On July 29, 2026, Meta Platforms reported second-quarter revenue of $60.8 billion, up 28% year-over-year — and net income of $15.8 billion, down 14%. That decline is the headline: it is the first year-over-year drop in quarterly profit since Meta began its AI-infrastructure spending ramp, and it happened in a quarter where the top line still grew nearly 30%. Total costs and expenses jumped 55% to $42.0 billion, operating margin collapsed from 43% to 31%, capital expenditures nearly doubled to $31.1 billion, and Meta both laid off staff and halted share buybacks to keep funding the build-out. "AI is accelerating our core business today," said CEO Mark Zuckerberg. The ad business agrees. The income statement is where the bill is starting to come due.
The Headline Numbers
Last quarter, Meta beat every estimate and the stock fell on capex guidance. This quarter, the tension moved out of the guidance and into the actual results: revenue growth stayed strong, but for the first time the cost curve bent the bottom line the wrong way.
| Metric | Q2 2026 | Q2 2025 | YoY Change |
|---|---|---|---|
| Total Revenue | $60,801M | $47,516M | +28.0% |
| Income from Operations | $18,775M | $20,441M | −8.2% |
| Operating Margin | 30.9% | 43.0% | −12.1pp |
| Net Income | $15,848M | $18,337M | −13.6% |
| Net Margin | 26.1% | 38.6% | −12.5pp |
| Provision for Income Taxes | $2,908M | $2,197M | +32.4% |
| Diluted EPS (GAAP) | $6.18 | $7.14 | −13.4% |
The story hides in the space between the first row and the second. Revenue grew $13.3 billion year-over-year; operating income fell $1.7 billion. That gap is entirely a cost story, and two of the costs are one-offs Meta itself flagged: a $2.40 billion charge for legal proceedings (sitting in G&A) and $1.18 billion of severance tied to a May 2026 headcount reduction. Add those $3.58 billion of one-off charges back and operating income would have been roughly $22.4 billion, up about 9% year-over-year — real growth, but at an adjusted operating margin near 37%, still well below last year's 43%. However you cut it, this is the first quarter in the AI-capex era where Meta's costs grew materially faster than its revenue.
Revenue Deep Dive: The Ad Machine Is Still Running
The top line is not the problem. Meta reports through two segments — Family of Apps (FoA) and Reality Labs (RL) — and Family of Apps delivered another near-30% quarter.
| Segment | Q2 2026 Revenue | Q2 2026 Income (Loss) from Ops |
|---|---|---|
| Family of Apps | $60,370M | $23,394M |
| Reality Labs | $431M | $(4,619)M |
| Total | $60,801M | $18,775M |
Within Family of Apps, advertising revenue was $59,363 million and "other revenue" (WhatsApp Business Platform fees, subscriptions) was $1,007 million. Ad revenue grew on a 14% increase in ad impressions and a 12% rise in average price per ad — the same AI-improved-ranking flywheel that drove Q1's 33%, still turning, just one notch slower. Family of Apps operating income of $23.4 billion was actually down slightly from $25.0 billion a year ago, because the legal and severance charges land inside the segment; the underlying ad engine remains extraordinarily profitable, throwing off operating income at a ~39% segment margin even after absorbing those hits.
Reality Labs is unchanged in character: $431 million of revenue against a $4.6 billion operating loss. Five years in, the metaverse-and-devices segment has never turned a profit, and Q2 2026 extends the streak — a roughly $4.6 billion quarterly drain that now runs alongside the far larger AI-infrastructure spend rather than instead of it. Reality Labs has lost $8.6 billion in the first half of 2026 alone.
The Cost Explosion: R&D Is the New Story
If Q1 2026's income statement was a tax story, Q2's is an expense story. Every cost line grew, but the mix is what matters:
| Expense | Q2 2026 | Q2 2025 | YoY Change |
|---|---|---|---|
| Cost of Revenue | $11,330M | $8,491M | +33.4% |
| Research & Development | $21,656M | $12,942M | +67.3% |
| Marketing & Sales | $3,431M | $2,979M | +15.2% |
| General & Administrative | $5,609M | $2,663M | +110.6% |
| Total Costs & Expenses | $42,026M | $27,075M | +55.2% |
Research and development — now Meta's single largest expense line at $21.7 billion, larger than cost of revenue — grew 67% year-over-year. That is the cash cost of the AI talent war and the compute to train frontier models: headcount in the AI organization, the depreciation beginning to flow off newly-placed data-center hardware, and a slice of the May 2026 severance. General and administrative more than doubled, but that line is distorted by the $2.40 billion legal charge; strip it out and G&A of roughly $3.2 billion grew a more ordinary ~20%. Cost of revenue's 33% rise is the most mechanically important for the future: it includes the depreciation of the $225 billion property-and-equipment base, and that number only goes up as more of the capex bill converts from "construction in progress" into depreciating assets.
The through-line is that Meta added roughly $15 billion of quarterly costs year-over-year to add $13 billion of revenue. Some of that is one-off (legal, severance); most of it — the R&D and depreciation — is structural, and structural costs do not reverse next quarter.
The Margin Story: The Lowest Operating Margin Since the Year of Efficiency
Placing Q2 2026 against the multi-year series shows just how sharp the compression is:
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | Q1 2026 | Q2 2026 |
|---|---|---|---|---|---|---|
| Operating Margin | 24.8% | 34.7% | 42.2% | 41.4% | 40.6% | 30.9% |
| Net Margin | 19.9% | 29.0% | 37.9% | 30.1% | 47.6% | 26.1% |
At 30.9%, Q2 2026's operating margin is the lowest Meta has posted since FY2022 — the "Year of Efficiency" trough, when $4.6 billion of restructuring charges and a stalled ad market crushed profitability. The difference is that FY2022's compression came from a revenue problem (growth went negative) and was fixed by cost-cutting. Q2 2026's compression comes from the opposite direction: revenue is growing 28%, and costs are growing 55% by choice, because Meta is pouring money into AI faster than even a re-accelerating ad business can absorb without margin giving way.
The net-margin line swings even more violently, but for a reason that flatters the comparison in the wrong direction: Q1 2026's 47.6% was inflated by an $8.03 billion one-time tax benefit (Treasury Notice 2026-7), so the sequential drop to 26.1% overstates the deterioration. The cleaner read is year-over-year: net margin fell from 38.6% in Q2 2025 to 26.1% in Q2 2026, a 12-point decline that mirrors the operating-margin compression almost exactly. Meta's tax provision returned to a normal $2.9 billion charge this quarter (management guided to a 15–17% rate for the rest of 2026), so the tax distortions that dominated the last three quarters are finally washing out — leaving the operating-margin compression as the clean, undistorted signal.
The One Big Question: Is the AI Spend Finally Outrunning the Ad Machine That Funds It?
For two years the bull case has been simple: Meta's AI capex is enormous, but the ad business that funds it is growing even faster, so margins hold and the spending is self-financing. Q2 2026 is the first quarter where the ledger stops cooperating with that story — and the cash flow statement shows why.
Capital expenditures, including finance leases, were $31.08 billion in the quarter, nearly double the $17.0 billion of Q2 2025. Operating cash flow was $31.86 billion. That leaves free cash flow of just $784 million — for a company that generated tens of billions in quarterly free cash flow as recently as last year. To keep funding the build-out, Meta made two capital-allocation moves it had avoided at this scale before: it stopped repurchasing stock entirely ($0 in buybacks in Q2 2026, versus $10.2 billion in Q2 2025 and $22.9 billion in the first half of 2025), and it issued $24.9 billion of net new long-term debt in the first half, pushing the balance from $58.7 billion at year-end 2025 to $83.7 billion. It also cut staff — the May 2026 headcount reduction that produced $1.18 billion of severance — even while spending $31 billion a quarter on infrastructure. Total headcount was 75,472 at quarter-end, down 1% year-over-year.
The bull case is that this is a timing mismatch, not a structural one: capex is front-loaded, depreciation lags, and the revenue payoff from better AI infrastructure shows up over years, not quarters. Management reaffirmed it expects full-year 2026 operating income above 2025's, and narrowed full-year capex guidance to $130–145 billion (raising the floor from $125 billion) alongside a $165–169 billion total-expense guide. The bear case is the one the ledger now makes visible: a $225 billion PP&E base generates a depreciation stream that grows for years regardless of what happens to revenue, and depreciation and amortization already rose 46% year-over-year this quarter (to $6.4 billion). If ad revenue growth decelerates from here while depreciation keeps climbing, the margin compression that showed up in Q2 2026 is not a one-quarter blip — it's the leading edge of a structural reset in how profitable Meta is.
Tracking the Capex Supercycle in Plain Text
Modeling Meta in Beancount, the open-source double-entry accounting system, makes this quarter's shift mechanically legible: revenue and costs are one transaction, and the balance-sheet build-out is a second set of rows you can watch grow period over period, every figure traceable to a 10-Q or 8-K.
Here is Q2 2026's income statement as a single zero-sum Beancount transaction — note the convention: Income accounts carry negative (credit) balances, Expenses carry positive (debit) balances:
; Q2 2026 Income Statement — three months ended June 30, 2026
; 1 MUSD = USD 1,000,000 | All figures in millions USD
; Check: −60,801 + 11,330 + 21,656 + 3,431 + 5,609 + 19 + 2,908 + 15,848 = 0 ✓
2026-06-30 * "Meta Platforms Inc" "Q2 2026 Income Statement"
Income:Revenue -60801 MUSD
Expenses:CostOfRevenue 11330 MUSD
Expenses:ResearchAndDevelopment 21656 MUSD
Expenses:MarketingAndSales 3431 MUSD
Expenses:GeneralAndAdministrative 5609 MUSD ; incl. $2.40B legal-proceedings charge
Expenses:OtherNet 19 MUSD ; interest and other income (expense), net (debit)
Expenses:IncomeTax 2908 MUSD
Equity:Adjustments 15848 MUSD ; net income offset (RE set by balance assertion)The two balance-sheet lines that define the era sit right next to each other and both jumped in a single quarter: Assets:NonCurrent:PropertyPlantEquipment reached $225,724M as of June 30, 2026, up from $176,400M at year-end 2025 — nearly $50 billion of net new infrastructure in six months — while Liabilities:NonCurrent:LongTermDebt climbed to $83,664M from $58,744M over the same span. Cash tells the mirror image: Assets:Current:Cash fell from $35,873M to $15,462M as the company spent and invested down its balance. This is a company converting cash and borrowed money into physical AI capacity as fast as it can build it.
The complete FY2021–Q2 2026 ledger — every balance sheet and income statement, sourced line-by-line from SEC filings — is open and auditable:
The Multi-Year Arc: The Balance Sheet Doubling Every Two Years
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | H1 2026 |
|---|---|---|---|---|---|
| Revenue | $116,609M | $134,902M | $164,501M | $200,966M | $117,111M |
| Net Income | $23,200M | $39,098M | $62,360M | $60,458M | $42,621M |
| Property & Equipment, net (period-end) | $79,518M | $96,587M | $121,346M | $176,400M | $225,724M |
| Long-Term Debt (period-end) | $9,923M | $18,385M | $28,826M | $58,744M | $83,664M |
The bottom two rows are the whole thesis. Property and equipment has gone from $79.5 billion at the end of FY2022 to $225.7 billion halfway through 2026 — nearly tripling in three and a half years — and long-term debt has gone from under $10 billion to $83.7 billion over the same span. First-half 2026 revenue of $117.1 billion (up 30% year-over-year) and net income of $42.6 billion (flattered by Q1's tax benefit) show the business is still growing into the spending. But the PP&E line is now larger than the company's entire annual revenue, and it is a line that must depreciate. The question the next several quarters will answer is whether that $225 billion of infrastructure produces enough incremental revenue to outrun its own depreciation — or whether Q2 2026's margin compression was the first honest look at what this era actually costs.
The Verdict: Bull vs. Bear
The Bull Case:
- Revenue grew 28% year-over-year to $60.8 billion on a $200B+ annual base — the ad flywheel (14% more impressions, 12% higher pricing) is still compounding, not stalling
- Roughly $3.58 billion of the cost increase was one-off (legal + severance); excluding it, operating income grew ~9% and adjusted operating margin held near 37%
- Management reaffirmed full-year 2026 operating income above 2025's and narrowed capex guidance to $130–145 billion, signaling confidence the spend is disciplined, not runaway
- The buyback pause and $24.9 billion debt raise are deliberate capital-allocation choices by a company with ample balance-sheet capacity, not signs of distress — Family of Apps still runs at a ~39% segment margin
- Tax distortions that muddied the last three quarters are washing out; the guided 15–17% rate makes forward earnings comparisons clean again
The Bear Case:
- Net income fell 14% year-over-year — the first profit decline of the AI-capex era — and operating margin (30.9%) is the lowest since the FY2022 "Year of Efficiency" trough
- Costs grew 55% versus revenue's 28%, and most of the increase (R&D +67%, rising depreciation) is structural, not reversible
- Free cash flow collapsed to $784 million as $31 billion of quarterly capex nearly consumed all operating cash flow; Meta halted buybacks and borrowed $25 billion to keep funding it
- Reality Labs lost $4.6 billion again ($8.6 billion year-to-date) with no profitability path, a permanent drag now dwarfed by the AI-infrastructure bet running beside it
- A $225 billion PP&E base means a depreciation stream that climbs for years regardless of revenue — D&A already rose 46% YoY — mechanically pressuring margins well beyond this quarter
Our Take: Q1 2026 was the quarter Meta beat every number and the market worried about the spending anyway; Q2 2026 is the quarter the spending showed up in the numbers. That makes it the more important print. The ad business is demonstrably healthy — 28% growth and a ~39% segment margin are not the marks of a company in trouble — so this is not a demand problem. It is a deliberate decision to trade near-term margin for AI capacity, financed by pausing buybacks, taking on debt, and, notably, cutting staff in the same quarter it spent $31 billion on infrastructure. The bull and bear cases don't actually disagree about the facts; they disagree about the lag. Bulls see a timing mismatch that resolves as the PP&E converts to revenue. Bears see the front edge of a permanently lower-margin Meta. The ledger will settle it the only way it can — by showing, quarter after quarter, whether that $225 billion of infrastructure grows revenue faster than it depreciates. Q2 2026 is the first data point, and it went to the bears.
