Meta Platforms is not an expensive stock, and that is not why it has fallen 31% from its 52-week high. At $545.83 the shares trade on roughly 22 times annualised second-quarter earnings — cheaper than the S&P 500. The problem is not the multiple. It is that in the June quarter Meta turned $31.86bn of operating cash flow into $784m of free cash flow, and its own guidance says the second half will be worse.
That is the number the market is trading and almost nobody is framing correctly. Meta’s operations are not deteriorating: operating cash flow grew 25% year on year. Capital expenditure grew 82%, to $30.12bn in a single quarter, and swallowed it. Reuters put the free cash flow decline at 91% in a single quarter. Run Meta’s own full-year capex guidance forward and the arithmetic gets sharper still: the company spent $49.1bn in the first half and guided to $130bn–$145bn for the year, which requires $81bn–$96bn in the second half — a 65% to 95% step-up. Hold operating cash flow at the current $32bn a quarter and Meta is free-cash-flow negative by roughly $8bn a quarter for the rest of 2026.
Our base case: a $790 bull case against a $430 bear case, with the stock at $545.83 as of the 20 August close. Note where the sell side sits — 62 analysts, a Strong Buy consensus, and a $754.14 average target. Not one of them has a Sell.
Key facts
- META closed at $545.83 on 20 August 2026, against a 52-week range of $520.26 to $790.80 — stockanalysis.com, 20 Aug 2026
- Q2 2026 revenue was $60.80bn, up 28%, but operating income fell 8% to $18.78bn — Meta 8-K, 29 Jul 2026
- Operating margin fell to 31% from 43% a year earlier; diluted EPS fell 13% to $6.18 — Meta 8-K, 29 Jul 2026
- Free cash flow was $784m on $31.86bn of operating cash flow and $31.08bn of capex — Meta 8-K, 29 Jul 2026
- 2026 capex guidance is $130bn–$145bn, implying $81bn–$96bn in H2 against $49.1bn in H1 — Meta CFO outlook, 29 Jul 2026
- Long-term debt was $83.66bn against $90.26bn of cash and securities — Meta 8-K, 29 Jul 2026
- 62 analysts polled by S&P Global rate the stock Strong Buy, average target $754.14, and zero rate it Sell — S&P Global via stockanalysis.com, 20 Aug 2026
The advertising business is fine
Start with what is working, because the bear case depends on separating it from what is not.
Revenue of $60.80bn grew 28% year on year, and 27% in constant currency. Ad impressions across the Family of Apps rose 14% and the average price per ad rose 12% — meaning growth is split almost evenly between more inventory and better monetisation of it, which is the healthy combination. Family daily active people reached 3.60 billion, up 3%. Operating cash flow of $31.86bn was up 25%.
Nothing in that paragraph describes a business in trouble. Meta is not Intel. The core advertising engine is compounding at a rate most companies a tenth its size cannot manage, and the AI investment is visibly improving ad targeting — a 12% increase in price per ad on 14% more impressions is what improved ranking models look like in the accounts.
“AI is accelerating our core business today, powering our next generation of products, and opening the door to entirely new enterprise opportunities,” said Mark Zuckerberg, Meta founder and CEO, in the results release. On the revenue line, the claim holds up.
What broke: the gap between profit and cash
Costs and expenses rose 55% against 28% revenue growth. That gap is the entire story of the quarter, and it dragged operating margin from 43% to 31% and net income down 14% to $15.85bn despite the revenue surge.
Two items inside the cost line are one-offs and should be treated as such: $2.40bn of charges related to legal proceedings and $1.18bn of severance from the May 2026 headcount reduction of roughly 8,000 people. Strip both and operating income is about $22.4bn, a margin near 37%. The legal charge has its own trajectory — a New Mexico judge added $567m to an existing $375m jury award in August — but these are not the structural problem.
The structural problem is depreciation, and it has barely started. Capital expenditure of $30.12bn in one quarter becomes depreciation expense over the following years, and Meta is guiding to $130bn–$145bn of it in 2026 alone. Every dollar of that build lands on the income statement later, whether or not the AI revenue arrives to meet it. Full-year 2026 expenses are now guided to $165bn–$169bn.
Meanwhile the balance sheet has changed character. Meta carried effectively no long-term debt until recently; it now reports $83.66bn, against $90.26bn of cash and marketable securities. Net cash is down to roughly $6.6bn on a company still valued near $1.4trn.
The debt that is not on the balance sheet
The $83.66bn understates the capital committed to Meta’s AI build, and the reason is worth understanding because it is becoming standard practice across the sector.
Meta’s Hyperion data centre campus in Louisiana sits inside a special purpose vehicle named Beignet, arranged by Morgan Stanley, which issued roughly $27bn of A+ rated debt alongside $2.5bn of equity, anchored by PIMCO and BlackRock. Blue Owl Capital owns 80% of the vehicle and controls its board; Meta holds 20%. Because Meta does not control it, the debt does not consolidate onto Meta’s balance sheet — and S&P has stated it will not treat the vehicle’s debt as Meta’s.
This is legal, disclosed, and rational: it moves construction risk to investors who want infrastructure yield. But an investor reading only the balance sheet sees $83.66bn of long-term debt when the capital actually committed to the AI build is materially larger. Reported estimates of Meta’s total AI-related obligations, including lease commitments, run into the hundreds of billions. Anyone underwriting Meta on reported leverage alone is underwriting the wrong number.
Why the market and the sell side disagree so violently
Here is the most striking fact in this entire analysis. Meta trades at $545.83. The average of 62 analyst price targets is $754.14 — 38% higher. The lowest of those 62 targets is $580, which is still above the current price. There are 47 Strong Buys, 8 Buys, 7 Holds and no Sells at all.
A 38% gap between the consensus target and the tape, with no analyst willing to post a Sell, is not a normal disagreement. It means the sell side is modelling the AI spend as an investment that earns a return, and the market is modelling it as a cost that does not — and one of them has to capitulate.
The market has been consistent about this for a year. When Microsoft and Meta reported within days of each other in July, Microsoft’s larger capex number was rewarded and Meta’s was punished, because Microsoft could point to Azure revenue landing against the spend and Meta could point only to better ad targeting. That asymmetry has not changed. It is also why the hardware suppliers on the other side of this trade — from Super Micro to Micron — are the visible beneficiaries of Meta’s capex line.
Analysts covering the stock have been cutting quietly rather than downgrading. Parkev Tatevosian, CFA, who has been publicly long, noted in a 15 August review that “my previous price target for Meta for 2027 was a range between 885 and 985 and I made that on January 20th, 2026” — a target he has since walked down. That pattern, trimming numbers while maintaining the rating, is what a consensus looks like shortly before it moves.
The bull case: $790
The bull case is that 2026 is the peak capex year and the depreciation is spread against revenue that keeps compounding.
Assume 2027 revenue near $297bn, roughly 20% growth on a 2026 base of about $248bn — slower than today, consistent with a maturing ad business plus early enterprise AI contribution. Assume total expenses of about $200bn, which absorbs the step-up in depreciation from the 2026 build while opex discipline holds after the 8,000-person reduction. That gives operating income near $97bn and, at the guided 16% tax rate, net income around $81bn, or roughly $31.80 per share on 2.56 billion diluted shares.
At 24.8 times — a market multiple, not a premium — that is $790, essentially the 52-week high and 45% above the current price. It is below the $1,000 street high and above the $754 average, which is the right place for a bull case: it requires the capex thesis to work, not to be exceeded.
The bear case: $430
The bear case does not need an advertising recession. It needs capex to keep climbing and the depreciation to arrive before the revenue does.
Assume 2027 revenue of $272bn — still 10% growth — and total expenses of $215bn, reflecting a 2027 capex year at or above 2026’s level feeding a much larger depreciation charge, plus continued AI compensation inflation. Operating income falls to about $57bn, net income to roughly $47.9bn, and earnings to $18.67 per share. At 23 times, the stock is $430.
That is 21% below the current price and below the 52-week low of $520.26. It is also below every one of the 62 published analyst targets, which is precisely why it is worth stating: if the market is right and the sell side is wrong, the price discovery happens somewhere the sell side has not modelled.
What decides it
Third-quarter capex, due late October. Meta’s own guidance implies $40bn–$48bn in the quarter. If it comes in at the low end or below, the guidance was conservative and the free cash flow scare was a one-quarter artefact. If it lands at $45bn or more against roughly $32bn–$36bn of operating cash flow, Meta will report a negative free cash flow quarter for the first time in its history as a public company, and the debate ends.
The 2027 capex number. Expect it with Q4 results in late January. This is the single most important disclosure for the stock. A number that flattens near $145bn signals a peak; a number starting with a 2 signals the build is open-ended, and the bear case becomes the base case.
Evidence of AI revenue, not AI capability. Zuckerberg’s “entirely new enterprise opportunities” needs to become a disclosed revenue line. Microsoft gets credit for its capex because Azure quantifies the return. Until Meta separates AI revenue from advertising in its reporting, the market will keep treating the spend as a cost centre — and on the evidence of the last four quarters, it is right to.
Our reading: the advertising business justifies a price well above $545, and the capex programme is why nobody will pay it. Meta has become a bet on capital allocation rather than a bet on advertising, and the market prices capital allocation bets at a discount until the return shows up in cash.
Frequently asked questions
What is the Meta stock prediction for 2027?
Our bull case is $790 and our bear case is $430, against a spot price of $545.83 on 20 August 2026. The bull case assumes 2027 revenue near $297bn with expenses around $200bn; the bear case assumes $272bn of revenue against $215bn of expenses as depreciation from the AI build lands. The 62-analyst consensus compiled by S&P Global is $754.14 with a Strong Buy rating.
Why did Meta stock fall after Q2 2026 earnings?
Revenue beat and grew 28%, but costs rose 55%, operating margin fell from 43% to 31%, and diluted EPS came in at $6.18 against expectations closer to $7.20. The decisive number was free cash flow of $784m, down roughly 91% year on year, as capital expenditure hit $31.08bn in a single quarter.
Is Meta still profitable?
Very. Net income was $15.85bn in Q2 2026 and operating cash flow was $31.86bn, up 25% year on year. The issue is not profitability but conversion: after capital expenditure, only $784m of that cash was left. Meta earns enormous profits and is currently reinvesting essentially all of them.
How much is Meta spending on AI in 2026?
Meta guided 2026 capital expenditure, including principal payments on finance leases, to $130bn–$145bn, narrowed from a prior $125bn–$145bn range. It spent $49.1bn in the first half, so the guidance implies $81bn–$96bn in the second half — a 65% to 95% increase on the first-half run rate.
What is the Beignet SPV and why does it matter?
Beignet is the special purpose vehicle holding Meta’s Hyperion data centre campus in Louisiana. It issued roughly $27bn of A+ rated debt plus $2.5bn of equity, with Blue Owl Capital owning 80% and controlling the board while Meta holds 20%. Because Meta does not control it, the debt stays off Meta’s balance sheet, and S&P has said it will not consolidate it. It matters because Meta’s reported $83.66bn of long-term debt therefore understates the capital committed to its AI build.
Is Meta stock cheap at 22 times earnings?
On earnings, yes — that is a discount to the S&P 500 for a business growing revenue at 28%. On cash, no. Annualising Q2’s $784m of free cash flow gives a free cash flow yield near 0.2%, and Meta’s own guidance points to negative free cash flow in the second half. Which multiple is the right one depends entirely on whether 2026 is the peak capex year, and Meta has not yet said that it is.
This article is analysis, not investment advice. Figures are drawn from Meta Platforms’ SEC filings and from S&P Global Market Intelligence data as of 21 August 2026.