Meta's Revenue Rose 28% and Left $784 Million in Cash

The easiest way to spot a bad quarter is to look at the revenue line. With the numbers Meta put out on the evening of July 29, that shortcut failed.

Revenue came in at $60.8 billion — up 28% year over year, and ahead of consensus. Daily users reached 3.6 billion. The advertising business, by any ordinary reading, was working.

The next day the stock fell 9%.

The answer was not at the top of the income statement. It was far below it. After paying every expense and funding every investment in the quarter, Meta had $784 million left. A year earlier, in the same quarter, that figure was $8.5 billion. The gap between those two numbers is the story of why a line item that looks small next to $60 billion of revenue was enough to reprice an entire company.

Put those four figures side by side and an odd picture emerges: revenue is compounding quickly, but the thing paying for that growth is no longer the company's own earnings.

A Month Earlier, the Same Company Knocked Down Chip Stocks

In early July, Meta announced it would sell compute capacity — rented GPU power — to outside customers under the name "Meta Compute." That single announcement wounded the scarcity thesis that had carried semiconductor stocks for three years, and more than a trillion dollars of value was erased across the sector. The details belong to a separate piece: Meta's One Announcement.

Meta's own stock barely flinched that day. The market read it charitably: if you have enough spare capacity to sell, that capacity eventually turns into revenue.

The July 29 report made that reading harder to hold. The capacity was certainly real — $31.1 billion of capex in the quarter, $50.9 billion in the first half of the year. The revenue attached to it was not. On the call, Mark Zuckerberg said Meta was fielding offers "meaningfully above what we're paying" for compute, then immediately added that selling all of it for near-term profit "would be unwise."

That sentence defends a strategy. What investors heard was simpler: the cash is not coming back this year either.

Both Sides of an Ad Company Got Heavier at Once

Three things happened simultaneously in Meta's earnings that quarter.

Costs grew faster than revenue. Total expenses rose 55% to $42 billion, with research and development up 68%. With revenue growing 28% and costs growing at roughly twice that rate, operating income fell 8%. Net income dropped 14% to $15.85 billion. Earnings per share came in at $6.18 against a $7.17 estimate.

Part of that miss is non-recurring: a $2.4 billion legal provision and $1.18 billion set aside for a roughly 8,000-person workforce reduction. Only part of it.

The cash side emptied out. Operating cash flow was $31.9 billion. Capital expenditure was $31.1 billion. The two very nearly cancelled each other out. What remained was $784 million.

The balance sheet took on debt. Long-term borrowings climbed from $58.7 billion to $83.7 billion, cash and equivalents shrank 57%, and no shares were repurchased in the first half of the year. A company that had bought back its own stock every quarter for years simply stopped.

The Arithmetic of Depreciation: Paid Today, Booked for Five Years

One piece of arithmetic explains most of Meta's story. The example below is deliberately simplified; the point is to convey scale, not precision.

The real fragility sits in the five-year assumption itself. The industry generally assumes a five- to six-year useful life for this equipment. If GPUs age faster than that — if the true life is three years — the same $140 billion books not $28 billion but roughly $47 billion a year. Nobody would have spent an extra dollar. One assumption changed, and a third of the profit went with it.

This is why capital expenditure is not a one-year question. An order signed today rewrites the profit estimates for the next five years immediately. That is why everyone running a valuation had to rebuild the model this quarter.

Same Week, Four Companies, Three Different Reactions

The last ten days of July were big-tech earnings season, and all four companies said the same thing: we are going to spend more. The market did not punish them equally.

What the bars show is that spending itself was not the offence. Amazon confirmed the largest budget of the group for 2026 — more than $200 billion — and its stock rose 15%, because cloud revenue grew 37% and beat expectations. Alphabet fell when it raised its budget on July 22, then recovered once Amazon's numbers landed; Google Cloud revenue had grown 82%.

Meta has no such line. The return on the money spent is measured only indirectly, through an advertising business that works somewhat better. That was precisely the phrase that recurred across the sell-side price-target notes: limited visibility into how the AI spending gets monetized outside of ads.

The charts below are live: they show where the stock trades today, not the event itself.

For comparison, the broader market:

Timeline: From Earnings Night to the Bounce

The whole episode fit inside thirty days, and the sequence matters: expectations for the sector had already cracked a week before Meta reported.

The fourth line shows where the real damage lives on days like this. The price-target cuts averaged 10% to 20% — meaning the banks were not lowering their profit forecasts so much as lowering the multiple they were willing to pay for those profits.

The Other Side: Two Readings of the Same Numbers

The only certainty sits between them: Meta is no longer an advertising company that generates cash, but an advertising company that routes its cash into a hardware program. Those two sentences describe the same revenue. They do not describe the same stock.

What Is Left: An Ad Company That Borrows

Meta trades roughly 26% below the peak it reached in August 2025 and is down 11% year to date. But the more durable change is not the drawdown. It is the change in what kind of company this is.

For years Meta fit a familiar description: low capital intensity, heavy cash generation, and a steady habit of buying back its own shares. Markets pay high prices for that profile, because growth is close to free.

The description is different now: $130 billion to $145 billion of annual spending, part of it financed with debt, and buybacks halted. That is the profile of an airline or a utility, and markets have never paid the same price for it.

The Lesson: Same Growth, Now Worth Less

Meta's revenue grew 28% this quarter. It grew at a similar pace the year before. What changed was not the rate of growth but what that growth costs.

The second lesson is more technical but more useful: calling a company profitable after reading only the income statement is a verdict handed down without looking at the bank account. Meta reported $15.85 billion of profit that quarter. It had $784 million left. Both were true — they were simply answers to different questions.

This article is based on reporting from CNBC, Fortune, The Motley Fool and market data providers as of August 2, 2026. Meta's quarterly figures come from the company's own release, and price-target changes from bank notes. Sources differ on the lower bound of the 2026 capex budget, placing it between $130 billion and $135 billion, and on Alphabet's one-day loss of July 22, which is reported as between 5% and 7%. The depreciation calculation is this article's own simplified illustration, not the company's accounting assumption.