Duolingo
Freemium: a free ad-supported tier funnels to Super Duolingo and Duolingo Max (AI) auto-renewing subscriptions (~80%+ of revenue); plus in-app advertising and the paid Duolingo English Test (DET). App-store-distributed, asset-light software.
Earnings, margins, COGS & capex
A rare profitable-and-cash-generative consumer app: FY2025 revenue crossed $1.0B (+38.7%), gross margin ~72%, ~35% FCF margin, >$1B net cash, zero debt. The problem is not the P&L today but the trajectory — total bookings (the forward-revenue lead indicator) decelerated from ~40%+ toward a guided 10-12% for FY2026 (Q1 FY2026 was +14%), and MAU growth collapsed from +33% to ~+6% YoY, so the market re-rated from a hyper-growth to a GARP multiple. Reported net income is also flattered by a one-time tax benefit (see margins).
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~28¢ is cost of goods and ~59¢ operating expense, leaving ~13¢ of operating profit (~40¢ net).
Revenue trend
Margins
Q1 FY2026 ~73.0% (per-unit AI cost optimization helped); guided down toward ~69% by Q4 2026 as more AI features roll out
15.5% in Q4 2025
FY2026 guided ~25.7% ($310M)
Q1 2026 50.6% ($147.8M, seasonally high on annual-plan cash timing); >$350M FCF implied FY2026
Flattered by a one-time $256.7M non-cash income-tax benefit from releasing a deferred-tax valuation allowance; underlying net margin is roughly mid-teens
COGS structure
Cost of revenue is dominated by app-store/payment-processing fees (Apple/Google 15-30% take), cloud hosting (AWS), and — increasingly — third-party LLM inference for Duolingo Max and AI-generated course content. AI features are the named driver of guided gross-margin compression (~73% Q1'26 -> ~69% by Q4'26), framed by management as deliberate product investment rather than a cost blowout.
Capex
Minimal — asset-light software; capex is a low-single-digit % of revenue, so FCF tracks operating cash flow closely.
Latest earnings
Beat consensus — revenue $292.0M (+27%), total bookings $308.5M (+14%), net income $43.5M, adj. EBITDA ~$83M (~28% margin), FCF $147.8M (50.6% margin), DAU +21% — yet the stock stayed depressed; the de-rating is about the decelerating forward guide, not the print
FY2026: revenue $1.205B (+~16%, range 15-18%), total bookings $1.28B (+~10-12%), adj. EBITDA $310M (~25.7% margin). Q2 2026 guide: revenue $295.5M, bookings $283.5M (~+6% on a tough comp), adj. EBITDA $71.0M (24.0%); a second-half bookings re-acceleration is embedded in the full-year guide. Medium-term target: 100M DAU by 2028.
- DAU (Q1 2026)
- 56.5M, +21% YoY
- MAU (Q1 2026)
- 137.8M, ~+6% YoY (decelerated from +33% a year earlier)
- Paid subscribers (Q1 2026)
- 12.5M, +21% YoY
- Paid penetration
- ~9% of MAU
- Adj. EBITDA (Q1 2026)
- ~$83M (~28% of revenue)
- Cash / debt
- >$1B / $0
- Diluted share count
- ~49.7M; guided up ~3.5-4% in 2026 on equity grants (net dilution, not net buyback)
Growth drivers
- Paid-subscriber growth — 12.5M paid subs (Q1 2026, +21% YoY) on ~9% paid penetration of MAU; Duolingo Max (AI) premium tier is the key upsell
- Daily-active-user growth (56.5M DAU Q1 2026, +21% YoY) via the streak/gamification habit loop feeding the freemium funnel; DAU/MAU deepening even as MAU growth slows
- AI-scaled content — ~20,500 course units published in Q1 2026, widening supply into non-English and adjacent subjects at low marginal cost
- Duolingo English Test (DET) institutional adoption as a cheaper TOEFL/IELTS alternative
- New verticals — Math, Music, Chess — extending the gamified-learning engine beyond languages
- Advertising revenue on the free tier and price/packaging (family, annual, longer-trial) monetization experiments
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-27. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
A de-rated, founder-led category leader that is already highly profitable and cash-generative — at ~16x FCF with 56M engaged daily users, >$1B net cash, and real AI/premium and DET optionality, the disruption fear may be over-discounted.
- Still growing: DAU +21%, paid subs +21%, bookings +14% (Q1'26) with revenue +27% — deceleration, not decline; a second-half bookings re-acceleration is embedded in the FY guide
- ~28% adj. EBITDA margin, ~35% FCF margin, >$1B cash, zero debt — funds experiments and opportunistic repurchases from a position of strength
- Valuation reset to ~15-16x FCF and ~5.4x sales — a reasonable-to-cheap multiple for a #1 consumer subscription franchise, versus a peak of many multiples higher
- AI is a tool it exploits, not only a threat: ~20,500 course units in one quarter, an AI video-call tutor deepening engagement and differentiating the paid tier
- DET and new verticals (Math/Music/Chess) are under-monetized optionality outside the core-language debate
- Habit/gamification moat (streaks) and brand keep top-of-funnel acquisition cheap; a general-purpose chatbot has no comparable daily-retention loop
The core product — gamified vocabulary drilling — is precisely what free general-purpose LLMs now do, growth has roughly halved with MAU up only ~6%, gross margins are set to compress under AI costs, and the headline earnings multiple is flattered by a one-time tax benefit, so 'cheap' may be a value trap on a structurally lower growth curve.
- Existential substitution risk: ChatGPT/GPT-5, Gemini and Claude deliver conversational practice and translation for free, eroding the reason to pay
- Growth already broke: bookings from ~40%+ to guided ~10-12%, and MAU growth from +33% to ~+6% YoY — the freemium top-of-funnel is stalling, which precedes paid-sub deceleration
- Gross margin guided down ~73% -> ~69% because its own AI features are expensive — the AI response costs margin without a proven pricing offset yet
- FY2025 net income ($414M) includes a $256.7M non-cash tax benefit; on underlying earnings the stock is not as cheap as the ~14x headline P/E suggests
- App-store tax (15-30%) permanently caps unit economics and leaves Apple/Google holding pricing power
- 'AI-first' memo and reliance on OpenAI/Microsoft models mean its supplier is also its disruptor — a strategically fragile position
- Diluted share count is guided UP ~3.5-4% in 2026 on SBC, so per-share support from buybacks is limited
- At a lower structural growth rate, even a de-rated multiple can compress further if the guided second-half bookings re-acceleration misses
What it is worth
Peer-relative multiple + reverse-DCF sanity check on FCF. At ~$5.65B market cap: ~5.4x FY2025 sales, ~15-16x FY2025 FCF ($360.4M) and ~16x the >$350M FY2026 FCF implied. Headline P/E (~14x on $414M FY2025 net income) is misleading — that net income includes a $256.7M one-time non-cash tax benefit, so the true earnings multiple is materially higher.
~$70-95/sh (~$3.3-4.5B)
substitution bites, MAU growth stalls, paid-sub growth decelerates and gross margin slips below guide; multiple compresses toward ~10-12x a flat/declining FCF as the market prices structural erosion.
~$125-150/sh (~$5.7-7B)
~10-12% bookings growth, modest MAU growth, ~69-70% gross margin, >$350M FCF; a de-rated but stable GARP name compounding at ~16-18x FCF, lightly offset by SBC dilution.
~$180-220/sh (~$8.5-10.5B)
bookings stabilize at ~12-15%, MAU growth re-accelerates, AI/Max monetization proves additive and margins hold near guide; re-rate toward ~22-25x a growing FCF stream.
The reverse-DCF read: at ~16x FCF the market is pricing low-single-digit to high-single-digit long-run FCF growth with a real terminal-disruption discount — i.e. it no longer credits Duolingo as a durable compounder. Whether that is too harsh depends on whether the guided second-half bookings re-acceleration and a MAU-growth stabilization arrive.
SWOT
Strengths
- Category-leading brand and scale — world's most-downloaded education app, 56.5M DAU, powerful streak/gamification habit engine
- Genuinely profitable and cash-generative consumer app — ~72% gross margin, ~35% FCF margin, >$1B net cash, zero debt
- Freemium funnel with proven upsell into subscriptions + an AI premium tier (Duolingo Max)
- Duolingo English Test — a structurally advantaged, high-margin institutional product with two-sided acceptance network effects
Weaknesses
- Growth roughly halved — total bookings decelerating from ~40%+ toward a guided ~10-12% (Q1'26 +14%); MAU growth collapsed from +33% to ~+6% YoY
- Core value prop (vocabulary drills, translation) is the exact task general-purpose LLMs now do for free
- Gross margin set to compress (~73% -> ~69% guided) as its own AI features raise inference costs
- Reported FY2025 net income inflated by a one-time $256.7M tax benefit — headline P/E understates the true earnings multiple
- Heavy app-store dependence (Apple/Google 15-30% take) caps unit economics and adds platform-rule risk
- SBC-driven dilution (~3.5-4% diluted-share growth guided for 2026) partly offsets the cash-return story
Opportunities
- Duolingo Max / AI video-call tutor as a differentiated premium tier (management cites users speaking materially more words YoY)
- AI-scaled content production to flood into more languages and new subjects at low marginal cost
- New verticals — Math, Music, Chess — and expansion of DET into more institutions
- Monetization experiments (longer trials, advertising, family/annual plans) to lift ARPU without adding user friction
- Re-rating optionality: a de-rated but cash-rich leader can compound if bookings and MAU growth stabilize
Threats
- Commoditization of language learning by free/cheap LLMs (ChatGPT/GPT-5, Google Gemini/Translate, Claude) — OpenAI's Aug-2025 GPT-5 app-building demo helped drive a ~38% decline over the three months from the May 2025 peak
- The same LLM vendors are both a key supplier and the competitive threat (strategic dependency)
- AI-native language tutors (Speak, ELSA) targeting the higher-value conversational-fluency segment
- Platform-fee and platform-rule changes by Apple/Google
- Key-person risk around founder-CEO Luis von Ahn; the 'AI-first' internal memo caused reputational backlash
Moats, dependencies & bottlenecks
Moats
World's most-downloaded education app; daily-streak psychology drives retention no general chatbot replicates — but habit does not equal learning outcome, which is the AI attack surface
56.5M DAU and years of learning-interaction data inform adaptive content; erodes if users defer to a single general-purpose AI assistant
Thousands of universities accept DET as a cheaper TOEFL/IELTS alternative — a two-sided acceptance network that is hard to replicate and high-margin
Free tier + virality keep customer-acquisition cost low; dependent on app-store ranking and continued top-of-funnel growth (now sharply decelerating)
~20,500 units/quarter is real supply, but the same AI that lets Duolingo make content lets anyone make it — low defensibility
Dependencies
Distribution & payments 15-30% platform fee is a permanent margin tax; ranking/rule changes directly affect acquisition and economics (AAPL, GOOGL)
AI model supplier Core premium/AI features depend on third-party frontier models — and the same providers commoditize the base product; cost + strategic exposure (MSFT; OpenAI private)
Cloud infrastructure Hosting/compute for the app and content pipeline (AMZN)
Key person / strategy Product vision and public brand are tightly tied to the founder; the 'AI-first' memo showed messaging risk
Subscriptions and in-app ads are both sensitive to consumer budgets and ad-pricing cycles
Advantages
- Profitable and FCF-positive at scale — rare among consumer apps, funds experiments internally
- Fortress balance sheet: >$1B cash, zero debt
- Best-in-class engagement/retention mechanics (streaks) and top brand recall in the category
- Multi-product optionality already built (Max, DET, Math, Music, Chess) rather than single-product exposure
- Low, asset-light capex so nearly all operating cash converts to free cash flow
Weaknesses
- Sharply decelerating bookings and ~+6% MAU growth signal the hyper-growth phase has ended
- Core language-drill product is the most AI-substitutable part of the offering
- Guided gross-margin compression from its own AI features
- Headline earnings quality dented by a one-time tax benefit
- Structural app-store fee drag and dependence on the very vendors disrupting it
- SBC dilution (~3.5-4% share growth guided 2026) offsets part of the cash-return case
Bottlenecks
- Top-of-funnel: MAU growth has fallen to ~+6% YoY, capping the pool the paid tier converts from
- AI cost curve — richer AI features raise cost of revenue faster than pricing can offset, driving the guided gross-margin compression
- Paid conversion — ~9% paid penetration — lifting it without adding user friction is the central monetization experiment
- App-store economics: 15-30% take caps the achievable operating leverage on subscription revenue
- Proving learning efficacy vs. a free chatbot — the differentiation the whole thesis now rests on
Top signals & trends
Top signals
Bull if it materializes · The full-year guide hinges on a 2H re-accel; a miss confirms the structural-decel bear case and risks further de-rating — next print expected early August 2026
The collapse from +33% to ~+6% is the clearest sign the funnel is slowing; a re-acceleration would relieve the substitution fear
AI-inference cost is the swing factor; falling below guide signals AI is a margin liability without a pricing offset
Management cites rising words-spoken per user; monetizing that at a premium is the proof AI can be additive, not just defensive
>$1B cash gives repurchase capacity, but diluted shares are guided up ~3.5-4% in 2026 — watch whether buybacks turn net-accretive
Trends
The defining structural threat — free LLMs replicate translation and conversational practice; de-rated the entire category
Lets Duolingo publish tens of thousands of course units cheaply and expand into languages/subjects it could not staff before
Favors AI-native tutors (Speak, ELSA) but also validates Duolingo's Max/video-call push if it executes
Any forced reduction in Apple/Google take-rate would directly widen Duolingo's margins
Users increasingly weigh a $7-13/mo app against a general AI subscription they already pay for
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
App Store distribution + payments; 15-30% take on in-app subscriptions
Android distribution + payments; same fee structure; also a competitor via Gemini/Translate
Cloud hosting and compute for the app and AI-content pipeline
LLM inference behind Duolingo Max and AI tutor features — key input and, simultaneously, the disruptor
56.5M DAU / 137.8M MAU; ~9% pay for Super Duolingo / Duolingo Max — the core revenue base
Thousands accept the Duolingo English Test as a TOEFL/IELTS alternative
Buy in-app ad inventory on the free tier
General-purpose AI that now does translation + conversational practice for free/low cost; its Aug-2025 GPT-5 app-building demo helped drive DUOL's multi-month slide
Free translation at massive scale plus a frontier assistant; also Duolingo's Android app-store landlord
Frontier LLM used for language help/tutoring; part of the free-substitution pressure
Owns Busuu, a subscription language-learning app; itself severely disrupted by AI — a cautionary comp for the category
Subscription language app focused on adult conversational learning; direct paid-tier competitor
Legacy immersion brand; direct competitor, diminished but still present
AI-native spoken-fluency tutor targeting the high-value conversational segment Duolingo's Max also chases
Incumbent English-proficiency tests that the Duolingo English Test competes against
Broader edtech peer (not language-specific) — comp for the AI-disruption-of-edtech multiple re-rating