Date: 2026-08-27 · Price: $571.10 (+5.05% vs. baseline) · Market cap: $1.45T · EV: $1.49T Baseline: Output/Stocks/Technology/META/analyze-2026-08-18.md, 2026-08-18 @ $543.67 — HOLD, conviction 6.0, FV $450–650

What this updates

The baseline is nine days old. Nine days produced no new quarterly print (Q3 arrives 2026-10-28) and no change to the capex guide, so almost every Structural and Trend claim in the baseline is untested-by-absence-of-data rather than re-verified — it simply has not had a chance to move. One real event happened, and it is large: on 2026-08-26 Meta settled the MDL 3047 multistate trial that was in opening statements the day the baseline was written. That is the whole delta. Price also moved +5.05% on no new fundamental information, which by itself makes the stock a worse buy than nine days ago even before judging the settlement.

Event list since 2026-08-18: 1. 2026-08-26 — MDL 3047 settled. 29-47 state AGs (reporting varies; corroborated as a large multistate group, not a single state) dropped the federal Oakland trial after roughly 1.5 weeks of testimony. Meta agreed to pay $16.68–18B over 10 years (sources cluster around $16.7B core / up to $17.1–18B all-in) and to implement negotiated product changes for US under-18 users: daily time limits, nighttime blocks, stronger age verification, parental tools. Meta admitted no wrongdoing. 2. 2026-08-24 (approx.) — Meta filed its FTC appeal response brief at the D.C. Circuit, on schedule against the 2026-08-20 trigger. Brief argues Meta is not currently a monopoly given Reels/TikTok/YouTube competition; no oral-argument date set. 3. Analyst reaction to the settlement — mixed-to-positive: Piper Sandler ($785, "clear positive… negligible impact to estimates"), Rosenblatt (raised to $886), BofA (reiterated Buy, $810, "reduced regulatory risk"); Truist trimmed to $763 despite a Buy rating; Wells Fargo and BMO flagged the teen-usage restrictions as a possible engagement headwind. Stock closed 2026-08-26 up ~1.1% at $576.14, then settled to $571.10 by 2026-08-27 — a muted reaction for an $17B number, and every source that covered the move flagged the same reason: capex, not litigation, is now the dominant overhang. 4. No news found confirming or denying the baseline's flagged "New Mexico #2 — trial 2026-09-08 seeking up to $62.85B." New Mexico's own case (the $942M ruling) was explicitly excluded from the multistate settlement and appears, from available reporting, to have already concluded with the two-phase trial that produced the $942M figure. Two independent searches found no trace of a distinct forthcoming NM trial. This is not strong enough to retract the baseline's row — it is UNTESTED, flagged for a primary-source docket check next pass. 5. Insider activity pattern unchanged — continued 10b5-1 selling (Susan Li, Javier Olivan, Andrew Bosworth), zero open-market buys, no acceleration. Same read as baseline. 6. No news of buyback resumption, dividend change, or new capex guidance. The $16.7–18B settlement, spread over 10 years (~$1.7B/yr pretax), is immaterial next to $130–145B/yr of capex and does not touch the FCF story.

The delta ledger

Lead items — the status changes:

#	Claim (baseline)	Type	Status	What moved it
15	"A live federal trial with unquantifiable injunctive risk is in opening statements today" — the conviction cap named explicitly in §6 Tension 4	Judgment/State	🔄 SUPERSEDED	Settled 2026-08-26 for $16.68–18B over 10yrs plus negotiated (not court-imposed) product changes for US minors. Bounded, quantified, well inside Meta's $130B/yr OCF, and Meta shaped the terms rather than a judge dictating a ranking-function change. Market reaction: muted positive (+1.1% on the day), several targets raised, one trimmed. This resolves the baseline's single named conviction cap in Meta's favour and roughly as the base case, not the bear case, anticipated — plaintiffs' own working anchor was ~$200B; the actual number is ~10x smaller.
18	"FTC appeal — response brief due 2026-08-20"	State	🔁 REFRESHED	Filed ~2026-08-24, arguing no current monopoly (Reels vs. TikTok/YouTube). No oral argument scheduled. Risk assessment unchanged — remote inside 12 months — but the open item is now closed and the argument is on record.
19	"Analyst mean target $754.14, low $580, zero sells — stale, embeds phantom EPS"	State	🔁 REFRESHED	Post-settlement moves are directional, not a full re-tally this pass: Piper $785, Rosenblatt $886, BofA $810 (raised); Truist $763 (trimmed). Consensus mean likely little-changed to modestly higher. Still zero sell ratings. The phantom-EPS critique still applies — see data traps below, unchanged.
17	"New Mexico AG trial 2, begins 2026-09-08, seeking up to $62.85B"	State	⏳ UNTESTED	Two searches found no confirmation this trial is still scheduled, was folded into the settlement, or was mis-scoped in the baseline (New Mexico's own case looks concluded via the $942M ruling already in the ledger). Needs a court-docket primary source next pass — do not assume resolved.
21	"Short interest 1.43% of float, −8.8%, shorts covering"	State	⏳ UNTESTED	Not re-pulled this pass — no FINRA/short-interest data fetched. Carried by default, not verified.

Carried — no new evidence, none expected in a 9-day window without a print (Structural + Trend, all 100% intact):

#	Claim	Type
1	Revenue +28% YoY, accelerating sequentially	Trend
2	FCF $41.0B TTM, +1.1%/yr CAGR since 2021	Trend
3	FCF/share $15.97, +3.3%/yr	Trend
4	Capital allocation 84.5% to capex, shareholder returns 6.7% of OCF	Trend
5	Shares outstanding +1.29% YoY, buybacks $0 for 3 straight quarters	Trend
6	Debt/Assets 25.0%, net debt/EBITDA 0.62×	State
7	ROIC 18.4% spot / 25.8% TTM, down from 30.9% a year ago	Trend
8	Depreciation-wall model — D&A converges to capex run-rate, margin troughs ~27–28%, op income dollars still grow	Structural
9	FY2026 FCF guided-implied roughly breakeven (−$8B to +$16B)	State
10	Ad engine strength — family impressions +14% AND price/ad +12% simultaneously; July's Tinuiti-panel bear read stays falsified	Structural
11	Moat decomposition — conversion-feedback loop genuinely owned and strong; compute is a barrier to entry, not a moat (admits every real rival)	Structural
12	Competitive landscape — every named attention rival (TikTok, Snap, X, Bluesky, YouTube-relatively) shrinking or sub-scale	Structural
13	Reality Labs — ~$92.2B cumulative loss, no disclosed hurdle rate or kill criterion	State
14	Evergreen rating 6.5/10	Judgment
16	New Mexico #1 — $942M total (jury $375M + $567M abatement/reform order), separate from and excluded from the multistate settlement	State
20	Insider activity — zero open-market buys, metronomic 10b5-1 selling, no acceleration	State

Redone from scratch (Price rows never carry forward):

#	Claim	Old	New
23	Fair value / entry / trim	$450–650, central $516, entry $455–495, trim 22× fwd	$460–660, central ~$528, entry $465–505, trim 23× fwd
24	Verdict / conviction	HOLD 6.0	HOLD 6.3

No rows are DRIFTED, RETRACTED, or NEW this pass.

How the close calls were decided

Is the settlement good news, bad news, or neutral? The forces: (a) it removes a genuinely unquantifiable risk — a federal court could have imposed a ranking-function constraint with no dollar ceiling and no negotiating room; (b) the actual number, $16.7–18B over a decade, is roughly 10% of the plaintiffs' own working anchor (~$200B) that the baseline used to size the tail, and is immaterial against $130B/yr of operating cash flow; (c) Meta negotiated the product-change terms itself rather than having a judge dictate them, and the changes are scoped to US minors — roughly 0.5% of the global user base per one analyst estimate; (d) the market's own verdict was a 1.1% pop that fully round-tripped within a day, and every outlet covering it independently concluded capex is the bigger issue now, not litigation. All four forces point the same direction. This is a genuine, if modest, upgrade to the legal-risk half of the thesis, and it is corroborated by primary reporting (CNBC, CNN, NPR, Reuters via fin.py's news feed) plus a second-order confirmation (analyst price-target moves), not a single source.

Why the conviction move is only +0.3, not larger. The baseline named three things removed at once as the reason conviction sat at 6.0 and not higher: the buyback, the falling-share-count reversal, and the disclosure needed to check management's return on the marginal $100B. None of those three moved. The settlement resolves a fourth, separate risk (litigation), which the baseline treated as a conviction cap, not the primary driver — Tension 4's own text said "this is correctly a conviction cap, not a valuation input." Removing a cap raises the ceiling; it does not itself supply new upside evidence on the capex question, which stays exactly where it was: unguided for 2027, with the Street's own range ($176B–$215B) still unresolved until Q3 (10-28) or Connect (09-23). Price also moved +5.05% on zero new fundamental information, which mechanically narrows the margin of safety and offsets part of the risk-reduction benefit for anyone deciding whether to add today.

Why fair value moved up modestly (central $516 → ~$528). The earnings-power model's bear scenario explicitly cited "MDL yields injunctive relief" as part of the 14× multiple; with that specific tail now bounded, the bear multiple deserves a small upward nudge (14×→15×) and the base case similarly (19×→19.5×, reduced-overhang argument from Piper Sandler et al.), while the bull case is unchanged (capex is still the constraint on the bull case, and nothing there moved). The reverse-DCF FV moved trivially — EV rose with the price but TTM FCF is unchanged (no new print), so the required-improvement multiple actually got slightly worse (2.1× → 2.2×), a small offsetting force. Net: fair value up about 2%, price up about 5% — the gap between price and central fair value widened, not narrowed, even though the news was net-positive. That is the correct read of "good news, bad entry point."

Thesis persistence and conviction delta

Thesis persistence: 100% of Structural + Trend claims (10 of 10: 4 Structural, 6 Trend) survived as CARRIED. This is expected and not a strong signal on its own — no new quarterly data existed this pass to test any of them against. The honest statement is: nothing has yet had the chance to break the thesis; nothing has broken it either. The next real test of Structural/Trend claims is the Q3 print (2026-10-28), which resolves capex trajectory, ROIC direction, buyback resumption, and the dilution streak all at once.

Conviction delta: 6.0 → 6.3. Driven specifically by row #15 (MDL 3047 settlement, SUPERSEDED in Meta's favour) resolving the baseline's named conviction cap on terms better than the plaintiffs' own working anchor implied. Offset by: price +5.05% on no new fundamental data (mechanical margin-of-safety erosion), and zero movement on the three factors the baseline actually weighted heaviest (buyback, share count, capex disclosure). Verdict stays HOLD — the legal de-risking is real but it was never the dominant variable; the capex/FCF/ROIC/dilution complex still is, and it is exactly as unresolved as it was nine days ago.

What is genuinely new

The settlement mechanism itself is new information about how this specific tail risk resolves: negotiated, bounded, scoped to a small user segment, paid over a decade. That is a data point worth carrying into the next legal-risk assessment on any name with similar youth-safety exposure (Snap, TikTok/ByteDance, Google/YouTube face parallel claims) — settlements of this type appear to price well below the state AGs' own headline asks, and the market appears to have already partly anticipated that (the muted 1.1% pop suggests this was not a surprise to positioning). This is a single observation on one ticker; it would need a second name to earn a knowledge-base pattern note (see below).
New Mexico's decision to go it alone (excluded from the multistate deal, recovered $942M vs. an estimated multistate per-capita share several times smaller per local reporting) is a governance data point about how state AGs may behave differently going forward — worth watching if it changes other states' incentive to settle vs. litigate on the remaining individual/bellwether track (3,137 personal-injury cases still active in the federal MDL, no MDL-wide settlement; California JCCP bellwethers late October 2026; federal school-district bellwethers reported for February 2027).

Updated verdict

HOLD — do not add here · Conviction 6.3 / 10 ⬆️ (from 6.0)

Fair value $460 – $660 · central ~$528 · Entry $465 – $505 · Strong buy < $450 · Trim 23× fwd (≈ $701 at FY2027 consensus EPS $30.50 — recomputed live from price ÷ current P/E each site build, not a frozen dollar)

The case in four lines, updated:

The single largest named risk cap from the baseline is resolved, and resolved well. MDL 3047 settled for $16.7–18B over 10 years — roughly a tenth of the plaintiffs' own working anchor — via terms Meta negotiated rather than a judge imposed. The market's own verdict (a 1.1% pop that mostly held) says this was already substantially priced.
Nothing else moved. Capex, FCF, ROIC, the dilution streak, and the missing 2027 number are exactly where the baseline left them. The next real test is the Q3 print (10-28) and Meta Connect (09-23) for the first Meta Compute pricing detail.
The stock got 5% more expensive on no new fundamental information, which is itself a small negative force independent of the settlement's merits — the gap to central fair value widened from ~5% to ~8%.
One open item needs a primary-source check next pass: whether the "New Mexico #2, up to $62.85B" trigger from the baseline is still live, already resolved, or was a scoping error. Do not assume it away.

What to do

Hold both positions. The thesis that supported HOLD nine days ago is fully intact; the one thing that changed argues for slightly higher conviction, not for action.
Do not add at $571.10. It sits further above central fair value than nine days ago, and the FCF/capex question that dominates the valuation range is unresolved until 10-28 or 09-23.
Entry $465–505, strong buy below $450 — both nudged up ~2% with fair value, not with price.
Trim 23× fwd (from 22×) — a one-point increase reflecting the bounded legal tail, still capped well below the bull-case multiple until capex clarity arrives.

The falsifiable tests (carried from baseline, unchanged)

🟢 Upgrade to 7.0+: a disclosed 2027 capex number with an ROI framework · buybacks resume at scale · a named Meta Compute customer with a dollar figure (most likely at Connect, 09-23) · FoA operating margin stabilises ≥45% for two quarters.

🔴 Downgrade to 5.0 or below: 2027 capex guided >$200B · ad impressions turn negative · FoA operating margin <35% · a second extension of server useful life · a fifth straight quarter of dilution with no buyback · a material adverse ruling in the remaining personal-injury/bellwether litigation track that the settlement did not cover.

What this pass did NOT test

No new quarterly data — Q2 remains the latest print; every Trend/Structural row is carried by absence of contradicting evidence, not by re-verification against fresh numbers. Treat the 100% persistence figure accordingly.
New Mexico #2 ($62.85B claim, 09-08 trigger) — UNTESTED. Needs a court-docket or primary-legal-source check, not just news search, before the next pass either carries or retires this row.
Short interest — UNTESTED this pass; the baseline's "shorts covering" read was not re-pulled.
Full analyst-target re-tally — only directional moves were captured (three raised, one trimmed); no comprehensive consensus mean was recomputed this pass.
EC WhatsApp API interim-measure appeal status — not re-checked this pass; carried from baseline by default, not verified.
The phantom forward-EPS vendor trap persists unchanged. fin.py today shows PE(fwd) 16.39× implying forward EPS ~$34.85 — matching no published consensus year, same magnitude as the 15.58×/$34.90 figure flagged nine days ago. pitfall-vendor-forward-eps-is-the-wrong-fiscal-year continues to fire on this name; no correction has been made to the vendor field.

Sources

Primary/legal: CNBC — settlement · NPR — settlement · mdlupdate.com — MDL 3047 status · Yahoo — $17B settlement clears legal overhang, analyst views · MediaPost — FTC appeal brief · Law360 — Meta tells DC Circuit to reject FTC appeal · Santa Fe New Mexican — NM opted out of multistate deal

Market reaction: TradingKey — shares rise on settlement · Investing.com — UBS reiterates Buy · 247wallst — Cramer commentary

Data: .mcp/fin.py META --news (2026-08-27) · Yahoo Finance snapshot (price $571.10, PE ttm 21.50, PE fwd 16.39)

Knowledge base: pitfall-vendor-forward-eps-is-the-wrong-fiscal-year (confirmed persisting) · principle-down-a-lot-is-not-cheap · pattern-ranking-function-as-policy-variable (partially defused by the negotiated, not court-imposed, settlement terms)

Baseline chain: Output/Stocks/Technology/META/analyze-2026-08-18.md (HOLD 6.0) · Output/Stocks/Technology/META/analyze-2026-07-29.md (HOLD 6.5) · Output/Stocks/Technology/META/analyze-2026-03-21.md