The $16.68 billion child-safety settlement that dominated Meta Platforms headlines in late August is the cheapest number on the company’s page, and the market spent the month re-pricing the wrong liability. Reuters reported on 26 August 2026 that Meta will pay up to $18bn over the next decade — maximum payments of about $16.7bn to 47 US states, Washington DC, Puerto Rico, American Samoa and the Northern Mariana Islands — of which roughly $12.7bn is guaranteed and about $5bn more is contingent on Snapchat, TikTok and YouTube imposing similar protections for children. Call the guaranteed portion $1.27bn a year. Against trailing-twelve-month operating income of $86.93bn — my calculation from FY2025 operating income of $83.276bn plus H1 2026 of $41.647bn less H1 2025 of $37.997bn — that is about 1.5%. Set the entire $16.68bn cap against the $349.31bn of non-cancelable contractual commitments disclosed in Meta’s Q2 Form 10-Q and it is 4.8%; against the $278.99bn of leases not yet commenced, 6.0%. The market noticed: on settlement day META rose as much as 4.1% intraday, then closed at $576.14 — up just 1.07% from $570.05.
Here is the part almost nobody priced. The settlement appears in no SEC filing at all. Meta has filed exactly one 8-K since July — the 29 July results release — and the 10-Q predates the announcement by nearly a month, so every figure above comes from press reporting rather than the company. And two states refused to sign at all: New Mexico and Florida were not part of Wednesday’s settlements. The liability that is in the filings is roughly four times larger and goes to trial in a week. Buried in the legal-proceedings note of that same 10-Q: “Trial in the New Mexico Attorney General’s case, which has expanded to include various claims related to content moderation issues, is scheduled to begin on September 8, 2026. The New Mexico Attorney General has indicated that they intend to seek up to $62.85 billion in penalties in this case.” That is 3.8× the capped, discounted, decade-long number that moved the tape, it is uncapped, it is disclosed by Meta itself, and it starts one week from today. The August re-rating was a story about the wrong document.
Key facts
- Spot $572.34 at the close on 31 August 2026; 52-week range $520.26–$790.80 — stockanalysis.com, 31 Aug 2026
- Q2 2026 revenue $60.801bn, +28% YoY, but operating income fell 8% to $18.775bn and margin compressed from 43% to 31% — Q2 2026 8-K, Exhibit 99.1, 29 Jul 2026
- FY2026 capex guided to $130–145bn against $72.215bn in FY2025 — roughly +90% at the midpoint — 8-K CFO Outlook Commentary, 29 Jul 2026
- $349.31bn of non-cancelable commitments, of which $53.52bn falls due in 2026 and $81.65bn in 2027 — Q2 2026 10-Q, 30 Jun 2026
- $278.99bn of leases not yet commenced, plus ~$68bn of further data-centre leases signed in July 2026 on 18–20 year terms — Q2 2026 10-Q
- Free cash flow $784m on $31.862bn of operating cash flow — a 2.5% conversion rate — Q2 2026 8-K
- Zero buybacks since December 2025 against $22.921bn in H1 2025, with $25.03bn authorised and untouched — Q2 2026 10-Q
The call: bear $450 < spot $572.34 < bull $700
Every scenario below holds Meta’s current multiple constant and flexes only earnings. At $572.34 the stock trades on 20.9× my base-case FY2027 EPS of $27.40. The method is the argument: if you cannot make the earnings work, no multiple story rescues the position.
The arithmetic, explicitly. The bear case of $450 sits 21.38% below the $572.34 spot — ($450 − $572.34) ÷ $572.34 = −21.38%. The bull case of $700 sits 22.30% above it — ($700 − $572.34) ÷ $572.34 = +22.30%. The base case of $585 is +2.21%. Bear sits below spot, bull above it, asymmetric to the upside by under one percentage point.
- Bull $700 — FY2027 EPS $32.95 × 20.9 = $688, rounded to $700, which implies 21.2× on the bull EPS.
- Base $585 — FY2027 EPS $27.40 × 20.9 = $572; the $585 base adds the roughly 2% of drift that a year of buyback resumption or modest multiple support would supply.
- Bear $450 — FY2027 EPS $21.85 × 20.9 = $456, rounded to $450, or 20.6× on the bear EPS.
Horizon is end-2027. Every FY2027 figure here is my model, not guidance: Meta has explicitly declined to guide 2027 capital expenditure, so anyone publishing a 2027 number — including me — is extrapolating.
What actually moved the stock — and it was never the lawyers
Rank the drawdowns on the chart. The worst single session was 30 October 2025: META closed at $666.47, down 11.33%. The second worst was 30 July 2026, at $539.03, down 7.95%. Both followed capital-expenditure guidance. Settlement day was a 1.07% up-close. This market forgives a legal charge and punishes a spending guide — the opposite of how August was written.
The Q2 print explains why. Revenue grew 28% to $60.801bn, yet total costs and expenses grew 55% to $42.026bn and operating income fell 8% to $18.775bn. Margin went from 43% to 31%. R&D alone rose 67% to $21.656bn. The quarter also absorbed $2.40bn of legal charges and $1.18bn of severance tied to a May 2026 headcount reduction affecting roughly 8,000 employees; headcount finished at 75,472, down 1%.
The demand side is not the problem. Family daily active people reached 3.60 billion in June 2026, up 3%. Ad impressions rose 14% and average price per ad rose 12% — and the price line matters most, because it says Meta is not buying growth with inventory. Meta discontinued its average-revenue-per-person disclosure, so here is a proxy I calculate rather than one the company publishes: quarterly revenue divided by reported DAP gives roughly $13.65 in Q2 2025 (on 3.48bn) against $16.89 in Q2 2026, a 23.7% increase. The monetisation engine is compounding. The cost base is compounding faster.
Family of Apps did $23.394bn of operating income on $60.370bn of revenue, a 38.7% margin. Reality Labs lost $(4.619)bn in the quarter and $(8.647)bn across the half — narrower than the prior-year half’s $(8.739)bn. The metaverse is no longer the story here. The data centres are.
Financed like a pipeline, valued like a software company
This is the cross-industry read the coverage keeps missing. Strip the ticker off Meta’s balance sheet, hand the liability structure to a credit analyst, and it does not look like a software company. It looks like midstream energy infrastructure.
There is $84.00bn of notes maturing between 2027 and 2066, carrying $4.40bn of short-term and $84.98bn of long-term scheduled future interest — $89.38bn of contracted coupon, summing the two disclosed figures. There are data-centre leases running 18 to 20 years, some up to 30. There is $349.31bn of non-cancelable commitments and a further $278.99bn of leases not yet commenced, landing between now and 2036. And there are the ventures: Meta holds a 20% interest in a Louisiana data-centre venture with roughly $27bn of estimated development costs, against which it has provided residual value guarantees with an aggregate threshold of approximately $28bn. A second, in El Paso, Texas, would carry guarantees of up to approximately $13bn — though that remains an exclusivity agreement subject to definitive documents, closing expected in Q3 2026. Together, roughly $41bn of residual value exposure attached to entities Meta does not consolidate. The 10-Q is precise: “RVG payments are not probable, and therefore no liability has been recorded to date.”
Minority stakes in capital vehicles, long-dated commitments, residual value guarantees keeping assets off the balance sheet, obligations funded with bonds maturing in 2066 — that is a pipeline company’s capital structure. Pipelines trade at 8–12× earnings because the obligations are contractual and the assets single-purpose. Meta trades at 20.9× because the market still prices it as an ad business with optionality. Both cannot stay true. Either the AI spend converts into earnings that justify a software multiple, or the multiple converges towards the balance sheet. That tension, not the settlement, is the actual investment question.
The cash-flow evidence is already visible: free cash flow of $784m on $31.862bn of operating cash flow is a 2.5% conversion rate. That is not a software margin profile. FinanceFeeds has tracked the same compression across the hyperscaler complex — see our coverage of Morgan Stanley putting a number on the AI capex problem and our Microsoft MSFT prediction at $675 bull versus $400 bear, where the identical bull/bear tension plays out on a business with far better free-cash-flow conversion.
The smallest tell: the buyback stopped
If you want one number that reveals what management actually believes about its own cash, it is not in the press release. Meta repurchased no shares at all in the six months to 30 June 2026. In H1 2025 it spent $22.921bn doing so. The authorisation was not cancelled or exhausted — $25.03bn “remained available and authorised” at both 31 December 2025 and 30 June 2026, sitting entirely unused.
Simultaneously, Meta raised $24.910bn of net proceeds from long-term debt in May 2026, issuing $25bn of notes maturing 2031 to 2066 at coupons of 4.55% to 6.45%. The company borrowed roughly the size of its untouched buyback authorisation rather than using cash to shrink the share count — and the count is going the other way, from 2,530 million shares at the start of 2026 to 2,548 million at 30 June.
Dilution is a headwind to every EPS scenario above. A company that believed its shares were cheap and its cash abundant would be buying them. Meta is issuing debt and letting the count drift up — rational if you need every dollar for compute, and revealing about which constraint binds. Compare Oracle’s ORCL bull/bear setup, where the leverage question is explicit rather than implied, or CoreWeave’s CRWV scenario band, where the entire equity story is a financing story.
The capex maths that decides the bear case
Meta guided FY2026 capital expenditure, including finance-lease principal, to $130–145bn, narrowed from $125–145bn. FY2025 was $72.215bn; H1 2026 was $50.918bn. So $79bn to $94bn — roughly $86.6bn at the midpoint — has to land in the second half alone. More than the entire prior year, in six months.
Then look at what is already contracted for next year. Of the $349.31bn of non-cancelable commitments, $53.52bn is due in 2026 and $81.65bn in 2027. That is $28bn more pre-committed spending in 2027 than in 2026, before Meta signs anything new — and it has kept signing, adding roughly $68bn of data-centre leases in July 2026 alone.
Full-year 2026 expenses are guided to $165–169bn against $117.690bn in FY2025 — the FY2025 figure being revenue of $200.966bn less operating income of $83.276bn. That is roughly 42% expense growth at the midpoint. Meta still expects 2026 operating income above 2025’s. For that and my base-case FY2027 EPS of $27.40 to hold, revenue growth must stay in the mid-twenties while expense growth decelerates sharply in 2027. The bear case is simply what happens if it does not: depreciation from $130bn-plus of 2026 capex lands in the 2027 P&L in full, expense growth stays above 25%, and EPS lands nearer $21.85. At an unchanged multiple that is $450.
The bull case is the mirror. If ad pricing keeps compounding at low double digits, if the May headcount reduction flattens opex growth into the teens, and if the enterprise AI revenue Zuckerberg points at contributes anything material, $32.95 of FY2027 EPS is reachable and $700 follows without a single turn of multiple expansion. Note what the bull case does not require: it does not need Meta to re-rate. It needs Meta to earn.
What management actually said
The framing from the top is unambiguous about where the money is going. “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 Q2 2026 results release on 29 July 2026. “The results are already showing, and I’m optimistic about the potential ahead.”
The CFO’s framing matters more for the model, because it isolates how much of the margin damage was one-off. Susan Li, Meta’s Chief Financial Officer, told analysts on the Q2 2026 earnings call: “Excluding the Q2 legal charges and severance expenses, our second quarter operating income would have increased 9% year-over-year.” That is a 17-point swing from the reported 8% decline, and the single most important adjustment in the bull case: back out $3.58bn of charges and the business is still growing profits, just far more slowly than revenue.
On 2027, Li declined to help. Asked directly about next year’s capital expenditure by Brian Nowak of Morgan Stanley, she replied: “we aren’t providing a specific outlook for 2027 CapEx at this time. Infrastructure planning remains highly dynamic and even this year, there are a range of outcomes embedded in our outlook.” That refusal is why the scenario band above is as wide as it is. The single largest input to 2027 earnings — the depreciation and operating cost of the infrastructure being built right now — is the one number management will not put a range on.
Regulatory tension: the capped number and the uncapped one
Treat the two legal exposures differently, because they are structurally different instruments.
The reported settlement is capped, staged over a decade, and partly contingent on rivals — roughly $5bn of it releases only if Snapchat, TikTok and YouTube impose comparable teen protections, which is to say Meta has bought the right to help write the rules its competitors must follow. Several party counts circulate without conflicting: Reuters counts 47 states plus DC and three territories, notes a separate Texas deal worth over $1bn, and describes federal COPPA claims brought by 29 states; Techdirt counts 52 state and local attorneys general across the judgment. US District Judge Yvonne Gonzalez Rogers, overseeing the trial since 18 August, approved it. Every one of these figures comes from reporting rather than from Meta, which has filed nothing on the matter. A capped, contingent liability of this size against a company generating $60bn of quarterly revenue is a rounding error dressed as a headline.
The New Mexico case is the opposite in every respect. New Mexico did not join the August settlement. Its case is disclosed by Meta, it is uncapped, the state has indicated it will seek up to $62.85bn in penalties, and trial is scheduled to begin on 8 September 2026 — next Tuesday. Meta also faces an FTC administrative proceeding seeking substantial changes to the 2020 modified consent order, including a proposed prohibition on using minors’ data for commercial purposes. That is the genuine tail risk, and it is not a fine: a structural restriction on youth data would touch the ad-targeting engine itself, the asset the entire 20.9× multiple rests on. Meta says it intends to defend vigorously.
My prediction, with the causal chain stated: the New Mexico trial generates materially more headline volatility between 8 September and year-end than the August settlement did, and unlike the settlement it carries a genuine tail. But it will not set the 2027 price. Capex will. The market has told us this twice in twelve months, in single sessions of −11.33% and −7.95%, and both times it was a spending guide that did it. Watch the Q3 print — expected in late October, though Meta has not yet confirmed a date; Q3 2025 landed on 29 October — and watch it for one thing above all: the first indication of a 2027 capital expenditure range. That number, whenever it arrives, is the one that resolves $450 versus $700.
FAQ
Is Meta stock a buy at $572?
This is analysis, not investment advice. The framework says risk is roughly symmetric: 22.30% to the $700 bull case, 21.38% to the $450 bear case. What tilts it is not the multiple, held constant at 20.9×, but whether FY2027 expense growth decelerates.
Why is the $16.68bn settlement not a bigger deal for the share price?
Because as reported it is a maximum rather than a bill, spread over ten years, with roughly $5bn of it contingent on Snapchat, TikTok and YouTube adopting comparable protections. The guaranteed portion works out near $1.27bn a year, about 1.5% of trailing operating income. It appears in no Meta SEC filing to date, so every detail comes from press reporting rather than the company.
What is the biggest risk to the bear case being too pessimistic?
Ad pricing. Average price per ad rose 12% year on year in Q2 2026 while impressions rose 14%. A business monetising that fast can absorb a great deal of depreciation. The bear case at $21.85 of FY2027 EPS assumes pricing decelerates while depreciation does not.
How much does Meta still have to spend in 2026?
Guidance is $130–145bn for the full year including finance-lease principal, against $50.918bn spent in H1 — leaving roughly $79–94bn for the second half, about $86.6bn at the midpoint, more than the $72.215bn spent in all of 2025.
When does Meta report Q3 2026 results?
Not confirmed. Meta has not announced a date — its investor events page still says “stay tuned” — and the equivalent quarter in 2025 was reported on 29 October, so late October is the reasonable expectation.
Why does the buyback matter so much for the price targets?
Every EPS scenario is a per-share number. Meta bought back nothing in H1 2026 versus $22.921bn in H1 2025, while shares outstanding rose from 2,530m to 2,548m. A rising denominator makes the bull case harder and signals that management is prioritising compute over capital returns.
Analysis and information only — not investment advice, and not a recommendation to buy or sell any security. Price scenarios are the author’s model and rest on assumptions that may prove wrong. Figures are sourced to Meta’s Q2 2026 Form 8-K and Form 10-Q and to market data as of 31 August 2026.
