Meta's AI build-out now consumes almost everything the business earns. In second-quarter results published July 29, Meta reported $60.80 billion of revenue, up 28% year over year, and $31.86 billion of cash from operations -- then spent $31.08 billion on capital expenditures in the same three months, leaving free cash flow of just $784 million. Operating margin fell to 31% from 43% a year earlier, and the company narrowed its 2026 capex guidance upward at the low end, to $130-145 billion.
Key facts
- The headline number: $784 million of free cash flow, from $31.86 billion of operating cash flow, after $31.08 billion of quarterly capital spending.
- When: July 29, 2026, for the quarter ended June 30.
- Who: Meta Platforms, in its own quarterly results release and CFO outlook commentary.
- Primary source: Meta Reports Second Quarter 2026 Results.
Start with what is going right, because it is substantial. Revenue grew 28%, which for a company of Meta's size is remarkable, and management guided third-quarter revenue to $61-64 billion. Mark Zuckerberg's framing in the release is the bull case in one sentence: "AI is accelerating our core business today, powering our next generation of products, and opening the door to entirely new enterprise opportunities. The results are already showing, and I'm optimistic about the potential ahead." Nothing in the numbers contradicts that on the demand side. The advertising machine is working, and it is working better because of the recommendation and ranking models running underneath it.
Now the other column. Total costs and expenses reached $42.03 billion, a 55% increase. Operating income fell 8% to $18.78 billion, dragging margin down twelve points. Long-term debt stood at $83.66 billion at quarter end, against $90.26 billion in cash and marketable securities. Headcount was 75,472, down 1%, and still includes roughly 8,000 people affected by the May 2026 reduction who will mostly be gone from the count by the end of the third quarter.
Be careful with the causation here, because the obvious story is slightly wrong. Two of the cost lines have nothing to do with data centers: $2.40 billion of charges related to legal proceedings and $1.18 billion of severance. Remove those and the operating-income decline largely disappears. So "AI capex crushed Meta's profits" overstates it. What survives without any adjustment is the cash-flow arithmetic, and that needs no interpretation at all: for every dollar of operating cash the business produced this quarter, about 98 cents went straight back out into buildings, chips, networking and land.
The useful way to think about this is the difference between an expense and a commitment. An expense is a choice you make each quarter. A gigawatt-scale data center is a multi-year obligation you sign years before it produces a token of revenue -- construction contracts, power agreements, chip orders, leases. Meta's own outlook makes the scale explicit: full-year expenses of $165-169 billion, capex of $130-145 billion, and a tax rate rising to 15-17%. Guidance also raised the low end of the expense range specifically to absorb the quarter's legal charges, which is a small window into how tight the planning envelope is.
This is the clearest quarterly evidence yet for something the whole industry has been circling: AI capacity has become an energy-and-contracts business rather than a software business. The physical constraints are real, and they show up in filings before they show up in products. It is also why efficiency work has become strategically important rather than merely tidy -- the same week Meta reported this, OpenAI published an account of cutting its serving costs by 20% with model-written GPU code, and Meta itself has been capping employee AI token spend internally. Every percentage point of inference cost avoided is a percentage point of capex that does not need financing.
The honest caveat cuts both ways. Bears should note that Meta still expects full-year operating income above 2025, that revenue growth is accelerating rather than fading, and that free cash flow is a choice here, not a symptom -- a company that wanted a prettier cash-flow line could simply build less. Bulls should note that $784 million of free cash flow is what optionality looks like when it is nearly gone, and that the bet only pays if AI revenue arrives on the schedule the spending assumes. Meta's biggest AI wins so far are improvements to an existing ads business, not a new franchise generating anything like $130 billion a year. The next number that matters is whether third-quarter capex growth decelerates, or whether free cash flow goes negative.
Originally published on Ground Truth, where every claim is checked against the primary source.
Top comments (0)