Meta grew 28%. Its profits went backwards.

Meta sold more advertising in the second quarter than in any quarter in its history. Revenue grew by more than a quarter against the same period last year.

It earned less money. Operating income fell. Net income fell further. Earnings per share fell with them.

Not less than analysts expected. Less than the company actually earned a year ago, on a quarter that was a quarter larger.

Where it went

Costs grew twice as fast as revenue. That gap is the whole of the profit decline, and no company outruns it for long.

Two lines did most of the work. Research and development, which carries the cost of building and staffing the AI effort, grew far faster than the business it serves. General and administrative expenses more than doubled. Cost of revenue, which carries the running cost of the data centres, grew faster than sales.

The operating margin lost twelve points in a year, at a company whose margins have been the envy of the industry for a decade.

Capital spending took roughly half of every dollar of revenue in the quarter, and came to about twice what the company earned. Meta raised fresh debt in May to help pay for it.

The number that is not in the accounts

None of that is the finding. Money already spent shows up in the income statement, and anyone reading the results could see the margin fall.

What is harder to find is what Meta has already agreed to spend. It appears once, in the risk factors, in a passage about whether the company can secure what it needs:

As of June 30, 2026, we had $349.31 billion of non-cancelable contractual commitments, comprising both short-term and long-term arrangements, most of which are related to third-party cloud capacity arrangements and other investments in technical infrastructure, and we continue to enter into additional significant contractual arrangements.

Non-cancelable is the company's own word. This is not capital Meta intends to deploy and could redirect if the returns disappoint. It is spending it has contracted for and cannot walk away from.

It is close to a year and a half of revenue. It is several times the cash the company holds, and several times its long term debt.

The tenant

What the money buys matters as much as the size. Most of it, the filing says, is third-party cloud capacity.

Meta spent fifteen years building its own data centres. Owning the infrastructure rather than renting it was the point of that, and it is part of why the margins were what they were. A company that buys capacity from someone else pays that someone else a margin too.

The passage those commitments sit in describes the risk in the company's own terms: that it may not secure enough components, equipment or services from third parties, or may have to buy them on unfavourable terms. That is a sentence written by a tenant, not a landlord.

What the number does not say

Three things should travel with this figure.

• It is not debt. It is contracted future spending, and it does not appear on the balance sheet.

• It spans several years and the filing gives no maturity breakdown. There is no contractual obligations table in this 10-Q, so nothing here shows how much falls due next year rather than in five.

• Commitments of some size existed before. No prior period figure is given, so this is a level and not a change.

The number survives all three. Whatever its shape across the years, spending on this scale that cannot be cancelled is a larger fact about the next several years than anything in the accounts it sits behind.

What would settle it

The third quarter filing will show whether the figure is still growing. A maturity breakdown, if one ever appears, would show how much of it lands soon. And the margin is the plainest test available: if revenue keeps growing at this rate while profit keeps falling, the question stops being whether Meta is spending too much, and becomes how long it has already agreed to keep doing it.