AppLovin
Rating
Strong Buy
High Conviction — Core Position
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
AppLovin's AXON AI engine is trained on in-app behavioral data from thousands of mobile apps mediated through MAX — a proprietary dataset that grew more valuable when Apple's ATT killed third-party tracking and that no competitor can replicate without first building AppLovin's publisher distribution. Q2 2026's 10-Q still shows the flywheel in the numbers: net revenue per install +58% YoY against install volume only −2%, and the SEC's voluntary inquiry closed with no recommended action.
AppLovin's moat rests on an AI data flywheel built on in-app behavioral data and network effects, not traditional enterprise lock-in:
- MAX Mediation → Proprietary In-App Data: AppLovin's MAX ad mediation platform sits between app publishers and ad networks, giving AppLovin first-party visibility into in-app user behavior across thousands of apps. When Apple's ATT framework killed third-party identifier tracking in 2021, AppLovin's behavioral pattern modeling (not tied to personal identifiers) became significantly more valuable than competitor approaches — creating a privacy-era data moat that has strengthened with each passing year as signal loss compounds for rivals. In Q2 2026 MAX publisher earnings still grew double digits sequentially and AppLovin's share of publisher waterfalls held steady even while model uplifts landed late.
- AXON AI Engine — Self-Reinforcing Flywheel: More ad spend on AXON → more conversion signal → better AXON predictions → better ROAS for advertisers → more ad spend. Since AXON 2.0 launched in Q2 2023, gaming advertiser spend on AppLovin has quadrupled to an estimated $10B annual run rate. In late June 2026 the engine opened to any advertiser as a self-serve product under the AppLovin Ads Manager brand, and the generative creative layer — interactive end-cards shipping, high-quality 30–60s video still the bottleneck — extends the flywheel into ad production. Q2 showed the flywheel's sensitivity to model cadence: lighter lifts in the quarter produced a soft print versus guide, and a material upgrade that landed just after quarter-end is what management says is already powering early Q3.
- Platform Bundle (MAX + AXON + Adjust): The combination of MAX mediation (publisher monetization), AXON demand-side (advertiser ROI), and Adjust (mobile attribution and measurement) into a single platform creates multi-sided lock-in. Publishers depend on MAX for revenue maximization, advertisers on AXON for install efficiency, and measurement partners integrate with Adjust — making the entire ecosystem self-reinforcing and costly to disassemble. Partnerships with analytics firms such as Triple Whale are now the preferred path to mid-market e-commerce advertisers rather than a long-tail SMB land grab.
Ten Moats Verdict
AppLovin is an AI-native business — AXON is its product, not a feature — making it a direct beneficiary of AI capability improvements and, as Q2 showed, of their timing: when model lifts land late, the guide is missed even with healthy demand. The June 2026 self-serve opening and the Q2 consumer spend record (28% above the Q4 peak) are the first real evidence that the beyond-gaming expansion is working; what is not yet demonstrated is that consumer is large enough to smooth gaming-model cadence. The SEC inquiry's close removes a regulatory overhang. The key AI risk remains that the Big Three (Google, Meta, Amazon) use their larger data assets to close the ROAS gap with AXON; AppLovin's proprietary in-app behavioral data is currently the moat, but that moat is narrower than the structural lock-ins of Cloudflare, Axon, or Microsoft. Overall, AppLovin's AI-era positioning is strong within ad-tech but more competitively exposed than platform businesses with regulatory or physical switching costs.
Performance marketers optimize AXON campaign bidding and creative parameters over months — institutional knowledge of what creative styles, bid strategies, and audience cohorts work on AXON is non-transferable to other platforms.
AXON is largely a black-box AI — advertisers don't configure deep business logic into the platform. Campaign structures are relatively simple to port. AI tools are accelerating migration testing, weakening this moat.
MAX mediation gives AppLovin visibility into in-app behavioral patterns across thousands of publisher apps — a first-party data stream that is less replicable than raw impression data but not entirely proprietary.
The ML engineers who built and iterate on AXON represent a scarce intersection of ad-tech domain knowledge and production-scale AI. Recruiting away from AppLovin is difficult given the equity upside and the unique data environment.
MAX mediation (publisher yield) + AXON demand (advertiser ROI) + Adjust attribution creates a three-sided bundle that cannot be easily unbundled without losing performance across all three surfaces simultaneously.
In-app behavioral data from MAX — how users interact with apps, session patterns, purchase propensity signals — is genuinely proprietary. When Apple killed IDFA, this behavioral signal became more differentiated, not less. 536 patents protect key algorithms.
Privacy regulations (Apple ATT, GDPR) accidentally created a moat for AppLovin's privacy-first behavioral approach — but this is not traditional regulatory lock-in. The SEC's voluntary inquiry into data-collection practices closed in Q2 2026 with no recommended action, removing the overhang; the accidental privacy-era advantage remains, without a formal lock-in.
Classic two-sided marketplace network effects: more publishers on MAX → more inventory → better advertiser outcomes → more advertiser spend → more publisher revenue. The AXON data flywheel adds a third dimension: more conversion signals → better predictions → better ROAS → more ad spend → more signals.
Every in-app ad auction, install event, and attribution event flows through AppLovin's infrastructure. Publishers cannot monetize without the platform; advertisers cannot track without Adjust. Deeply embedded in the mobile transaction layer.
Adjust serves as a system of record for mobile attribution events, but this is a narrower system of record than financial or legal data. AppLovin's ad server records are not authoritative for external reporting purposes.
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
AppLovin's AXON AI engine is trained on in-app behavioral data from thousands of mobile apps mediated through MAX — a proprietary dataset that grew more valuable when Apple's ATT killed third-party tracking and that no competitor can replicate without first building AppLovin's publisher distribution. Q2 2026's 10-Q still shows the flywheel in the numbers: net revenue per install +58% YoY against install volume only −2%, and the SEC's voluntary inquiry closed with no recommended action.
Growth Score
Q2 2026 delivered $1.92B revenue (+53% YoY), $1.61B adj. EBITDA (84% margin) and $863M FCF — essentially at the low end of the $1.915–1.945B guide after a quarter where gaming model uplifts arrived late. Net revenue per install still rose 58% YoY while install volume fell only 2% (from −18% in Q1), so the price term continues to carry growth. The consumer/e-commerce vertical hit a record: advertiser spend finished 28% above the seasonally strong Q4 2025 peak in what is normally retail's weakest quarter. Q3 is guided to $2.055–2.085B (+46–48%) at 83% adj. EBITDA margin, embedding the post-quarter model uplift and continued consumer ramp. Management reiterated a long-term framework of roughly 30% annual revenue compounding with low-80s EBITDA margins; the SEC inquiry closed with no recommended action.
Valuation Score
APP trades at ~$336 after a ~20% single-day drop on August 6 following a Q2 print that grew 53% but landed near the low end of guidance on late model uplifts — leaving it ~55% below its December 2025 high near $746 and ~16% above the bear ($290). At ~21× 2026 consensus EPS near $16 and ~16× 2027's ~$21, the price sits ~35% below the base case ($520). The de-rating is pricing model-cadence risk and an unproven self-serve ramp; it is not pricing a broken flywheel — net revenue per install still rose 58% while installs fell only 2%, consumer spend set a record, and the SEC inquiry closed with no action.
The AXON Data Flywheel
AppLovin's moat rests on an AI data flywheel built on in-app behavioral data and network effects, not traditional enterprise lock-in:
- MAX Mediation → Proprietary In-App Data: AppLovin's MAX ad mediation platform sits between app publishers and ad networks, giving AppLovin first-party visibility into in-app user behavior across thousands of apps. When Apple's ATT framework killed third-party identifier tracking in 2021, AppLovin's behavioral pattern modeling (not tied to personal identifiers) became significantly more valuable than competitor approaches — creating a privacy-era data moat that has strengthened with each passing year as signal loss compounds for rivals. In Q2 2026 MAX publisher earnings still grew double digits sequentially and AppLovin's share of publisher waterfalls held steady even while model uplifts landed late.
- AXON AI Engine — Self-Reinforcing Flywheel: More ad spend on AXON → more conversion signal → better AXON predictions → better ROAS for advertisers → more ad spend. Since AXON 2.0 launched in Q2 2023, gaming advertiser spend on AppLovin has quadrupled to an estimated $10B annual run rate. In late June 2026 the engine opened to any advertiser as a self-serve product under the AppLovin Ads Manager brand, and the generative creative layer — interactive end-cards shipping, high-quality 30–60s video still the bottleneck — extends the flywheel into ad production. Q2 showed the flywheel's sensitivity to model cadence: lighter lifts in the quarter produced a soft print versus guide, and a material upgrade that landed just after quarter-end is what management says is already powering early Q3.
- Platform Bundle (MAX + AXON + Adjust): The combination of MAX mediation (publisher monetization), AXON demand-side (advertiser ROI), and Adjust (mobile attribution and measurement) into a single platform creates multi-sided lock-in. Publishers depend on MAX for revenue maximization, advertisers on AXON for install efficiency, and measurement partners integrate with Adjust — making the entire ecosystem self-reinforcing and costly to disassemble. Partnerships with analytics firms such as Triple Whale are now the preferred path to mid-market e-commerce advertisers rather than a long-tail SMB land grab.
Ten Moats Verdict
AppLovin is an AI-native business — AXON is its product, not a feature — making it a direct beneficiary of AI capability improvements and, as Q2 showed, of their timing: when model lifts land late, the guide is missed even with healthy demand. The June 2026 self-serve opening and the Q2 consumer spend record (28% above the Q4 peak) are the first real evidence that the beyond-gaming expansion is working; what is not yet demonstrated is that consumer is large enough to smooth gaming-model cadence. The SEC inquiry's close removes a regulatory overhang. The key AI risk remains that the Big Three (Google, Meta, Amazon) use their larger data assets to close the ROAS gap with AXON; AppLovin's proprietary in-app behavioral data is currently the moat, but that moat is narrower than the structural lock-ins of Cloudflare, Axon, or Microsoft. Overall, AppLovin's AI-era positioning is strong within ad-tech but more competitively exposed than platform businesses with regulatory or physical switching costs.
Performance marketers optimize AXON campaign bidding and creative parameters over months — institutional knowledge of what creative styles, bid strategies, and audience cohorts work on AXON is non-transferable to other platforms.
AXON is largely a black-box AI — advertisers don't configure deep business logic into the platform. Campaign structures are relatively simple to port. AI tools are accelerating migration testing, weakening this moat.
MAX mediation gives AppLovin visibility into in-app behavioral patterns across thousands of publisher apps — a first-party data stream that is less replicable than raw impression data but not entirely proprietary.
The ML engineers who built and iterate on AXON represent a scarce intersection of ad-tech domain knowledge and production-scale AI. Recruiting away from AppLovin is difficult given the equity upside and the unique data environment.
MAX mediation (publisher yield) + AXON demand (advertiser ROI) + Adjust attribution creates a three-sided bundle that cannot be easily unbundled without losing performance across all three surfaces simultaneously.
In-app behavioral data from MAX — how users interact with apps, session patterns, purchase propensity signals — is genuinely proprietary. When Apple killed IDFA, this behavioral signal became more differentiated, not less. 536 patents protect key algorithms.
Privacy regulations (Apple ATT, GDPR) accidentally created a moat for AppLovin's privacy-first behavioral approach — but this is not traditional regulatory lock-in. The SEC's voluntary inquiry into data-collection practices closed in Q2 2026 with no recommended action, removing the overhang; the accidental privacy-era advantage remains, without a formal lock-in.
Classic two-sided marketplace network effects: more publishers on MAX → more inventory → better advertiser outcomes → more advertiser spend → more publisher revenue. The AXON data flywheel adds a third dimension: more conversion signals → better predictions → better ROAS → more ad spend → more signals.
Every in-app ad auction, install event, and attribution event flows through AppLovin's infrastructure. Publishers cannot monetize without the platform; advertisers cannot track without Adjust. Deeply embedded in the mobile transaction layer.
Adjust serves as a system of record for mobile attribution events, but this is a narrower system of record than financial or legal data. AppLovin's ad server records are not authoritative for external reporting purposes.
Growth Analysis
Growth Drivers
Key Risk
If gaming model uplifts stay lumpy enough that another quarter misses the company's own guide while the consumer vertical is still too small to smooth it — or if self-serve mid-market cohorts fail to convert durable spend once the post-launch novelty fades — FY2027 revenue lands closer to +20% than the ~25–30% consensus and management's ~30% frame carry, visible by the Q4 2026 print in February 2027
Score Derivation
90.1 base + 2.7 trajectory − 5 risk = 88
Base ~90 on a 30.5% midpoint — the 3–5yr blended rate, decayed from the +53% Q2 print and +46–48% Q3 guide toward management's ~30% long-term frame and the ~25–30% 2027 consensus implies, in line with how every equity in coverage decays a current rate toward terminal. + 2.7 trajectory (two accelerating drivers of three; e-commerce restored to accelerating on the Q2 spend record) + 0 margin (stable in the low-80s; Q3 guided 83%) − 5 key risk (moderate) = 88. The Q2 guide miss is charged as model-cadence lumpiness in the key risk rather than as a cagr cut — the measured series still compounds in the high-40s into Q3 and management's long-term 30% frame is unchanged.
Research Covering This Name
Price Scenarios (12–24 Months)
Valuation Multiples
| Trailing P/E (GAAP) | ~25× |
| Forward P/E (2026E) | ~21× |
| Forward P/E (2027E) | ~16× |
| PEG Ratio | ~0.7× |
| Price / Sales (2026E) | ~14× |
| Price / FCF (2026E) | ~22× |
The August 6 de-rating has taken APP from ~25× forward earnings in July to ~21× on 2026 and ~16× on 2027, against a business that grew 53% last quarter at an 84% EBITDA margin and guided Q3 to +46–48%. The PEG of ~0.7× is deep GARP territory on consensus arithmetic — but the Q2 guide miss showed that model-cadence timing can gap a quarter even when demand is healthy, and consumer is still not large enough to smooth it. The multiple is no longer demanding; the question is whether the earnings line compounds at the ~30% frame management reiterated, or whether another soft model quarter forces that frame down.
Approximate figures as of August 6, 2026.
Where We Are vs Targets
Loading live price…
Model uplifts stay lumpy through H2, the self-serve cohort fails to compound durable spend, and gaming decelerates into tougher comps — 2027 growth lands near 15–20% and the multiple compresses to the mid-teens on forward earnings.
- Another quarter where material model lifts land after the period — as in Q2 — causes a second guide miss while consumer is still too small to offset, and the self-serve mid-market cohort churns after first campaigns, taking FY2027 revenue toward ~$9.5B (+15%) against ~$10.5B+ consensus
- Google's Privacy Sandbox and Meta's Advantage+ close the behavioral targeting gap through 2027, cutting AXON's ROAS premium below 15% and returning e-commerce budgets to Meta and Google
- Generative video creative remains unsolved into 2027, so e-commerce advertisers without video assets never reach the spend ceilings management frames — the consumer vertical stalls as a niche rather than becoming the second engine
- FCF conversion settles below the guided ~75% of adj. EBITDA as creative-generation and model compute scale, so 2027 FCF lands near $5B rather than the $7B the bull case needs
AppLovin delivers roughly $8.2B 2026 revenue at low-80s EBITDA margins, the post-quarter model uplift and consumer ramp support the Q3 guide, and 2027 earnings arrive near the ~$21 consensus — a ~25× multiple on that, in line with where the stock traded through the mid-2026 de-rating before the August print.
- FY2026 revenue lands near $8.0–8.3B with H2 carrying the post-Q2 model uplift and the first full quarters of public self-serve; FY2027 reaches ~$10.5B (+25–30%), consistent with management's ~30% long-term frame
- The consumer vertical keeps growing past seasonal peaks and begins to smooth gaming-model timing by H1 2027, validating the beyond-gaming expansion without requiring the bull case's 100,000-customer frame
- Adj. EBITDA margins hold in the low-80s and FCF conversion settles at the guided ~75%, giving ~$5B+ of 2026 FCF; buybacks continue against the ~$1.8B remaining authorisation after Q2's $551M
- The SEC inquiry's close with no recommended action removes the regulatory discount that sat on the multiple since late 2025
Self-serve becomes the default ad-buying surface for mid-market e-commerce, video auto-generation unlocks one-click campaigns, and 2027 earnings clear consensus by a wide margin — supporting ~30× on 2027 EPS well above ~$21.
- Analytics-partner lead flow and mid-market mix push average first-year spend above the previously cited ~$70K bogey — incremental auction demand phasing in through 2028 that sits outside consensus
- Generative video removes the production bottleneck that caps e-commerce spend: 30–60s video auto-generation graduates from bottleneck to default, lifting both the share of qualified leads that reach live spend and the spend per advertiser
- The ROAS advantage on contextual, non-identifier signal holds against Privacy Sandbox and Advantage+, forcing e-commerce brands to move a structural share of Meta and Google budgets rather than testing at the margin
- FY2027 revenue reaches $12B+ at low-80s EBITDA margins, and even at the guided ~75% conversion FCF clears $7B — a ~$113B cap today paying ~16× that, with buybacks compounding the per-share effect