Meta Platforms META

Reference table

1. What the business actually is

The 10-Q is explicit about the structure: "We report our financial results for our two reportable segments: Family of Apps (FoA) and Reality Labs (RL). FoA includes Facebook, Instagram, Messenger, WhatsApp, and other services. RL includes our virtual and augmented reality related consumer hardware, software, and content." And on the revenue model: "Substantially all of our revenue is currently generated from advertising on Facebook and Instagram."

Revenue by segment (from seg_prod, which reconciles exactly to reported revenue):

Reality Labs is 1.1% of revenue and consumed $8.65 billion of operating profit in six months. The 10-Q states: "During the six months ended June 30, 2026, our RL segment reduced our overall operating profit by approximately $8.65 billion, and we expect our full-year 2026 RL operating losses to remain similar to 2025." The March-quarter figure was $4.03 billion. Critically for the Q2 diagnosis in Section 2, Reality Labs is not deteriorating — the 10-Q says its loss "increased $89 million, or 2%" in the June quarter and "decreased $92 million, or 1%" for six months.

The June-quarter operating detail from the 10-Q, which is the most important paragraph in this dive:

> "FoA income from operations in the three months ended June 30, 2026 decreased $1.58 billion, or 6%, compared to the same period in 2025, driven by higher FoA costs and expenses partially offset by an increase in revenue. The increase in costs and expenses was primarily due to increases in employee compensation, including severance expenses; infrastructure expenses related to our data centers, technical infrastructure, and third-party cloud services; legal-related costs; and third-party AI token costs."

Family of Apps revenue rose 28% and Family of Apps profit fell 6%. Four named causes, none of them Reality Labs.

Advertising operating metrics, June quarter (all from the 10-Q): Family daily active people 3.60 billion, +3%. Ad impressions delivered +14% (against +11% in the same period of 2025). Average price per ad +12% (against +9%). Advertising revenue +27% in the quarter and +30% for six months. Family of Apps other revenue — "paid messaging from WhatsApp and subscriptions" — +73%. Reality Labs revenue +16%, "driven by higher sales of AI glasses, partially offset by lower Meta Quest sales."

Both the volume and the price of advertising are accelerating simultaneously. That is the strongest fact on this page and it sits underneath a 26% drawdown.

Geography — and a vendor defect that must be corrected before the table can be used. The seg_geo block for FY2025 lists five entries: US & Canada $78.866B, Asia Pacific $53.817B, Europe $46.569B, Rest Of World $21.714B — and separately "UNITED STATES" $74.780B. The first four sum to exactly $200.966B, the reported total. The "UNITED STATES" line is a separate single-country disclosure and must NOT be added to the regional split; the vendor has merged two different disclosure bases into one dictionary. FY2024 contains only the four regions and sums correctly to $164.501B.

Foreign exchange helped. The 10-Q quantifies it: at constant rates, June-quarter revenue "would have been $60.12 billion ... which was $685 million ... lower than actual" — so roughly 1.1 points of the 28% growth was currency, and 4.5 points of the six-month figure ($2.43B).

Headcount 75,472 at 2026-06-30, "a decrease of 1% year-over-year," and the 10-Q adds that this "includes approximately 8,000 employees impacted by the May 2026 headcount reduction, the majority of whom will no longer be reflected in our headcount by the end of the third quarter of 2026." The vendor's profile.employees field of 78,865 is the 2025-12-31 figure from the 10-K and is now stale; we use 75,472.

2. The two distorted quarters — and what Meta actually earns

Neither of the last two headline earnings numbers means what it appears to mean, and the vendor file gives no indication of that.

Q1 2026: EPS $10.44 against an estimate of $6.70 — an apparent 56% beat. The 10-Q filed 2026-07-30 explains it:

> "Our effective tax rate for the six months ended June 30, 2026 was (5)%, compared to the U.S. federal statutory rate of 21%. The rate was lower than the statutory rate primarily due to a discrete income tax benefit of $8.03 billion recognized in the first quarter of 2026 in connection with U.S. Corporate Alternative Minimum Tax (CAMT) transitional relief provided under Treasury Notice 2026-7 ... The income tax benefit of $8.03 billion partially offsets the $15.93 billion discrete charge recognized in the third quarter of 2025 upon enactment of the One Big Beautiful Bill Act."

$8.03 billion on approximately 2.564 billion diluted shares is $3.13 a share. Strip it and Q1 EPS was roughly $7.31 against a $6.70 estimate — a real beat, of 9%, not 56%. The tell was on the face of the income statement and the vendor file carries it without comment: Q1 net income of $26.773B EXCEEDED operating income of $22.872B by $3.9 billion, which for a company with $562M of interest expense is only possible through a negative tax provision.

The mirror image sits in Q3 2025. The vendor's inc_q reports GAAP diluted EPS of $1.05 for the September 2025 quarter on operating income of $20.535B; the vendor's earn_cal reports $7.25 for the same quarter. The difference is exactly the $15.93 billion discrete charge: $15.93B ÷ 2.572B shares = $6.19, and $1.05 + $6.19 = $7.24. The vendor stripped the one-off charge from Q3 2025 but did NOT strip the one-off benefit from Q1 2026 — the two blocks are on inconsistent bases and nothing in the file says so (Section 8).

Q2 2026: EPS $6.18 against an estimate of $7.19 — a clear 14% miss, on revenue of $60.801B that BEAT the $60.224B estimate. The cause is not tax (the quarter's effective rate was a normal 16%) and it is not Reality Labs (loss up $89 million). It is Family of Apps costs:

• Severance: $1.18 billion. The 10-Q: "Employee compensation for the three months ended June 30, 2026 also includes $1.18 billion of severance expenses related to the May 2026 headcount reduction of approximately 8,000 employees which impacted both FoA and RL segments." At a 16% tax rate that is roughly $0.39 a share.
• Research and development +67% year over year in the quarter (+$8.71 billion) and +57% for six months (+$14.26 billion), "primarily due to higher employee compensation, infrastructure expenses related to our data centers, technical infrastructure, and third-party cloud services, and third-party AI token costs."
• Cost of revenue +33% (+$2.84 billion), "primarily due to higher infrastructure expenses."
• Legal-related costs, named explicitly as a driver in both the consolidated and the Family of Apps discussion.

Building the real run-rate:

Two conclusions we carry through the rest of the dive. First, the honest trailing multiple is about 19.9x, not 22.2x — the stock is cheaper than it screens. Second, the FY2026 consensus of $32.005 is NOT a clean earnings number: roughly $3.13 of it is a tax event that has already occurred and will not repeat. We therefore anchor the valuation on FY2027, which contains no such distortion.

Forward tax guidance, half-weighted as management framing but useful: the 10-K guided "our effective tax rate for the full year 2026 to be in the range of 13-16%"; the 10-Q now states "we expect our effective tax rate for the remaining quarters of 2026 to be between 15-17%." Gross unrecognised tax benefits are $18.74 billion (from $16.45 billion at year-end), with $12.73 billion that would benefit the provision if realised, and $2.97 billion of accrued interest and penalties. The long-running transfer-pricing dispute with the tax authority covering 2010 onward remains open.

3. Growth, margins and the inversion of operating leverage

Read the capex column. From 20.2% of revenue in 2023 to 34.7% in 2025 to a guided $130-145 billion in 2026 — roughly 51-57% of revenue. For comparison, the most capital-intensive cloud vendor in this batch spends 34.9%. Meta is now more capital-intensive, relative to revenue, than the companies it is buying the capacity from.

The quarterly series, where the inversion is visible:

The June quarter is the whole argument in one row. Revenue grew 28.0% year over year. Operating income fell 8.2%, from $20.441B to $18.775B. The operating margin fell 1,210 basis points, from 43.0% to 30.9% — and from 40.6% one quarter earlier. Gross margin is essentially unchanged at 81.4%, so this is entirely an operating-expense event, not a cost-of-revenue event.

Consensus forward path (analyst counts in brackets): revenue $253.950B 2026 (39), $304.548B 2027 (43), $358.937B 2028 (40), $428.582B 2029 (31). EPS $32.005 (41), $34.293 (41), $39.716 (34), $50.444 (30). Analyst coverage on this name is unusually deep and consistent — 30 to 43 estimates on every row through 2029 — which is a meaningful contrast with most names in this batch and one reason we lean on the FY2027 and FY2028 rows with some confidence.

Note the shape of the consensus, though: EPS growth of just +7.2% in 2027 against revenue growth of +19.9%. The street already models margin compression continuing through next year. The FY2027 estimate is not a heroic number; it is a cautious one.

4. Balance sheet — the vendor is wrong by $45.7 billion, and the real risk is off it

Correction one, and it is the largest single vendor error in this batch. The vendor reports netDebt of $48.024B. That is total debt ($83.897B) less cash and cash equivalents only ($35.873B). It silently discards $45.719B of short-term marketable securities. Meta's own 10-K states the correct figure in its highlights: "Cash, cash equivalents, and marketable securities were $81.59 billion as of December 31, 2025" and "Long-term debt was $58.74 billion as of December 31, 2025."

Counted properly:

> Cash and marketable securities $81.592B − total debt $83.897B = net debt of $2.305B, not $48.024B.

>

> On the filing's own "long-term debt" measure of $58.744B, Meta had NET CASH of $22.848B at 2025-12-31.

The vendor overstates net leverage by $45.719 billion. For a company with a $1.5 trillion market capitalisation this is the difference between "carries meaningful debt" and "essentially unlevered," and it propagates into enterpriseValueTTM and netDebtToEBITDATTM.

Correction two — enterprise value, the largest residual in this batch. enterpriseValueTTM reports $1,594.625B. Rebuilt (market cap $1,497.769B + total debt $83.897B − cash and marketable securities $81.592B) it is $1,500.074B. The vendor figure is $94.551B — 6.3% — too high, and every derived multiple carries the error: evToSalesTTM of 6.99x should be 6.57x, evToEBITDATTM of 14.89x should be roughly 14.01x, and netDebtToEBITDATTM of 0.905x should be approximately 0.043x.

Correction three — the finance-lease line is real and is NOT a double-count. We tested for the known failure mode: longTermDebt ($58.744B) and capitalLeaseObligationsNonCurrent ($22.940B) are different numbers, and the 10-K's separate statement of $59.0 billion of aggregate senior-notes principal confirms that the $58.744B is bonds and the $25.153B (current $2.213B plus non-current $22.940B) is genuinely finance leases. The composition is therefore: bonds $58.744B, finance leases $25.153B, total $83.897B. Meta's finance-lease book is large and growing because it leases data-centre capacity — but the vendor has recorded it correctly.

Debt composition and the trajectory:

Read the notes row. Senior notes went from $18.385B to $84.00 billion of principal in thirty months, maturing "from 2027 through 2066". The 8-K of 2026-05-04 documents the most recent tranche — six series maturing in 2031, 2033, 2036, 2046, 2056 and 2066 — and the 10-Q records $24.91 billion of net proceeds from it. Future interest obligations have risen from $59.72B to $89.38B in six months. Meta has gone from a net-cash company financing itself out of operations to a serial issuer of forty-year paper.

And $10.80 billion of the cash is no longer available. The 10-Q: "In connection with escrow requirements under certain multi-year infrastructure purchase agreements, $10.80 billion of money market funds was reclassified as restricted cash equivalents as of June 30, 2026. These funds are restricted from general corporate use and are expected to be released between 2028 and 2030." Adjusting for it, the $90.26B of cash and securities is $79.46B of usable liquidity against $84.00B of notes principal — Meta crossed into net debt during the June quarter on any strict reading.

Correction four — the numbers that are in no ratio at all, and they are the largest in this batch.

The 10-K, at 2025-12-31:

> "In addition to lease liabilities included in our consolidated balance sheets, we have leases that have not yet commenced, with total lease obligations of approximately $103.77 billion, mostly for data centers, colocations, and network infrastructure ... These operating and finance leases will commence between 2026 and 2030 with lease terms of greater than one year to 30 years."

The 10-Q, six months later:

> "In addition to the lease liabilities that are included on our balance sheet, we have operating and finance leases that have not yet commenced as of June 30, 2026. These lease obligations were approximately $278.99 billion, consisting of data centers, colocations, and certain network infrastructure, which will commence during the remainder of 2026 through 2036 with lease terms ranging from greater than one year to 30 years. In July 2026, we entered into additional data center leases with lease obligations of approximately $68 billion, which are expected to commence in 2027 and 2028, with lease terms of 18 to 20 years."

$103.77 billion → $278.99 billion → approximately $347 billion in seven months. And separately:

> "As of June 30, 2026, we had $349.31 billion of non-cancelable contractual commitments ... These commitments mostly relate to third-party cloud capacity arrangements and investments in servers and network infrastructure, data centers, and consumer hardware products in Reality Labs, with approximately $53.52 billion and $81.65 billion due in 2026 and 2027, respectively. In addition ... we have contingent obligations to purchase up to $14.72 billion of cloud capacity over a five-year period."

Against $131.05 billion of contractual commitments disclosed in the 10-K just six months earlier, with $30.63 billion due in 2026. The 2026 due amount alone has risen from $30.63B to $53.52B.

Total off-balance-sheet obligation: roughly $347 billion of uncommenced leases plus $349.31 billion of contractual commitments, against a market capitalisation of $1,497.8 billion and annual revenue of $200.966 billion. None of it appears in any vendor field, any leverage ratio or any risk score. This — not the debt — is the reason the risk score is 6.

Capital returns, and the suspension. The 10-Q:

> "Our board of directors has authorized a share repurchase program of our Class A common stock, which commenced in January 2017 and does not have an expiration date. We did not repurchase any shares of Class A common stock during the six months ended June 30, 2026. As of June 30, 2026, $25.03 billion remained available and authorized for repurchases."

Meta repurchased $30.125B of stock in 2024 and $26.248B in 2025 — 40 million shares — and then zero in the first half of 2026, with $25.03 billion of authorisation sitting unused. The dividend continues at $0.525 per quarter ($2.10 annualised, a 0.36% yield), $2.70B paid in the half.

Other balance-sheet facts, 2025-12-31: total assets $366.021B (from $276.054B); property, plant and equipment $196.804B, from $136.268B — a $60.5 billion increase in one year; goodwill $24.534B and intangibles $3.692B (only 7.7% of assets — a genuinely tangible balance sheet, unlike most of this batch); total equity $217.243B; retained earnings $121.179B. Book value per share $102.72 (5.72x), tangible book $93.52. Current ratio 2.23. Interest coverage 38.1x. Return on equity 29.7%, return on invested capital 17.1%.

Contingencies are NOT immaterial here. The 10-Q discloses an EUR €1.2 billion administrative fine from the lead European data-protection regulator over transatlantic data transfers, under appeal; a stayed proceeding by the domestic competition regulator to modify a consent order; and extensive privacy class actions. It states: "we have accrued significant amounts for loss contingencies related to these inquiries and investigations in Europe, and we believe there is a reasonable possibility that additional accruals for losses related to these matters could be material individually or in the aggregate." "Legal-related costs" are named as one of four drivers of the June-quarter operating-income decline. No specific aggregate accrual figure is extractable from the filing text, so we make no valuation adjustment — but we do not describe this as immaterial.

5. Cash flow — from $46 billion of free cash flow to approximately nothing

Operating cash flow is superb and rising — $64.09 billion in six months, which the 10-Q attributes to "an increase in cash collections from our customers driven by higher revenue, and a decrease in cash tax payments" (cash taxes paid were only $2.00 billion in the half against $7.58 billion for all of 2025, another consequence of the CAMT benefit). Free cash flow is collapsing anyway.

Do the arithmetic on the guidance. First-half capital expenditure was $50.92 billion including finance-lease principal ($49.11 billion of pure property and equipment purchases). Full-year guidance is $130-145 billion. That implies $79-94 billion in the second half — a 55-85% step up from the first. Even on a generous full-year operating cash flow of $140 billion, free cash flow lands somewhere between negative $5 billion and positive $10 billion, against $46.109 billion last year.

And the guidance itself has been raised. The 10-K, filed 2026-01-29: "We anticipate making capital expenditures of approximately $115 billion to $135 billion in 2026." The 10-Q, filed 2026-07-30: "We anticipate making capital expenditures of approximately $130 billion to $145 billion in 2026." A $15 billion increase at both ends of the range, in six months.

The financing response is unambiguous. The 10-Q: cash provided by financing "mainly consisted of $24.91 billion net proceeds from the issuance of the Notes in May 2026, partially offset by $8.70 billion of taxes paid related to net share settlement of RSUs, and $2.70 billion of payments of dividends" — and it names the swing factor directly: "the increase ... was mostly due to net proceeds from the May 2026 Notes and the absence of share repurchases in the current period."

Meta is now funding a portion of its capital programme with forty-year debt and with the cash it used to return to shareholders. That is a legitimate corporate decision. It is also, precisely and completely, why the stock has de-rated — and the knowledge base said so in March (Section 7).

Total shareholder yield is now 0.36% — the dividend alone. With the buyback suspended and $13.69 billion of six-month share-based compensation still vesting, the diluted share count (2.566B in the June quarter against 2.570B a year earlier) is essentially flat and will drift up if repurchases do not resume.

6. Valuation — priced in or room?

At $587.94 (market cap $1,497.8B, rebuilt enterprise value $1,500.1B):

Note the crossover in the two EPS rows and understand why it happens. On the reported basis, the multiple falls smoothly from 22.2x to 18.4x to 17.1x. On the clean basis it rises from 19.9x trailing to 20.4x for 2026 before falling to 17.1x in 2027. That is because the reported 2026 consensus banks a tax windfall that has already occurred. The genuine year of earnings growth is 2027, and that is where we anchor.

Peer context. The vendor peer set for a social-media and advertising company is close to useless: Applied Materials ($434.0B), Broadcom ($1,989.4B), Cisco ($479.8B), Micron ($1,008.2B), Taiwan Semiconductor ($2,163.6B), Garmin ($59.3B) and IBM ($221.5B) are semiconductor, networking and hardware businesses. Only Alphabet ($4,570.4B) is a genuine comparable, and no peer multiples are supplied in the file, so no peer-multiple comparison is drawn.

The vendor's scoring model rates Meta B+ with an overall score of 3 — return-on-equity and return-on-assets sub-scores of 5, debt-to-equity 3, price-to-earnings 2 and price-to-book 1. The price-to-earnings sub-score of 2 is difficult to justify at 17.1x forward and reads as a mechanical trailing-multiple output; the price-to-book of 1 reflects a 5.72x multiple on a book value that now contains $196.8B of data-centre assets.

6a. What today's price assumes (the inversion)

At $587.94 — 17.1x FY2027 consensus EPS of $34.293, and roughly 19.9x clean trailing earnings — the price embeds roughly the following falsifiable claims:

• Revenue compounds ~20% annually to $304.548B in 2027 and $358.937B in 2028. (Consensus-derived: 43 and 40 analysts.) This is the best-supported assumption on the page — advertising revenue grew 27% in the June quarter with impressions +14% and price per ad +12%, both accelerating against the prior year.
• The operating margin stabilises somewhere near the low thirties rather than continuing to fall. (Our number, from the quarterly series: 43.0% in Q2 2025 → 40.6% in Q1 2026 → 30.9% in Q2 2026.) This is the most fragile assumption in the price. Consensus EPS growth of just +7.2% in 2027 against revenue growth of +19.9% implies the street already expects further compression — but a 1,210-basis-point decline in a single year has no obvious floor, and the depreciation from a $130-145 billion capital programme has barely begun.
• Capital expenditure plateaus after 2026. (Our reading.) Nothing in the filings supports or contradicts this; no 2027 guidance has been given. What the filings do show is $81.65 billion of contractual commitments already due in 2027 and $68 billion of leases signed in July that commence in 2027-2028.
• Free cash flow returns. (Our number, derived from H1 actuals and the capex guide.) 2026 free cash flow is heading toward approximately zero against $46.109B in 2025. The price of a business is ultimately the cash it distributes; a permanent zero would justify a far lower multiple than 17x.
• The ~$347 billion of uncommenced leases and $349.31 billion of commitments convert into revenue-generating capacity rather than stranded cost, and the legal accruals in Europe and the state-court product litigation do not become material. (Filing-derived; the 10-Q states additional European accruals "could be material individually or in the aggregate.")

6b. The return bridge (why the multiple moves)

Expected return over 12-24 months decomposes as: EPS growth (+7.2% in FY2027 and +15.8% in FY2028 on consensus, from a clean 2026 base of ~$28.87 the two-year growth is closer to +37%) + multiple drift (we assume EXPANSION, from 17.1x to ~20.1x forward) + shareholder yield of 0.36% (dividend only, because the buyback is suspended).

Be explicit that the base case requires multiple expansion, and that this is the fragile leg. From $587.94, if 2027 EPS lands at the consensus $34.293 and the multiple simply holds at 17.1x, the price is $588 — flat. Our base of $690 requires 20.1x, i.e. three turns of expansion.

The argument for it is that the compression has already happened and was driven by a specific, dateable fear. Meta traded at $790.00 as recently as within the last twelve months and has fallen 25.6% while revenue grew 22-33% a quarter. That is definitionally a de-rating, and our knowledge base names the cause precisely and contemporaneously: "Already seeing multiple compression with Meta, the biggest of the CapEx names; rising debt/CapEx and falling free cash flow raise the bar, hurt multiples" (2025-11-16, skill 2.0). A de-rating driven by a fear about future free cash flow reverses when free cash flow stops falling — not when it becomes wonderful, only when it stops falling.

Cross-checks. $690 is 17.4x FY2028E of $39.716 and 23.3x clean trailing earnings of $29.61 — neither demanding. It sits 4.7% BELOW the street consensus of $723.69 and 4.8% below the median of $725. We are the conservative party relative to 65 covering analysts, and deliberately so.

The bull case at $840 requires 24.5x FY2027E — below the street high of $883 — and needs capital expenditure to plateau, the margin to stabilise, and the buyback to resume.

6c. Variant perception (where we differ, what would surprise)

• We differ from the vendor file, and from anything built on it, on the balance sheet — by $45.7 billion. Screens show net debt of $48.024B and an enterprise value of $1,594.625B. The correct figures are net debt of $2.305B and an enterprise value of $1,500.074B, confirmed by the 10-K's own statement that "cash, cash equivalents, and marketable securities were $81.59 billion" against "long-term debt ... of $58.74 billion." Meta is not a levered company; it is an unlevered company making an enormous discretionary investment. That is a completely different risk with a completely different remedy — it can be stopped. Watchable number: cash and marketable securities against notes principal at each filing; $90.26B against $84.00B at 2026-06-30, with $10.80B of the former now restricted.
• We differ from the headline earnings on both of the last two quarters, in opposite directions. The Q1 "beat" was 56% reported and roughly 9% real, once the $8.03 billion discrete tax benefit is removed. The Q2 "miss" of 14% included $1.18 billion of severance for a permanent cost reduction of approximately 8,000 employees — an expense that improves 2027. Clean trailing EPS is ~$29.61 against a reported $26.54, so the stock screens 11.6% more expensive than it is. Watchable number: the effective tax rate, guided to 15-17% for the remaining quarters after negative 5% in the half.
• We differ from the more relaxed framings on the size of the off-balance-sheet commitment, and we think it is under-appreciated relative to the debt. Uncommenced leases went $103.77B → $278.99B in six months, plus ~$68B in July; contractual commitments went $131.05B → $349.31B. That is roughly $696 billion of forward obligation against $200.966B of annual revenue — and, unlike the cloud vendors buying the same hardware, Meta discloses no contracted revenue backlog against it at all. Watchable numbers: both disclosures at the 2026-10-28 filing.
• We do NOT have a variant perception on the destination, only on the path and the accounting. Base $690 against a street consensus of $723.69 is a 4.7% gap — essentially agreement. Our edge here is the balance-sheet correction, the earnings-quality correction and the willingness to buy at an RSI of 26.2, not a differentiated view of 2028. That should moderate position size.
• Positive surprise that would force a re-rate: the share repurchase programme restarting — $25.03 billion is authorised and unused, and its resumption would be the single clearest statement that management sees a ceiling on the capital cycle. Alternatively, 2027 capital-expenditure guidance coming in flat or below 2026, or Family of Apps operating income returning to growth. Watchable numbers: repurchase activity in the Q3 10-Q; Family of Apps income from operations; any 2027 capex framing.
• Negative surprise that would break the thesis: a third consecutive increase in the capital-expenditure guide; Family of Apps operating income declining for a second and third quarter; ad impression growth or price-per-ad growth decelerating (both are currently accelerating); a material European legal accrual; or an adverse verdict in the state-court product-addiction litigation that one tracked source describes as capable of snowballing. Watchable numbers: impressions and price per ad at each print; the legal-contingency language in each filing.

Synthos fair values

All three anchors are multiples of FY2027 consensus EPS of $34.293 (41 analysts) — chosen over FY2026 because the FY2026 consensus embeds roughly $3.13 a share of non-recurring tax benefit — cross-checked against FY2028E of $39.716 (34 analysts) and clean trailing EPS of ~$29.61. Not a discounted cash flow; stated arithmetic.

• Bear ~$480 — 14.0x FY2027E. Cross-checks: 12.1x FY2028E; 16.2x clean trailing. The scenario: 2027 capital expenditure is guided higher again, free cash flow is negative for a full year, Family of Apps operating income declines for three more quarters, the uncommenced-lease book keeps compounding, and the market prices a permanently capital-intensive advertising business. Note $480 is 8.7% below the 52-week closing low of $525.72 — a genuine bear case, not a token one. −18.4%.
• Base ~$690 — 20.1x FY2027E. Cross-checks: 17.4x FY2028E; 23.3x clean trailing; 13.7x the 2029 consensus of $50.444. Sits 4.7% below the street consensus of $723.69 and 4.8% below the median. The scenario: revenue compounds near 20%, the operating margin stabilises in the low-to-mid thirties as the severance action takes effect, capital expenditure plateaus in 2027, free cash flow re-emerges, and the compression of the last twelve months partially unwinds. +17.4%, plus 0.36% dividend.
• Bull ~$840 — 24.5x FY2027E. Cross-checks: 21.2x FY2028E; below the street high of $883. The scenario: capital expenditure guidance for 2027 comes in flat, the buyback resumes with the $25.03 billion authorisation, margins recover toward 40%, and the market pays a premium multiple for the fastest-growing advertising franchise at scale. +42.9%.

Base is 17.4% above spot; asymmetry roughly 2.3:1 to the upside (18% down, 43% up) — the best risk-reward ratio in this batch, which is what a 26.2 RSI is supposed to produce. The destination supports ownership; the absence of any catalyst for 85 days argues for accumulating rather than committing. Buy — Tactical, staged.

7. Knowledge base — 109 claims, and a dated bear call that came true

Tracked knowledge-base claims naming Meta: 139 raw hits, of which 109 are entity-level. Searches covered META (case-sensitive, 69 matches), Facebook (22), Instagram (2), Zuckerberg (2), Meta Platforms, WhatsApp, and the company's open model-family name (14) across all 51,928 records.

Discards and weighting, stated precisely because the pattern set was broad. Thirty raw hits were thesis- or category-level only — the company was named inside an argument about something else — and are excluded from the entity count. A further and important qualification: 40 of the 139 pattern matches came from product-, person- or model-level terms rather than the ticker or company name. In particular, fourteen matched on the name of Meta's open model family, and most of those claims are about open-weight model development, distillation and the competitive landscape between model developers — they are claims about a technology programme, not about the equity, and they carry no valuation weight even where they are entity-tagged. We identify them wherever they appear and do not cite them as investment evidence. Sixteen management-lane claims (skill 0.5) are half-weighted: fourteen from Meta's chief executive, one from a competing chip company's management, one from a rival model developer's management.

Stance mix: 68 bullish, 18 neutral, 23 bearish across 31 distinct sources, spanning 2016-10-07 to 2026-08-01. The full-history mix reads positive, and we do not use it that way. The lane is fifteen years long and most of the bullish claims are historical observations about what a good business Facebook has been. The 2026 lane is close to evenly split, and it is the one that matters, because it argues about exactly the variable that has moved the stock.

The single most valuable claim in this file is dated, specific, bearish, and has now largely been proved right:

> 2026-03-30 · anthony_pompliano_show · bearish · conviction 72

> "Meta flipped from cash cow to capital-intensive cash burner; its $50-70B of free cash flow is going down to about $5B."

Made 2026-03-30. Free cash flow in 2025 was $46.109 billion; first-half 2026 free cash flow was roughly $13.2 billion against full-year capital-expenditure guidance of $130-145 billion, which implies a full-year figure close to zero. That claim was a forecast when it was written and it is now nearly a description. A dated, falsifiable, correct call is worth more than a dozen general endorsements, and it is a direct input to the risk score.

The highest-skill voice in the file (skill 2.0) has been bearish or non-committal for nine consecutive months and has been right:

> 2025-11-16 · jordi_visser (skill 2.0) · bearish · conviction 65

> "Already seeing multiple compression with Meta, the biggest of the CapEx names; rising debt/CapEx and falling free cash flow raise the bar, hurt multiples."

> 2026-02-14 · jordi_visser (skill 2.0) · bearish · conviction 70

> "The four hyperscalers spending $650B this year are the line in the sand; if AI capex revenues don't materialize, their equity falls via multiple compression."

> 2026-07-04 · jordi_visser (skill 2.0) · bearish · conviction 72

> "Meta pivoting to a compute/data-center company after overspending on talent is throwing stuff at the wall; it's a cash-flow floor on the stock, not a reason for multiple expansion."

> 2026-07-11 · jordi_visser (skill 2.0) · neutral · conviction 45

> "Not convinced on Meta yet—it doesn't have the same technical picture as Apple; watching the hyperscalers much more closely instead."

The sharpest analytical objection in the entire file, and the one we have not been able to refute from the filings:

> 2026-05-04 · andreas_steno · bearish · conviction 62 · skill 0.8

> "Unconvinced by Meta's capex: unlike [the cloud vendors] it has no backlog justifying the spend, and its ad-price-from-AI story fails an empirical test since [its assistant] is basically irrelevant in consumer usage."

This is correct on the first half and we verified it. The cloud vendors covered elsewhere in this batch disclose contracted remaining performance obligations against their capital programmes. Meta discloses none — no backlog of any kind — against approximately $347 billion of uncommenced leases and $349.31 billion of contractual commitments. It has an advertising business instead, which is a different and in some ways better answer, but it is not a contracted one.

The bull lane's strongest current claims are about price and about the quality of the underlying franchise:

> 2026-07-28 · Michael Batnick · bullish · conviction 55

> "The business is on fire and trades at 16x forward earnings; if they merely maintained current capex guidance the stock would go up 10% or more."

> 2026-07-28 · Michael Batnick · bearish · conviction 60

> "Reels monetization will be insane, up 30% YoY, a monster number — but it won't matter; investors only care about capex and free cash flow, so they'll punch the stock at the open."

These two claims were made by the same source on the same day, the day before the print, and both were right. The business was on fire — advertising revenue grew 27% — and the stock was punched anyway. That pair is the most accurate single description of this equity we have, and it also tells you what has to change for the stock to work: not the business, the capex.

> 2026-07-31 · we_study_billionaires · bullish · conviction 55

> "Facebook is the best business in history — makes ~$50-60 per user on ads in North America and prints money; still the best ad ROI, where small budgets should go first."

> 2026-06-15 · gavin_baker · bullish · conviction 55

> "Meta bull case: potentially the greatest business in history not threatened by [artificial intelligence] but accelerated by it — though its actual super-intelligence/coding strategy remains unclear and uncommunicated."

> 2026-03-31 · compound_and_friends · bullish · conviction 60

> "Meta at 16x earnings after a 9% single-day drop is mispriced; buyers came back with a 7% rebound."

Note the "16x" figure appearing in two independent sources four months apart. Both are forward multiples on a consensus that includes the tax benefit; on our clean basis the comparable number is nearer 20x for 2026 and 17.1x for 2027. The tracked lane is systematically quoting a multiple about three turns too cheap, and we correct it rather than adopt it.

The most useful neutral claim, because it is the honest middle:

> 2026-07-03 · compound_and_friends · neutral · conviction 45

> "Not a believer in Meta's AI reinvestment (open [model], workplace tools, renting out compute)—best ad business in the world but bad at redeploying cash; only tempted as a ~16x value play."

> 2026-06-15 · gavin_baker · neutral · conviction 50

> "Market's low EV/net-PP&E on Meta reflects warranted skepticism about monetizing its asset base, but Zuckerberg will likely pivot to external GPU monetization."

One risk the filings understate and the lane surfaces:

> 2026-07-31 · compound_and_friends · bearish · conviction 45

> "Meta's addiction lawsuits, playing out in state courts, are roughly 13% of costs and could snowball once one plaintiff wins — a very big problem going forward."

We half-weight this — the "13% of costs" figure is not verifiable from the filings and looks like a misstatement — but the underlying point is corroborated: the 10-Q names "legal-related costs" as one of four drivers of the June-quarter operating-income decline, and states that further European accruals "could be material individually or in the aggregate."

Management-lane claims, half-weighted and not used as valuation inputs. Fourteen claims from the chief executive describe multi-gigawatt data-centre construction "funded entirely from company cash flow, unlike competitors who must raise capital" (2025-07-16) — a statement that the 10-Q now contradicts: $24.91 billion of net note proceeds in May 2026 and no buyback. They also describe compute-per-researcher advantages, wearable form factors, and open-weight distillation. We record the contradiction because it is the clearest available illustration of why management framing is an input and not evidence.

Why Moderate-High rather than High. The lane is deep (109 claims, 31 sources, ten years), the sources are strong, and it contains a dated bear call that has been vindicated and a dated bull valuation call that is live. What holds it below High is that the highest-skill voice has been negative or non-committal on this name for nine months and has been correct throughout, and the bull lane systematically quotes a forward multiple about three turns too cheap. The lane supports buying the drawdown; it does not support buying it aggressively.

8. Data integrity — what we rejected from the vendor file and why

Nine items. Five are direct filing-versus-vendor contradictions, and in every one the filing wins.

1. FILING OVERRIDES VENDOR — netDebt is overstated by $45.719 billion. The vendor reports $48.024B, computed as total debt ($83.897B) less cash and cash equivalents only ($35.873B), discarding $45.719B of short-term marketable securities. The 10-K states in its own highlights: "Cash, cash equivalents, and marketable securities were $81.59 billion as of December 31, 2025" and "Long-term debt was $58.74 billion." Correctly computed: net debt of $2.305B on total debt including finance leases, or NET CASH of $22.848B on the filing's own long-term-debt measure. This is the largest single vendor error in this batch. netDebtToEBITDATTM of 0.905x should be approximately 0.043x.

2. FILING OVERRIDES VENDOR — enterprise value is 6.3% too high, the largest residual in this batch. enterpriseValueTTM of $1,594.625B against a rebuild of $1,500.074B (market cap $1,497.769B + total debt $83.897B − cash and marketable securities $81.592B) — a $94.551B overstatement carrying directly from item 1. We use $1,500.074B throughout, which corrects evToSalesTTM from 6.99x to 6.57x and evToEBITDATTM from 14.89x to approximately 14.01x.

3. FILING OVERRIDES VENDOR — the payload contains no record of ~$347 billion of uncommenced leases or $349.31 billion of contractual commitments. The 10-K discloses "leases that have not yet commenced, with total lease obligations of approximately $103.77 billion" at 2025-12-31 and "$131.05 billion of contractual commitments." The 10-Q discloses $278.99 billion of uncommenced leases at 2026-06-30 "which will commence during the remainder of 2026 through 2036 with lease terms ranging from greater than one year to 30 years," plus "additional data center leases with lease obligations of approximately $68 billion" entered in July 2026, and $349.31 billion of non-cancelable contractual commitments with $53.52B due in 2026 and $81.65B in 2027, plus $14.72B of contingent cloud-capacity obligations. None of this appears in any vendor field or leverage ratio, and the six-month rates of change — $103.77B to $278.99B, and $131.05B to $349.31B — are the single most important facts about the risk profile.

4. FILING OVERRIDES VENDOR — the earnings blocks treat two comparable one-off tax items inconsistently, and nothing labels either. For Q3 2025 the vendor's inc_q shows GAAP diluted EPS of $1.05 while earn_cal shows $7.25; the $6.20 difference is precisely the $15.93 billion discrete tax charge ($15.93B ÷ 2.572B shares = $6.19) disclosed in the 10-Q. For Q1 2026 both blocks show $10.44 — the vendor did NOT strip the offsetting $8.03 billion discrete tax benefit ($3.13 a share) disclosed in the same footnote. The result is that the "enormous beat" in Q1 and the "collapse" in Q3 2025 are artefacts of inconsistent treatment. We use a clean trailing EPS of ~$29.61 against the reported $26.54 and state both.

5. FILING OVERRIDES VENDOR — seg_geo merges two different disclosure bases into one dictionary. The FY2025 entry contains four regions (US & Canada $78.866B, Asia Pacific $53.817B, Europe $46.569B, Rest Of World $21.714B) that sum exactly to reported revenue of $200.966B, and a fifth entry, "UNITED STATES" $74.780B, which is a separate single-country disclosure. Adding all five overstates revenue by 37%. We use the four-region split and flag the United States figure as a memo item only.

6. est net-income and EBIT rows are REJECTED as internally incoherent; epsAvg used exclusively. The FY2024 row shows netIncomeAvg of $63.434B against ebitAvg of $60.885B — net income exceeding operating income, which is not possible for a taxpayer. More usefully, the implied share counts from dividing netIncomeAvg by epsAvg are incoherent across rows: 2.510B (FY2026), 2.698B (FY2027), 2.803B (FY2028), 2.574B (FY2029), 2.574B (FY2030) — a count that rises 12% and then falls back. The epsAvg and netIncomeAvg series are not built on a common share base. We use epsAvg and revenueAvg only. The EBITDA and EBIT rows pass the elementary ordering test in every forward year, which is uncommon in this feed and worth stating, but we still do not quote them.

7. Quarterly depreciation and amortisation is MISSING for the most recent quarter. The vendor's inc_q shows D&A = 0.00 for the quarter ended 2026-06-30 — a null, not a zero. The 10-Q states $12.35 billion of depreciation and amortisation for the six months, and the vendor's Q1 figure is $5.999B, so the June quarter was approximately $6.35B. We use the filing.

8. profile.employees of 78,865 is stale. That is the 2025-12-31 figure from the 10-K. The 10-Q states headcount of 75,472 at 2026-06-30, "a decrease of 1% year-over-year," including "approximately 8,000 employees impacted by the May 2026 headcount reduction, the majority of whom will no longer be reflected in our headcount by the end of the third quarter of 2026." We use 75,472 and note that the effective figure by the September quarter will be materially lower.

9. Two minor internal inconsistencies, flagged once. (a) quote.yearHigh of $796.25 and yearLow of $520.26 are intraday extremes; tech.hi52 of $790.00 and tech.lo52 of $525.72 are closing extremes. We use the closing series throughout, so the drawdown is 25.6% rather than 26.2% and the distance above the low is 11.8% rather than 13.0%. (b) quote.sma50/sma200 ($602.18/$634.49) differ trivially from tech.sma50/sma200 ($601.50/$633.25); we use the tech block.

Not defects, correctly reported: freeCashFlowYieldTTM of 2.74% and priceToFreeCashFlowRatioTTM of 36.6x look poor for a business with an 82% gross margin and are correct — they reflect the genuine collapse in free cash flow documented in Section 5, and the forward figures are worse, not better. capexToOperatingCashFlowTTM of 0.686 and capexToRevenueTTM of 0.391 are likewise correct and are the numbers the whole thesis turns on.

Non-equity tripwire: checked and clear. META is common stock (Class A), beta 1.246, average volume 17.9M shares, a fixed but recently increased dividend ($0.525 quarterly, raised from $0.50 in the first quarter of 2025), and a 52-week closing range of $525.72-$790.00 — a 50% band. Nothing resembling a preferred or baby-bond profile. Note the dual-class structure: the 8-K of 2026-05-29 records that shareholders rejected a proposal to eliminate it, so voting control remains concentrated.

9. Technicals — the deepest drawdown and the lowest RSI in the batch

• Price $587.94, down 0.39% on the day, on volume of 16.7M against a 17.9M average — a quiet session. Intraday range $578.25-$592.79.
• −25.6% from the 52-week closing high of $790.00; only +11.8% above the 52-week low of $525.72. Maximum drawdown from peak over the trailing year is −25.6% on closing prices — the deepest in this batch.
• BELOW both moving averages: −2.3% under the 50-DMA of $601.50 and −7.2% under the 200-DMA of $633.25. The 50-day is itself below the 200-day, a downtrend structure.
• RSI 26.2 — deeply oversold and the lowest reading in this batch by a wide margin. MACD −8.96. Readings below 30 are uncommon in mega-cap equities and historically mark points of maximum discomfort rather than points of maximum risk.
• Relative performance, and it is the defining fact: 12-month −21.6% against SPY +24.3% and QQQ +30.8% — deficits of 45.9 and 52.4 points; 6-month −16.8% against SPY +11.1% and QQQ +15.7%; 3-month −3.7% against SPY +7.6% and QQQ +7.7%. The underperformance is continuous across all three windows, so this is a trend and not a single bad print.
• The street's position is worth stating against the chart. Sixty-five analysts produce a consensus of $723.69 (+23.1%), a median of $725, a high of $883 and a low of $595 — 1.2% above spot. The entire published target band sits above the price. That is either a collective failure to mark to market or a genuine dislocation; the honest reading is some of both, since the low target is only just above the price and will likely be cut if the October print disappoints.
• Practical read, and it is the operative point of this dive. Meta has fallen for twelve months while its revenue accelerated. The fall is explained — capital expenditure, free cash flow, a suspended buyback — and the explanation does not resolve for 85 days. An RSI of 26.2 in a business growing advertising revenue 27% is where accumulation happens, not where confirmation happens. The 50-DMA at $601.50 is 2.3% above and would be the first evidence of stabilisation; the 52-week closing low at $525.72 is 10.6% below and is where the bear case would be priced rather than argued. This argues for a staged purchase — an initial tranche now, a second on any test of the low, and a third reserved for the 2026-10-28 print — rather than a single trade into a downtrend. That is what makes the verdict Tactical rather than Core.

10. Insiders — one officer, tiny amounts, and a conspicuous absence

The honest reading is that this table contains almost nothing. Every record is the same officer, on the same two dates, selling the same four parcels (837, 408, 82 and 57 shares) across direct and indirect holdings — 2,768 shares in total across both dates, worth roughly $1.7 million against a $1.5 trillion company. The identical, repeating parcel sizes at weekly intervals are the signature of a pre-arranged Rule 10b5-1 plan, not a decision. We do not read it as a view.

Two observations that are worth one sentence each. First, both dates executed above today's price — $645.85 and $607.85 against $587.94 — so the shares have fallen 9.0% and 3.3% since. Second, and more notably, through a twelve-month period in which the stock fell 21.6% and 25.6% from its high, the insider file contains no transaction by the chief executive, the chief financial officer or any director at all. For a founder-controlled company with a long history of programmatic executive selling, an absence of selling into a drawdown is weak positive evidence — and we weight it as weak. The alternative, heavy insider selling into a 26% decline, would have been meaningful; it did not occur.

11. Verdict, kill-criteria and flip conditions

Buy — Tactical.

The business is performing. The stock is not. The reason is documented, dateable, and does not resolve for 85 days.

What the filings established, five and six days ago. Advertising revenue up 27% in the June quarter and 30% for six months, with ad impressions delivered up 14% (against 11% a year earlier) and average price per ad up 12% (against 9%) — both accelerating simultaneously. Family of Apps revenue up 28%; Family of Apps other revenue — paid messaging and subscriptions — up 73%. 3.60 billion daily active people. Gross margin 81.4%. Return on equity 29.7%. Net debt of $2.305 billion, not the $48.024 billion the vendor reports. Sixty-five covering analysts with a consensus of $723.69 and a low target of $595 — the entire band above the price. RSI 26.2.

What you are accepting, stated without softening. Family of Apps income from operations FELL 6% while its revenue rose 28%. Total operating margin fell from 43.0% to 30.9% in a year. Free cash flow was $46.109B in 2025 and roughly $13.2B in the first half of 2026, against full-year capital-expenditure guidance of $130-145 billion — raised from $115-135 billion six months earlier. The buyback stopped completely: "we did not repurchase any shares of Class A common stock during the six months ended June 30, 2026", with $25.03 billion authorised and unused. Uncommenced lease obligations went from $103.77B to $278.99B in six months, with ~$68B more signed in July; contractual commitments went from $131.05B to $349.31B; and Meta discloses no contracted revenue backlog against any of it. Senior notes went from $18.4B to $84.0 billion of principal in thirty months, now maturing out to 2066. And Meta's own risk factors state that these investments "have the effect of reducing our cash flows, operating margin and profitability." Our knowledge base called this in March 2026 — "flipped from cash cow to capital-intensive cash burner" — and was right.

Why Tactical rather than Core, and why not Watch. Base fair value of $690 is 17.4% above spot with roughly 2.3:1 asymmetry — the best risk-reward in this batch — and sits 4.7% below a street consensus of $723.69, so we are the conservative party. The reason this is not Watch is that an RSI of 26.2 on a business growing advertising revenue 27%, with a corrected balance sheet showing essentially no net debt, is a price at which accumulation is the correct action rather than observation. The reason it is not Core is that there is no catalyst for 85 days, the operating-margin decline has no demonstrated floor, and the single variable that matters — 2027 capital expenditure — has not been guided. Buy it in tranches into the weakness.

Pre-registered KILL criteria — what would break the thesis:

• A third consecutive increase in capital-expenditure guidance, or a 2027 guide above the 2026 range of $130-145 billion. Two raises is a build; three is a business-model change.
• Family of Apps income from operations declining for a second and third consecutive quarter. One quarter is a severance charge; three is inverted operating leverage.
• Full-year 2026 free cash flow below zero, or any indication that 2027 free cash flow will also be negative.
• Ad impression growth or price-per-ad growth decelerating. Both are currently accelerating (14% and 12% against 11% and 9%) and they are the load-bearing facts in the bull case.
• The uncommenced-lease disclosure rising materially above ~$350 billion, or contractual commitments above ~$400 billion, without any disclosure of contracted revenue against them.
• A material adverse development in the state-court product-addiction litigation, or a European accrual large enough to be quantified separately.
• A capital raise in equity, or notes issuance at a scale that takes principal materially above $100 billion.

Pre-registered FLIP TO BUY — CORE (upsize) criteria:

• The share repurchase programme restarting. $25.03 billion is authorised and unused; resuming it would be the single clearest statement that management sees a ceiling on the capital cycle, and it is the highest-signal event available.
• 2027 capital-expenditure guidance at or below the 2026 range.
• Family of Apps income from operations returning to year-over-year growth, which the ~8,000-person reduction should assist from the September quarter onward.
• Free cash flow positive and rising for two consecutive quarters.
• A test of the 52-week closing low of $525.72 with the advertising metrics intact — this is the second most likely path to a full-size position and the explicit reason a tranche is held back.
• Any disclosure of contracted revenue against the infrastructure investment — the absence of one is the most articulate bear argument in the tracked lane, and its arrival would remove it.

Where META fits in the Synthos Framework Portfolio. The mega-cap platform sleeve, at a 2.5-3% target weight built in three tranches over 8-12 weeks — roughly one third now at $587.94, one third on any test of the $525-545 area, one third reserved until the 2026-10-28 print. The house standing rule that mega-cap incumbents are guilty until proven innocent on a forward basis applies with full force here, and Meta's defence is partial: the forward evidence on the revenue line is excellent and accelerating, and the forward evidence on the earnings line is absent. What earns it a place is that the price already reflects a great deal of the second problem — a 46-point twelve-month deficit to the index while revenue accelerated — and that the balance sheet, correctly computed, gives management complete freedom to stop spending whenever it chooses. That optionality is not available to a levered company, and the vendor file conceals it.

On overlap within the batch. Meta and Broadcom (also covered 2026-08-04) are the two sides of the same capital cycle and are held for opposite reasons. Broadcom sells the infrastructure, is up 44.9% over twelve months, gapped 6.61% today, trades at 69.7x trailing GAAP earnings, and discloses a $164.6 billion contracted order book. Meta buys it, is down 21.6% over twelve months, has an RSI of 26.2, trades at roughly 19.9x clean trailing earnings, and discloses no backlog at all against roughly $347 billion of uncommenced leases. We rate both Buy — Tactical, at 12.4% and 17.4% base upside respectively. They are deliberately not redundant: holding both is a hedge against being wrong about which end of the capital cycle captures the economics, and if the spending proves malinvestment, the seller keeps the cash and the buyer keeps the depreciation. Logged as a tracked Synthos call (Buy — Tactical) as of 2026-08-04 at $587.94.

Single biggest risk: that approximately $696 billion of forward obligation — ~$347 billion of uncommenced leases plus $349.31 billion of contractual commitments — has no contracted revenue behind it. Every cloud vendor buying the same hardware discloses a signed backlog against its capital programme. Meta discloses none, because its answer is an advertising business rather than a customer contract. That answer may well be better — advertising revenue is growing 27% and both the volume and the price of ads are accelerating — but it is an inference, not a commitment, and it is not testable in advance. The tracked knowledge base states the objection exactly and we have not been able to refute it from the filings.

Most fragile assumption in the price: that the operating margin stabilises near the low thirties. It was 43.0% in the June quarter of 2025, 40.6% in the March quarter of 2026, and 30.9% in the June quarter of 2026 — a 1,210-basis-point decline in twelve months, of which only about 190 basis points is the $1.18 billion severance charge. Every element of the valuation — the 17.1x forward multiple, the consensus EPS path, the base case of $690 — assumes that slide arrests within a quarter or two. The depreciation from a $130-145 billion capital programme has barely begun to flow through the income statement, and there is no disclosed floor.

Provenance & disclosures

• Traceability: 109 entity-level knowledge-base claims name Meta (139 raw hits; 30 thesis- or category-level mentions excluded; breadth 31 distinct sources; net conviction mixed). Stance mix 68 bullish / 18 neutral / 23 bearish across the full 2016-10-07 to 2026-08-01 range — a positive full-history reading that we explicitly decline to use, because the 2026 lane is close to evenly split and is the relevant one. A further and material qualification: 40 of the 139 pattern matches came from product-, person- or model-level search terms rather than the ticker or the company name — 22 on the flagship application's name, 14 on the open model family's name, 2 on the chief executive's surname and 2 on a second application's name. The fourteen model-family matches in particular are claims about an open-weight technology programme rather than about the equity; they are identified, excluded from valuation reasoning, and not quoted as investment evidence. Sixteen management-lane claims (skill 0.5) are half-weighted — fourteen from the chief executive, one from a competing chip company's management, one from a rival model developer's management — and none is used as a valuation input; one is explicitly contradicted by the 10-Q and we say so. Searches covered META (case-sensitive, 69 matches), Meta Platforms, the flagship application name (22), a second application name (2), the messaging application name, the chief executive's surname (2) and the open model family name (14) across all 51,928 records. All quoted claims are verbatim from the stored claim text, with bracketed substitutions only where a house naming restriction applies.
• Data as-of: fundamentals — the quarter ended 2026-06-30, reported 2026-07-29 and 10-Q filed 2026-07-30, five and six days before this dive; the last complete fiscal year is 2025, 10-K filed 2026-01-29 · estimates 2026-08-04 · prices 2026-08-04 ($587.94, −0.39%; 50-DMA $601.50; 200-DMA $633.25; RSI 26.2; MACD −8.96) · knowledge-base claims 2026-08-04. Every figure comes from the Synthos vendor data file for META or from the SEC filings in filings/META/; no figure comes from memory, recall or external retrieval.
• Filings consulted: 10-K filed 2026-01-29 (fiscal year 2025); 10-Q filed 2026-04-30 (quarter ended 2026-03-31); 10-Q filed 2026-07-30 (quarter ended 2026-06-30 — the primary source for this dive); 8-K filed 2026-05-04 (senior-notes offering, six tranches maturing 2031 through 2066); 8-K filed 2026-05-29 (annual meeting results, including rejection of a proposal to eliminate the dual-class structure); 8-K filed 2026-07-29 (second-quarter results announcement).
• Filing-versus-vendor overrides (five, all resolved in favour of the filing): (1) netDebt — the vendor reports $48.024B; the 10-K states "cash, cash equivalents, and marketable securities were $81.59 billion" against "long-term debt ... of $58.74 billion", so correctly computed net debt is $2.305B (or net cash of $22.848B on the filing's own debt measure) — a $45.719 billion error; (2) enterprise value — vendor $1,594.625B against a rebuilt $1,500.074B, a 6.3% overstatement, and we use the rebuilt figure; (3) ~$347 billion of uncommenced lease obligations ($103.77B at 2025-12-31 → $278.99B at 2026-06-30, plus ~$68B signed in July 2026) and $349.31 billion of non-cancelable contractual commitments (from $131.05B six months earlier), none of which appears in any vendor field; (4) the earnings basis — the vendor's earn_cal strips the $15.93B discrete tax charge from Q3 2025 but does not strip the offsetting $8.03B discrete tax benefit from Q1 2026, an inconsistency nothing in the file discloses, so we use a clean trailing EPS of ~$29.61 against the reported $26.54; (5) seg_geo — the FY2025 entry merges a four-region split (which sums exactly to revenue) with a separate single-country "UNITED STATES" disclosure of $74.780B, and adding all five overstates revenue by 37%. Two further minor overrides: quarterly depreciation and amortisation is null (0.00) for the June quarter against the 10-Q's $12.35B for six months, so the derived Q2 figure of ~$6.35B is used; and profile.employees of 78,865 is the year-end figure against the 10-Q's 75,472.
• Other data rejections (detailed in Section 8): est net-income and EBIT rows rejected — the FY2024 row shows net income exceeding EBIT, and the share counts implied by dividing netIncomeAvg by epsAvg swing incoherently between 2.510B and 2.803B across adjacent years, so the two series are not built on a common base; forward valuation runs on epsAvg and revenueAvg only. The est EBITDA and EBIT rows do pass the elementary ordering test in every forward year — uncommon in this feed, and stated — but are not quoted. Intraday 52-week extremes ($796.25/$520.26) rejected in favour of the closing series ($790.00/$525.72).
• Vendor fields specifically VERIFIED as correct: capitalLeaseObligationsNonCurrent ($22.940B) is not identical to longTermDebt ($58.744B) and is not a double-count — the 10-K's separate statement of $59.0 billion of aggregate senior-notes principal confirms the split, so Meta's finance-lease book of $25.153B is genuine and correctly recorded. seg_prod reconciles exactly to reported revenue. freeCashFlowYieldTTM (2.74%), capexToRevenueTTM (0.391) and capexToOperatingCashFlowTTM (0.686) all look alarming and are all correct.
• Estimate coverage: FY2026 41 analysts on EPS and 39 on revenue; FY2027 41 and 43 — the primary forward anchor; FY2028 34 and 40; FY2029 30 and 31; FY2030 16 and 31. This is the deepest and most consistent analyst coverage of any name in this batch, which is one reason the FY2027 and FY2028 rows carry more weight here than elsewhere. The FY2028 EPS dispersion is nonetheless wide ($29.708 low to $59.443 high on 34 estimates), reflecting genuine disagreement about the capital cycle.
• Peer note: the vendor peer set is unusable for this company. Applied Materials ($434.0B), Broadcom ($1,989.4B), Cisco ($479.8B), Micron ($1,008.2B), Taiwan Semiconductor ($2,163.6B), Garmin ($59.3B), Oracle ($419.7B) and IBM ($221.5B) are semiconductor, networking, hardware and enterprise-software businesses. Only Alphabet ($4,570.4B) is a genuine comparable, and no peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $480 / $690 / $840 anchors are multiples of FY2027 consensus EPS of $34.293 (14.0x / 20.1x / 24.5x), cross-checked against FY2028E of $39.716 and against clean trailing EPS of ~$29.61. FY2027 was chosen deliberately over FY2026 because the FY2026 consensus of $32.005 embeds roughly $3.13 a share of non-recurring tax benefit already banked in the first quarter. Stated arithmetic, not a discounted cash flow. The base case requires roughly three turns of multiple expansion from 17.1x to 20.1x, and that is flagged as the fragile leg; at an unchanged 17.1x, FY2027 consensus earnings imply approximately today's price.
• Timing: the second quarter of 2026 was reported 2026-07-29 and the 10-Q filed 2026-07-30, five and six days before this dive, so the most recent quarter is fully reflected. The next print is 2026-10-28, 85 days away — no earnings banner applies, and the absence of any near-term catalyst is an explicit and stated input to the staged-entry structure rather than an oversight.
• Accessibility note: all tables are labelled, all emphasis is carried by text rather than colour, and every figure is stated in full rather than encoded in a chart.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.