APP › analyze
APP · Analyze
2026-09-28 · Price $309.375 · Market cap $103.9B (diluted) · EV $104.4B 52wk $297.50 – $745.61 (−58.5% off high, −56.4% over 12 months) · Beta 2.49
This supersedes analyze-from-before-2026-09-10.md (ACCUMULATE, 6.5, FV $300-430).
Requested as a fresh /analyze — the watchlist's own analysis queue lists APP under
"Fresh /analyze", and this pass went to primary filings rather than re-testing the prior
claim ledger. Three prior claims did not survive that contact.
Today's tape is a tell. APP opened $322.74 (+3.9%) on "buyers returning after a year-long slide" coverage and faded all day to $309.38 (−0.44%). The bounce that the financial press wrote up this morning did not hold. Treat the "sentiment has turned" narrative as unproven.
Verdict — ACCUMULATE · Conviction [6.5] (unchanged)
Hold the rating, hold the verdict, and stage the remaining size around 4 November.
The rating is unchanged but the composition of it has shifted materially. The business is in better shape than the last two reports described — the margin scare that drove the August de-rating was miscast, and I can now show that from the company's own 8-K rather than a vendor feed. Against that, two legal items I did not previously have are now dated and live, and the headline quality metrics the prior reports leaned on are substantially a divestiture artifact. Those roughly offset, which is why 6.5 stands rather than moving.
A portfolio-specific passage was removed from the public build.
The one-line case: a business compounding revenue ~50% at an 84% adjusted-EBITDA margin with near-zero capex, trading at 18.6× this year's earnings, where the price embeds only 14% FCF growth — held back by a regulatory matter the company has never disclosed in a filing and a deceleration it blames on its own release cadence.
What changed since 2026-09-10 — and what I got wrong before
| # | Claim in prior reports | Status after this pass |
|---|---|---|
| 1 | "adj EBITDA margin 85% → 83%, the step-down is guided as permanent" | ❌ RETRACTED. Primary source: 81% (Q2'25) → 84% (Q2'26), Q3 guided 83%. Margin expanded 300bps YoY. |
| 2 | "ROIC 14.8% → 42.0% → 70.3%" as proof of quality | ⚠️ HEAVILY QUALIFIED. ~50-60% mechanical — the Apps divestiture cleaned numerator and denominator. |
| 3 | "R&D is only ~5.1% of revenue" as a red flag | ❌ RETRACTED as a criticism. R&D is +127% YoY ($44.0M → $99.9M), 3.5% → 5.2% of revenue. Rising hard. |
| 4 | Securities class action, unspecified | ✅ NOW DATED. Brownback MTD fully briefed since Feb 2026 — undated binary, pending 7+ months. |
| 5 | — | 🆕 SECOND class action (Rosen), class period Feb 12 – Aug 5 2026, alleging the AI-creative/model-cadence story was itself misleading. |
| 6 | "Q3 prints early Nov (unconfirmed)" | ✅ CONFIRMED 2026-11-04 (Yahoo isEarningsDateEstimate: false). |
| 7 | Short interest unclear / "$12B notional" | ✅ RESOLVED and benign. 10.94M sh, 3.76% of float, 2.36 days to cover, falling (11.37M prior month). |
| 8 | "Insiders sold ≈$305.7M, zero buys" | ✅ CARRIED. Still zero open-market buys ever. No sales recorded after 2026-07-06. |
1. The margin correction — the most important finding in this report
The August 21% single-day drawdown was driven by a narrative of permanent margin impairment. That narrative does not survive the 8-K. From AppLovin's own Q2 2026 earnings exhibit:
| Q2 2025 | Q2 2026 | Q3 2026 guide | |
|---|---|---|---|
| Revenue | $1,258.754M | $1,923,686K | $2,055 – $2,085M |
| Adjusted EBITDA | $1,018.347M | $1,613,823K | $1,710 – $1,740M |
| Adj EBITDA margin | 81% | 84% | 83% |
| Free cash flow | — | $863.317M | — |
| Buyback | — | 1.1M sh / $551.3M | — |
The margin did not step down from 85%. It stepped up from 81% to 84% year over year, and is guided to 83% — roughly 200bps above the year-ago Q3. The prior report's framing ("the step-down is guided as permanent") compared a sequential peak against a guide and read a 1-point give-back as structural impairment.
The GAAP line corroborates this independently. Operating margin by quarter:
| Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 | |
|---|---|---|---|---|---|
| Revenue ($M) | 1,259 | 1,405 | 1,658 | 1,842 | 1,924 |
| QoQ growth | — | +11.6% | +18.0% | +11.1% | +4.4% |
| GAAP op margin | 76.1% | 76.8% | 76.9% | 78.2% | 77.7% |
| R&D % of revenue | 3.5% | 3.1% | 5.0% | 5.1% | 5.2% |
| Diluted shares (M) | 342.2 | 341.0 | 339.9 | 338.7 | 337.0 |
GAAP operating margin is flat-to-up across the whole window and sits at 77.7%. There is no margin deterioration in this business on any measure I can construct. The watchlist break trigger "EBITDA margin <82% for 2 quarters" is nowhere near firing — it is at 84% with 83% guided.
The deceleration is real; the margin story attached to it was not. Sequential revenue growth genuinely halved (+11.1% → +4.4%), and that remains the open question. But the market repriced this name on a margin thesis that the filings contradict.
The Q2 FCF drop is tax timing, not deterioration
FCF fell from $1,291M (Q1) to $869M (Q2), −32.7%, while revenue rose 4.4%. Cause, from the cash-flow statement: cash taxes paid $106.7M → $533.1M (+$426M) and working capital −$14.8M → −$490.0M. Receivables built $214M. DSO went 95.6 → 102.7 days — but a year ago it was 114.3 days, so this is a partial give-back within a multi-quarter improvement, not a new credit or collection problem. TTM FCF $4.499B, FCF/share $13.39.
2. Health — the numbers are not in question
| Metric | Value | Read |
|---|---|---|
| TTM revenue | $6.829B (+52.8% YoY) | ✅ |
| TTM FCF | $4.499B · FCF/sh $13.39 | ✅ FCF conversion ~66% of revenue |
| Gross margin | 88.5% | ✅ |
| GAAP operating margin | 77.7% | ✅ flat, not eroding |
| Capex | ~$28M/yr on $6.8B revenue (0.4%) | ✅ genuinely asset-light |
| Net debt / EBITDA | ($3.515B − $3.053B) / $5.41B = 0.09× | ✅ effectively unlevered |
| Current ratio | 4.30 | ✅ |
| Shares (diluted) | 342.2M → 337.0M, −1.5%/yr | ✅ real, self-funded |
| Buybacks | $551.3M in Q2 (63% of FCF), $0 debt issued, $0 M&A | ⚠️ see below |
| SBC | $85.6M/qtr = 4.5% of revenue, 9.9% of FCF | ✅ genuinely low for the sector |
| Employees | 876 → $7.8M revenue each | ✅ |
| Equity | $1.167B → $3.163B over 4 quarters | ✅ book repairing fast |
Capital allocation — one genuine criticism. Q2's buyback retired 1.1M shares for $551.3M, an average of ≈$501/share. The stock is $309. That is a 62% markup to today's price, and Q1's $981.7M was executed into a similarly elevated tape. Management bought aggressively into the top and has bought less as the price fell. This is the ordinary hazard of a buy-back-the-whole-way-down program, but it is a real mark against the capital-allocation record and it is not something the prior reports flagged.
Cloud commitment (from the 10-Q, not previously surfaced): a $1.3B three-year minimum spend with a cloud provider signed August 2024; $953.5M paid by 30 June 2026. That is 73% of the commitment consumed in ~61% of the term — they are running ahead of it. Compute is the one real variable cost in this model and it is scaling with revenue.
3. Moat — 6.5/10, held
The three claimed edges have very different durabilities, and only one is a moat.
- Distribution (MAX mediation + SDK footprint) — load-bearing and durable. A genuine two-sided network effect with switching costs: publishers integrate once, advertisers bid where the inventory already is. This survives intact even under an adverse SEC outcome.
- Data (proprietary signal) — contested, and possibly borrowed rather than owned. This is exactly what the SEC probe and four short reports target. If the edge depends on signal extracted contrary to platform terms, it is not a moat — it is an arbitrage that any one of Apple, Google or Meta can close unilaterally.
- Model (better ML on the same signal) — a treadmill, not a wall. Q2 proved this: growth is release-dependent by management's own admission. A 2-year lead maintained by continuous R&D and compute spend.
The ROIC number needs a health warning. Prior reports cited ROIC 14.8% → 42.0% → 70.3% and GM 80.8% → 88.9% as evidence of a strengthening moat. The Apps/studio divestiture (Tripledot, closed 30 June 2025) removed a low-margin, goodwill-heavy segment — cleaning both the P&L and the invested-capital base. Roughly 50-60% of that improvement is compositional, a one-time step that has now fully played through; there is no low-margin segment left to shed. Anyone underwriting further ROIC expansion off this base is extrapolating an artifact.
The honest version: the post-divestiture trend is still good — every quarter from Q3'25 to Q2'26 is a clean pure-advertising quarter, and across them GAAP op margin went 76.8% → 77.7% and adj EBITDA margin 81% → 84%. That is genuine incremental operating leverage. It is just far less dramatic than "14.8% → 70.3%" implies.
Adversarial test on e-commerce. Meta/Google/Amazon have more data, more capital and the demand relationships; if they choose to fight directly on price they win. The mitigating fact is that they have not had to — a Jefferies advertiser survey puts APP's e-commerce budget share at 11.1%, up 169bps, but sourced from advertisers diversifying incremental budget rather than defecting from the walled gardens. APP is currently additive, not yet a share-taker. That is comfortable near-term and means the thesis is untested under real competitive response. Moloco is the one like-for-like competitor that could contest AXON on model quality rather than capital.
Retention is the load-bearing unknown. Management claims near-zero churn past 30 days of advertiser spend and $70K+ first-year spend per new advertiser. That is company-disclosed, on its own newest and most story-dependent segment, with no third-party corroboration, and the self-serve platform only opened to all advertisers in June 2026 — three months of live data. Too early to call this a recurring book. The $1B annualized run-rate figure is from April 2026 and has not been refreshed since.
Evergreen: no. This company has re-underwritten its core once already (Apps divestiture and model pivot) and has a second existential test running concurrently. A business whose core data-sourcing mechanism is under federal investigation is not a forever business yet. It fails the 🏛 Evergreen test and belongs in a growth sleeve, not the compounder sleeve.
4. The legal stack — now dated, and larger than before
I read Part II Item 1 of the Q2 FY26 10-Q and the FY2025 10-K directly rather than through a summarizer. Three findings:
(a) The SEC investigation is disclosed in NO SEC filing. Not in the FY2025 10-K (filed 19 Feb 2026), not in the Q2 FY26 10-Q (filed 5 Aug 2026). Searches for "SEC investigation", "investigation by the", "subpoena", "Wells", "informal inquiry" return zero hits in the 10-K. Item 1 mentions only generic "governmental and other regulatory investigations and proceedings" boilerplate, and Note 5 states: "As of June 30, 2026 and December 31, 2025, the Company had no material loss contingencies related to legal proceedings for which accrual or disclosure was required."
This cuts both ways and should be read carefully. Companies routinely do not disclose non-public investigations absent a Wells notice or a materiality determination — this is normal and not improper on its face, and the continued absence of disclosure is weak evidence against imminent enforcement. But it also means there is zero company-sourced status information, and the last independent confirmation that the probe is even active is a February 2026 Bloomberg report. We are seven months past the last data point on the single item capping conviction. That is the epistemic position, and it should not be dressed up as anything firmer.
(b) Brownback Action — a pending, undated binary nobody is watching. Verbatim from the 10-Q: class period 7 Nov 2024 – 27 Mar 2025; Amended Complaint 12 Sep 2025 adding CTO Basil Shikin; "The defendants filed a motion to dismiss the Amended Complaint in November 2025, and that motion was fully briefed as of February 2026." A fully-briefed MTD has been awaiting ruling for seven months. It can land any week, and a denial would be a materially negative surprise into an already-fragile tape. Parallel shareholder derivative suits track it.
(c) A second class action, and its allegation is management's own excuse. The Rosen action (solicitation releases 17-24 Sep 2026, lead-plaintiff deadline 16 Nov 2026) alleges APP overstated its generative-AI video ad-creative feature and the consistency of the model improvements underpinning the "virtuous cycle" pitch, over a class period of 12 Feb – 5 Aug 2026. The significance is not the suit itself — stock-drop suits are cheap to file. It is that the allegation directly targets the explanation management gave for the Q2 miss. If the Q3 print misses again on the same stated cause, this suit acquires real substance. Its class period ends the day of the Q2 print, which post-dates the 10-Q, so its absence from that filing is expected, not suspicious.
5. Sentiment, insiders, positioning
Insiders — the streak is intact and it is the cleanest bear signal on this name. Applying
the RSU pitfall discipline (pitfall-yahoo-insider-purchases-counts-rsu-grants) row by row
across the full history back to Dec 2024: zero open-market purchases, ever. Every
dollar-value entry is a disposition; option exercises were at legacy $5.05/$25.55 strikes
(cashless exercise-and-sell, not buys). Genuine open-market selling Jun–Jul 2026 totals
≈$148.4M — Director Vivas $82.6M, CEO Foroughi $51.1M across three sales, Officer
Valenzuela $11.3M, Director Webb $3.4M. Cumulative Mar–Jul ≈ $324M.
⚠️ No insider transaction appears after 2026-07-06. That is either a genuine pause (plausible — blackout around the print plus a collapsed price) or a lagging feed. I cannot distinguish these and am not going to guess.
Analysts — total capitulation on price, total refusal on ratings. 31 analysts, mean $498.37, median $475, low $325, high $790. Every target is above the $309 spot. Since August essentially every firm cut targets 10-25% while holding the rating: Morgan Stanley $650→$450, Evercore $630→$510, Goldman $585→$465, RBC $700→$575, Citi $710→$650. One downgrade (BofA, Buy→Neutral, 11 Aug). Still zero Sell ratings. A name down 56% in twelve months with no sell rating anywhere is a crowded long that has not finished unwinding — do not anchor on the $498 mean.
Short interest — resolved and benign. 10.94M shares, 3.27% of shares out / 3.76% of float, 2.36 days to cover, down from 11.37M the prior month. The "$12B notional, largest outside Mag7" claim circulating in secondary sources is not supported — 10.94M × $309 is $3.4B. There is no squeeze setup and no large standing bear position here.
Governance — worth naming. Yahoo/ISS-style risk deciles: board 9, compensation 10, shareholder rights 10, overall 10 (10 = highest risk). Dual-class structure with Class B super-voting shares. This is a founder-controlled company; minority holders have limited recourse if the SEC matter is mishandled.
Platform risk — no news is not good news, it is no news. I found no Apple or Google policy action touching APP's data access since August 2026. That is an absence of evidence from a search pass, not a confirmed all-clear. This remains the largest tail risk in the name — the ATT precedent shows platforms act unilaterally, on their own timetable, and that the damage is multi-year targeting impairment rather than a fine. Meta absorbed ATT at $1.5T; APP at $104B with a single revenue engine has far less room.
6. Valuation
The four vendor traps — all four still firing. (Pitfalls re-verified, not assumed.)
- Forward P/E is the wrong fiscal year (
pitfall-vendor-forward-eps-is-the-wrong-fiscal-year, 4th firing on this ticker). YahooforwardPE 14.75usesepsForward $20.98= FY2027. Yahoo's ownepsCurrentYearis $16.67 → FY2026 P/E is 18.6×, not 14.7×. - 3yr revenue CAGR 24.8% is corrupted by the divested Apps segment
(
pitfall-divested-segment-corrupts-multiyear-cagr). The 2022 base ($2.82B) is apps-inclusive. Clean advertising-only: $1.84B → $3.22B → $5.48B = +75.0%, +70.2%. - Graham IV is VOID. √(22.5 × 13.02 × 9.43) = $52.57 against a $309 price. BVPS $9.43 on a capital-light business that has bought back $2.2B+ of stock. Graham is meaningless here and reporting it as a fair value would be malpractice. (Noting the direction though: BVPS rose from $7.03 to $9.43 as equity nearly tripled — the hollowing is reversing.)
sharesOutstanding305.73M is Class A only. The 10-Q balance sheet: Class A 305,084K + Class B 30,208K = 335,292K outstanding; diluted weighted-average 337.0M. Yahoo'simpliedSharesOutstanding335.94M is the right field. Using 305.73M understates market cap by ~9%.
Models applied conditionally. No dividend → DYT and DDM are N/A. Graham void (above). That leaves FCF-based methods and multiples, which is correct for a capital-light high-growth name.
Reverse-DCF — the core of the case. At EV $104.4B on TTM FCF $4.499B (10% discount, 5 years then 3% terminal), the price embeds 14.0% FCF growth. Against a company guiding +46-48% revenue growth for Q3. (Note: this hurdle is slightly higher than the prior report's 13-14% at $351, because the Q2 tax/working-capital swing shrank the TTM FCF base — the denominator moved, not just the price.)
Scenario fair value (same model, per share, net debt included):
| Scenario | 5yr FCF CAGR | Value/share |
|---|---|---|
| Bear — deceleration is a ceiling | +8% | $242 |
| Base — decelerates hard but compounds | +15% | $321 |
| Bull — e-commerce works | +25% | $472 |
| At guidance | +35% | $676 |
Multiple cross-check: FY26E EPS $16.67 at 18-25× = $300-417. FY27E EPS $20.98 at 18-25× = $378-524.
Bogle expected return: 0% yield + earnings growth (~25-35% near-term, decaying) ± multiple change. At 18.6× FY26 against a decelerating-but-still-50% grower, the multiple is more likely to expand than compress if Q3 holds — which is exactly the binary.
Fair value $300 – $440 · central ≈$370
Entry $280 – $320 · strong buy < $270 · Trim 25× fwd (FY2027) ≈ $525
Spot $309.375 sits inside the entry zone. The trim is expressed as a forward multiple per the standing valuation rule — at FY27E EPS $20.98 that is ≈$525 today and rises on its own as earnings grow. Do not convert it to a frozen dollar.
A portfolio-specific passage was removed from the public build.
8. Tensions I am not resolving
Per the framework, these are named rather than buried.
- Cadence or ceiling — genuinely unresolved, and it is the whole thesis. For cadence: R&D +127% YoY, e-commerce reaccelerating while gaming stalled (selective, not broad, saturation), management says material improvements went live early Q3. For ceiling: management explicitly declined to promise a quarterly cadence, calling it "fundamental R&D" — which reads as "we do not control when the next unlock arrives." Only the 4 November print settles this.
- Quality: real or compositional? Fundamentals sees flat 78% GAAP operating margins, rising R&D and 66% FCF conversion — a clean, healthy business. Moat sees 50-60% of the headline ROIC/margin improvement as a divestiture artifact. Both are correct. The current earning power is real; the rate of improvement is not repeatable. Do not extrapolate the trend line.
- The SEC silence. Non-disclosure is normal pre-Wells and weakly bullish; it also means we have no company-sourced status at all and are relying on a seven-month-old press report. I am treating "active and unresolved" as the working assumption, explicitly flagged as not confirmed as of today.
Data gaps — stated, not papered over
- SEC probe status since Feb 2026 — unconfirmed in either direction.
- Brownback MTD ruling date — unknowable; fully briefed since Feb 2026.
- E-commerce run-rate is an April 2026 figure; advertiser count, churn and self-serve progress have no September update. Retention rests entirely on management assertion.
- Insider feed shows nothing after 2026-07-06; genuine pause vs. lagging feed indistinguishable.
- Apple/Google policy: no change found, which is not a confirmed all-clear.
- Yahoo's
freeCashflowfield ($3.18B) disagrees with the sum of reported quarters ($4.499B) and with the company's own Q2 figure ($863.3M). I used the company's numbers.
Break triggers & upgrade conditions
🔴 Break (any one forces re-underwrite): sequential revenue <+5% for a second consecutive quarter · adj EBITDA margin <82% for 2 quarters (currently 84%, guided 83% — far from firing) · SEC enforcement action or Wells notice · any Apple/Google data-access restriction · Brownback MTD denied · e-commerce run-rate flat or disclosed as declining · buyback paused or debt-funded · GM <85%
🟢 Upgrade to [7.0]: SEC investigation closed without enforcement · Q3 ≥+8% sequential with margin ≥83% · any open-market insider purchase, at any price · e-commerce run-rate disclosed >$2B · Brownback MTD granted · third-party corroboration of the advertiser-retention claim
⏳ Next hard date: 2026-11-04 — Q3 FY26 print (confirmed). Watch sequential revenue against the +7-8% guide and the 83% margin guide — not the YoY headline, which will read ~+46% regardless and tell you nothing.