S&P turns Meta's credit outlook negative over AI spending and cash flow

Meta has spent the last three years telling investors that its artificial intelligence bet will pay for itself. On Friday evening, October 30, 2026, one of the people whose job it is to check that claim said it is no longer so sure. S&P Global Ratings revised its outlook on Meta Platforms to negative from stable, citing what it called significant uncertainty around the company's capital spending and its free operating cash flow. The rating itself stays at AA-, the fourth highest grade on S&P's scale. But a negative outlook is a formal warning: S&P is telling bond investors that a downgrade has become a real possibility over the next one to two years.

The reason is simple to state and hard to fix. Meta is spending so much on AI data centers that its cash flow has turned negative, and it is signing data center leases so large that its debt like obligations are climbing fast. According to the S&P research update, Meta's adjusted leverage could rise above 1.0x, the downgrade threshold for its rating, as early as 2027. In 2025 it was about 0.1x.

What S&P actually decided

There are two parts to every credit rating decision, and it helps to keep them apart. The rating is S&P's current view of how likely a company is to pay its debts. The outlook is S&P's view of where that rating is heading. On October 30, S&P affirmed Meta's AA- rating, so nothing changes for Meta's existing bonds today. What changed is the direction. A stable outlook says the agency expects no change. A negative outlook says that the next move, if there is one, is more likely to be down than up.

S&P spelled out what would trigger that move. It could lower the rating if Meta's adjusted leverage rises above 1.0x on a sustained basis and its free operating cash flow stays negative beyond 2028. It could also act if it takes a negative view of Meta's AI strategy because of ongoing uncertainty about whether those investments can be monetized. That last condition is unusual. Credit agencies normally look at balance sheets and cash flows, not at whether a product strategy will work. Here S&P is explicitly saying that its view of Meta's AI business could move the rating on its own.

CNBC put the scale in context: S&P's highest rating is AAA, and anything below BBB counts as junk. Meta is still very far from junk. AA- is a grade most companies in the world would be happy to have. But Meta has used that strong credit to borrow cheaply for its buildout, and in its own financing announcement last week it pointed to its AA- and Aa3 ratings as an advantage. A downgrade would make that borrowing more expensive at exactly the moment Meta needs the most money.

The numbers behind the warning

The core of S&P's report is a new forecast for Meta's capital spending, and it is much higher than before. S&P now expects Meta's capex to reach $140 billion in 2026, in the middle of the company's own guidance of $135 billion to $145 billion. For 2027, S&P forecasts more than $225 billion, up from $163 billion in its previous model. For 2028 it now expects $252 billion, up from $178 billion. S&P still thinks spending growth will flatten from 2028, but it says it is less certain that Meta will cut back as previously expected, because Meta has not yet disclosed its plans for 2027 and its multi year AI ambitions are, in S&P's words, very ambitious.

That spending eats the cash that Meta's advertising business produces. In the third quarter, Meta's capex, including principal payments on finance leases, rose to $34.9 billion, and the company reported a free operating cash flow deficit of about $2.1 billion. It was the first time Meta's free cash flow went negative. S&P now forecasts deficits of $22 billion in 2026, $71 billion in 2027 and $39 billion in 2028. Before this update, S&P had expected Meta to return to positive free operating cash flow in 2028. Added up, that is about $132 billion of cash going out the door in three years, more than Meta's entire cash pile of roughly $82 billion at the end of September.

A note on the numbers, because they have been reported inconsistently. A Reuters story that was widely syndicated said S&P forecasts a $39 billion deficit in 2026. The S&P report itself gives $22 billion for 2026 and $39 billion for 2028. The figures in this article follow S&P's own text.

Leases are the hidden part of the bill

The capex numbers get the headlines, but S&P's bigger concern is something that does not show up as spending at all. Meta increasingly leases data center capacity instead of building and owning all of it. In its third quarter 10-Q filing, Meta disclosed that its future lease obligations rose to $355.7 billion as of September 30, up from roughly $279 billion one quarter earlier. The new figure includes leases for a proposed data center in El Paso, Texas, which S&P calls Sopaipilla, and S&P also names a second mega data center project called Beignet.

Rating agencies treat long term leases much like debt, because a company has to pay them whether or not the data center earns money. S&P says it treats the El Paso project exactly the way it treated Meta's Louisiana joint venture with Blue Owl. That is why Meta's leverage can jump even though the company has issued relatively little debt. Including the $37.5 billion of stock Meta sold last week, S&P expects adjusted leverage of 0.3x at the end of 2026. Then it climbs to 1.1x in 2027 and 1.3x in 2028 when the new leases start. Before the update, S&P had expected 0.4x, 0.7x and 0.8x for those three years.

There is one escape hatch in S&P's math. Meta announced on October 26 that it plans an at the market stock program worth about $60 billion once its new non voting Class C shares start trading in 2027. S&P did not include that program in its base case. If it did, Meta's leverage would be 0.7x in 2027 and 1.1x in 2028. That is better, but 2028 would still be above the threshold. In other words, even a historic amount of new stock does not fully cover the lease bill in S&P's model.

Why S&P singles out Meta

Meta is not the only hyperscaler spending heavily on AI, so why does it get the warning? S&P's answer is about payback. In the agency's view, Meta's ability to monetize its AI investments is more limited than that of its peers. Several of those peers already earn money from their data centers by renting compute capacity to frontier model labs and enterprise customers. Microsoft has Azure, Amazon has AWS, and Google has Google Cloud. Meta has no public cloud. Its AI spending has to pay off through better ad targeting, more engagement and new products such as the Muse assistant, the Business Agent and the new Meta Enterprise Platform.

S&P does not dispute that the core business is strong. In the third quarter, Meta's revenue grew 26.1 percent and ad revenue grew 25.9 percent, with double digit gains in both ad impressions and average price. But costs grew faster, helped by a one time charge of $9.89 billion for the resolution of litigation with a group of state attorneys general, and Meta raised its 2026 expense guidance to $175 billion to $179 billion. S&P says it is unclear when the new revenue streams will make material contributions that exceed their rising costs over the next few years.

Meta disagrees. "We respectfully disagree with S&P's assertion about Meta's AI monetization capabilities," a spokesperson told CNBC and Reuters. "We see significant product and business opportunities ahead of us, supporting our very strong credit profile."

Wall Street sees a smaller hole

S&P's forecast is notably darker than the consensus among stock analysts. A TIKR analysis published on Yahoo Finance on November 1 found that the average analyst estimate for Meta's free cash flow is minus $9.55 billion in 2026, minus $43.32 billion in 2027 and minus $8.45 billion in 2028. That adds up to about $61 billion over three years, less than half of S&P's $132 billion. The two use different cash flow definitions, so the comparison is rough, but the direction is clear. The biggest gap is in 2028, the year on which S&P's rating decision depends. Analysts see Meta close to breakeven by then. S&P sees a $39 billion hole.

The difference comes mostly from capex. According to the same analysis, analysts model about $210 billion for 2027 and about $227 billion for 2028, at least $15 billion and about $25 billion below S&P. The next real data point arrives with Meta's fourth quarter results, when the company is expected to give its first spending guidance for 2027. If that number lands near S&P's $225 billion, the agency's case gets stronger.

A bad week for AI spenders

The timing adds weight to the decision. When Meta reported its third quarter results in the week before, the stock fell about 7 percent despite a revenue beat because investors focused on the negative free cash flow. Meta shares ended October down 15 percent, their worst month in four years, according to CNBC. On Monday, October 26, Meta announced a stock sale of up to roughly $95 billion to fund its buildout, which we covered in detail. S&P's action came after the close on Friday, so Monday, November 2 will be the first trading day with the warning priced in.

Meta is also not the only name S&P touched. On Thursday, the agency lowered its outlook on Broadcom to stable from positive, saying the chipmaker's rapid AI growth increases its exposure to a few large customers. Earlier this year S&P cut Oracle's rating to BBB-, one notch above junk, because of its heavily leveraged balance sheet. And the same week, OpenAI cut the target for its new funding round from $30 billion to $20 billion after tepid interest from some large investors. "If capex remains elevated and continues to grow at a faster pace than we expect, credit profiles will continue to weaken and cash flows will be negative," Naveen Sarma, S&P's US media and telecom sector lead, said according to Reuters.

Why it matters

For years, the largest tech companies paid for their data centers out of cash flow and still had money left for buybacks. That era is ending. Meta is now funding AI with a mix of operating cash, bonds, giant leases and a historic stock sale, and the credit agencies are starting to say out loud that the mix has limits. A negative outlook does not cost Meta anything today, but it raises the price of the next dollar it borrows, and it gives the board a clear signal that the bond market is watching the 2027 spending plan.

It also matters beyond Meta. The AI buildout depends on a small number of companies being willing and able to spend hundreds of billions of dollars a year. If the most profitable social media company in the world gets a credit warning because its AI payback is uncertain, every lab, neocloud and data center developer that relies on those budgets inherits some of that uncertainty. The question S&P is asking, when does AI start paying for itself, is the same question equity investors have been asking since the selloff at the end of October.

Dany's take

I think S&P is doing exactly what a rating agency should do here. It is not saying that Meta is in trouble. Meta still has an advertising machine that grows 26 percent a year and a cash pile most countries would envy. It is saying that the plan has outgrown the cash, and that the gap is being closed with leases that look a lot like debt. That is a fair point, and Meta's own decision to sell close to $100 billion of new stock says the same thing in different words.

The part I find most interesting is the comparison with the cloud companies. Microsoft, Amazon and Google can rent out every GPU they buy if their own products fall short. Meta cannot. Its whole bet rests on AI making ads, feeds and agents much better than they are today. If the 2027 capex guidance comes in near S&P's $225 billion and Muse revenue is still small, I would expect the downgrade to follow. If Meta shows real revenue from agents and enterprise customers by then, this warning will look overcautious. Either way, the next earnings call just became a lot more important.

Sources: S&P Global Ratings research update, S&P Global Ratings bulletin, CNBC, Reuters via CNA, Investing.com, TIKR via Yahoo Finance.