Phase Space AI

Valuation

Meta Platforms [META]

Meta Platforms [META] — Valuation, re-underwritten 2026-07-29

Framework v1.5.1 · supersedes archive/03_Valuation_and_Factor_Scorecard_2026-07-27.md Spot $544.34 (post-print, 2026-07-29 after-hours; regular-session close $587.39, prior close $593.30) This memo issues no position verdict. It scores Criteria and outputs an analysis; the book decides.

Everything below is computed from Q2 2026 actuals filed today (8-K, 2026-07-29, EX-99.1) — a quarter that did not exist when the prior memo was written. The prior memo's anchor was FY2025.

---

1. What changed and why

Objectively-positive corrections: the share count, the net-cash rebuild, the exit-multiple identity, and the single committed terminal margin. Unproven methodology judgment: the choice of 39.5% rather than 41.4%, and the 40th-percentile exit multiple for the target.

---

2. Company state — A (mature-stable), with a declared caveat

Evidence for A: 19 consecutive fiscal years of revenue growth with one exception (FY2022, −1.1%); gross margin has sat in an 78.3–86.6% band for twelve years; operating margin has never been negative; the business is a single-product advertising monopsony with 3.60bn daily active people. Nothing here is venture-like (D) or scaling-observable-but-unproven (C).

Why not B (cyclical), despite the capex cycle: a cyclical designation requires a demonstrated peak-to-trough in demand. META has never had one — revenue grew 28% in the quarter just filed. What is behaving cyclically is free cash flow: $8,549m in Q2 2025 → $784m in Q2 2026, a 91% collapse, with FY2026 capex guided to $130–145bn against ~$140bn of operating cash flow. So:

State A on revenue and margin; the FCF series is behaving cyclically. This is precisely why the admissible instrument is EV/EBIT with a reverse-DCF cross-check, and not an FCF-based DCF. An FCF-based DCF anchored on a $784m quarter values META near zero; anchored on FY2024 it ignores the capex. Both are artefacts of where in the cycle you start. EBIT is the stable line.

---

3. Terminal EBIT margin — 39.5%, and why it is not below the trailing actual

Computed as operatingIncome / totalRevenue from Alpha Vantage normalized statements (66 quarterly, 19 annual periods) and cross-checked against the primary filings. AV's ebit field was not used — on META it runs +0.7pp to +2.6pp above true operating income (see §7).

Own demonstrated operating margin:

Opex bridge to the terminal year (FY2031), all lines taken from the primary filing, not AV:

m_EBIT,T 39.5% ≤ m_gross,T 76.0% ✓. Bridge closes: 76.0 − 26.0 − 5.5 − 5.0 = 39.5 ✓.

Is 39.5% below the trailing actual? It is above the TTM actual of 38.08% and 1.9pp below the FY2025 actual of 41.4%. That gap is deliberate and has a named cause: the depreciation load. META is converting ~$137bn/yr of capex into an asset base that must be depreciated through cost of revenue and R&D. FY2025's 41.4% was earned on a $176bn PP&E base; the terminal year will carry a base several times larger. A terminal margin at 41.4% would implicitly assume the AI asset base generates revenue at the same capital intensity as the 2019-vintage ad-serving fleet, which is the opposite of what the company says. 1.9pp below a cyclical peak is not the 19.5pp-below-trailing defect the audit found; that figure came from the prior memo's 19.0% Bear case being read as the terminal parameter.

---

4. Net cash — rebuilt from the primary balance sheet. EV is lease-inclusive.

Balance sheet, 2026-06-30 (8-K EX-99.1):

Declared: EV is lease-inclusive. At 2,566m shares the lease treatment is worth $11.17/share and $28.7bn of EV (2.0%). A lease-exclusive anchor previously invalidated a seven-name ladder in this project; this memo is on the inclusive basis and says so.

Deliberately excluded from net cash, each with a reason: - Non-marketable equity investments $30,157m — illiquid, unmarked, no observable exit. Adding them would take net debt to +$8.1bn net cash and flip the sign of the anchor. - Restricted cash $13,809m ($702m current + $13,107m in other assets, up from $1,662m a year ago). It is restricted; it is not available to retire debt. The 8x increase is itself notable and is flagged in the research doc as a disclosure to chase in the Q2 10-Q. - Finance lease liabilities — META does not break these out on the condensed balance sheet; they sit inside accrued expenses and other liabilities. Principal payments were $962m in the quarter ($1,805m H1), so the liability is material but unquantified in the source available today. Including it would make net debt more negative. Stated as a known omission, not resolved.

---

5. Reverse DCF — what $544.34 requires

reverse_dcf.py --spot 544.34 --shares 2566 --net-cash -22058 --revenue 228248 --years 5 --wacc 0.085 --terminal-margin 0.395 --exit-multiple 17.4

EV implied by the price: $1,418,834m = 6.2x TTM revenue = 16.3x TTM EBIT.

Exit multiple from the identity, not a free parameter: EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g) with t=0.17 (company's own guided 15–17%), g=0.045, ROIC=28%, WACC=8.5% → 17.42x. At g=0.035 / WACC=9.0% / ROIC=25% the identity gives 13.0x.

Sensitivity on the terminal margin, the disputed variable, at 13.5x: 39.5% → 12.4% required; 33.0% (FY2026 guided margin held forever) → 15.98% required, still +2.5pp inside the demonstrated 18.5%. The PASS survives holding the trough margin in perpetuity.

5a. The interim-cash-flow correction — and a live tooling defect

Every figure in the table above was solved without interim free cash flow, which overstates required CAGR in proportion to cash generation. That is a documented bug in this project's own reverse_dcf.py, recorded in DATA_DEFECTS.md as "Fixed; pass fcf_margin".

It is not fixed on the interface. fcf_margin is plumbed through project_ev() and solve() in the library, but it was never registered with argparse — --fcf-margin 0.10 exits with error: unrecognized arguments. Every agent invoking the documented CLI silently receives the old terminal-only answer. The framework's own pinned copy, framework/v1.7.0/assets/reverse_dcf.py, has no fcf_margin at all; the two copies have diverged, so framework_version does not identify which arithmetic a memo used. Both are filed as defects 12 and 13. The figures below were produced by importing the module and calling solve() directly.

META's demonstrated FCF margin: FY2025 21.7% (OCF $115,800m − capex $69,691m − finance-lease principal $2,524m = $43,585m on $200,966m); TTM 16.6%.

5% is adopted rather than the 16.6% trailing figure, and this is the conservative choice on purpose: FY2026 capex of $130–145bn against roughly $135bn of operating cash flow puts near-term FCF at approximately zero, recovering only later in the five-year window. Using the trailing margin would flatter the case by a further 1.8pp.

The defect's magnitude on META is 0.80pp of required CAGR at the adopted margin, 2.61pp at the trailing one. Note the direction: it inflates required CAGR and therefore understates the valuation margin. That is why it has survived three separate discoveries — it can never manufacture a false PASS, only suppress a true one.

Verdict: Valuation PASS by +13.0pp (+12.2pp on the uncorrected solve). The margin narrows from the prior +22.9pp because the prior figure was inflated by a forbidden peer multiple and a lease-exclusive EV, not because the business deteriorated. The PASS is robust across the entire grid: the worst cell tested — 13.0x exit, zero interim FCF, terminal margin cut to the guided 33.0% trough and held in perpetuity — still requires only 16.85% against 18.5% demonstrated, +1.6pp.

---

6. 12-month target — $734 (+34.9%), anchored on META's own multiple history

META's EV/EBIT on a lease-inclusive EV, weekly, 261 observations since 2021-08-02:

META trades at the 22nd percentile of its own five-year EV/EBIT range. That is the single most important valuation fact on the name and it is a fact about META, not about peers.

The two independent routes to the exit multiple corroborate each other. The identity-derived 17.4x sits at the 26th percentile of this same distribution, a shade above its p25 of 17.2x. A multiple derived from (1−t)(1−g/ROIC)/(WACC−g) and a multiple read off 261 weeks of META's own tape agree to within 1.1x. That is the check the prior memo's 31.6x peer anchor failed outright — 31.6x sits above the 95th percentile of META's own history, so the peer set was not describing this company.

Target build: - FY2027 revenue $300bn (+20% on FY2026's ~$249bn; the company just printed +28% and guided Q3 to $61–64bn) - FY2027 operating margin 32.0% — below FY2026's ~33% because 2027 carries the depreciation of the 2026 capex year. FY2027 EBIT $96.0bn. - Exit multiple 20.1x = the 40th percentile, not the median. Justification: the FCF collapse and a capex programme with no committed 2027 number are genuine reasons the market should pay less than its own median. A median-reversion target ($791) is stated but not adopted. - Net debt at target date $(45,000)m — capex ≈ OCF, leases grow to ~$35bn.

Target = (20.1 × 96,000 − 45,000) / 2,566 = $734, +34.9%.

Downside, named cause, MEASURED (does not reject the name): 2027 is the depreciation-recognition year for a $137bn capex programme while ad impression growth normalises. At p10 (10.6x) on a FY2027 EBIT of $75bn (28% margin), the target is $265, −51%. Precedent is META's own FY2022: operating margin 39.6% → 24.8% in four quarters.

---

7. Criteria

---

8. Defects found in this re-underwriting

1. AV longTermDebt returns NONE for META in the latest quarter while the true figure is $83,664m. Read naively → a $83.7bn EV error (5.9% of EV). AV's shortLongTermDebtTotal = $112,318m is usable and is exactly debt $83,664m + operating leases $28,654m, i.e. it is lease-inclusive.
2. AV commonStockSharesOutstanding = 2,566m is the diluted weighted-average share count, not shares outstanding (true outstanding 2,538.4m at the Q1 cover). Coincidentally the right number for a diluted EV, for the wrong reason.
3. AV OVERVIEW.SharesOutstanding = 2,196m is Class A only — it omits all 342.4m Class B shares, a 13.5% understatement. MarketCapitalization ($1,506bn) is simultaneously computed off ~2,564m shares. The two AV fields are internally inconsistent by 16.8%.
4. AV ebit ≠ operating income on META: +$334m to +$1,355m per quarter, +0.70pp to +2.64pp of operating margin, and NONE for Q2 2026. Not used anywhere here.
5. AV ebitda for Q2 2026 = $18,756m is BELOW ebit/operating income of $18,775m with positive D&A of $6,356m — arithmetically impossible. True EBITDA = 18,775 + 6,356 = $25,131m, a $6,375m / 25.4% understatement. Computed from the cash-flow-statement D&A as instructed.
6. AV capitalLeaseObligations and otherNonCurrentLiabilities return NONE for META; operating leases are recoverable only from the primary filing.
7. Prior memo internal inconsistency: shares_diluted_m = 2,538.0 in the manifest against "2,564,000,000 diluted" in its own analysis_summary — 1.0%, ~$15bn of EV.
8. Prior memo's DCF put 1.77% of EV in the terminal value and produced $78.01/share against a $593 price. For a State-A company that is a broken model, not a bearish one.
9. Prior memo's exit multiple (31.6x, peer growth-matched, n=30) is forbidden by the brief and violates the terminal-value identity at every defensible parameter set. It accounts for roughly 10.7pp of the prior +22.9pp valuation margin.
10. av_vs_edgar.py reports VERIFIED for META on 4 comparisons — and correctly so for revenue and net income — while AV's ebit and ebitda fields for the same quarters are wrong. The tool does not compare those fields. Empirical confirmation of the brief's warning that a consistency check has no power over the fields it does not test.
11. SPLITS endpoint, queried explicitly: META has NO recorded splits, ever (data: []). No split-basis risk on this name. Price history was pulled with adjustment=split regardless.