 billion this year. We covered those results, and Meta’s new Class C share plans, in our Meta Q3 2026 earnings story.

Meta disagrees, and the payoff is the open question

Meta pushed back. “We respectfully disagree with S&P’s assertion about Meta’s AI monetization capabilities,” spokesperson Kamran Mumtaz said, according to Reuters and 24/7 Wall St. He pointed to “significant product and business opportunities” and a “very strong credit profile”.

Both sides have a point. Meta’s advertising business is still growing, and its revenue rose 26% last quarter. But rivals such as Microsoft, Amazon and Google rent out their AI computing power to other businesses through their cloud services. Meta mostly uses its AI inside its own apps, so the return is harder to see on paper. This is our reading of why S&P draws the comparison.

The concern is not new. On 22 October, the cost of insuring Meta’s debt hit a record, as we reported in our look at Meta’s US$825 billion of AI commitments.

If you own Meta stock or an index fund

Meta stock closed at US$614.99 on Friday, down 1.5%, after its worst month in four years. If you hold it directly, or through a fund that tracks big US companies, this is a reason to watch, not to panic. Meta still has strong cash income from ads and a grade most companies would envy.

If you simply use Facebook, Instagram or WhatsApp, nothing changes for you.

Why patience, not panic, fits this moment

On balance, S&P’s move reads as a fair, early warning rather than a judgment that Meta’s AI plan has failed. Big building cycles often look expensive before they pay off, and Meta has the ad business to carry a lot of weight. The harder truth is that the spending keeps growing faster than forecasts, and debt is now doing more of the work. The likelier test is not whether Meta can afford AI, but whether it can show, in numbers, what AI earns.

Meta plans to give its 2027 spending forecast with its fourth-quarter results, usually due in late January. If that number keeps climbing without a clearer return, the quiet note in the file may turn into a lower grade.