AppLovin APP

⚠ EARNINGS IN ONE DAY — 2026-08-05

AppLovin reports second-quarter 2026 results on 2026-08-05, the morning after this dive is struck. Vendor consensus is revenue $1,941.897M and EPS $3.76. Against a corrected prior-year base of $1,258.754M that is +54.3% year on year, and +5.4% sequentially on the March quarter's $1,842.449M. The company has beaten the consensus EPS estimate in each of the last six reported quarters (by 3.5%, 9.8%, 2.9%, 15.3%, 15.9% and 16.0%). Beta is 2.529. No verdict struck on 2026-08-04 survives contact with that print intact, and this one is set with the timing as an explicit input rather than in spite of it.

Reference table

1. What the business is, and what the disclosure will and will not tell you

AppLovin describes a platform, not a product line. Its advertising solutions, per the 10-K, are Axon Ads Manager, MAX, Adjust and Wurl:

• Axon Ads Manager — the self-serve platform advertisers use to set goals and let the engine optimise spend. The 10-K states flatly that it "comprises the vast majority of revenue" and that this revenue "is determined dynamically based on advertisers' campaign goals." The Axon branding was launched publicly in 2025 alongside the self-serve platform.
• MAX — the mediation tool advertising networks use to buy app inventory; revenue is a percentage of client spend.
• Adjust — measurement, attribution and fraud prevention, acquired April 2021; revenue is primarily an annual software subscription.
• Wurl — connected-television content distribution and advertising, acquired April 2022; usage- and CPM-based.

No revenue figure is disclosed for any of the four. Since the Apps divestiture the 10-K states: "Following the sale of the Apps business, we operate as a single operating and reportable segment." The seg_prod block reflects that honestly — the FY2025 entry is a single line, "Reportable Segment": $5,480,717,000, which reconciles to reported revenue exactly.

That is the central disclosure fact about this company and it cuts both ways. There is no way, from public filings, to size Axon Ads Manager against MAX, to size mobile gaming against the e-commerce vertical the company has been expanding into since 2024, or to size Wurl's connected-television business. The 10-K discusses e-commerce in the risk factors and in the strategy narrative — "since 2024 we have begun expanding our customer base to include web-based e-commerce advertisers" — and never once quantifies it.

Geography is the only mix disclosure that exists, and it reconciles exactly:

Read the totals row before reading the shares. FY2025 sums to $5,480.7M and matches the income statement exactly. FY2024 sums to $4,709.2M and the FY2024 income statement says $3,224.1M, because the geographic table for FY2024 and earlier is on the total-company basis including the divested Apps business, while the income statement has been restated to continuing operations. The same is true of FY2023 ($3,283.1M against a restated $1,841.8M). The FY2024 and FY2023 geographic shares in this file are therefore not comparable with FY2025's and no year-on-year geographic growth rate is computed anywhere in this dive. Details in Section 8.

The concentration the 10-K discloses is structural, not customer-level. No single customer is named as exceeding 10% of revenue. What is disclosed is platform dependence: "a significant amount of our total revenue is derived through mobile applications subject to a limited number of third-party distribution platforms, such as the Apple App Store, the Google Play Store, and Meta. Because Meta and Google are also significant partners of AppLovin and Adjust, a deterioration in our or Adjust's relationship with such companies would have a greater impact on our business." Elsewhere the same filing lists "some of the largest global internet platforms, such as Meta and Google" among its advertising clients. Meta and Google are simultaneously the distribution layer, large customers, and the most capable competitors. That is the long-run risk in one sentence, and it is the company's own.

Governance: a controlled company. Three share classes. Class A has one vote, Class B has 20 votes, Class C has none. Co-founder and chief executive Adam Foroughi and director Herald Chen hold all 30,208,000 Class B shares under a voting agreement and, per the 10-K, "collectively held approximately 67% of the voting power of our outstanding capital stock" at 2025-12-31. AppLovin is a "controlled company" under Nasdaq rules and may elect not to comply with certain corporate-governance requirements.

And there was a lot of change at the top in April 2026. The 8-K filed 2026-04-07 discloses, all dated 2026-04-02: Chief Technology Officer Basil Shikin stepping down effective 1 July 2026 (staying as a Distinguished Engineer); Chief Administrative & Legal Officer Victoria Valenzuela retiring effective 1 August 2026, then consulting to May 2027 for $150,000 and being nominated to the board; Giovanni Ge promoted from Chief Product and Engineering Officer to CTO; Corina Cacovean promoted to Chief Legal Officer; Craig Billings appointed independent Chairperson, succeeding Adam Foroughi, who remains chief executive and a director; and director Alyssa Harvey Dawson not standing for re-election, explicitly not over any disagreement. An independent chair replacing a founder-chair is a governance improvement. Three C-suite changes in a single filing is a lot of institutional memory moving at once. We report both.

2. Growth — and the vendor's quarterly income statement, which is broken

This is the most consequential data finding in this dive, so it goes here rather than buried in Section 8.

The vendor's inc_q block reports 2024 quarterly revenue of $1,058.115M, $711.015M, $835.186M and $1,372.779M — a total of $3,977.095M. The vendor's own inc_a block reports FY2024 revenue of $3,224.058M. The quarters exceed the year by $753.0M and the file contradicts itself.

The filings resolve it exactly. The 10-Q filed 2025-11-05 reports third-quarter 2024 revenue of $835,186 thousand and nine-month 2024 revenue of $2,224,571 thousand, both on the restated continuing-operations basis. Working from those two anchors and the 10-K's FY2024 total:

Two of the four 2024 quarters are on the pre-divestiture total-company basis and two are restated. The 2025 and 2026 quarters are clean — they sum to $5,480.717M, matching FY2025 exactly, and Q1 2025's $1,158.974M is confirmed verbatim by the March 2026 10-Q's comparative column.

What that does to the growth picture is not cosmetic. Computed naively from the vendor file, AppLovin's year-on-year revenue growth reads +9.5%, +77.0%, +68.2%, +20.8%, +59.0% across the last five quarters — an erratic series with two apparent collapses. Computed correctly, it reads:

A clean, monotonic deceleration from 77% to 59%, with margins expanding the whole way. That is a far more informative — and far more benign — picture than the vendor file produces, and it is only available because the filings carry the restated comparatives. The vendor reports 2024 quarterly revenue totalling $3,977.1M; the filings show $3,224.1M; we use the filings, and every year-on-year growth rate in this dive is computed against the filing-derived base.

The annual series has the same disease one level up. The vendor's inc_a block reports FY2021 revenue of $2,793.1M and FY2022 of $2,817.1M — both of which reconcile to the total-company segment tables for those years — then FY2023 of $1,841.8M, which is continuing operations only. Read straight down the column it looks as though revenue fell 34.6% in 2023. It did not: on the restated basis the 10-K reports, continuing-operations revenue rose from $1,049.2M in FY2022 to $1,841.8M in FY2023, +75.5%. The FY2021 and FY2022 rows of the income statement are not comparable with anything below them and are excluded from every growth calculation in this dive.

The usable annual series, all continuing operations, all from the 10-K:

Every figure in that table is verbatim from the 10-K's own non-GAAP reconciliation, not derived. Adjusted EBITDA margin of 82.3% and free cash flow of $3.95B on $5.48B of revenue — a 72.1% free-cash-flow margin — is what the business actually converts.

The one variable

Here is the mechanism, in the company's own words across three filings.

> FY2024 (10-K): "our revenue increased by $1.4 billion, or 75%, from the prior year period primarily due to improved Axon Ads Manager performance, where the volume of installations increased 50% and net revenue per installation increased 22%."

> FY2025 (10-K): "our revenue increased by $2.3 billion, or 70%, from the prior year period primarily due to improved Axon Ads Manager performance, where the volume of installations increased 3% and net revenue per installation increased 72%."

> Q1 2026 (10-Q): "our revenue increased by $683.5 million, or 59%, compared to the same period in the prior year due primarily to improved Axon Ads Manager, where net revenue per installation increased 93%, partially offset by a decrease in the volume of installations of 18%."

Volume has gone from +50% to +3% to −18% in two years while price has gone from +22% to +72% to +93%. The composition of AppLovin's growth has completely inverted, and the total has held up only because price accelerated faster than volume decayed.

There are two readings and we do not have the evidence to choose between them.

The benign reading is that this is what a better model does. Axon targets more precisely, so each installation it delivers is worth more to the advertiser, so the advertiser bids more and AppLovin serves fewer, better-qualified installations. Under this reading, falling volume is a feature: the engine is discarding low-value inventory. The margin expansion — gross margin from 83.9% to 88.9%, operating margin from 59.3% to 78.2% — is consistent with it, because serving fewer, higher-value impressions costs less infrastructure per revenue dollar.

The adverse reading is that the mobile-gaming install base is contracting, that AppLovin is harvesting price from a shrinking pool of advertisers who have nowhere better to go, and that the price series has a ceiling somewhere. Under this reading, the 93% figure is not evidence of strength but of how hard the company is now having to push a single lever.

What makes this the risk rather than a curiosity is that net revenue per installation is not a market price. It is the realised outcome of an auction AppLovin operates, into an inventory pool AppLovin mediates, measured by a metric AppLovin defines and does not further disaggregate. There is no external series that corroborates or contradicts it. For a company where this one number is now the whole growth story, that is a genuine analytical gap and we are not going to pretend otherwise.

The 2026-08-05 print is the next observation. Whether installations fall again, and whether revenue per installation holds near 90%, is more informative than the revenue headline.

3. Margins, costs and the quality of the earnings

Trailing twelve months to 2026-03-31, computed from the four clean quarters:

Cost discipline is the second half of the story, and it is severe. In the March 2026 quarter, against revenue up 59%:

• Cost of revenue rose 34% to $203.6M — "primarily an increase of $39.9 million in expenses associated with operating our network infrastructure." Eleven percent of revenue, down from 13%.
• Sales and marketing rose 2% to $60.8M. Three percent of revenue, down from 5%. A $16.7M increase in advertising programmes almost entirely offset by a $15.3M decrease in personnel costs.
• Research and development rose 67% to $94.1M — "an increase of $36.5 million in personnel-related expenses related to an increase in stock-based compensation-related payroll costs." Five percent of revenue, flat.
• General and administrative fell 15% to $44.0M — lower bad-debt expense and lower transaction costs. Two percent of revenue, down from 4%.

Sales and marketing at 3% of revenue for an advertising company is the number worth staring at. It is the arithmetic signature of a self-serve product: advertisers configure campaigns themselves through Axon Ads Manager and the salesforce does not scale with revenue. It is also why operating margin can rise 1,890 basis points in two years.

Stock-based compensation is falling, which is unusual and creditable. $342.6M (FY2023), $357.4M (FY2024), $208.0M (FY2025) per the 10-K's own reconciliation — down 41.8% in a year while revenue rose 70%. As a share of revenue that is 18.6%, 11.1%, 3.8%. The vendor's stockBasedCompensationToRevenueTTM of 3.77% corroborates.

The tax rate is the one item that flatters the trend and it should be named. FY2024's provision for income taxes was $22.4M on $1,611.9M of pre-tax income — a 1.4% effective rate, driven by stock-compensation benefits. FY2025's was $519.7M on $3,952.9M, a 13.1% rate. Trailing is 14.7%. So the FY2024-to-FY2025 net income growth of +111% understates pre-tax progress (pre-tax grew +145%), and equally, the current 14.7% rate is not a durable assumption — the 10-Q attributes the March-quarter increase partly to "a decrease in stock-based compensation benefits." Any normalisation toward a low-20s rate is roughly a 9% headwind to EPS, and consensus does not obviously carry it.

Earnings quality is good. The vendor's incomeQualityTTM of 1.118 — operating cash flow above net income — is the right sign, and the FY2025 cash-flow bridge in the 10-K is clean: $3.97B of operating cash flow from $3.3B of net income, adjusted for $210.4M of stock compensation, $194.8M of amortisation/depreciation/write-offs, $188.9M of goodwill impairment, $50.0M of impairment of non-marketable equity securities, and $77.9M of working-capital release, less a $106.2M gain on the Apps divestiture.

Two of those deserve naming rather than netting. The $188.9M goodwill impairment in FY2025 is a write-down against businesses AppLovin bought to build the optionality it now talks about — the acquisition stack is Adjust (April 2021), MoPub (January 2022) and Wurl (April 2022). Goodwill fell from $1,803.4M to $1,540.0M across FY2025 and to $1,523.1M by March 2026. The $106.2M divestiture gain flatters FY2025 net income by roughly 3.2% and is not operating income — it sits below the line and does not contaminate the $4,151.9M operating figure, which we verified. The FY2025 operating margin of 75.8% is clean of it.

4. Balance sheet and enterprise value — and the vendor is right, for once

At 2026-03-31, from the 10-Q balance sheet, filing-verified:

Net debt is $755.4M ($3,514.0M less $2,758.7M). The 10-Q states "we had a total of $3.6 billion in aggregate principal amount of senior unsecured notes outstanding" at 2026-03-31, with the difference to the $3,514.0M carrying value being unamortised discount and issuance costs, plus $1.0 billion of commitments under the revolving credit facility with no outstanding borrowings and a $100M letter-of-credit sublimit. The 10-Q confirms "our Senior Notes are currently rated investment-grade by various rating agencies."

The enterprise-value rebuild — and it confirms the vendor:

> Market cap $140.994B

> + total debt $3.514B

> − cash $2.759B

> = enterprise value $141.749B

The vendor reports $141,749,174,259. That is an exact match, to the dollar, against the 2026-03-31 balance sheet. Given that this defect class — vendor enterprise value built on a stale or partial balance sheet — has appeared repeatedly across this programme, a clean pass is worth recording as a finding rather than passing over in silence. The vendor's netDebtToEBITDATTM of 0.153x is likewise correct against our computed trailing EBITDA of $4.944B.

Leverage is immaterial. Net debt of $755.4M against FY2025 adjusted EBITDA of $4,512.5M is 0.17x. Interest coverage is 23.1x. There is no maturity disclosed as near-term in the file.

Two balance-sheet items are worth flagging as judgement calls rather than facts.

Equity is $2,363.5M against $7,707.7M of assets — a 30.7% equity ratio, and the vendor's priceToBookRatioTTM of 59.9x and returnOnEquityTTM of 222% are both arithmetically correct and both economically meaningless. AppLovin has repurchased $2,191.9M of stock in FY2025 and $1.0B more in the March quarter; book equity is a residual, not a capital base. returnOnInvestedCapitalTTM of 65.2% is the honest return measure and is what we use. Note that the vendor's own composite rating scores AppLovin 1 out of 5 on both price-to-book and debt-to-equity, which is what drags its overall grade to B / 3 — a rating artefact from buybacks, not a judgement about the business, and we reject those two sub-scores as lenses.

Days of sales outstanding is 115.9 and days of payables outstanding is 355.0, producing a cash conversion cycle of −239 days. AppLovin collects from advertisers in about four months and pays publishers in about twelve. That is a large, structural, negative-working-capital float — it is a genuine source of the cash generation, and it also means that any slowdown in revenue releases cash more slowly than it appears, because the payables balance unwinds with it. Receivables of $1,958.0M are 31.8% of trailing revenue.

5. Cash flow and capital returns

The 10-K's free-cash-flow definition nets purchases of property and equipment ($473 thousand in FY2025) and principal payments of finance leases ($18,669 thousand), giving $3,951,952 thousand. The vendor's cash-flow block reports capital expenditure of $28,318 thousand for FY2025 — a different composition, which includes purchases of intangible assets. The vendor's $3,942.8M and the 10-K's $3,952.0M differ by $9.2M, or 0.2%, purely on definition. We use the 10-K's figure and note the reconciliation.

Capital expenditure of $473 thousand on $5.5 billion of revenue — 0.009% — is the single most striking line in the financial statements. This is a company that rents its compute and owns almost nothing.

There is no dividend and there is no indication of one. dividendPerShareTTM is 0 and lastDividend is 0.

The buyback is the entire capital return and it is running hard. $2,191.9M in FY2025 — 1.55% of the current market capitalisation — and $1.0 billion in the March 2026 quarter alone, at which pace an annualised $4B would be 2.8% of the market capitalisation. The 10-Q states: "During the three months ended March 31, 2026, we repurchased and retired 2.2 million shares of Class A common stock for $1.0 billion. As of March 31, 2026, $2.3 billion remained available for repurchases under the program." That works out to an average of roughly $455 per share — 8.4% above today's close.

Net share count is actually falling, which for a company of this profile is unusual. Shares issued and outstanding went from 338,313,000 at 2025-12-31 to 336,294,000 at 2026-03-31 — down 0.6% in a single quarter, net of all equity issuance. Diluted weighted-average shares went from 371.6M (FY2022) to 342.0M (FY2025), a 7.9% reduction. Total shareholder yield is therefore roughly 2.8% and it is entirely buyback, at a level that is a real, not cosmetic, contributor to per-share growth.

6. Valuation — priced in or room?

At $419.70 (market cap $140.994B, enterprise value $141.749B):

Estimate coverage is the best in this batch — 19 to 23 analysts on the FY2026 and FY2027 rows, 15 to 17 on FY2028. The FY2029 and FY2030 rows are excluded from every conclusion for reasons set out in Section 8: FY2029 revenue ($13.684B, 14 analysts) prints below FY2028 ($13.820B, 17 analysts) while FY2029 EPS rises 20.7%, and the FY2030 row carries epsAvg and netIncomeAvg of exactly zero.

Per the data contract and our own verification, est.ebitdaAvg and est.ebitAvg are rejected outright here — not merely treated with caution. In every one of the five forward years, ebitdaAvg is exactly 52.17% of revenueAvg and ebitAvg is exactly 36.40%, to four decimal places. That is a fixed ratio applied mechanically to the revenue line, the same fabrication signature the data contract documents at 50.000%/33.000%. It is also nowhere near reality: AppLovin's actual FY2025 adjusted EBITDA margin, from its own 10-K, was 82.3%, and its GAAP operating margin was 75.8%. Every forward valuation figure in this dive runs on epsAvg and revenueAvg only.

Peer context is unusable and we say so. The vendor peer set is C3.ai, Asana, Bill.com, Datadog, DocuSign, JFrog, GitLab, Intuit, monday.com, Shopify, Snowflake, Atlassian, Workday and Zoom. Not one is an advertising business. These are subscription software companies with 70-80% gross margins and 0-25% operating margins; AppLovin has an 88.4% gross margin and a 77.1% operating margin from an auction. No peer multiple comparison is drawn, because the comparison would be meaningless.

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

At $419.70 — 26.1x FY2026 consensus and 19.5x FY2027 — the price embeds the following falsifiable claims:

• Revenue reaches $8.203B in calendar 2026 and $10.727B in 2027. (Consensus-derived; 22 and 23 analysts.) Against FY2025's $5,480.7M that is +49.7% then +30.8%. The nearest test is tomorrow: consensus wants $1,941.897M for the June quarter, +54.3% on the corrected prior-year base of $1,258.754M. Because the first quarter already printed $1,842.449M, the full-year consensus requires roughly $4.419B across the September and December quarters, an average of $2.21B per quarter against $1.84B just delivered.
• EPS reaches $21.570 by 2027 — an 83.6% increase on the trailing $11.75 in under two years. (Consensus.) Because share count is falling and margins are near their structural ceiling, the great majority of that has to be revenue.
• Net revenue per installation keeps compounding at something like its current rate. (Our derivation, from the 10-K and 10-Q disclosures.) With installations at −18% year on year, revenue growth of 30% in FY2027 requires price per installation to rise roughly 50-60% again — on top of +72% and +93%. This is the most fragile assumption in the price, and it is fragile in a specific way: it is not that the number is implausible, it is that no one outside the company can observe or verify it, and it has no external market reference.
• The effective tax rate does not normalise. (Our number.) The trailing rate is 14.7% and FY2024's was 1.4%. A move to a low-20s rate is roughly a 9% EPS headwind that the consensus EPS path does not visibly carry.
• The market keeps paying roughly 20x or better on the forward year. (Our number.) At 15x FY2027E the stock is $324; at 25x it is $539. The multiple is a wider source of variance than the earnings, and the stock has already de-rated once this year — from $733.60 to $419.70 with the estimates going up, not down.

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

Expected return over the next twelve months decomposes as: EPS growth (+34.3%, from FY2026E $16.066 to FY2027E $21.570) + multiple drift (COMPRESSION, from 26.1x on the forward year to roughly 21-22x on the then-forward year, about −17%) + shareholder yield (+2.8%) ≈ +12% to +20%.

Our base case assumes multiple compression, and the reason is deceleration, not distrust. Consensus revenue growth runs +49.7% (FY2026), +30.8% (FY2027), +28.8% (FY2028). Corrected quarterly growth has already fallen from 77.0% to 59.0% in four quarters. A multiple appropriate to 50% growth is not appropriate to 30% growth, and the market has spent the last several months applying exactly that logic — this stock fell 42.8% from its high while forward estimates were rising. Assuming the de-rating simply stops here would be assuming the crowd has finished a process it is visibly mid-way through.

Almost all of the expected return is earnings growth, and the multiple is working against it. If the multiple held at 26.1x on FY2027E, the price would be $563 (+34.1%). If it compressed to 15x — the level a 20%-growth software business might carry — the price would be $324 (−22.8%). The entire outcome band sits inside a multiple we do not control and, on a name with a knowledge-base lane of one claim, have no edge on.

The bear case at $300 requires no earnings miss at all — it is 14.7x the FY2027 consensus low of $20.393, i.e. a de-rating to a mature-software multiple on a modestly disappointing estimate. The bull case at $675 requires no earnings beat beyond the top of the current range — it is 29.3x the FY2027 consensus high of $23.036, roughly today's FY2026 multiple carried forward.

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

• We differ from the street by 27.6%, and downward. Consensus is $649.53 with 23 of 26 analysts at buy; our base is $470. The gap is not a disagreement about the business — we score growth 9 and exponential 8 — it is a disagreement about what multiple a decelerating single-variable growth story earns. When our number sits nearly 30% below a near-unanimous buy consensus, that is a signal to be cautious about the consensus and about ourselves in equal measure, and it argues for waiting rather than for shorting the crowd.
• We differ from the vendor data, provably, on the growth rate — and the correction runs in AppLovin's favour. The vendor file makes recent quarters look erratic (+9.5%, +77.0%, +68.2%, +20.8%, +59.0%). The filings show a clean deceleration from 70.8% to 59.0%. Anyone screening this name off the vendor's quarterly series is seeing two collapses that did not happen. That is a real informational edge, and it is a mildly bullish one — which is worth saying, because most of our data corrections in this programme have cut the other way.
• We think the installation decline is the disclosure that matters and is being under-weighted. The market and the sell-side discuss AppLovin in terms of revenue growth and e-commerce expansion. The 10-Q's "a decrease in the volume of installations of 18%" is the first hard evidence in the filings that the substrate is contracting, and it is a single clause in an MD&A paragraph. Watchable number: the same disclosure in the 2026-08-05 10-Q. A second consecutive double-digit decline in installations, with revenue per installation decelerating below roughly 70%, would be the concrete falsifier.
• Positive surprise that would force a re-rate: the June quarter beating $1,941.897M with installations returning to growth and revenue per installation holding above 70% — that would resolve the ambiguity in the benign direction and make the 19.5x FY2027 multiple look like a mistake. Or the first quantified disclosure of e-commerce revenue at a material scale, which would give the growth story a second leg and a second observable variable.
• Negative surprise that would break the thesis: a June-quarter revenue miss with installations down again; net revenue per installation growth decelerating below roughly 50%; a further goodwill impairment against the acquisition stack after FY2025's $188.9M; or a platform-policy change at Apple, Google or Meta, which the 10-K names as simultaneously the distribution layer and the largest clients. Any of those hits a stock with a 2.529 beta.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $21.570, low $20.393, high $23.036, 20 analysts) — the best-covered forward row with a full year of visibility — cross-checked against FY2026E of $16.066 and FY2028E of $28.493.

• Bear ~$300 — 14.7x the FY2027 consensus LOW of $20.393. Cross-check: 18.7x FY2026E; 10.5x FY2028E. The scenario needs no catastrophe: installations keep falling, revenue per installation decelerates, growth settles nearer 20%, the tax rate normalises, and the market pays a mature-software multiple. −28.5%.
• Base ~$470 — 21.8x the FY2027 consensus MEAN of $21.570. Cross-check: 29.3x FY2026E; 16.5x FY2028E; enterprise value of $158.6B on FY2027E revenue of $10.727B is 14.8x sales. Sensitivity, stated openly: 20x gives $431 — 2.8% above spot — and 24x gives $518, 23.4% above. The whole answer lives inside a 20-24x band on one estimate row. The scenario: consensus revenue roughly lands, margins hold near their current structural level, the buyback keeps shrinking the count, and the multiple compresses modestly with the growth rate. +12.0%.
• Bull ~$675 — 29.3x the FY2027 consensus HIGH of $23.036, which is approximately today's FY2026 multiple carried forward a year. Cross-check: 42.0x FY2026E; 23.7x FY2028E. The scenario: the benign reading of the volume/price inversion is correct, e-commerce becomes a disclosed second leg, and the market re-rates the name back toward where it traded when growth was 70%. +60.8%, and still 3.9% below the street's $649.53 plus its own dispersion.

Base is 12.0% above spot; asymmetry roughly 2.13:1 to the upside (28.5% down, 60.8% up). That ratio is genuinely attractive and it is the strongest argument in this dive for a higher tier. What holds it at Watch is that the ratio is struck the evening before a print on a 2.529-beta stock, that our base is 27.6% below a near-unanimous street, and that the knowledge base contributes one eighteen-month-old claim at conviction 50.

7. Knowledge base — fourteen hits, one claim, and the discards are the finding

Raw hits: 14. Entity matches: 9. Text matches: 6 (one overlapping). Used: 1. Discarded: 13.

The sweep ran entity terms AppLovin, Applovin, APP and App Lovin with word-boundary matching, plus a free-text sweep on applovin, axon ads, adtech and ad tech, across all 51,928 distilled claims. It returned 14 hits. Thirteen of them are noise, and the noise pattern is exactly what the batch brief warned about for short tickers.

Discard class 1 — the "Cash App" homograph (8 claims). A case-insensitive, word-boundary match on the ticker APP matches the word App inside the entity "Cash App" — Block's consumer product. Eight claims across business_breakdowns (3, dated 2023-05-20, on Cash App unit economics and Block's product advantage), a16z (3, dated 2026-04-01, on Block's AI-driven restructuring, generative UI, and Cash App as ~60% of Block gross profit), real_vision (1, 2026-02-12, on stablecoin rollouts) and arthur_hayes (1, 2020-11-12, on Bitcoin on-ramps) are entity-tagged to Block, Square, Cash App and Afterpay. None of them has anything to do with AppLovin. Re-running the entity sweep case-sensitively on APP eliminates all eight.

Discard class 2 — the "ad tech" substring inside ordinary English (5 claims). A free-text search on ad tech matches the letters inside: "bad technology" (invest_like_the_best, 2024-06-03 and 2021-06-11, both on blockchain), "broad tech basket" (lyn_alden, 2020-12-09), "baseload technologies" (macrovoices, 2022-10-13, on renewables), and "a bad technical sign" (jordi_visser, 2026-06-21, on Microsoft's chart). Five claims, five different subjects, none of them advertising technology. This is the same defect class as the "Chevron deference" and "Amex Exploration" catches recorded elsewhere in this programme, and it is why we report discards with reasons rather than counts.

That leaves one claim. Here it is, verbatim.

> 2025-01-21 · bullish · conviction 50 · horizon: thesis · entities: APP · speaker: Joel Schuman · role: independent · channel: darius_dale · skill 0.8

> "AppLovin is a long-held position, owned nearly two years, consistent with their entrepreneurial-growth model; cites it as last year's hottest Nasdaq name."

What this claim is, stated without inflation. It is a position disclosure, not an analysis. It says a named, independent speaker owned the stock and had owned it for about two years as of January 2025. It offers no mechanism, no valuation, no falsifiable test, and it carries conviction 50 — the lowest band we record, which is the distillation layer's own assessment that this is a weak signal. The speaker role is independent, so the half-weight management discount does not apply and no management voice contaminates the lane.

And it is stale in a way that matters. Dated 2025-01-21, it predates: the sale of the Apps business (30 June 2025); the public launch of the Axon branding and self-serve platform (2025); the $188.9M goodwill impairment (FY2025); the executive turnover of April 2026; the inversion of installation volume from +50% to −18%; and a 42.8% drawdown. Every fact this dive turns on postdates the only claim in the lane.

Conclusion, stated as a finding. The Synthos knowledge base has no usable view on AppLovin. There is no concentration sensitivity test to run, because a single claim from a single channel is a concentration of 100% by construction and stripping it leaves nothing. Breadth is 1, net conviction is positive-negligible, and this is a direct input to the Watch verdict: a name with a 2.13:1 payoff ratio and no independent conviction behind it is a name to size small and enter on price, not on view.

8. Data integrity — what we rejected and why

Eleven findings. AppLovin's vendor file is a mixed case: the balance sheet, enterprise value and margin ratios are clean and filing-confirmed, while the quarterly and annual revenue series and the entire forward EBIT/EBITDA block are not.

1. inc_q 2024 quarterly revenue is on two different bases — REJECTED and rebuilt from the filings. The four 2024 quarters sum to $3,977.095M against the vendor's own FY2024 revenue of $3,224.058M, an internal contradiction of $753.0M. The 10-Q filed 2025-11-05 gives third-quarter 2024 revenue of $835,186 thousand and nine-month 2024 revenue of $2,224,571 thousand on the restated continuing-operations basis. Working from those: Q1 2024 is $678.370M, not $1,058.115M (vendor overstates by $379.7M); Q4 2024 is $999.487M, not $1,372.779M (overstates by $373.3M); Q2 and Q3 are correct. The vendor reports 2024 quarterly revenue totalling $3,977.1M; the filings show $3,224.1M; we use the filings. Consequence: every year-on-year growth rate computed naively from inc_q for the 2025 quarters is wrong, in two cases severely — Q1 2025 reads +9.5% when it is +70.8%, and Q4 2025 reads +20.8% when it is +65.9%.

2. inc_a switches basis mid-series between FY2022 and FY2023 — the FY2021 and FY2022 rows are DISCARDED. FY2021 ($2,793.1M) and FY2022 ($2,817.1M) reconcile exactly to the total-company seg_prod tables for those years, which include the Apps business. FY2023 onward ($1,841.8M) is continuing operations only, matching the 10-K's restated presentation. Read as a series it shows revenue falling 34.6% in 2023; the 10-K shows continuing-operations revenue rising 75.5%. No growth rate in this dive touches FY2021 or FY2022.

3. seg_geo and seg_prod for FY2024 and earlier are on the pre-divestiture basis — used for FY2025 only. FY2024 seg_geo sums to $4,709.2M against income-statement revenue of $3,224.1M; FY2023 sums to $3,283.1M against $1,841.8M. FY2025 sums to $5,480.717M and matches revenue exactly. Only the FY2025 geographic split is used; no geographic growth rate is computed. The FY2025 seg_prod entry is a single line, "Reportable Segment", which correctly reflects the 10-K's statement that "Following the sale of the Apps business, we operate as a single operating and reportable segment."

4. est.ebitdaAvg and est.ebitAvg carry the fabrication signature — REJECTED in full. In every forward year — FY2026 through FY2030 — ebitdaAvg is exactly 52.17% of revenueAvg and ebitAvg is exactly 36.40%. That is a fixed ratio applied to the revenue line, the same mechanism the data contract documents as a fabrication signature at 50.000%/33.000%. The figures are also plainly wrong: AppLovin's 10-K reports an FY2025 adjusted EBITDA margin of 82.3% and a GAAP operating margin of 75.8%. Corroborating defect: netIncomeAvg exceeds ebitAvg in every forward year — $7.426B against $3.904B in FY2027, $9.141B against $5.030B in FY2028, $11.757B against $4.981B in FY2029 — which is arithmetically impossible. All forward valuation uses epsAvg and revenueAvg only.

5. The FY2029 estimate row is internally incoherent and the FY2030 row is empty — both EXCLUDED. FY2029 revenueAvg of $13.684B (14 analysts) is below FY2028's $13.820B (17 analysts), while FY2029 epsAvg of $34.38 is 20.7% above FY2028's $28.493. Revenue falling while EPS rises 21% is not a forecast, it is two different analyst panels being averaged. FY2030 carries epsAvg, epsLow, epsHigh and netIncomeAvg of exactly zero against a revenue estimate of $15.304B from 8 analysts. Neither row supports any conclusion in this dive.

6. epsdiluted is null in every row of inc_a and inc_q — worked around. The vendor supplies eps (basic) and weightedAverageShsOutDil but no diluted EPS anywhere in the file. FY2025 eps of 9.84 against net income of $3,333.8M implies 338.8M basic shares; diluted, using the supplied 341.97M, gives $9.75. All trailing per-share figures in this dive are computed and labelled.

7. inc_a.netIncome and cf_a.netIncome disagree for FY2025 — both are right, on different definitions. The income statement reports $3,333,764 thousand (total, after the $99,444 thousand loss from discontinued operations); the cash-flow statement reports $3,433,195 thousand (continuing operations). The 10-K confirms both figures. We use $3,333.8M as net income and note that continuing-operations net income was $3,433.2M, a 62.6% net margin on the company's own presentation.

8. Free cash flow differs from the 10-K by $9.2M on definition — the 10-K's figure is used. The vendor computes FY2025 free cash flow as $3,942.8M (operating cash flow less capital expenditure of $28.3M, which includes purchases of intangible assets). The 10-K defines it as operating cash flow less purchases of property and equipment ($473 thousand) and principal payments of finance leases ($18,669 thousand), giving $3,951,952 thousand. A 0.2% difference. We use the 10-K's, and note that the vendor's capexToRevenueTTM of 0.46% is on the broader definition.

9. quote.yearHigh and yearLow disagree with tech.hi52 and lo52 — we use tech. The quote block reports $745.61 and $359.00; the computed technical block reports $733.60 and $366.91, discrepancies of 1.6% and 2.2%. We use the tech figures throughout, for consistency with the moving averages, drawdown and relative-return series computed from the same close history.

10. Enterprise value, net debt and the margin ratios are CORRECT and filing-confirmed — recorded as a clean pass. enterpriseValueTTM of $141,749,174,259 reconciles exactly to market cap $140.994B plus long-term debt $3,514.022M less cash $2,758.671M from the 2026-03-31 10-Q balance sheet. netDebtToEBITDATTM of 0.153x, grossProfitMarginTTM of 88.37%, operatingProfitMarginTTM of 77.09%, netProfitMarginTTM of 64.29%, stockBasedCompensationToRevenueTTM of 3.77% and interestCoverageRatioTTM of 23.15x all reconcile independently against the quarterly statements. Given that stale or partial balance sheets have been the single most common defect across this programme, a clean, current, exactly-reconciling enterprise value is worth stating as a finding.

11. Book-equity metrics are economically meaningless — REJECTED as lenses. priceToBookRatioTTM and priceToFairValueTTM (both 59.9x), returnOnEquityTTM (222%) and debtToEquityRatioTTM (1.487x) are arithmetically correct and distorted by the same thing: $4.3B of buybacks across three years against $2.36B of remaining equity. The vendor's composite rating scores AppLovin 1 out of 5 on price-to-book and 1 out of 5 on debt-to-equity, dragging the overall grade to B / 3. That is a buyback artefact, not a judgement about a business with 65.2% return on invested capital and 0.17x net leverage.

Non-equity tripwire — checked and passed. APP is Class A common stock, $0.00003 par value, Nasdaq-listed, confirmed on the cover page of every filing in the archive. Price of $419.70 is not par-like; beta is 2.529; there is no dividend at all; volume was 7.00M shares (~$2.9B of turnover); and the 52-week band of $366.91 to $733.60 is a 100% range. This is common equity, and it is a high-beta one.

Verified against the filings and correct — worth recording: the share count (10-Q cover 336,294,000 at 2026-03-31 against an implied 335.9M from market cap and price, a 0.1% difference fully explained by post-quarter buyback); the FY2025 buyback of $2,191.9M; the FY2025 and FY2024 revenue, gross profit, operating income and net income lines; the geographic split for FY2025; and the single-segment presentation.

9. Technicals

• Price $419.70. −42.8% from the 52-week high of $733.60; +14.4% above the 52-week low of $366.91. Position within the annual range: 14th percentile — near the bottom, not the middle.
• Both moving averages are overhead and falling: −14.0% below a 50-day average of $487.77 and −19.0% below a 200-day average of $518.45. The 50-day sits below the 200-day. This is a downtrend on every timeframe the file measures.
• RSI 39.0 — weak, not oversold. MACD −23.48 — negative.
• Maximum drawdown from peak over the trailing year: −42.8%, i.e. the current price is close to the worst point of the drawdown, with the low only 12.6% below.
• Relative performance, and it has inverted: 3-month −11.6% against SPY +7.6% and QQQ +7.7%; 6-month −13.1% against SPY +11.1% and QQQ +15.7%; 12-month +10.7% against SPY +24.3% and QQQ +30.8%. AppLovin has underperformed the S&P 500 by 13.6 points over twelve months and by 19.2 points over three months — a stock that was among the market's best performers now lagging it materially.
• Sentiment is the outlier in the whole file: 23 buy, 2 hold, 1 sell across 26 analysts; consensus target $649.53 (+54.8%), median $660, high $798 (+90.1%), low $340 — 19.0% BELOW spot. A near-unanimous buy consensus at a 55% implied upside on a stock in a 43% drawdown is a large disagreement between the coverage and the market, and the low target of $340 shows the coverage itself is not unanimous about the downside.

Today's move and what it does to the entry

APP closed 2026-08-04 at $419.70, up 3.33% or $13.54 from $406.16. It opened at $419.67 — a 3.32% gap — traded $405.33 to $426.90, and closed mid-range on 7.00M shares against a 5.72M average, roughly 22% above normal volume, which is what the session before a print usually looks like.

The honest read: this is not a breakout and it is not an entry. A 3.33% move on a 2.529-beta stock is a normal day. The price remains 14.0% below the 50-day average and 19.0% below the 200-day, both declining, and the entire session is best understood as positioning ahead of tomorrow morning's report.

What we are telling the reader plainly: we are not asking you to buy the night before a print on a stock that can move 20% on it. The setup that would change this is one of two things, and they are pre-registered as flip conditions in Section 11:

• A price near $360-380 — roughly 17x FY2027E — which would put our base $470 at a 24-31% premium and make the risk/reward decisively asymmetric. That is 9-14% below today's close, and the 52-week low of $366.91 sits inside that band.
• Or the 2026-08-05 print showing installations returning to growth with net revenue per installation still above 70%, which would resolve the single ambiguity this whole dive turns on. At that point paying up is defensible, because the thing being paid for has been observed rather than assumed.

10. Insiders — eight transactions, one director, and it is not a signal

The reading, carefully. The five sales on 2026-07-06 total 1,142 shares for approximately $591,000 — five tranches at five prices between $516.25 and $520.60, all coded indirect ownership, against a holding that ends at 122,378 shares. That is 0.93% of the position, sold across a single day at incrementally rising prices, which is the signature of a pre-arranged trading plan executing in small clips, not a discretionary decision. The director's remaining stake is worth approximately $51.4M at today's close.

Note the prices. Every sale executed between $516 and $521 — 23-24% above today's close of $419.70, four weeks ago. That is a fact about the stock's decline, not about the director's judgement.

The two A-Award entries (41 and 40 shares, at $0) are routine director equity grants. The Form 3 for Corina Cacovean, filed 2026-08-03, is her initial statement of beneficial ownership on becoming Chief Legal Officer — the appointment disclosed in the April 8-K — and reports zero transacted.

What the file contains is therefore: zero open-market purchases, and one director's plan-like disposal of under 1% of his holding. That is a neutral signal and we report it as neutral. What is notable by its absence: there is no transaction by the chief executive, the chief financial officer, or any executive officer anywhere in the file. With the stock 43% off its high and the founder-CEO holding Class B super-voting stock, the absence of any discretionary purchase is worth one line — not because it is bearish, but because a founder buying his own 43% drawdown would have been a genuine signal and it is not there.

11. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely excellent, and none of it is in dispute: an 88.4% gross margin, a 77.1% operating margin and a 64.3% net margin; $6.164B of trailing revenue from 898 full-time employees; free cash flow of $3,952.0M on the company's own definition, a 72.1% margin, against $473 thousand of property-and-equipment spending; adjusted EBITDA margin rising 67.1% → 74.8% → 82.3% across three years; net debt of $755.4M against $4,512.5M of adjusted EBITDA; stock compensation falling 41.8% in a year while revenue rose 70%; a share count that is actually shrinking; and a return on invested capital of 65.2%. Revenue from continuing operations grew 75.0% and then 69.9% in consecutive years, and the filings — not the vendor file — show the last five quarters at 70.8%, 77.0%, 68.2%, 65.9% and 59.0%.

What we are declining to underwrite today: a business whose entire growth is now one internally-set price variable, with the volume variable at −18% year on year; a base fair value of $470 that sits 27.6% below a near-unanimous street consensus of $649.53, which means either we are wrong or 23 analysts are; a 14.7% effective tax rate that the company's own 10-Q says is already normalising; a knowledge-base lane of one claim, eighteen months old, at conviction 50; three C-suite changes in a single April 8-K; a controlled-company structure with 67% of votes under a two-person agreement; and an earnings report tomorrow morning on a stock with a beta of 2.529.

The distinction that matters. This is not a case where the business is bad or the price is absurd — at 19.5x FY2027 consensus with a 2.13:1 payoff ratio, AppLovin is the most interesting risk/reward in this batch, and if the entry conditions below are met it should be owned. What we are declining is the timing and the conviction, not the company. In the Synthos frame, a name with a strong payoff ratio, a fair value well below the crowd's, an empty conviction lane and a binary event inside twenty-four hours is a Watch with a named entry, and converting it to a Buy tonight would be paying for information we will have for free tomorrow.

Pre-registered KILL criteria — what would take this to Avoid:

• Installation volume down again in the June quarter AND net revenue per installation growth below roughly 50%. Both legs of the growth decomposition decelerating together is the thesis failing, and the 10-Q will disclose both.
• June-quarter revenue below roughly $1.80B — below the March quarter, i.e. the first sequential decline in the file.
• A further goodwill impairment against the remaining $1,523.1M, after FY2025's $188.9M. Two in two years against the acquisition stack would say the optionality was bought badly.
• A platform-policy change at Apple, Google or Meta materially affecting mobile attribution or in-app advertising, which the 10-K names as the concentrated dependency.
• Gross margin below 85%, which would break the three-year expansion and say the cost of serving each dollar is rising as volume falls.
• The chief executive selling stock in size on the open market, which — given the Class B structure and the absence of any executive transaction in the current file — would be a materially new fact.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• A price in the $360-380 band with no deterioration in the fundamentals — roughly 17x FY2027E, putting the base case at +24% to +31%. This is the single most likely path to a Buy and it is only 9-14% below today's close; the 52-week low of $366.91 is inside it.
• The 2026-08-05 print showing installations returning to growth with net revenue per installation still rising above 70%. That resolves the one ambiguity the entire dive turns on, in the benign direction, and would justify paying up.
• A first quantified disclosure of e-commerce or connected-television revenue at material scale, which would give this company a second observable growth variable for the first time since the Apps sale — worth roughly a tier on its own, because it converts an unverifiable story into a measurable one.
• A reclaim of the 200-day moving average at $518.45, which would confirm the de-rating has ended — though on its own this is a technical flip only and would need the disclosure leg with it.
• Genuine name-level knowledge-base conviction appearing. At present there is one stale claim at conviction 50, and that alone is worth roughly one tier.

Where APP fits in the Synthos Framework Portfolio. The AI-applications sleeve — not the AI-infrastructure sleeve, because AppLovin buys compute rather than selling it — at 0% today with a 1.5-2.5% target on a fill in the $360-380 band or a confirming June-quarter disclosure. This is a position to be sized for a 2.529 beta and a 43% realised drawdown, which means small and scaled. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $419.70, with the fair-value anchors, the kill criteria and the flip conditions above all gradeable.

Single biggest risk: net revenue per installation stops rising. It is the only growth variable left. Installations are falling 18% year on year, so revenue growth of 30% in FY2027 requires the price per installation to compound roughly 50-60% again after +72% and +93%. It is set inside a proprietary auction, measured by a company-defined metric, with no external market reference and no segment, product or vertical disaggregation that would let anyone outside the company check it. If it decelerates to 30%, revenue growth goes to roughly zero on a shrinking install base, and a stock at 19.5x a forward estimate that no longer holds is not a 19.5x stock. That is the $300 bear case, and it requires no fraud, no scandal and no competitive event — only for one number to normalise.

Provenance & disclosures

• Traceability: 1 tagged knowledge-base claim names AppLovin out of 51,928 distilled claims (raw hits 14, used 1, discarded 13; breadth 1, net conviction positive-negligible). The used claim: darius_dale channel, 2025-01-21, speaker Joel Schuman, role independent, bullish, conviction 50, horizon thesis, skill 0.8 — a position disclosure, not an analysis, and it predates every fact this dive turns on. The 13 discards, with reasons: 8 are "Cash App" entity collisions on a case-insensitive boundary match of the ticker APP (channels business_breakdowns ×3, a16z ×3, real_vision, arthur_hayes; all entity-tagged to Block/Square/Cash App/Afterpay); 5 are substring collisions on "ad tech" inside "bad technology" (invest_like_the_best ×2), "broad tech basket" (lyn_alden), "baseload technologies" (macrovoices) and "a bad technical sign" (jordi_visser). Re-running the entity sweep case-sensitively on APP eliminates all eight Cash App collisions, which is the standing remedy for short tickers. No concentration sensitivity test is reportable — one claim from one channel is 100% concentration by construction. No claim in the lane carries a management speaker_role, so the half-weight management discount does not apply. The quote is verbatim from the stored claim text.
• Data as-of: fundamentals — income statement, balance sheet and cash flow through 2026-03-31, all filing-verified from the 10-Q filed 2026-05-06; annual figures for FY2025, FY2024 and FY2023 from the 10-K filed 2026-02-19 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873601 = 2026-08-04T20:00:01Z ($419.70, +3.33%; 50-DMA $487.77; 200-DMA $518.45; RSI 39.0; MACD −23.48) · knowledge-base claims 2026-08-04. AppLovin's fiscal year is the calendar year. All figures come from the Synthos vendor data file for APP or from the SEC filings in the APP archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-19 (fiscal 2025); 10-Q filed 2025-11-05 (September 2025 quarter — the source of the restated 2024 comparatives that resolve the quarterly defect); 10-Q filed 2026-05-06 (March 2026 quarter); 8-K filed 2026-04-07 (executive and board transitions); 8-K filed 2026-05-06 (first-quarter results announcement — the cover document only; Exhibit 99.1, the press release carrying the financials, was not extracted); 8-K filed 2026-06-05 (annual meeting voting results). All six carry preserved [TABLE] statement data. No filing in the archive covers the June 2026 quarter, which reports 2026-08-05.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the 2024 quarterly revenue series — the vendor's four quarters sum to $3,977.1M against its own annual $3,224.1M, and the 10-Q's restated comparatives ($835,186 thousand for Q3 2024, $2,224,571 thousand for nine months) prove Q1 2024 is $678.4M not $1,058.1M and Q4 2024 is $999.5M not $1,372.8M, which changes reported year-on-year growth for Q1 2025 from +9.5% to +70.8% and for Q4 2025 from +20.8% to +65.9%; the FY2021 and FY2022 annual revenue rows, which are on the total-company basis and are discarded, having made FY2023 look like a 34.6% decline when continuing-operations revenue rose 75.5%; seg_geo and seg_prod for FY2024 and earlier, on the pre-divestiture basis and therefore not comparable with FY2025; est.ebitdaAvg/ebitAvg, rejected in full for carrying a fixed-ratio fabrication signature (exactly 52.17% and 36.40% of revenue in every forward year) against an actual FY2025 adjusted EBITDA margin of 82.3%; the FY2029 estimate row, incoherent (revenue below FY2028 while EPS rises 20.7%); the FY2030 row, with EPS and net income of exactly zero; free cash flow, $3,942.8M vendor against $3,951,952 thousand on the 10-K's own definition; and quote.yearHigh/yearLow ($745.61/$359.00) rejected for tech.hi52/lo52 ($733.60/$366.91). Where vendor and filing AGREED — recorded because this defect class usually bites: the enterprise value of $141,749,174,259 reconciles exactly to the 2026-03-31 10-Q balance sheet (market cap $140.994B + long-term debt $3,514.022M − cash $2,758.671M); the share count (10-Q cover 336,294,000 against 335.9M implied, a 0.1% gap explained by post-quarter buyback); the FY2025 revenue, gross profit, operating income and net income lines; the FY2025 buyback of $2,191.9M; the FY2025 geographic split summing exactly to revenue; and every trailing margin and coverage ratio tested.
• Basis note: the vendor's earn_cal "actual" EPS figures reconcile to GAAP diluted EPS from continuing operations (March 2026 actual $3.56 against net income $1,205.613M on 338,729,000 diluted shares = $3.56), so consensus in this file appears to be struck on a basis close to GAAP rather than a heavily adjusted non-GAAP. epsdiluted is null in every row of the vendor's income statements, so all trailing per-share figures here are computed and labelled. Forward multiples use est.epsAvg. Given the small and shrinking stock-compensation add-back (3.8% of revenue in FY2025), no conclusion turns on the choice.
• Estimate coverage: 22 analysts on FY2026 revenue and 19 on FY2026 EPS; 23 and 20 on FY2027 — the best-covered rows in this batch and the anchor for all three fair values; 17 and 15 on FY2028. FY2029 (14 revenue analysts) and FY2030 (8) are excluded from every conclusion for the incoherence documented in Section 8.
• Peer note: the vendor peer set — C3.ai, Asana, Bill.com, Datadog, DocuSign, JFrog, GitLab, Intuit, monday.com, Shopify, Snowflake, Atlassian, Workday, Zoom — contains no advertising business at all and is rejected as a comparison universe. AppLovin's own 10-K names no competitors by name in a comparable list; it names Meta and Google as simultaneously distribution platforms, significant partners and advertising clients. No peer-multiple comparison is drawn anywhere in this dive.
• Fair-value caveat: the $300 / $470 / $675 anchors are multiples of the FY2027 consensus EPS distribution — 14.7x the low of $20.393, 21.8x the mean of $21.570, and 29.3x the high of $23.036. Stated arithmetic, not a discounted cash flow. The base is deliberately sensitivity-disclosed: 20x gives $431, essentially spot; 24x gives $518. The whole answer lives inside a 20-24x band on one estimate row, and the multiple is a wider source of variance than the earnings.
• Timing: second-quarter 2026 results are released 2026-08-05, ONE DAY after this dive, against consensus revenue of $1,941.897M and EPS of $3.76. This is the dominant near-term fact and the verdict is set with it as an explicit input: a Watch struck the evening before a print on a 2.529-beta stock is a statement about the calendar as much as about the price. The last report was 2026-05-06 (EPS $3.56 against $3.44 estimated, revenue $1,842.449M against $1,773.727M), the sixth consecutive beat in the file.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• 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.