APP · Analyze — Financebot

2026-08-05 · Price ~$351 (post-print, after hours) · Market cap ~$118.9B (diluted) · Communication Services / Advertising Agencies · First analysis file ever written on this name

A portfolio-specific passage was removed from the public build.

Verdict — WATCH · Conviction [6.0]

This is a genuinely elite business at a price that is finally defensible, carrying one unresolved risk that goes to the source of the moat itself.

The knee-jerk read on tonight's −22% is that the market overreacted to a 0.6% revenue miss. That read is right about the headline and wrong about the cause. The market did not reprice a rounding error; it repriced sequential growth halving from +11.1% to +4.4%, a self-guided margin step-down, and management's explanation that the shortfall came from "a lighter-than-normal pace of advertising-model improvements." That last sentence is the most important thing in the release, and it is not a good sentence.

What keeps this at 6.0 rather than 4.0 is that the business quality is not in question and the valuation hurdle is now genuinely low. ROIC has gone 14.8% → 42.0% → 70.3%, gross margin 80.8% → 88.9%, operating margin 41.9% → 78.2% — all rising, all in three years. SBC consumes 7.6% of FCF (the disqualifier that killed most of the March 2026 software cohort). Buybacks are fully self-funded at ~50-56% of FCF. And a reverse-DCF says $351 embeds only ~13-14% FCF growth for five years against a company guiding +46-48%.

What keeps it from 7.0+ is the SEC. An active investigation into AppLovin's data-collection practices — confirmed February 2026, still open, whistleblower-triggered, preceded by four separate short-seller reports — is not a headline risk. The alleged conduct is the data advantage. If the data acquisition is curtailed, the thing being valued at 70% ROIC changes character. No price fully compensates for an unresolved binary of that shape, and it is why the entry zone below sits meaningfully under spot rather than at it.

Action: Hold the stub. Do not add here. The position is ~$435 and costs nothing to carry. The sizing question is legitimate and live — but it should be answered after the Q3 print and with more clarity on the SEC matter, not into an after-hours gap on the night of a guidance reset.

⚠️ Data Integrity — Four Vendor Traps On This Name

Per Knowledge/Playbook/pitfall-*, checked before trusting any figure. Four fired.

#	Trap	Vendor says	Truth	Impact
1	Forward P/E is the wrong fiscal year	`forwardPE` 19.35x, `forwardEps` $21.5875	Yahoo's own `epsCurrentYear` is $16.56 (FY2026). $21.59 is FY2027. Reconstructed: Q1'26 GAAP dil EPS $3.56 + Q2'26 adj $3.76 = H1 ~$7.32 — consistent with ~$16.5, impossible for $21.59	The real FY26 multiple is 25.2x at $417.80 and 21.2x at $351, not 19.4x/16.3x. ⚠️ My own 8/05 screen quoted the wrong one. → `pitfall-vendor-forward-eps-is-the-wrong-fiscal-year` fires a third time
2	Revenue CAGR corrupted by a divested segment	`fin.py` prints "Revenue CAGR 24.8% (3y)"	FY2022's $2,817M includes the Apps/games business; FY2023-25 are continuing operations only. Like-for-like: $1,842M → $3,224M → $5,481M = +75.0%, +70.0%	The vendor CAGR understates growth by ~3x and makes an exceptional business look ordinary. New pitfall note written
3	Graham IV is void	Graham $42.64 vs a $351 price	√(22.5 × $11.50 × $7.03). BVPS is $7.03 because ~$2.2B/yr of buybacks against $1.93B retained earnings hollows the book	The model punishes the exact capital allocation creating the value. Discard, do not report as "92% overvalued" — same call as GDDY
4	Share count is Class A only	`sharesOutstanding` 305.7M	Dual-class. Yahoo's own market cap implies ~336M; diluted is 338.7M	Using 305.7M understates market cap by ~10% → `pitfall-yahoo-share-count-dual-class-fpi`

✅ Checked and did NOT fire: pattern-net-margin-above-operating-margin-is-a-tripwire — Q1'26 net income $1,206M sits below operating income $1,440M. Earnings are operating earnings; there is no unrealized-mark inflation here, unlike GOOGL/AMZN/AMAT. This is a real point in APP's favour and worth stating explicitly.

The Print — What Actually Happened

Metric	Q2 2026	Guide / Consensus	Read
Revenue	$1.924B, +52.9% YoY	Company guide $1.915-1.945B; consensus ~$1.935B	✅ INSIDE its own guided range, below the $1.930B midpoint. −0.6% vs consensus
Adj EPS	$3.76, +57.3% YoY	$3.67	✅ BEAT
Net income	$1.27B, +55% YoY	—	✅
Adj EBITDA	$1.61B, +58% YoY	Guided margin 84-85%	🚩 Margin 83% — BELOW the guided range
Q3 revenue guide	$2.055-2.085B (+46-48% YoY, +7-8% seq)	~$2.068B	🚩 ~0.6% light at midpoint
Q3 adj EBITDA guide	$1.71-1.74B (+48-50%)	—	🚩 ~83% margin — the step-down is guided as permanent

The number that actually moved the stock

Not the 0.6%. This:

	Q4'25 → Q1'26	Q1'26 → Q2'26	Q2'26 → Q3'26E
Sequential revenue growth	+11.1%	+4.4%	+7-8% (guided)
YoY revenue growth	+59.0%	+52.9%	+46-48% (guided)
Adj EBITDA margin	85%	83%	~83% (guided)

Sequential growth halved in one quarter. A 0.6% miss on the headline is noise; a 670bp collapse in sequential momentum, arriving with a permanent-looking 200bp margin reset, is a different quarter than the one consensus modelled. The market repriced the second derivative, and it was right to notice.

Management's explanation — and why it is the crux

Management attributed the shortfall to "a lighter-than-normal pace of advertising-model improvements during the quarter," while stating explicitly that there was no indication of weaker advertiser demand and no change in the competitive environment, and that the next model improvement is now live, has driven a strong start to Q3, and is reflected in guidance.

Take that at face value and it is reassuring: demand is intact, the miss was internal and already fixed. But read what it concedes. AppLovin is telling you its revenue in any given quarter is a function of how fast its own model improved — an input that is not externally observable, not on a schedule, and not something an outside investor can verify or forecast.

This is the honest core of the bear case, and it is structural rather than cyclical: the company has just demonstrated that a quarter can miss for reasons visible only from inside the model. Combined with the R&D figure below, it means the growth rate is levered to a black box.

§1 · Fundamentals — Health Scorecard

The growth engine (continuing operations)

	FY2023	FY2024	FY2025	TTM / Q1'26
Revenue	$1,842M	$3,224M +75.0%	$5,481M +70.0%	$6,829M
Gross profit	$1,489M	$2,700M	$4,820M	—
Gross margin	80.8%	83.7%	87.9%	88.9% (Q1'26)
Operating income	$772M	$1,911M	$4,152M	—
Operating margin	41.9%	59.3%	75.8%	78.2% (Q1'26)
Net income	$357M	$1,580M	$3,334M	~$4,413M
FCF	$998M	$2,072M	$3,939M	$4,402M
Diluted shares	362.6M	347.8M	342.0M	338.7M
FCF / share	$2.75	$5.96	$11.52	$12.99

FCF CAGR (FY23→FY25): +98.7%/yr. FCF/share CAGR: +104.7%/yr. The framework asks for a 5-8yr CAGR; it is not available on a comparable basis because the Apps divestiture breaks the series before FY2023 — flagged rather than fabricated.

Revenue/share FY25: $16.03. FY2026E revenue ~$8.09B (+47.5%) on the Q3 guide.

Capital allocation of FCF — clean, and rare

Use	FY2025	TTM
Buybacks	$2,190M (55.6% of FCF)	$2,173M (49.4% of FCF)
Dividends	$0	$0
Net debt repayment	−$19M	~flat
M&A	$0	$0
Cash accumulation	+$1,790M	+$1,566M

No dividend, no acquisitions, no debt-funded anything. Roughly half of free cash flow retires stock and the other half builds the balance sheet. Cash went $697M → $2,759M in five quarters while debt stayed flat at ~$3.5B.

Share count is genuinely falling: 371.6M → 338.7M, −2.7%/yr over three years (−1.8% most recent). Per §0 — "growth funded by dilution isn't growth for me" — this passes cleanly, and it is the opposite of the dilution that disqualified FSLY, U and NVTS.

Balance sheet

Metric	Value	Read
Total debt	$3,514M	Flat for two years
Cash	$2,759M	Rising fast
Net debt	$755M	~0.13x TTM adj EBITDA — effectively unlevered
Debt / assets	45.6% (from 48.8%)	🚩 Headline looks leveraged…
Interest coverage	28.1x	…and the coverage says it isn't
Goodwill	$1,523M	20% of assets, 64% of equity — from Adjust/MoPub/Wurl, stable, no impairment
Equity	$2,363M	Low by choice — buybacks, not losses (retained earnings +$1,927M and rising)

⚠️ Do not read Debt/Assets 45.6% as a risk flag in isolation. It is high only because buybacks keep the asset base small. At 0.13x net debt/EBITDA and 28x coverage, with $2.76B of cash against $3.51B of debt, the balance sheet is a non-issue. Interest expense is ~$205M/yr (~5.8%) against ~$5.8B of annualized operating income.

The SBC test — passes, decisively

SBC ~$83.4M/quarter ≈ $334M/yr = 7.6% of FCF, 4.5% of revenue. For comparison from this repo's own files: TEAM 96% of FCF, ZS ~82%, DOCU 59%, NOW 47.7%, ADBE 19.7%. AppLovin's headline FCF is close to real cash. SBC-adjusted FCF ≈ $4.07B → $12.01/share → 29.2x at $351.

§2 · Moat & Competitive Advantage

Quantitative base — the strongest ROIC trend in this repo

	FY2023	FY2024	FY2025
NOPAT	$704M	$1,885M	$3,606M
Avg invested capital	$4,746M	$4,488M	$5,124M
ROIC	14.8%	42.0%	70.3%

ROIC rising from below WACC to ~70% in two years, while gross margin expands 800bps. Per §2, "rising/stable high ROIC = a real moat throwing off returns; margin compression = moat erosion." This is the inverse of erosion, and it is unambiguous. Note also that invested capital grew only 8% while NOPAT grew 412% — this is not capital being poured in.

What the moat actually is

Axon (rebranded AppLovin Ads) is a proprietary deep-learning bidding and prediction engine — not an LLM — that forecasts each user's conversion value and bids accordingly. The durable asset is the training data, assembled deliberately: the 2021 acquisitions of Adjust (attribution/measurement) and MoPub (exchange + mediation) gave AppLovin sight of the entire funnel — impression → auction → install → post-install monetization — across app categories and device types.

Moat sources, named per §2: - Cost advantage / scale economies in data — the real one. Each additional advertiser improves predictions for every other. 876 employees producing $5.48B of revenue ($6.25M per employee) is what a data flywheel looks like on an income statement. - Switching costs — moderate, not high. Advertisers chase ROAS and will move. - Network effects — genuine between the supply (mediation) and demand (Axon) sides. - Intangibles / brand — weak. Nobody buys AppLovin for the brand.

Adversarial stress-test — "you are a well-funded rival; how do you attack?"

In mobile gaming: you don't, and this is settled. Unity — the one rival with structurally comparable assets — has fumbled the ironSource integration through executive turnover and lost the monetization war decisively despite owning the dominant game engine. That is close to the strongest possible evidence: the best-positioned competitor tried and lost.

In e-commerce, where growth now has to come from, the answer is very different. The competitors become Google, Meta and Amazon — each with more first-party data, more direct merchant relationships, and no dependency on third-party mobile signal. AppLovin is leaving the pond it dominates to swim where the whales live. Early ROAS is reported as comparable to Meta's Audience Network, which is a real datapoint and also a modest bar — Audience Network is not Meta's crown jewel.

The most effective attack is not a product — it is the platform layer. Apple and Google control the signal that feeds Axon. ATT already reshaped this industry once. A privacy or policy change at the OS level is the single fastest way to damage AppLovin, and AppLovin gets no vote. That risk is materially correlated with the SEC matter below: regulatory findings about data practices are precisely what prompts platform-level restriction.

Revenue-stream map

Stream	Scale	Durability
Mobile gaming advertising	The large majority of revenue	High — won, defensible, but a maturing market
Web / e-commerce advertising	~$1B annual run-rate, from an Oct 2025 referral launch	Unproven but REAL. Self-serve opened globally June 2026
CTV (Wurl)	Small	Optionality; unproven

✅ principle-story-vs-revenue — passes. The e-commerce thesis is not a TAM claim: the web advertising line reached a $1B annualized run-rate within months. That is realised revenue growing, which is exactly what the principle demands before crediting a growth story. ⚠️ But management's "non-gaming TAM is 5-10x gaming" IS a TAM claim and should be given no weight whatsoever.

Disruption forecast (5-10 year arc)

Generative AI in creative — currently a tailwind. Axon 3.0 generating ads in real time lowers the advertiser's cost to participate. Also lowers everyone else's — it is not obviously proprietary.
🚩 The deep-learning bidding advantage is the kind of edge that compresses. APP spends ~5.1% of revenue on R&D ($94.1M on $1,842M in Q1'26 — itself up 67% YoY). Meta and Google spend 15-30%. AppLovin is defending a machine-learning moat on a fraction of the research budget of the firms it just chose to compete with. Management's own explanation for the Q2 miss — model-improvement cadence — is evidence this constraint binds.
Regulatory/platform change — the live one. See below.

Evergreen assessment — No

This fails the 🏛 Evergreen test ("would I hold through a 40% drawdown without re-underwriting the thesis?"). A 40% drawdown just happened and this document is a re-underwriting. The moat is real, high-returning, and conditional — on data access, on platform policy, and on sustained model-improvement velocity. That is a ⚡/🔧 profile, not a 🏛 profile.

Moat rating: 6.5/10 — very strong in gaming, unproven in e-commerce, and resting on an input the company does not control.

§3 · Valuation

Applied conditionally per §3. No dividend → DDM and DYT are N/A, stated rather than faked.

Multiples at $351 (and at the $417.80 pre-print close)

Metric	At $351	At $417.80
P/E on FY2026E ($16.56)	21.2x	25.2x
P/E on FY2027E ($21.59)	16.3x	19.4x
P/E trailing ($11.50)	30.5x	36.3x
P/FCF (TTM $12.99/sh)	27.0x	32.2x
P/FCF (SBC-adj $12.01/sh)	29.2x	34.8x
EV / TTM revenue	17.5x	20.8x
EV / FY26E revenue (~$8.09B)	14.8x	17.6x
FY26E FCF yield	~4.4%	~3.7%

⚠️ Use the FY2026 line. The widely-quoted "19x forward" is FY2027 — see trap #1.

Graham's Intrinsic Value — VOID

√(22.5 × $11.50 × $7.03) = $42.64. Discard. BVPS of $7.03 exists only because AppLovin has retired ~$2.2B of stock per year; tangible book is barely positive. Graham weights book value, and this company's book value is a residual of good capital allocation, not a measure of assets. Reporting "$42.64 vs $351" would be reporting the buyback as a defect.

Bogle Expected Return

dividend yield (0%) + earnings growth ± P/E change

Scenario	Earnings growth	Terminal P/E (from 21.2x)	Bogle ER
Bull — growth holds ~35%	+35%/yr	22x (flat)	~+36%/yr
Base — decelerates to ~22%	+22%/yr	18x over 5yr (−3.2%/yr)	~+19%/yr
Bear — decelerates to ~12%	+12%/yr	14x over 5yr (−7.9%/yr)	~+4%/yr

The spread is entirely the growth term. No valuation judgement resolves this; only the next few prints do.

Reverse-DCF — the most useful model here

At $351: EV ≈ $119.6B, FY26E FCF ≈ $5.26B (65% FCF margin on ~$8.09B). Discounting at r = 10% with a 3% terminal, solving for the 5-year FCF growth rate that justifies today's price:

$351 embeds roughly 13-14% annual FCF growth for five years, then 3% in perpetuity.

The company is guiding +46-48% revenue growth for the very next quarter. The market is pricing a ~70% deceleration from the guided rate, immediately and permanently.

This is the bull case stated precisely, and it is a strong one — you do not need AppLovin to keep growing 50%; you need it to beat 14%. It is also exactly why the stock is not a screaming buy: a business whose quarterly result depends on unobservable model-improvement cadence, with an open SEC investigation into its data supply, is one where a collapse below 14% is a live scenario rather than a tail.

Fair value range

Case	Basis	Value
Bear	Growth → ~15%, multiple → 15x FY26E $16.56	~$250
Severe bear	SEC action curtails data / platform restriction, 12x	~$200
Base	22-25x FY26E $16.56	$364-414
Base (FY27 cross-check)	18-20x FY27E $21.59	$389-432
Bull	Growth holds ~40%, 25x FY27E	~$540

Fair value: $300 – $450 (central ~$375)

Spot ~$351 sits inside the range, modestly below centre.

Entry $290-330 · Strong buy <$270 · Trim 25x fwd (≈$540 — a FY2027 multiple, since the site's fwd field is FY2027 on this ticker; see trap #1)

Conservative bias applied per §3: the entry zone deliberately sits below spot. This is not a name where paying fair value is adequate compensation — the SEC binary requires a discount, in the same way the Taiwan risk on TSM argued for a smaller position rather than a higher price.

§4 · Sentiment & Intelligence

🚩 The SEC investigation — the single most important open item

February 2026: the SEC confirmed an active investigation into AppLovin's data-collection practices, reportedly triggered by a whistleblower complaint. Still active as of May 2026, unresolved today.
Preceded by four separate short-seller reports — Fuzzy Panda, Culper Research, Muddy Waters, and CapitalWatch. Allegations: improper user-data collection, breach of platform partners' terms of service, forced/undisclosed app installs inflating engagement metrics, and tracking of minors without consent.
The February 2025 reports took ~$13.7B of shareholder value in days; multiple securities class actions followed.

Why this is a moat question and not a headline question. The entire valuation rests on a data advantage assembled through Adjust and MoPub. The allegations are that some of that data is obtained in ways regulators and platform owners may not permit. An adverse finding does not merely produce a fine — it invites the platform-level response (Apple/Google policy) that is the fastest existing route to damaging Axon. The company has no vote in that outcome.

⚖️ Stated fairly: short-seller reports are adversarial documents from parties positioned to profit, and an SEC investigation is not a finding of wrongdoing. AppLovin has continued to grow ~50%+ throughout, which is not what a business does when its core practice has been curtailed. But "unresolved for 18 months" is not the same as "resolved favourably," and this file will not pretend otherwise.

🚩 Insider activity — verified row-by-row, and it is one-directional

Per pitfall-yahoo-insider-purchases-counts-rsu-grants, every row was checked; all zero-price rows are Stock Award (Grant) or Stock Gift and are excluded — they are not purchases.

Insider	Role	Sales (Mar-Jul 2026)	Price range
Vivas, Eduardo	Director	$166.5M	$446.88-520.30
Foroughi, Arash Adam	CEO	$78.6M	$449.38-516.51
Shikin, Vasily	CTO	$39.0M	$477.71-493.66
Valenzuela, Victoria	Officer	$11.3M	$555.63-586.49
Stumpf, Matthew	CFO	$5.4M	$600.00
Webb, Maynard Jr	Director	$3.4M	$516.25-593.00
Dorosh, Dmitriy	Officer	$1.4M	$452.66-453.60
TOTAL		≈ $305.7M
Open-market purchases		$0.00 — ZERO

~$306M of selling across the CEO, CFO, CTO and two directors in roughly four months, with not a single open-market purchase, concentrated at $450-600 against a stock now at $351. The CFO's sale printed at exactly $600.00; officer and director sales clustered at $555-593.

⚖️ The honest caveat: 10b5-1 plan status is not verifiable from this data, and executives of a stock that ran to $745 have ordinary diversification reasons to sell. This is not evidence of foreknowledge. But §6.3 treats insider buying as a golden flag and its absence as informative — and there has been no buying at any price, including none disclosed after the stock fell by half from the high.

Analyst positioning — unanimity as a risk, not a confirmation

	Count
Strong buy	7
Buy	22
Hold	3
Sell / Strong sell	0

32 ratings, zero sells, mean target $656.20 — a figure set before tonight's print and now ~87% above spot. UBS raised its target to $798 on August 3rd, two days before the quarter. Jefferies had already cut $860 → $700 earlier in the year.

This is the MU setup from 8/04 repeating in a different sector: a consensus with no dissenting voice is a positioning fact, not an analytical one. Expect a wave of target cuts, and do not use the $656 mean as an anchor — it has not yet absorbed the guide.

Short interest is modest: 4.3% of float, 2.13 days to cover. Whatever tonight was, it was not a short squeeze unwinding, and the shorts are not crowded here.

§5 · Synthesis — Where the Agents Disagree

Fundamentals vs Moat. Fundamentals returns a near-perfect scorecard: 70% ROIC, expanding margins, real FCF, minimal SBC, self-funded buybacks, falling share count, unlevered balance sheet. Moat returns 6.5/10 with an explicit warning that the advantage is conditional on third-party data access and on out-researching Google and Meta at ~5% of revenue.

Resolution — Moat is weighted higher, and §5 says why. For high-growth technology, "Moat + Sentiment outweigh Graham's static valuation." The Fundamentals scorecard is a backward-looking measurement of an advantage that was working; it cannot detect the advantage's dependency. The Q2 miss is the first observation where the moat's fragility showed up in the numbers before the scorecard deteriorated — margins and ROIC still look excellent while sequential growth halved. That ordering matters: on this business model, the scorecard will be the last thing to break, not the first.

Valuation vs Sentiment. Valuation says a 13-14% reverse-DCF hurdle against 47% guided growth is a wide margin of safety. Sentiment says an unresolved SEC investigation into the data supply, $306M of one-way insider selling, and 32 ratings with no sells are not conditions under which a low hurdle should be trusted.

Resolution — split the difference honestly, which is what the entry zone does. The valuation case is real and is why this is a 6.0 and not a 4.0. The sentiment case is why the entry sits at $290-330 rather than at spot: the discount has to pay for the binary, not just for the growth deceleration.

🚩 Risk register

SEC investigation into data-collection practices — unresolved, ~18 months old. Goes to the source of the moat. Worst realistic case is not the fine; it is a platform-policy response.
Growth deceleration is confirmed, not feared: +59% → +53% → +46-48% guided; sequential +11.1% → +4.4%.
Margin step-down guided as permanent — 85% → 83%, with Q3 guided at ~83%.
Results depend on unobservable model-improvement cadence, by management's own explanation.
R&D at ~5.1% of revenue while entering Google/Meta/Amazon's market.
~$306M insider selling, zero buying, clustered near the highs.
Beta 2.53 — the most volatile position in either portfolio.
Platform dependency — Apple/Google control the signal; ATT reshaped this industry once.
Consensus has no dissent — 0 sells out of 32.

🟢 Golden flags

ROIC 14.8% → 42.0% → 70.3% with invested capital up only 8%.
Gross margin expanding four straight years — no price erosion.
SBC 7.6% of FCF — near-real cash earnings.
Buybacks ~50-56% of FCF, self-funded, zero debt issued, share count −2.7%/yr.
Net debt/EBITDA 0.13x, coverage 28x.
Q2 revenue landed inside the company's own guidance range.
Web advertising at a ~$1B run-rate within months — realised revenue, not TAM.
Unity — the best-positioned rival — tried and lost.
Earnings are operating earnings; no unrealized-mark inflation.

Portfolio Fit & Sizing — The Real Question

A portfolio-specific passage was removed from the public build.

The watchlist instruction was never a decision. "+1010%. Let it run. Trim $650+" is written in the register of a meaningful position. At ~$435 — against, for scale, 76 QQQ shares (~$54,500) in the Beta account — APP is a rounding error that has been generating the feeling of a decision without the substance of one. The correct question was never "when do I trim"; it is "does this deserve to be a position at all."

Correlation check against the Concentration Map: - TTD ($18.96, −79% off high, do-not-add) — direct adtech peer. Owning both is one bet. - META ($588.77, in the add zone) — APP's competitor in e-commerce advertising. Partially offsetting, and worth noting that META at 16.6x is the cheaper way to own digital advertising. - U (Unity, $35.47) — the rival AppLovin beat. Held. Awkward, and arguably APP is the better expression of the same industry view. - GOOGL, AMZN — the firms AppLovin has chosen to compete with.

Adding APP concentrates advertising exposure, which is already represented by TTD (broken), META (strong), U (weak) and indirectly GOOGL/AMZN. It does not fill a gap.

Recommendation on sizing: if the user wants APP to be a real position, the honest sequence is (1) wait for the Q3 print in early November to see whether sequential growth returns to the guided +7-8%, and (2) require the entry zone $290-330 to pay for the SEC binary. Starting a position into an after-hours gap on guidance-reset night is the opposite of the framework's conservative bias. A [6.0] with an unresolved regulatory investigation warrants ≤2% of the portfolio even when it is in zone.

Break Triggers & Upgrade Path

Downgrade to [4.5] — thesis broken: - Sequential revenue growth below +5% again in Q3 (guide is +7-8%) — would confirm Q2 was trend, not cadence - SEC enforcement action, or any Apple/Google policy change restricting AppLovin's data access - Adjusted EBITDA margin below 82% for two consecutive quarters - Web/e-commerce revenue run-rate flat or declining — the entire growth thesis - Buyback paused, or funded with new debt - Gross margin below 85%

Upgrade to [7.0]: - SEC matter closed without enforcement — the single largest re-rating event available - Q3 delivers ≥+8% sequential with EBITDA margin back to ≥84% - Any open-market insider purchase at these prices — there has been none at any price - Web advertising run-rate above $2B with disclosed advertiser counts

⏳ Next dated events: Q3 FY26 print, early November 2026 — watch sequential revenue growth and EBITDA margin, not the YoY headline. The YoY number will look strong regardless because the comparison base is small; the sequential line is where the truth is.

Sources

Quantitative data: Yahoo Finance via .mcp/fin.py and yfinance, 2026-08-05. Statements are GAAP as reported; adjusted figures are labelled as such.