S&P Just Flagged Meta's $225 Billion AI Bet

Meta spent $2.12 billion¹ more cash than it brought in during the third quarter of 2026, Reuters reports. A year earlier, its free cash flow was positive. On October 30, S&P Global Ratings kept Meta at AA-, near the top of the credit scale, but revised its outlook to negative from stable. It also cut Meta's outlook to negative.

The math behind the call is short. Meta guided to $135 billion² to $145 billion of capital spending for 2026. S&P forecasts more than $225 billion³ in 2027. Against the $140 billion midpoint, that is a jump of about 1.6 times in one year.

On October 26, four days before S&P cut its outlook on October 30, Meta pushed ahead with a plan to sell up to $95 billion⁴ of stock to pay for AI. Benzinga reports the plan has since grown to $97.5 billion⁵. OpenAI reportedly trimmed both its funding target and its valuation in those same days.

We cannot tell you whether Meta's AI bet pays back. S&P cannot either, and that uncertainty is exactly what it priced. The bill now arrives before the answer does.

How Leases Become Leverage at Meta

Meta is taking on debt without bonds. S&P wrote that the rise in Meta's leverage "primarily stems from its strategy of leasing data center capacity." It named two mega data center projects, Beignet and Sopaipilla. A lease signed this year is a payment owed for many years, whatever AI revenue does.

S&P's adjusted leverage takes debt, counts lease obligations as debt too, and divides by yearly cash earnings. By that measure S&P puts Meta at about 0.1x³ in 2025 and expects 1.1x in 2027 and 1.3x in 2028. Its downgrade threshold for the AA- rating is 1.0x. Meta reported $80.92 billion⁶ of long-term debt at September 30, 2026, a figure that leaves out much of what S&P now counts.

The facts sort into three buckets.

The spend is already large and still rising. Meta's third-quarter capital spending, including principal paid on finance leases, was $34.91 billion⁶, according to its results. S&P sees about $252 billion³ of spending in 2028.

Lease commitments begin to bite as each data center comes online. That is why S&P's leverage line keeps climbing into 2028, after the buildings are done.

The gap shows up in free operating cash flow, which is cash from operations minus capital spending. Benzinga reports S&P expects deficits of $22 billion⁵ in 2026 and $71 billion in 2027. S&P says the deficit could last beyond 2028.

Put the buckets together and Meta's profile starts to look like a power company's. The concrete and the chips get paid for first. Revenue comes later, if it comes. A credit analyst prices that shape the same way whether the logo belongs to a utility or a social network.

Reading Meta Through Its Cash Flow

Meta still reports healthy profits. Cash flow is the better judge of a business, though, and Meta's cash flow went negative last quarter. Accounting profit spreads the cost of a data center across years of depreciation. Cash counts the check on the day it clears.

The clearest tell is how S&P's own forecasts moved. Its earlier estimate for 2027 capital spending was $163 billion³. It now expects more than $225 billion, a $62 billion upward revision. Its 2028 figure rose from $178 billion to about $252 billion, a $74 billion jump.

When a forecaster raises its numbers after watching a company spend, it is telling you something about the company. Benzinga's headline quotes S&P as "less certain that it will curtail its spending." That phrase matters more than any single ratio. Credit analysts can live with a big build. What makes them nervous is management signaling the build has no ceiling.

For now, the evidence supports one decision. Treat Meta's AI program as an infrastructure project with a cash payback date, and judge it quarter by quarter. Three facts would change that call. A 2027 capital spending guide well below S&P's $225 billion³ would ease the pressure. Ad revenue growing faster than capital spending would show the payback has started. Leverage holding below 1.0x at the end of 2027 would keep the rating safe.

The bull case deserves a fair hearing. A ratio of 1.1x³ to 1.3x is modest by the standards of telecom and utility borrowers. Meta does not need a standalone AI subscription to earn a return. Better ad ranking and recommendation across its existing apps could lift revenue on a very large base.

S&P's counterpoint is that Meta's ability to monetize AI looks more limited than that of some peers. Cloud sellers can bill a customer for compute. Meta has to prove that compute shows up later as higher ad prices or more engagement. That chain is longer, and harder to measure.

The stock sale moves the risk around without removing it. Equity funding shields bondholders, which helps the rating. Existing shareholders pay through dilution instead. An at-the-market program sells shares into the market over time, so a falling stock price means more shares sold for each dollar raised.

We think S&P has this right. The core risk is timing. It is unclear whether AI-driven ad gains will arrive before 2028, the year S&P expects Meta could fund its cash shortfall entirely with debt if spending stays high.

Meta's Leverage May Cross 1.0x by 2027

Line up S&P's dated figures and the arc is plain. Leverage runs from 0.1x³ in 2025 to 0.3x at the end of 2026, then hits 1.1x in 2027 and 1.3x in 2028. Capital spending runs $140 billion, then $225 billion, then $252 billion.

Add the three spending years and you get $617 billion. The growth rate is slowing: 2027 adds $85 billion, while 2028 adds $27 billion. The 2028 step is about 12% ($252 billion³ divided by $225 billion). S&P has not forecast 2029 or 2030, so what follows is our own napkin math.

If 12% growth holds, 2029 spending lands near $282 billion and 2030 near $316 billion. The five years from 2026 through 2030 would then total roughly $1.2 trillion. A calculator will confirm it in under a minute. Even if growth stalls at zero after 2028, the five-year total still clears $1.1 trillion.

Here is where the asymmetric risk sits. Bondholders carry the downside of idle data centers and collect none of the upside from a hit AI product. Shareholders keep the upside and pay through dilution. The side of the market with the most to lose and the least to gain spoke first.

The pattern compounds, too. Each giant spends partly because being late looks more dangerous than overspending. Higher spending needs more financing, and financing adds fixed obligations. Those obligations shrink how long a company can wait for revenue.

Capital markets are starting to price AI ambition like a power grid rather than an app. That could mean the next downgrade wave in tech comes from lease footnotes, long before any bond default. And chips age faster than power plants. A new accelerator generation can cut the value of last year's hardware well before the lease on its building runs out.

Track Lease Commitments Before the Next Filing

You can watch this risk build yourself with a spreadsheet and the SEC's free filing database. It takes about an hour the first time and fifteen minutes each quarter after that.

Step one. Go to SEC EDGAR and open Meta Platforms' most recent 10-Q⁶. A 10-Q is the quarterly report most domestic US public companies file with the Securities and Exchange Commission for the first three fiscal quarters.

Step two. Make one row per quarter. Record cash from operations, purchases of property and equipment, principal payments on finance leases and long-term debt. All four sit in the cash flow statement and the balance sheet.

Step three. Compute free operating cash flow. Take cash from operations and subtract both spending lines. Meta's third quarter should land near the $2.12 billion¹ deficit Reuters reported. If yours is far off, check which lease lines you included.

Step four. Find the lease footnote. Record future payments on leases that have been signed but have not yet started. Those payments are next year's leverage, sitting in plain sight.

Step five. Add two peers, such as Alphabet and Amazon, on separate tabs. The comparison shows whether Meta is an outlier or the leading edge of a sector trend.

Step six. Set two flags in the sheet. The first turns red when capital spending exceeds cash from operations. The second turns red when your rough leverage figure climbs toward 1.0x³, S&P's threshold for Meta.

Your leverage number will not match S&P's adjusted figure. S&P makes its own adjustments, and a homemade version will drift from it. That is fine. Track the direction across quarters and note where your figure departs from the agency's published ones.

Rerun the sheet the day each filing lands. Meta's annual report for 2026 is next. If the not-yet-started lease line keeps growing while cash from operations stays flat, you will see the rating pressure before the agency writes it up.

Practise it step by step: Practice: Put AI to Work on Documents