Full Report

Industry — Mobile-First Consumer Education

Duolingo competes in consumer digital language learning — a slice of the global education economy where the product looks and behaves like a mobile game, the customer is a self-paying individual, and distribution runs through the Apple App Store and Google Play.

Two facts frame the industry. First, standalone language-learning apps generated only about $1.11B of revenue in 2024 out of a $6.34B total category, with the rest sitting offline in classrooms, tutoring, and corporate training (Straits Research / Business of Apps via quantumrun.com). Second, Duolingo captured roughly two-thirds of that app revenue pool by itself — a level of dominance that is rare in any consumer software category and is the single most important industry fact in this report.

1. Industry in One Page

Consumer mobile education sells a gamified daily habit to a freemium funnel. A few hundred million people install the app, the majority engage for free, and a single-digit percentage subscribes (Duolingo: 9.2% of MAUs in 2025). Subscriptions account for the dominant share of revenue — at Duolingo, $873M of $1,038M FY2025 revenue (84%) is subscription, with the rest split across advertising, in-app purchases of virtual goods, and the Duolingo English Test. Profits exist because the marginal cost of an additional learner is near zero (cloud hosting + app-store fees), while content and brand build a moat that compounds with scale.

Cycle in one sentence: good cycles are driven by smartphone penetration, English as an economic ladder in emerging markets, New Year resolutions, and viral cultural moments; bad cycles surface in slowing free-user growth, falling DAU/MAU, ad-rate compression, and — the structural risk highlighted by Chegg's 64% subscriber collapse — substitution by free generative-AI tutors.

2. How This Industry Makes Money

The revenue engine is a freemium subscription with three monetization layers stacked on a single free product. The unit of pricing is an annual or monthly subscription per user, around $84/yr for Super Duolingo individual and $20/yr per seat on the six-seat family plan (dealnews.com, May 2026). Advertising on the free tier and one-off test fees are secondary; everything else is experimentation. Cost of revenue is small and dominated by third-party payment processing fees charged by app stores and increasingly by AI inference (10-K: "third-party payment processing fees… hosting fees and Artificial Intelligence costs").

The economics rest on three structural features. First, gross margins compress slowly because every paid subscriber routed through an app store costs the platform 15-30%; subscribers who pay on the web keep more economics. Second, R&D is the largest line on the income statement, not COGS — FY2025 R&D of $306M was 29% of revenue, reflecting the fact that "supply" in this industry is engineering and curriculum, not capacity. Third, bargaining power sits at two ends: with the consumer (who can churn any month and now has a free AI alternative), and with Apple and Google (who set commission rates and app-store policies). The middle layer — content production — is being commoditized by generative AI, which is good for incumbents with brand and bad for any aspiring entrant.

3. Demand, Supply, and the Cycle

Two cyclical episodes matter for base-rating this industry. The COVID upcycle (2020-2021) pulled installs and engagement forward — Duolingo downloads jumped from 385M cumulative in 2019 to 500M in 2020 and the company's valuation tripled to $6.6B at IPO. The 2022-2023 EdTech reset then compressed multiples across the entire category as growth normalized and AI substitution narratives took hold; Duolingo's valuation fell to $3.4B in 2022 before recovering. The cycle here does not look like commodities (volume + price) or housing (rates + builders) — it looks like consumer mobile software, where the dominant cycle drivers are app-store rank, app-store policy, and shifts in how attention is captured. The current chapter, written across 2024-2026, is AI: rapidly falling content-production cost (good for incumbents with data and brand) and rising substitution risk (bad for anyone with a thin moat — see Chegg).

4. Competitive Structure

In strict GICS terms, Duolingo is filed under "Diversified Consumer Services / Education." The economically relevant universe is much narrower: mobile-first DTC language apps competing for share of daily attention, with adjacent threats from generative AI, online tutoring marketplaces, and offline language schools. Within the mobile-app slice, the market is closer to winner-take-most than fragmented.

The public comparables above are a deliberately mixed bag because Duolingo's closest economic peers are private (Babbel, Busuu, Memrise, Preply). The map: COUR and CHGG are the direct EdTech reference points and both are cautionary tales — Coursera's enterprise-cert engine is decelerating and Chegg's homework-help franchise has been cut to ribbons by free AI (subscriber base from 8.1M to 2.9M, a ~64% decline). SPOT and RBLX are the engagement and freemium templates — they tell you what scaled consumer engagement is worth when monetization works (Spotify $91B market cap on $17B revenue; Roblox $32B on $4B revenue). LRN is the lone profitable US-listed EdTech name but operates an entirely different model — government-funded virtual K-12 schools, not consumer apps. Within this set Duolingo is the only company that combines fast growth (39%), software-grade margins (72% gross), positive GAAP operating income, and an undisputed consumer brand.

5. Regulation, Technology, and Rules of the Game

External rules in this industry are not classroom or curriculum rules — they are platform, data, and AI rules. The most economically important rule is the unstated one: how much commission Apple and Google will continue to charge.

Generative AI is the only rule-change above that is also a product axis. Management has publicly committed to an "AI-first" operating model — Q1 2026 disclosed 20,500 course units published in the quarter, more than ten times the historical rate — and management is guiding gross margin down to ~69% by end of 2026 (from 72%) as the AI inference cost of features like Video Call and Roleplay flows through cost of revenue. That is the deliberate trade: pay incremental compute cost today to widen the engagement and content lead before challengers can match it.

6. The Metrics Professionals Watch

Reading a freemium consumer-subscription business well means watching seven numbers. The mistake is to fixate on revenue growth alone, because revenue is a lagging blend of bookings, conversion, ARPU, and FX. The metrics below are what sell-side and buy-side analysts triangulate.

Two metrics deserve a step beyond the table. Bookings vs. revenue is the single most misread line in any subscription P&L: a customer who pays $84 upfront produces $7/month of GAAP revenue for twelve months but $84 of cash today. When Duolingo guides "softer bookings," sell-side downgrades come fast (the February 2026 stock drop on FY26 bookings outlook is a recent example) — even when revenue growth is intact. S&M intensity is the canary for the engagement flywheel: if Duolingo had to push S&M from 12% to 20% of revenue to keep DAU growth above 15%, the operating-leverage story would fade and the multiple would compress. As of FY2025, the flywheel is still working.

7. Where Duolingo, Inc. Fits

The summary read is that Duolingo is the only public way to own the category leader of consumer mobile language learning, and the category itself is a small but rapidly compounding slice of the broader education economy. Treat the rest of the report accordingly: the moat conversation belongs in Warren's tab, but the industry tab's verdict is that there is a real industry here, Duolingo sits at its center, and the structural risks worth tracking are AI substitution and app-store economics — not curriculum, accreditation, or enrollment cycles.

8. What to Watch First

The seven signals below would tell you, within a single quarter, whether the industry backdrop is improving or deteriorating for Duolingo specifically. They are deliberately ordered from most to least likely to move the multiple.

The Chegg base rate matters here. CHGG went from $776M revenue and ~8M paid subs in 2022 to a roughly $115M market cap by mid-2026 — a collapse driven not by an income-statement event but by a slow loss of free-user habit to ChatGPT. Duolingo's defense is not better content, it is better gamification and a brand that has become part of pop culture. Multiple watchlist signals flashing simultaneously — DAU growth slowing, S&M intensifying, paid % flat, and gross margin slipping for non-AI reasons — would be the Chegg pattern asserting itself; the analyst playbook on that pattern is multiple compression ahead of the print.

Know the Business

Duolingo is a consumer mobile-software business dressed up as EdTech: a freemium subscription engine where word-of-mouth distribution, a gamified daily habit, and ~9% paid conversion produce 72% gross margins, ~30% Adjusted EBITDA margins, $360M of free cash flow, and zero debt. Daily active engagement is the lever that compounds into paid subscribers; the two most commonly mis-weighted variables are the non-discretionary toll paid to Apple and Google inside cost of revenue, and franchise durability against zero-cost generative-AI substitutes at the lower-skill end of the funnel.

1. How This Business Actually Works

This is a freemium consumer subscription. Anyone can use the app for as long as they like at no cost; about 9% of monthly users pay roughly $84 a year for an ad-free experience and extra features. Subscription is 84% of revenue; the remaining 16% is advertising on the free tier, the Duolingo English Test, and small in-app virtual-good sales. Cost of revenue is overwhelmingly third-party payment processing fees charged by Apple and Google (15–30% of in-app subscription proceeds) plus cloud hosting and a rising AI-inference line. Everything else on the income statement is people — and the largest people line is R&D, not sales and marketing.

The economics tilt unusually far toward the product and the platform tax. FY2025 S&M was only 12% of revenue (including stock comp) because the brand grows by going viral, not by buying impressions; R&D ran 30% of revenue and is best read as the cost of supply, since the "factory" producing supply is engineers, designers, and the AI tooling that now generates 20,500 course units a quarter — more than 10× the 2024 baseline. Pricing power is shallow at the high end: management does not push price aggressively because the freemium funnel works only as long as the paywall feels optional. The two parties with real bargaining power are the consumer (instant churn, free LLM alternatives) and the duopoly distribution platforms.

2. The Playing Field

Duolingo has no clean public peer. Its closest economic substitutes — Babbel, Memrise, Busuu, Rosetta Stone, Preply — are private. The listed comparable set therefore blends three different things: direct consumer EdTech that is shrinking (COUR, CHGG), profitable but structurally different K-12 EdTech (LRN), and large freemium-engagement platforms used as templates for what scaled monetization can look like (SPOT, RBLX). Read the table accordingly: Duolingo is the only name combining 30%+ growth, 70%+ gross margins, positive GAAP operating income, and material FCF generation.

What the peer set reveals: the base rate for "EdTech consumer subscription" is bad (COUR decelerating to single digits, CHGG losing two-thirds of subscribers to free LLMs), and the only listed name with Duolingo-like growth-plus-margin is Roblox — but Roblox burns cash on stock comp at 23% of revenue and runs negative GAAP operating income. Stride is profitable but answers to school-district procurement cycles, not consumer App Store rankings. Spotify is the relevant template for what scaled freemium subscription "looks like" at maturity (~10% growth, low-teens operating margin, high-teens FCF margin); the bull case for Duolingo is that better gross margins and a more defensible niche let it stay above Spotify's terminal economics. The bear case sits one row up — Chegg shows what happens when a thin-moat consumer EdTech name meets free generative AI.

3. Is This Business Cyclical?

Duolingo is secular, not cyclical. There is no inventory cycle, no rate-sensitive customer, no capex pulse, no commodity input. Revenue compounds at 35–45% off a flywheel — more users → more data → better product → more users → higher paid conversion. The thing that looks cyclical in the history is the post-IPO consumer-tech reset (2022) and the COVID demand pull (2020–21), both driven by the equity-market environment rather than the underlying business.

What is sensitive to outside conditions is mix and timing, not direction. EM ARPU moves with FX and local consumer discretionary income (the average revenue per user has trended down as the install base shifts toward lower-ARPU geographies). Bookings, which are collected upfront, show up before revenue does — that is why the stock sold off in February 2026 on a soft FY26 bookings guide even though revenue growth was intact. App-store policy can shift gross margin by hundreds of basis points overnight if EU DMA-style web-billing relief expands. None of these is a classical macro cycle; they are mostly platform and regulatory exposures inside an otherwise secular trajectory.

The structural risk that would create a real cycle for this business is AI substitution. The Chegg base rate — subscriber count collapsing from ~8M to ~3M as ChatGPT/Gemini replaced paid homework help — is the cautionary template, even if Duolingo's defenses (gamification, brand, social streaks, parent–child trust signals) are materially stronger.

4. The Metrics That Actually Matter

Read this business through five numbers. Revenue growth is a lagging blend of all of them and tells you less than people think.

Two further notes belong here. S&M intensity (12% of revenue including SBC) is the canary for the word-of-mouth flywheel — if Duolingo had to ramp S&M to keep DAUs growing, the operating leverage story would unwind quickly. And SBC at 13% of revenue, guided to ~15% in FY26, is the largest non-cash cost in the P&L; treating Adjusted EBITDA at face value (29.5% margin) without subtracting SBC overstates true cash earnings by half. The cash-comp-adjusted FCF margin is closer to 22%, not 35%.

5. What Is This Business Worth?

Duolingo is best valued as one economic engine, not a sum of parts. The Duolingo English Test is real (6,100+ accepting institutions) but is bundled inside "Other" revenue, ran roughly flat at $11M in Q1 FY26, and is not a separately listed asset. Math, Music, and Chess are early-stage product extensions that share the same app, the same data flywheel, and the same paying user. There is no listed subsidiary, no holdco discount, no regulated/unregulated mix to disaggregate. Force-fitting SOTP here would manufacture false precision.

The right lens is price-to-free-cash-flow with a growth-runway overlay — the same lens you would apply to Spotify or any scaled consumer-software cash compounder — adjusted for stock-based compensation because SBC is a real cost.

On headline EV/FCF, Duolingo screens as cheap relative to comparable engagement platforms. On a stock-comp-honest basis (subtracting ~$140M of annual SBC from cash flow), the multiple is roughly in line with where Spotify and Roblox trade. A premium versus those names is supportable only if (a) DAU growth stays above 20% into 2027–28, (b) per-unit AI compute costs continue to fall, and (c) the AI-substitution narrative does not start showing up in retention curves. A discount is warranted if any one of those breaks.

6. What I'd Tell a Young Analyst

Don't be impressed by GAAP net income. The FY2025 figure of $414M includes a $256.7M one-time tax benefit from releasing the valuation allowance on deferred tax assets. Underlying GAAP operating income was $136M (13% margin). Adjusted EBITDA was $306M (29.5% margin). Free cash flow was $360M (35% margin). After subtracting the $137M of real SBC cost, owner earnings are closer to $220M — still excellent, but only about half of what the headline net income line suggests.

Watch bookings before you watch revenue. Annual subs collect cash upfront and are recognized monthly, so revenue lags reality by 6–12 months. The February 2026 stock drop happened on a soft FY26 bookings guide while the revenue print was clean. If you wait for revenue, the multiple has already moved.

Watch DAU growth and S&M intensity together. DAU growth above 20% with S&M at 12% of revenue (roughly today) is the flywheel working. DAU growth at 12% and S&M at 18% would mean the flywheel has stopped compounding and management is buying users to cover the gap — that is the moment the multiple compresses.

The gross-margin path is a policy choice, not an outcome. Management is choosing to let gross margin drift from 72% to 69% to fund AI-feature differentiation. If gross margin breaks below 69% for non-AI reasons (mix shift to advertising, app-store fee changes), that is a real signal. Inside the guidance, it is not.

The real moat is gamification and brand, not curriculum. Anyone with a credit card and an LLM API key can teach Spanish. What is hard to replicate is a daily streak that 15M people have held for over a year, a brand that has become an internet meme, and a 2-billion-exercise-per-day data flywheel. The day to worry is the day those streaks stop showing up in App Annie engagement data — that is the leading indicator of AI substitution, and it leads everything else by a quarter or two.

Three things would change the thesis. (1) DAU growth dropping below 15% for two consecutive quarters. (2) Paid % of MAU flattening or falling year-over-year. (3) Gross margin slipping below 68% with management blaming "mix" rather than "AI investment." Any one of those, in isolation, is noise. Two in the same quarter is when the de-rating playbook typically starts to run.

Long-Term Thesis

The long-term thesis is that Duolingo is the category-defining consumer brand of mobile language learning that, over the next 5 to 10 years, can convert a still-deepening daily habit into 200M+ MAUs, 35-40M paid subscribers, and a $3-4B revenue base at 30%+ Adjusted EBITDA and 30%+ FCF margins — only if the behavioural moat (streaks, gamification, brand) survives the generative-AI substitution cycle without forcing S&M intensity above 15% of revenue. The 5-to-10-year case rests on three durable conditions that are independent of any single quarter: (1) DAU compounds toward management's 100M-by-2028 target while paid % of MAU trends up rather than flat, (2) AI inference cost falls faster than feature use grows so gross margin stabilises in the high-60s/low-70s, and (3) capital allocation continues to neutralise SBC dilution with buybacks because the founder PSU caps new equity grants through 2031. This is not a long-duration compounder unless the Chegg substitution vector — paid conversion stalling while engagement holds — is decisively absent in third-party retention data over the next 18 months. The biggest single risk to the long-duration thesis is not the next bookings print but a slow erosion in paid % of MAU that confirms behavioural defences are being silently arbitraged by free LLMs at the lower-skill end of the funnel.

1. The 5-to-10-Year Underwriting Map

The thesis decomposes into seven durable drivers. Each has to be true individually — together they describe the compounder this name could be over the next decade. The driver that matters most is paid-conversion durability: the single line item that distinguishes "freemium platform that monetises a daily habit" from "thin-moat consumer subscription losing the funnel to free LLMs."

The single most important driver is row 2 — paid conversion deepening. Drivers 1, 3, 5, 6 and 7 are all defences and optimisations; they protect or amplify what already works. Driver 2 is the load-bearing test of whether the moat survives the generative-AI substitution cycle. Engagement (row 1) can hold while paid % stalls — that is the Chegg base rate. If paid % of MAU keeps trending up, every other driver gets the runway it needs. If paid % flattens, no amount of AI margin tailwind or multi-subject optionality matters at the multiple.

2. The Compounding Path

A consumer-software compounder turns engagement into bookings, bookings into revenue, revenue into operating leverage, and operating leverage into owner cash. Duolingo has now demonstrated each step at least once. The question over a full 5-to-10-year cycle is whether the curves compound or just amortise.

The compounding mechanics are unusually clean for a consumer name. Capex is structurally light (~1.7% of revenue) so OCF and FCF track within a few hundred basis points. Working capital is a cash source because annual subscriptions prepay (deferred revenue $496M at YE25). Real cost of growth is engineering R&D, which scales sublinearly as the platform matures. The balance-sheet capacity — $1.05B net cash, zero debt — gives management complete flexibility to absorb a year of negative FCF, a recession-driven EM ARPU shock, or a major M&A move without raising capital. None of those are likely required; all are available.

The FCF-margin chart is the central long-term picture. FY26 is a self-described investment year — Adj EBITDA margin guided down from 29.5% to ~25.7%, gross margin guided down from 72% to ~69%. If this is a transition rather than a permanent reset, FCF margin should re-expand toward the original 30-35% target over FY27-FY30 as AI compute costs ease and operating leverage on R&D returns. That trajectory would generate roughly $8-10B of cumulative free cash flow over the next decade — more than the current $5.3B market cap — entirely from the existing business with no incremental optionality required.

3. Durability and Moat Tests

A 5-to-10-year thesis is only as strong as the tests it sets for itself. The five tests below are deliberately structured so that each has a measurable validation signal and a refutation signal that a portfolio manager could check against quarterly disclosures, third-party panel data, or annual filings. At least one is competitive, at least one is financial, and all are observable.

Tests 1 and 2 are competitive — the AI substitution and brand-recovery readings together account for whether the moat actually exists in 2028 the way the analyst day deck describes it. Tests 3 and 4 are financial — they validate that the moat translates into cash and returns on capital. Test 5 is the optionality leg; the long-term thesis works even if it fails, but the upside scenario needs at least one win here.

4. Management and Capital Allocation Over a Cycle

The 5-to-10-year underwriting case depends materially on who allocates the cumulative ~$8-10B of free cash flow the business is likely to throw off through 2035. The track record is mostly positive with one fresh blemish.

The strategic record is unusually strong. Luis von Ahn called the AI inflection two months after ChatGPT shipped — Duolingo Max launched March 2023 with GPT-4 — and the company scaled AI-generated content velocity 10x in one year while most consumer EdTech peers were still debating whether to use the technology. Through FY22-FY25 management beat the initial revenue, bookings, and EBITDA-margin guide every single year, with EBITDA margin beating by 200-700 bps annually. The transition from -24% operating margin (FY21) to +13% (FY25) and from negative cash flow to $360M FCF is among the cleanest profitability pivots in recent consumer-software history.

The communications record is where credibility took the recent hit. The April 2025 AI-first memo went viral on LinkedIn the wrong way, DAU growth decelerated over the next four quarters, and the memo is now cited in the FY25 10-K risk factors as something that "may have contributed to … a deceleration in user growth" — companies rarely flag their own communications inside a filed risk factor. CEO von Ahn walked back the memo's most controversial pieces twice in 12 months, ultimately calling it a "stupid memo" on the Q4 FY25 call. The brand moat is real but proved more mood-sensitive than the consensus narrative assumed; the question for the next decade is whether this was a one-time miss or a recurring pattern of founder communications fragility.

Capital allocation has been disciplined, not creative. Through FY25 buybacks were small (~$13M FY25, mostly employee withholding mechanics) and no dividend was paid — the right call when the business was scaling 35-40% annually. The February 2026 $400M buyback authorisation, sized at ~8% of float at the price trough, was timed well and is uniquely shareholder-friendly because the founder PSU framework caps new CEO/CTO equity grants through 2031. That means every dollar of buyback shrinks the share count rather than refilling a comp pool. Acquisitions have been small tuck-ins ($33M FY25, mostly engineering teams), with no transformational M&A and no related-party transactions. The risk to watch is what happens after 2031, when the existing PSU framework expires and a new pay package has to be designed — that is the structural decision that most defines the 2031-2035 portion of the long-term thesis.

The structural caveat for a long-duration thesis is dual-class control plus classified board. Public shareholders carry 86% of the economic risk and 24% of the vote. The behaviour to date has been shareholder-friendly, but the structure gives no recourse if behaviour changes. Over a 10-year horizon, this is the governance risk to underwrite, not a near-term concern: dual-class structures most often turn problematic when founders age out and the structure persists for a successor who didn't earn the original alignment.

5. Failure Modes

The failure modes below are the ones most likely to compress the long-term thesis. They are deliberately structured around durable, observable signals rather than generic "execution risk" complaints. The single highest-severity failure mode is the one that is hardest to see in real time — paid-conversion erosion from AI substitution — because it can persist for two-to-four quarters before showing up in the headline numbers.

6. What To Watch Over Years, Not Just Quarters

Five multi-year milestones determine whether the long-term thesis is on track. Each is observable in public disclosure or established third-party data, each has a meaningful time horizon, and each has explicit validation and refutation thresholds. These are deliberately structured to filter out single-quarter noise — none of them moves on a single bookings print.

Competitive Position

Competitive Bottom Line

Duolingo's moat is real but uneven. On the dimensions that actually generate cash today — daily engagement, word-of-mouth distribution, freemium conversion, brand virality — it is the category leader by a wide margin and the only public name combining 39% revenue growth, 72% gross margin, positive GAAP operating income, and 35% FCF margin. On the dimension that determines whether the moat survives the next investment cycle — defensibility against zero-price generative-AI tutors at the lower-skill end of the funnel — the moat is narrower than the narrative suggests, and the public peer set provides a brutal cautionary template in Chegg. The one competitor that matters most is not Babbel, Coursera, or Roblox — it is a free large-language-model chat session. Everything in this tab is organized around that asymmetry.

The Right Peer Set

There is no clean public comparable for Duolingo. The four closest direct competitors by product — Babbel, Memrise, Busuu, Rosetta Stone — are private, owned by other companies, or both. The listed comp set therefore deliberately mixes three things: direct consumer EdTech that has decelerated or collapsed (COUR, CHGG); the only profitable listed US EdTech (LRN), which operates a completely different government-funded model and is included to anchor the profitable end of the EdTech spectrum; and scaled freemium-engagement platforms (SPOT, RBLX) that Wall Street uses as templates for what monetization looks like once a daily-habit product matures. Read the table as a triangulation, not a like-for-like.

The private language-app competitors deserve a separate row because Duolingo's most natural product-substitution risks live there, not in the public set above. Their absence from the table understates how many people study a language on something other than DUOL.

Duolingo is the only listed name combining 30%+ growth with 30%+ FCF margin. Roblox matches the growth but not the cash conversion; Spotify matches the cash conversion but not the growth; Stride is a different business model entirely. The chart is the single best argument for why DUOL deserves a premium versus public peers — and why the bear case has to come from outside this set.

Where The Company Wins

Four advantages show up in the data, in descending order of how hard they are to replicate.

The S&M chart is the single most important slide in this tab. Coursera spends roughly 4x as much as Duolingo, as a share of revenue, on go-to-market — and still grows at one-quarter the rate. That gap is brand and habit doing the work that paid marketing has to do at every other listed EdTech name. It is also what would unwind first if the AI-substitution narrative began to bite: if Duolingo had to push S&M to 20%+ of revenue to keep DAUs growing, the operating-leverage story would collapse before any direct revenue impact showed up.

Where Competitors Are Better

Four areas where peers have something Duolingo demonstrably does not. Each is specific to a named competitor, not a generic "competition is intense" complaint.

The four gaps cluster into a single observation: Duolingo's economics are uniquely good for a 100%-consumer-DTC business, but they are also uniquely exposed to the one channel they sit on. Apple and Google take ~$1 of every $3-7 of in-app subscription revenue; that is structural and unavoidable. None of LRN, COUR, or even SPOT (which has won regulatory carveouts via EU DMA and the Epic ruling) carries that exposure to the same degree.

Threat Map

Eight threats ordered by severity. Severity reflects probability times magnitude of impact on the next 24 months of revenue and gross margin, not theoretical concern.

Moat Watchpoints

Five measurable signals that would tell an investor — within one or two quarters and before the revenue print — whether the competitive position is improving, holding, or deteriorating. Ordered by how quickly they would move the multiple.

The synthesis: Duolingo has the strongest moat of any listed consumer-EdTech name and one of the better moats in scaled freemium consumer software, but the moat is behavioral (engagement, brand, gamification) rather than technical (curriculum, content, or distribution). That is the kind of moat generative AI cannot replicate — but can erode at the edges by removing reasons to keep the habit alive. The watchpoints above are the cleanest way to monitor that erosion in real time.

Current Setup & Catalysts

The stock is sitting just above the $89.71 52-week low after a 78% drawdown from the May 2025 peak, and the market is mostly watching whether the +21% Q1 FY26 DAU print is the floor of a "deliberate transition" or the start of a new mid-teens growth profile. The last six months reset the entire frame: a 23-point bookings deceleration (FY25 +33% → FY26 guide +10.5%), a CFO change in the same window as the worst guidance event in company history, a $400M buyback authorized at the bottom, and a bulge-bracket downgrade cascade that left consensus 12-month targets clustered near $95-105. The near-term calendar is thin but unusually high-stakes: only one hard-dated print in the next six months really matters (Q2 FY26 on August 5), and it is the first test of whether the "bookings cliff" reaccelerates or confirms structural. Outside the print, three soft windows can still move the multiple — Annual Meeting (June 3), the UK Home Office DET tender decision, and a class-action consolidation event from the six open securities probes.

1. What Changed in the Last 3-6 Months

The current setup is fully described by what happened between Nov 5, 2025 and May 5, 2026 — five months that reset the multiple, the consensus PT, the CFO seat, and the audit committee in roughly that order. Everything before November is now context.

The narrative arc across these six events is tight. Eight months ago the market was paying ~$540 a share for a "category-defining consumer compounder" on a quasi-Spotify multiple plus an AI-tailwind premium. Today it is paying $113 for a "consumer subscription business mid-reset" with a structural-deceleration overhang. The debate has moved from engagement durability and Max attach rates to paid-conversion erosion, brand fragility, and CFO succession quality. August 5 is the next time the data votes.

2. What the Market Is Watching Now

Five debates currently set the multiple. The first three move price; the next two move underwriting confidence.

The watch list is unusually concentrated on one structural KPI (paid % of MAU) and one corporate-action KPI (buyback execution). Everything else — DAU growth, bookings cc, legal overhang — is a function of either of those two. If paid % of MAU holds above 9% through Q3 FY26 and the buyback pace stays at the current cadence, the multiple has stopped compressing. If either drifts, there is room for a re-test of the $89.71 52-week low.

3. Ranked Catalyst Timeline

Ten items rank-ordered by decision value. Hard-dated events come first only when their impact justifies the rank. The Q2 FY26 print is the highest because it is the single event that can move the multiple in either direction by 20%+; the AGM is third despite being the soonest because the agenda is mostly procedural.

The calendar is thin but front-loaded. Three events inside 90 days (AGM, the start of buyback execution disclosure in the Q2 10-Q, and the Q2 print itself) bracket the next decision window; everything after August 5 is either soft-window optionality (UK DET, Duocon) or continuous (buyback pace, technical levels). Items 7 and 8 (lawsuits, app-store policy) deserve attention only if a discrete event lands; otherwise they continue to weigh on the multiple as background overhangs rather than catalysts.

4. Impact Matrix

Five items actually resolve the investment debate. Everything else either confirms what is already known or adds information without changing the rating.

5. Next 90 Days

Three dated items, one continuous item, one window. The next 90 days are dominated by the August 5 print; everything else is either soft (Duocon timing, DET tender), procedural (AGM), or background (buyback disclosure cadence).

The 90-day window is functionally one event. Investors who need optionality through August 5 should be positioned by mid-July when sell-side previews and third-party panel data start landing. The AGM is a check-the-box event unless a withhold vote prints an outlier; Duocon is real but slips just past the 90-day fence and is more relevant to the Q3 FY26 narrative than to Q2.

6. What Would Change the View

Three observable signals would force the investment debate to update over the next six months. First, the Q2 + Q3 FY26 paid-% of MAU sequence — two prints flat or down YoY simultaneously turns the Chegg substitution thesis (Failure Mode 1) from an open question into a confirmed pattern and breaks the long-term compounder framing; the same sequence reaccelerating above 9.5% does the inverse and refutes the bear's primary structural claim. Second, the UK Home Office DET tender outcome — a win is the first concrete validation of Driver 5 (DET institutional franchise) inside any reasonable underwriting horizon, and would re-rate the optionality leg of the bull case from "could happen by 2030" to "is happening now." Third, buyback execution pace — cumulative repurchase clearing $200M by the Q3 print at a falling diluted-share count would close out the SBC-dilution overhang that today caps the multiple, while a slowing cadence into the Q2 print would tell PMs that management is preserving optionality (likely for M&A) at the exact moment the founder PSU caps say capital should be returned. These three signals together cover every leg of the long-term thesis the next six months can plausibly update; everything else on the calendar is either confirmation or background noise.

Bull and Bear

Verdict: Watchlist — the engagement core is intact and AI unit economics are inflecting in DUOL's favor, but the FY25→FY26 bookings cut (33% → 10.5% guide) is a 23-point step-down that the bull thesis explicitly needs the next one to two prints to refute. Both advocates anchor their primary trigger on the same variable: bookings growth in Q2/Q3 FY26 and whether paid % of MAU keeps rising off the 9.3% Q1 FY26 reading. That symmetry is the tell — there is no fundamental edge yet, only an observable evidence path. The valuation is genuinely cheap on headline cash earnings (11.8x EV/FCF) but only fair on SBC-honest owner earnings (~18x EV); reasonable people can land on either side of that lens.

The single piece of evidence that would force a decision is a Q2 FY26 bookings print of 13%+ in constant currency with paid % of MAU rising again — that combination would convert this from Watchlist to Lean Long. A print under 8% with paid % flat-or-down converts it to Avoid.

Bull Case

Bull-case fair value $195 on 22x EV/FCF applied to FY27E free cash flow of ~$400M (in line with SPOT's 26.5x, discount to RBLX's 21x). Timeline 12–15 months, anchored to Q2 and Q3 FY26 bookings prints plus visible buyback execution. Primary catalyst: Q2 FY26 bookings re-accelerating above the implied 10.5% trajectory — a 13%+ cc print would force the sell-side to re-underwrite the structural-deceleration thesis. Disconfirming signal: paid % of MAU prints flat-or-down YoY in any FY26 quarter, or DAU growth drops under 15% for two consecutive quarters — either is the Chegg pattern entering and would force a full exit.

(Bull's fourth point on AI gross-margin expansion was dropped — the +190 bps Q1 FY26 print is real but a single quarter is thin support for a structural claim, and the 72% FY25 gross margin already absorbs the read.)

Bear Case

Bear-case fair value $75 (~34% below the May 18, 2026 close of $113.24) on peer-multiple compression: FY27 revenue ~$1.30B at 22% SBC-adjusted FCF margin = $286M owner earnings × 12x = $3.43B EV + $1.05B net cash ÷ 49M diluted = $91, less an 18% haircut for execution and class-action overhang. Timeline 12–18 months. Primary trigger: Q2 or Q3 FY26 bookings growth below 8%, or paid % of MAU flat-to-down YoY for two consecutive quarters. Cover signal: Q2 FY26 bookings above 18% YoY combined with paid % of MAU re-accelerating above 9.5% — that combination would refute the structural ceiling read.

(Bear's fourth point on peer-comp / technicals was dropped — the technical action is corroborating, not load-bearing, and the SBC-adjusted comp work is already inside point 3.)

The Real Debate

Verdict

Watchlist. The bull and bear case rest on the same decisive variable — Q2/Q3 FY26 bookings and the paid % of MAU trajectory — and right now the engagement data leans bullish (paid % rose to 9.3%, DAU +21%, gross margin +190 bps) while the headline bookings cut leans bearish (33% to 10.5% is not a small revision). The most important tension is whether AI is a moat-widener or the same substitution vector that hollowed Chegg; the bull side carries slightly more weight here because the freemium-conversion line is moving in the wrong direction for the Chegg analogy, but the bear is right that brand-mediated behavioral moats can fade in a single news cycle and the April 2025 memo proved that channel exists. The bear could still be right: a structural mid-teens grower with 22% SBC-honest cash margins is fairly priced near current levels, not cheap, and the governance friction (Audit Committee chair to CFO, MAU disclosure demoted as growth slowed, five law-firm investigations) extends the discount window even if the operating thesis holds. The durable thesis breaker is two consecutive quarters of paid % of MAU flat-or-down YoY — that converts the verdict to Avoid because it confirms the Chegg substitution pattern. The near-term evidence marker is the Q2 FY26 bookings print — 13%+ cc with paid % rising converts the verdict to Lean Long; under 8% with paid % flat converts it to Avoid. Until those data points exist, position the name on the watchlist and let the prints decide.

Moat — What Protects Duolingo, If Anything

A moat is a durable economic advantage that lets a company protect returns, margins, share, or customer relationships against competition. The verdict here is Narrow moat. The defensible part of Duolingo is behavioural — a daily-habit engagement loop, a culturally embedded brand, and a 2-billion-exercise-a-day data flywheel that no listed peer can match. The undefended part is structural — Apple and Google take an estimated 15–30% of every in-app subscription, curriculum is increasingly commoditised by free large-language-model chat, and Duolingo competes for share of attention with infinite-scroll video, not just with other language apps. The strongest single piece of moat evidence is third-party retention data: Duolingo had a 28% Western-market churn rate in late 2023, the lowest of any EdTech app tracked by Sensor Tower, versus 58% at Babbel and 64% at Simply Piano. The strongest single piece of evidence against a wide moat is the Chegg base-rate — a thin-moat consumer EdTech subscription that lost roughly two-thirds of its paying users to free LLMs in three years. Duolingo's defences are materially better; the structural exposure is the same.

1. Moat in One Page

The conclusion is Narrow moat for three reasons.

Why it qualifies as a moat at all. Duolingo is the only listed consumer-subscription business combining +39% revenue growth, 72% gross margin, positive GAAP operating income, and 35% FCF margin. That financial profile cannot be produced by execution alone — it requires a structural advantage. The advantage that shows up in the data is engagement: a DAU/MAU ratio of roughly 40%, which is a level only the best gaming and social platforms reach, and a Sensor Tower-measured churn rate that is roughly half that of the closest direct competitor (Babbel). The third-party data and the company's own KPIs agree on the same point: paying users stick.

Why it is narrow rather than wide. Two structural exposures cap the moat. First, the platform tax: ~28% of FY2025 cost of revenue is third-party payment processing paid to Apple and Google — a toll that does not exist at Stride (B2G), Coursera (B2B/web), or in any of the institutional incumbents. Duolingo has no leverage over this line. Second, AI substitution: the exact tasks Duolingo monetises (vocabulary drilling, grammar correction, conversational practice) are the tasks where a free LLM chat session is now competitive on quality and superior on flexibility. The defence is behavioural (streaks, brand, social) — not curriculum or technology — and behavioural moats erode invisibly until the day they don't.

Why it isn't "no moat." The Chegg base rate exists because Chegg never had real engagement. Its DAU/MAU was low, its product was task-based not habit-based, and its brand was a utility. Duolingo's product is consumed daily and emotionally; the green owl is an internet meme; the company runs the largest mobile A/B testing platform outside of Big Tech. None of that is replicable in 18 months by a well-funded entrant. The moat is real — it is just not infinitely deep.

2. Sources of Advantage

Every claimed source below has to clear two tests: does it show up in numbers, and is it specific to Duolingo or merely a feature of the industry?

The two sources that pass both tests are brand + cultural footprint and behavioural switching costs. Data flywheel, scale economies, and app-store ranking are real but second-order; they protect against subscale challengers and not against the substitution risk that actually matters. Network effects are an external-rating cliché that does not hold up against the company's own disclosures — Duolingo is not a marketplace and not a UGC platform.

3. Evidence the Moat Works

A claimed moat is only useful if it shows up in actual business outcomes. The table below pulls together the strongest external and filing-based evidence — for and against.

The five "supports" items establish that an advantage exists; the three "refutes" items establish that it is narrower than the bull case implies. Investors who weight only the first five end up at "wide moat" — which the April-2025 brand event and the FY26 bookings cut have already shown to be optimistic.

The churn chart is the single best evidence point in this tab. It is the only direct, peer-comparable measurement of whether the alleged moat actually does what a moat is supposed to do: keep customers from leaving. Duolingo's churn is roughly half that of the closest direct substitute (Babbel) and a third of the closest gamified-EdTech substitute (Simply Piano).

4. Where the Moat Is Weak or Unproven

Three areas where the moat is meaningfully thinner than the consensus narrative.

The brand has been measurably damaged in the last twelve months. The April 2025 AI-first CEO memo is named in the FY25 10-K risk factors as a realised event that "may have contributed to unfavorable publicity, adverse impacts on the Company's brand and social media presence, and a deceleration in user growth." DAU growth went from +49% in Q1 FY25 to +21% by Q1 FY26. Companies rarely name their own communications as the cause of a realised risk. A brand moat that fades on a single LinkedIn post is not the kind of brand moat investors should be paying a premium for; it is closer to a consumer-sentiment moat, which is real but mood-sensitive.

The data flywheel is being commoditised. Frontier LLMs ship with high-quality multilingual capability out of the box, with no Duolingo-style A/B testing required. The Birdsong experimentation platform and the 2-billion-exercise-per-day data flywheel are advantages over a subscale rival like Babbel — but they are not advantages over OpenAI or Google. Management has accepted this implicitly by guiding gross margin down ~300 bps to fund AI features that ride on third-party model APIs; the cost structure now includes a variable inference line whose unit economics are set by Anthropic, OpenAI, and Google, not by Duolingo.

App-store distribution is the moat that nobody wants to talk about. Roughly 28% of FY25 cost of revenue is payment processing fees to Apple and Google. The advantage of being #1 in the App Store Education category is enormous — but it sits inside a duopoly that owns 100% of mobile distribution and that has historically not been a value-accretive partner for any subscription business. Spotify won a multi-year regulatory campaign to weaken Apple's hold; Duolingo has not. The moat against Babbel is real; the moat against the platform on which Duolingo and Babbel both run is zero.

The unproven moat sources — patents, true network effects, and curriculum quality — should be excluded from any underwriting case. Patents are a defensive line item in this industry; "network effects" as cited by external rating services confuses social features with two-sided markets; curriculum is the one place where a free LLM is structurally competitive.

5. Moat vs Competitors

Duolingo's moat has to be sized relative to the alternatives, not in absolute terms. The comparison below uses the listed peer set from Competition tab, plus the most relevant private substitutes and the AI-tutor reference point that does not appear on any public peer table.

The peer table tells two different stories about the same name. Among listed consumer-EdTech, Duolingo has the strongest moat — by a wide margin. Among scaled freemium consumer-engagement platforms, Duolingo's moat is mid-pack: weaker than Roblox's true two-sided network, weaker than Spotify's regulatory progress on app-store billing, but with better unit economics than either. Where Duolingo is genuinely alone is in its financial profile — the upper-right of growth versus margin — not in its moat depth. That gap between "unique financial profile" and "narrow moat" is the source of investor confusion when valuing the business.

6. Durability Under Stress

A moat only matters if it holds up under pressure. The table below tests Duolingo's defences against the specific stresses likely to test them over the next 24 months.

The stress map is unbalanced by design. AI substitution sits above every other risk because the base-rate (Chegg) is severe, recent, and directionally applicable. Every other stress is manageable inside an engagement-led moat; the AI scenario is the one stress that could compress the moat itself rather than just rotating around it.

7. Where Duolingo Fits

The moat does not apply uniformly across the business. The strongest defences are concentrated in the consumer language-app core. The weakest defences are at the edges.

The honest reading is that the moat is overwhelmingly concentrated in the consumer language-app Super tier, which carries roughly four-fifths of revenue. Max, advertising, DET, and the multi-subject expansion are all reasonable optionality but should not be weighted heavily in any moat underwriting. The moat is one product, not a portfolio.

8. What to Watch

Five signals would tell an investor whether the moat is strengthening, holding, or fading — and would do so before the revenue print and before sell-side price-target cuts.

The first moat signal to watch is paid % of MAU. The Chegg pattern started exactly there — engagement held while paid conversion flattened, then fell. If Duolingo's paid % of MAU drifts down even one quarter while DAU growth holds, the worst-case AI-substitution scenario is already starting to play out, and every other moat metric will follow within two quarters.

Financial Shenanigans — Duolingo, Inc. (DUOL)

Duolingo's reported FY2025 net income jumped from $88.6M to $414.1M, but $231.7M of that lift is a one-time non-cash income-tax benefit from releasing a deferred-tax valuation allowance — without it, "core" GAAP net income is closer to $182M. Layered on top are three additional pressure points: (1) management has demoted MAUs to a "supplementary" metric in the Q1/FY2026 letter at the same time MAU growth decelerated to under 6% YoY, (2) the Adjusted EBITDA and Free Cash Flow definitions have both been amended in the last twelve months, and (3) multiple plaintiff law firms (Faruqi, Pomerantz, Portnoy, Levi Korsinsky) opened investigations in 2025–2026 alleging user-growth disclosure issues. We grade the file Elevated (45/100). There is no restatement, no auditor change, no material weakness — Deloitte has audited since 2018, the balance sheet is cash-rich, and underlying cash conversion is genuine. The single data point that would most change the grade is whether DAU growth diverges meaningfully from third-party panel data in 1H 2026.

1. The Forensic Verdict

2. Breeding Ground

The structural conditions that make accounting shenanigans more likely are present, but they are largely classic founder-led tech-company features rather than evidence of misconduct. The combination that matters: dual-class control, an inside-board CFO promotion in early 2026, repeated non-GAAP definition changes, and active plaintiff-law-firm investigations — these would not individually move the needle, but together they raise the underwriting bar.

The dual-class structure is the most consequential. With the two co-founders controlling 76% of voting power against 1.3% of Class A economic interest, ordinary shareholder vote outcomes — including auditor ratification and Say-on-Pay — are effectively pre-decided. The Section 16 late filings are administrative noise; the Munson CFO appointment, however, is a real governance event: the new CFO chaired the Audit Committee until January 8, 2026, less than seven weeks before assuming the CFO role. That is not prohibited and not unusual at small-cap tech companies, but it does compress the independence buffer between the audit oversight function and the financial reporting function.

The compensation structure is unusually founder-friendly: the CEO has taken no annual stock grants since 2023 because the 2021 IPO-era Founder Awards (1.8M PSUs across ten stock-price hurdles over ten years) still provide the upside. As of 12/31/2025, eight of ten stock-price hurdles have already been achieved, meaning the incentive to hit the final two tranches via aggressive reporting is in principle present — but those hurdles are based on 60-day trailing share-price averages, not accounting metrics.

3. Earnings Quality

The headline FY2025 net income is mechanically inflated. Excluding the $256.7M valuation allowance release (net $231.7M benefit on the tax line), GAAP net income would be approximately $182M — high quality, but a fraction of the $414M printed. Gross margin and operating margin both look clean; the issue is below the operating line.

FY2023 was the one period where receivables grew much faster than revenue (90% vs 44%) — a textbook revenue-recognition flag, but in this case the receivables are predominantly amounts due from Apple App Store and Google Play, not customer credit. By FY2025 the relationship had reversed: AR grew 26% against 39% revenue growth, a clean test. DSO has nonetheless drifted up over five years — 39 days (FY21) → 54 days (FY25) — which is consistent with mix shift toward annual subscriptions, but worth tracking in 2026.

Below-the-line items have done meaningful work. Interest income on the ~$1.1B cash pile contributed $45M in FY2025; FX gain $1.6M; and the one-time tax benefit $231.7M. In FY2024, interest income alone was about two-thirds the size of operating income. This is a real business — operating income tripled from $62.6M to $135.6M in FY2025 — but headline earnings should not be read at face value.

R&D capitalization is modest ($9.3M in FY2025, 0.9% of revenue) and consistent with prior periods; capex-to-depreciation ran 1.26x in FY2025, in a normal range. No big-bath impairments, no restructuring charges. Stock-based compensation of $137M represents 13.2% of revenue — declining from 20.0% in FY2022 — but is excluded from every non-GAAP measure management leads with.

4. Cash Flow Quality

Operating cash flow is genuine and growing — but it leans hard on two structural levers: deferred revenue from annual subscription billings and stock-based compensation add-backs. After stripping the tax benefit from net income, multi-year CFO/NI converts at about 2.9x — a strong number that is driven less by working-capital lifelines than by the gap between cash-basis billings and ratable revenue recognition.

The FY2025 visual is misleading because net income is propped by the deferred-tax-asset release. The cleaner test is FCF: $369.7M, up from $273.4M and $150.4M — three consecutive years of FCF growth above 40% with declining capex intensity (capex/revenue of about 1.7%). Acquisitions accelerated to $33.1M in FY2025 (NextBeat London team) but FCF after acquisitions still rose to $336.6M.

Two cash-flow flags are worth naming:

• Deferred revenue as a structural CFO contributor. The MD&A states explicitly: "A substantial source of our cash from operations comes from deferred revenue." Deferred revenue stood at $496.2M at Dec 2025 against $513.3M at Mar 2026. Bookings ($1,158M) exceeded revenue ($1,038M) by $120M in FY2025, a meaningful timing benefit on a 12-month-subscription business. If renewal cohorts deteriorate — which is exactly what the FY2026 guidance for 10.5% bookings growth implies (versus 16% revenue growth) — operating cash flow will compress before revenue does.
• Stock-based compensation as a non-cash add-back. SBC of $137M in FY2025 is added back to compute CFO. Adjusted for SBC, "real" cash earnings power is closer to $250M, not $387M. This is standard SaaS accounting, but the gap between reported CFO and SBC-adjusted CFO matters for valuation — and management's free-cash-flow definition does not include the $41.6M in cash paid for tax withholding on share-based awards (which sits in financing). That $41.6M is economically equivalent to a buyback to neutralize dilution and should be subtracted when assessing organic cash generation.

The SBC/revenue ratio is declining — from 20.0% (FY22) to 13.2% (FY25) — which is a positive forensic signal. Guidance for FY2026 sets SBC at approximately 15% of revenue, suggesting it may stop declining. No receivable sales, factoring, supplier finance, or securitizations are disclosed. Accounts payable is $8M against $288M of cost of revenue — there is no working-capital lifeline from stretching vendors.

5. Metric Hygiene

This is where the forensic weight of the case sits. Management's preferred metrics — Adjusted EBITDA, Free Cash Flow, DAUs, total bookings — have been redefined or reprioritized within the last twelve months, and the redefinitions are not trivial.

Bookings have led revenue lower for two straight years and management's own FY2026 guidance has bookings growing only 10.5% versus revenue of 16.1%. Because Duolingo recognizes subscription revenue ratably over the 12-month subscription life, bookings are the better forward indicator. The bookings-revenue scissor opening in FY2026 means revenue growth will follow bookings down in FY2027 unless retention sharply improves.

The MAU demotion is the most pointed editorial signal. Through Q4/FY2025 the 10-K reports MAUs first and prominently (133.1M, +14% YoY); in the Q1/FY2026 shareholder letter just three months later, MAUs are footnoted with the language "DAUs… with MAUs as a supplementary metric." The change is timed exactly with MAU growth dropping into the single digits. The risk-factor language acknowledges measurement uncertainty: "real or perceived inaccuracies in those metrics may negatively affect our reputation and our business" and "we may improve or change our methodologies… prior periods may not be as accurate or comparable."

6. What to Underwrite Next

The forensic risk does not invalidate the business — Duolingo generates real subscriber growth, $336M of FCF after acquisitions, and has $1.14B of net cash on the balance sheet. It does, however, justify (a) using "ex-tax-benefit" net income for valuation, (b) discounting non-GAAP comparability while definitions remain in motion, and (c) building disconfirming evidence around user metrics into any long thesis.

Specific items to track over the next two quarters:

• MAU disclosure. Watch whether Q2/FY26 shareholder letter restores MAUs to a primary table or keeps them footnoted. If MAU growth re-accelerates above 10%, the prominence shift looks like an editorial choice; if MAUs continue decelerating while DAUs grow, the prominence shift was load-bearing.
• Bookings-revenue scissor. FY26 guidance of 10.5% bookings vs 16.1% revenue implies the deferred-revenue tailwind reverses in FY27. Track quarterly bookings growth — anything below 8% would mean FY27 revenue growth in single digits.
• Deferred tax asset realization. $227.3M deferred tax asset, net is on the balance sheet as of Dec 2025. If FY26 pre-tax results fall materially below the 18–20% effective-rate guide, the valuation allowance test resets the other way — a partial DTA write-down would be a one-time GAAP charge of similar magnitude.
• CFO continuity. Munson is in seat from 2026-02-23. The first full quarter under her signature is Q1/FY26 (already filed). Watch for any change in non-GAAP definitions, KPI disclosures, or accounting policies in Q2/FY26.
• Securities investigation status. No complaint has been filed publicly; if any of the five firms file a class action by Q3/FY26 the discovery-driven disclosure risk steps up. Track 8-K filings for Item 8.01 "Other Events" or any updates to legal-proceedings note.

Signals that would downgrade the grade to Watch (≤30): restoration of MAUs to the primary engagement table, a stabilization of bookings growth above 12% in Q2/Q3 FY26, no class action filed by year-end 2026, and an FY26 effective tax rate that matches the 18–20% guidance.

Signals that would upgrade the grade to High (≥65): any restatement, auditor change, material weakness, SEC subpoena, formal complaint that survives a motion to dismiss, further redefinition of Adjusted EBITDA or FCF, or another KPI demotion.

Bottom line. The accounting risk here is not a thesis breaker. It is a valuation haircut and a comparability liability. An institutional investor should (i) use approximately $182M as core FY2025 net income for multiples work, (ii) apply a normalized 18–20% tax rate going forward, (iii) require SBC-inclusive earnings or FCF for fair-value bridges, and (iv) build the user-metric class-action and bookings-deceleration scenarios into the position-sizing limiter. This is a Watchlist file with elevated underwriting friction — not a short, not a clean long.

The People

Governance grade: B. Founder-led with exceptionally tight skin in the game and almost no cash compensation at the top, but a dual-class structure that pins ~76% of voting power on two people, a classified board, and an audit committee weakened by losing its financial expert to the CFO seat. The behaviour is shareholder-friendly; the structure is not.

The People Running This Company

Duolingo is a true founder-operator company. Both co-founders are still in their executive seats fifteen years after starting the business, and three of the six executive officers have been with the company for at least seven years. The new CFO is a recent and consequential change: Gillian Munson moved from being the Audit Committee Chair to running finance, replacing Matthew Skaruppa after his January 2026 resignation.

What They Get Paid

The headline is that the CEO is paid almost nothing in cash and gets no new equity grants — and that is by design. Von Ahn received $767,500 in 2025 (mostly the $750,000 base salary plus a $17,500 401(k) match), the same package he has received every year since 2023. The Board decided at the 2021 IPO that the founder PSU award would be the exclusive equity grant to both co-founders for the entire ten-year post-IPO period. The CEO pay ratio is 2.6 to 1 versus the median employee — one of the lowest of any large-cap US technology company.

The non-founder NEOs are paid at competitive but not eye-popping levels for a $5B technology company: roughly 85-90% of total pay is delivered in time-based RSUs that vest quarterly over four years. There is no annual cash bonus, no PSUs for non-founders, and no perquisites — a clean, equity-heavy structure that suits a growth company but offers no operating metric to tie pay to outcomes.

The pay-versus-performance chart tells the real story. Reported CEO comp is a flat ~$767K line. "Compensation Actually Paid" — the mark-to-market value of the founder PSU — swings from +$140M in 2023 to negative $44M in 2025 as the stock fell. Von Ahn's economic outcomes are dominated almost entirely by share price, which is exactly what shareholders should want.

The Founder PSU — What's Left

Eight of the ten tranches have already been earned. With shares at $126 (May 2026), the last two — at $612 and $816 — would require a roughly five-to-six-fold appreciation by the July 2031 expiry. The remaining 420,000 CEO PSUs and 210,000 CTO PSUs represent enormous upside-only motivation that can only be realised if outside shareholders also win big.

Directors are paid fairly, not generously

Director pay is roughly $220-235K each, ~75% delivered in RSUs that vest annually — standard mid-cap-tech construction.

Are They Aligned?

This is the section where Duolingo is genuinely exceptional, with one structural caveat that limits how much credit it should get.

Ownership and Control

Class B common stock carries twenty votes per share and converts one-for-one into Class A on transfer. The two co-founders together hold 6.76M of the 6.36M outstanding Class B shares (the gap reflects vesting PSUs counted as outstanding) and command ~75.8% of total voting power despite holding only ~15.6% of economic interest. Class A holders — including every major institutional investor — collectively own 86% of the economic upside but control only 24% of the vote.

Insider Selling — Real but Programmatic

Insider monetisation in 2025 was substantial — the CEO realised $115M from PSU vesting and $11M from option exercises — but it was mechanical, tied to the pre-agreed PSU vesting calendar and option expirations. There is no evidence of opportunistic dumping above the published trading windows, and the dollars roughly tracked the stock-price hurdles being satisfied. Von Ahn also gifted 50,000 shares and converted Class B to Class A during the year, the latter of which actually reduces his voting power.

Dilution and Capital Allocation

The 2021 Plan has a 5% annual evergreen feature through 2031, which is large. The Board waived the equivalent automatic ESPP increase for 2026 and beyond — a small but real shareholder-friendly act. The combined overhang of 12.5M unissued shares plus 859K exercisable options is ~28% of basic Class A share count, so dilution risk is real but spread over multiple years. The company does not pay a dividend and has not bought back stock — both capital is being reinvested into the business and into share-based comp.

Related Party Transactions

The proxy discloses only indemnification agreements as related-party transactions. There is no founder-controlled supplier, no related-party real estate, no family employment, and no sweetheart financing. For a founder-controlled company, this is genuinely clean.

Skin-in-the-Game Scorecard

9 / 10. The behaviour and pay structure are about as aligned as a public company can be. The one point off reflects the fact that founder voting control is materially higher than founder economic ownership — alignment within the founder pair is perfect, but alignment between founders and the rest of the cap table is one-sided.

Board Quality

The board has nine directors: two founders and seven independents. That is a reasonable mix on paper. Below is how each director scores on independence, tenure, expertise, and committee load — and where the gaps are.

What's Strong

• Independent majority — 7 of 9 directors qualify as independent under Nasdaq rules.
• Real expertise, especially Bing Gordon (Kleiner Perkins, Zynga, EA — consumer/product/gaming), John Lilly (Greylock, Mozilla — open-source consumer tech), and Mario Schlosser (founder/operator at Oscar). Jim Shelton brings genuine education-sector experience.
• No interlocking directorships among the founders' other roles; von Ahn left the Root, Inc. board in 2022.
• All committee chairs are independent, all members of audit and compensation committees are independent, and the board has a written clawback policy, anti-hedging policy, and an Insider Trading Policy that requires window-based pre-clearance.
• Auditor independence is solid — Deloitte since 2018, $1.79M in audit fees, only $0.56M in tax fees (a healthy ratio that does not look like fee-stretching).

What's Weaker

• Audit Committee continuity loss. Gillian Munson — Audit Chair and designated financial expert — became CFO in February 2026. Sara Clemens has assumed both roles; the committee now relies on a single director for SEC-filing financial expertise and chair leadership. There is no immediate plan disclosed to recruit a replacement financial-expert director.
• Classified board with three-year staggered terms makes director change very slow even if Class A holders wanted to push it.
• Bing Gordon is 76 and has served since 2020. Not a problem yet, but a typical mandatory-retirement threshold would already have been triggered at peer companies.
• Sara Clemens carries an unusual committee load: she is Audit Chair, on Compensation, on Mergers and Acquisitions — that is three of the four standing committees, which is a lot for one person.
• Mario Schlosser is concurrently President of Technology and CTO at Oscar Health — a demanding day job that limits the time available for Comp-Committee chair duties.
• Eight late Form 4s in one year suggests the company's Section 16 compliance process is not tight.

The Verdict

Grade: B. Behaviour-wise, this is an A-minus governance set-up — a founder team with massive ownership, almost no cash compensation, fully stock-price-linked equity, no related-party self-dealing, an independent audit committee with a credible auditor, and a competent board with operating experience that actually matches the business. The downgrade is structural: the dual-class share scheme and classified board mean public shareholders have very limited recourse if behaviour ever changes, and the recent loss of the Audit Committee's financial-expert chair to the CFO chair has created an immediate gap that has not yet been filled.

Strongest positives

Founders hold ~16% of the economics and ~76% of the vote, take a fixed $750K cash salary, and have no annual bonus or non-PSU equity awards through 2031. The CEO's 2025 "Compensation Actually Paid" was negative $44M when the stock fell. Capital allocation has been clean — no buybacks, no dividend, no related-party transactions.

Real concerns

Class A holders carry 86% of the economic risk and only 24% of the vote, with a classified board on top. The Audit Committee lost its financial-expert chair when Munson took the CFO seat. Eight directors filed late Form 4s in 2025. The 2021 equity plan has a 5% annual evergreen feature through 2031.

The single thing that would most likely change the grade

Upgrade trigger: recruit a new audit-committee financial expert with a credible CFO/CPA background, AND signal that the dual-class structure has a sunset (e.g., a future seven-year sunset like Snap's). Either would move this toward a B+ or A-minus. Downgrade trigger: if von Ahn or Hacker were to receive a new equity award before 2031 outside the existing PSU framework, or if a material related-party transaction appeared in a future proxy, governance would slip to C territory because the only thing keeping the dual-class structure tolerable is the explicit ten-year pay commitment.

The Story of Duolingo

Duolingo went public in July 2021 as an organic-growth, freemium darling — DAU growth above 50% for nearly four straight years, EBITDA margin expanding from negative to ~29%, and a co-founder CEO (Luis von Ahn) who built the story brick by brick. Then in April 2025 a CEO memo on AI-first operations triggered consumer backlash, DAU growth halved over three quarters, the company guided FY2026 bookings growth to 10–12% (from 33% in FY2025) and the stock fell roughly 70% peak-to-trough by May 2026. The current chapter is a deliberate, self-confessed reset: prioritize user growth and "teaching better" over near-term monetization. Management's strategic credibility — they did transform the P&L and they did call the AI inflection — is intact. Their communication credibility took a real hit: they spent every quarter from Q2 FY22 through Q3 FY25 raising guidance, then walked into a steep FY26 cut after publicly insisting "things are going great."

1. The Narrative Arc

Three transitions matter, and they all happened on Luis von Ahn's watch:

1. The profitability pivot (FY2022 → FY2024). The IPO story was growth-at-all-costs. By Q4 FY22 management started guiding EBITDA margin to 10–12%, then re-rated to 21–23% (FY24) and 27–29% (FY25). Actuals beat guidance every year — this is the most credible thing on the page.
2. The AI declaration (Q4 FY22 letter, Feb 2023). Two months after ChatGPT shipped, von Ahn put AI at the center of the strategy. Duolingo Max launched as the second monetization tier in March 2023. Through FY24, AI was a tailwind narrative — Video Call (Q3 FY24) was the marquee feature.
3. The AI-first reset (April 2025 → today). A CEO LinkedIn memo telling employees AI would replace contractors and become a performance-review input went viral the wrong way. By the time Q3 FY25 results landed in November, DAU growth had decelerated from +49% (Q1) → +40% → +36%. Q4 FY25 results crashed FY26 guidance and the stock with it.

2. What Management Emphasized — and Then Stopped Emphasizing

The shareholder letters from Q2 FY22 to Q1 FY26 are formulaic enough that the words management chose to stop using are as revealing as the ones they added. Below is the rough frequency of headline themes in the quarterly letters and press releases (0 = absent, 3 = dominant).

Four patterns stand out:

• "Take the long view" was one of nine operating principles in 2021 and never appeared at the top of a letter for thirteen quarters. It became a defining phrase only in Q3 FY25 — exactly when management needed cover for slowing DAU growth and a coming guidance cut. The phrase carried through Q4 FY25 and Q1 FY26.
• Word-of-mouth/brand marketing was a recurring brag in FY22–FY24 (Cannes Gold Lion, Super Bowl "Superb Owl," Dead Duo at 1.7B impressions in Q1 FY25). By Q2 FY25 management noted they had toned down "unhinged" social content "as we listened to community feedback" — and the brand bragging never came back to its prior intensity. The same channel that powered organic growth had become a liability after the AI-first memo.
• Profitability / margin expansion dominated letters from Q4 FY22 through Q2 FY25 — every quarter they raised the FY EBITDA-margin guide. Starting Q3 FY25 the margin narrative recedes; by Q1 FY26 the company is guiding FY26 gross margin down (to ~69% from 73%) and EBITDA margin down (to ~25.7% from ~29%).
• Multi-subject (Math, Music, Chess) is the most volatile theme. Math launched as a separate iOS app in 2022, got folded into the main app in 2023, then disappeared from headline coverage. Music followed the same arc. Only Chess (launched Q1 FY25) sustained attention — "fastest-growing subject we've ever launched" — but by Q1 FY26 it was grouped generically with Math and Music as a "next user growth engine."

3. Risk Evolution

The structural top-3 risks (engagement, competition, brand/product) are unchanged across all five 10-Ks. The interesting moves are at the edges — what got added, what got softened, and what the company finally admitted in plain English.

Reading the risk-factor changes year by year:

• AI risk went from zero to dominant in three years. FY2021 had no AI disclosure. FY2022 added a single regulatory paragraph. FY2023 added the full operational framework after Max launched, naming OpenAI/GPT-4 dependency and content-training-data litigation. FY2024 added third-party AI vendor risk. FY2025 made it personal — the April 2025 memo is named as a contributing cause of user-growth deceleration.
• Apple concentration is the slow-burn risk. App Store revenue share rose from 50% (FY21) to 62% (FY25). The "generally 15–30%" fee disclosure has been identical for five years despite the entire epic-vs-apple regulatory backdrop. Management has started testing direct web purchase flows for US iOS users — material to FY26 gross margin if it works.
• DET shrunk from 6.1% to 4.0% of revenue between FY24 and FY25 and the risk language now explicitly cites "a country's decision to cap or reduce immigration." This is the first quantitative decline in DET disclosure.
• The pre-IPO risks faded on schedule — COVID-19, Brexit, "limited operating history," "operating losses each year since inception" all disappeared as the company matured. By FY25, "operating losses" became "maintain or increase profitability."
• The 10-K reads tougher in FY25. Adding GenAI competitors (chatbots, wearable translators), AWS us-east-1 outage example (October 2025, 12+ hours of downtime), Texas AI Act, and Trump's December 2025 AI EO all add real surface area. The brand risk language is also the most candid it has ever been.

4. How They Handled Bad News

Three episodes test the framing.

The DAU deceleration (Q4 FY23 → Q3 FY25)

Management telegraphed the deceleration honestly. The Q4 FY23 letter (Feb 2024) warned: "so far in Q1 2024, our YoY DAU growth is closer to the mid-50s than the 60%+ it was throughout 2023." That softened the landing for the next eighteen months. The break came in Q3 FY25 — the deceleration accelerated, and management connected it to the AI memo for the first time.

The AI-first memo backlash (April–May 2025)

This is the messiest episode. The Q2 FY25 letter (August 2025) — published four months after the memo went viral on LinkedIn and triggered consumer-press coverage from PCMAG, Inc, and others — contains zero direct reference to the memo. The only acknowledgment is buried in the Q3 FY25 letter (November 2025): "Growth was slightly slower than Q2 in part because we posted less 'unhinged' content on our English-speaking social media accounts as we listened to community feedback and prioritized building long-term brand sentiment." By the time the FY2025 10-K was filed in February 2026, the company finally said it plainly in the risk factors.

CEO von Ahn then walked back the memo's most controversial pieces twice — first publicly retreating from contractor replacement in May 2025, then scrapping the AI-usage performance-review rule in April–May 2026. Both walk-backs were covered externally; neither got top-line treatment in a shareholder letter.

The FY26 guidance cut (February 2026)

This was handled the way it should have been: pre-announced as a deliberate, strategic decision. The Q3 FY25 letter set the stage ("we're prioritizing user growth over monetization in the A/B tests that get launched"). The Q4 FY25 press release (Feb 26, 2026) bundled the FY26 cut (10–12% bookings vs 33% the year prior) with a $400M buyback authorization and a new CFO (Gillian Munson, replacing Matt Skaruppa, who never appeared by name in shareholder letters despite being CFO since the IPO). The stock fell ~14% on the day, ~22% intraday — investors did not buy the "deliberate" framing.

5. Guidance Track Record

The promises that mattered, plain reading:

• Financial guidance: kept and exceeded, every year. Revenue, bookings, and EBITDA margin beat the initial annual guide for FY22, FY23, FY24, and FY25 — no exceptions. The EBITDA margin path beat initial guide by 200–700 bps each year. Long-term margin guide of 30–35% (set Q4 FY22) is still notionally outstanding but is no longer the headline.
• The DAU promise is where the credibility damage sits. Management spent 2023–early 2025 implying that organic, word-of-mouth growth was structurally durable. They warned in Q4 FY23 that 60%+ was unsustainable, but Q1 FY25 still printed +49% and the Q1 letter projected continued strength. Three quarters later DAU growth was +21% and FY26 was guided to a 10–12% bookings deceleration.
• The AI promise was double-edged. AI as a tailwind (Max, Video Call, content scaling) is real — Max contributes meaningful bookings, content production accelerated ~10× by Q1 FY26. AI as a cost lever has been managed. But AI as a communications topic has been mismanaged: the April 2025 memo was a self-inflicted growth headwind that the company now formally cites in its risk factors.

Credibility: 6 of 10 — financial guidance pristine, communications damaged. Financial guidance has been a clean beat for four consecutive years — that is genuinely above average. Strategic foresight was real on AI (early Max launch, content automation, Video Call) and on multi-subject extension (Chess is the most interesting new product since the IPO). The downgrade is entirely about communication: management let an internal-facing AI memo go viral, then waited until the 10-K to acknowledge it had cost the company users; they kept raising guidance into Q2 FY25 and then cut FY26 hard; and the unannounced CFO transition (Matt Skaruppa to Gillian Munson) arrived in the worst possible market window. A 6 reflects "competent operators we believe on numbers, less on narrative."

6. What the Story Is Now

The current story is a previously-loved compounder mid-reset. The reset is part voluntary (lean into AI; tone down brand voice; sacrifice near-term monetization for long-term DAU) and part involuntary (the AI memo permanently changed the consumer brand-perception risk profile; immigration policy permanently lowered the DET ceiling; the AI-cost path means gross margin trades 200–400 bps lower for now).

What has been de-risked: the P&L. Duolingo is profitable, generates real free cash flow, has $1B+ of cash, just announced a $400M buyback, and has now disclosed honestly that the AI memo cost it users. The financial guidance machine — the most robust thing about the company — has not given investors a reason to disbelieve it for four years.

What still looks stretched: the long-term 30–35% EBITDA margin target, the assumption that DAU growth re-accelerates without further heroics, and the implicit narrative that this was a "deliberate" reset rather than a forced one. Reasonable people can read Q1 FY26's +21% DAU as either the cycle bottom or the new normal — and that uncertainty is what the ~70% drawdown is pricing.

The next two prints (Q2 and Q3 FY26) decide which.

Financials - What the Numbers Say

Duolingo crossed $1B in annual revenue in FY2025, with operating margin lifting to 13.1% from negative territory just two years earlier and free cash flow reaching $370M (35.6% FCF margin). The reported FY2025 net income of $414M, however, is inflated by a one-time release of the deferred-tax valuation allowance worth ~$232M; on a pretax basis the business earned $182M, which is the cleaner number to underwrite. Cash conversion is excellent (FCF is 0.95x operating cash flow), the balance sheet holds $1.05B of net cash with effectively no debt, and FY2025 returns on invested capital cleared 50%. The valuation tells a different story than the income statement: after a 65% drawdown from the FY2024 close, DUOL now trades at ~3.8x EV/Revenue and ~17x forward earnings — multiples that price in the recent deceleration (revenue growth slowing from 41% to 27% year on year and bookings guidance stepping down to ~10.5% for 2026) rather than the underlying margin expansion. The single financial metric to watch is bookings growth, which leads revenue by two quarters and is the line analysts have used to reset price targets.

Financials in One Page

Reading guide for financial scores you may have seen elsewhere: Quality Score is a 0-100 composite of profitability, financial strength, and growth durability; Fair Value is a model estimate of what the stock is worth on its history of returns and growth. Neither is currently available in our dataset for DUOL, so this page leans on first-principles ratios and peer multiples instead.

Revenue, Margins, and Earnings Power

Revenue compounded at 71% CAGR from FY2019 to FY2025, accelerating after the Q4 2022 generative-AI scare and through the 2023-2024 "Family Plan + Super Duolingo" subscription cycle. The shape of the income statement flipped in 2023-2024: gross margin sits steadily in the low-70s (typical of software with App Store and AWS costs in COGS), while operating margin moved from -23.9% in FY2021 to +13.1% in FY2025 — almost entirely on operating leverage over S&G and R&D, not gross-margin gains.

The FY2025 net margin spike to 39.9% is a tax-driven optical illusion. Operating margin — the line management actually steers — moved from 8.4% in FY2024 to 13.1% in FY2025, a +4.7pp expansion. R&D as a percent of revenue fell from 31.4% to 29.5%, and S&G from 33.0% to 29.6%; gross margin gave back 50 bps as AI inference costs rose. Earnings power is still expanding but at a slower pace than 2023→2024 (+11pp).

This is the chart the sell-side reacted to in February and May 2026: revenue growth held near 40% through Q3 FY2025, decelerated to 35% in Q4, then stepped down to 27% in Q1 FY2026. Management framed 2026 as a "transition year" with deliberate prioritization of daily-active-user growth over near-term monetization, which is consistent with the gap between still-strong revenue and softer bookings.

Cash Flow and Earnings Quality

Free cash flow is cash from operations minus capital expenditures — the cash the business produces after running and growing itself, before paying lenders or shareholders. For Duolingo, FCF turned positive in FY2020 and has compounded faster than revenue ever since.

For FY2019-FY2022, net income was deeply negative while operating cash flow was positive — the classic signature of a subscription software business where the cash from annual prepayments lands before the revenue is recognized. That dynamic still helps: in FY2025 cash from operations of $388M was driven by $137M of stock-based compensation (a non-cash expense) plus working-capital tailwinds from deferred revenue. Capex stays light at ~1.7% of revenue.

Major cash-flow distortions in FY2025

Balance Sheet and Financial Resilience

Duolingo carries no traditional debt — the $93M of liabilities classified as "debt" in the ratios file is operating-lease obligations on its Pittsburgh and global offices. Cash and equivalents reached $1.14B at year-end FY2025, retained earnings turned positive for the first time at +$288M, and total equity stands at $1.35B with only $35M of goodwill on the balance sheet.

The decline in current and quick ratios from 5x to 2.6x is a function of growing deferred-revenue balances (Super subscriptions sold but not yet recognized), not balance-sheet weakness — they remain firmly above 2x. Interest expense is immaterial. The balance sheet gives management complete strategic flexibility: a year of negative FCF could be absorbed without raising capital.

Returns, Reinvestment, and Capital Allocation

Returns on capital crossed into structural-compounder territory in FY2025. Return on invested capital reached 52%, return on equity 38%, and return on assets 19%. These would be world-class for any business; for one that was burning cash three years ago, they are an unusual signal of operating leverage finally arriving.

The bulk of capital "spent" by Duolingo is stock-based compensation, not cash. Buybacks have been minimal and largely offset employee-withholding mechanics. Diluted share count has crept from 39.5M (FY2022) to 48.3M (FY2025) — a roughly 7% increase over three years, or ~2.4% per year, which is at the lower end of what large-cap software peers absorb but is real dilution that the FCF-per-share figure has to outrun.

The verdict on capital allocation: management is reinvesting almost everything operating cash flow generates back into R&D and content (AI tutoring, new courses, the English Test). With ROIC over 50%, that is the right call. The open question is whether ROIC stays high once growth normalizes — historically, software ROIC compresses as growth slows.

Segment and Unit Economics

Duolingo does not report multiple operating segments in its public filings; the company operates as a single reportable segment with revenue split between subscriptions (~80%+ of revenue), advertising (low-double-digit %), Duolingo English Test, and other in-app purchases. Geographic detail is also limited in the financial statements. The most useful unit-economic disclosure management provides — and the lever the stock now trades on — is the relationship between Daily Active Users, Monthly Active Users, Paid Subscribers, and Bookings.

Public disclosure summarized from FY2024 and FY2025 shareholder letters:

Bookings (cash collected for subscriptions, recognized over the subscription term) is the leading indicator analysts now anchor on. The gap between resilient DAU growth (~21%) and slowing bookings growth (~10.5% guide for 2026) is the central debate: is this AI-driven engagement substitution that will eventually monetize, or a structural ceiling on how much language learners will pay?

Valuation and Market Expectations

The stock traded between $87 and $540 over the last 52 weeks — a 6x range. As of May 18, 2026 at $113.24, market cap is ~$5.28B, enterprise value ~$4.23B after netting $1.05B of net cash.

The valuation reset is real: EV/Sales went from 18.4x at the FY2024 close to ~3.8x today, and on a trailing P/E basis Duolingo went from 648x (FY2023, when GAAP net income was just $16M) to 173x (FY2024) to 20.5x (FY2025) to 17x on forward consensus EPS — a multiple that almost no one would have written down in early 2025. Two interpretations:

Bear: The 27% Q1 revenue growth plus 10.5% bookings guide imply the long-term steady-state growth rate is mid-teens, not 30%+. At that growth level a mid-teens P/E is fair, not cheap.

Bull: The deceleration is a deliberate, management-controlled investment cycle (AI tutoring rollout, DAU push), not a competitive break. Pretax earnings of $182M FY2025 plus continued operating leverage suggests an underlying earnings power closer to $250-300M in 2026-2027, putting the true forward P/E in the 18-22x range — reasonable for a 15%+ grower with 35% FCF margins.

Consensus on Bloomberg-style trackers shows analyst targets clustering between $90 and $275, a wide spread reflecting genuine disagreement on whether 2026 marks a strategic pause or a structural deceleration. Recent cuts (JP Morgan to $95, Morgan Stanley to $95, Evercore to $97, Citi to $101) cluster well below the older $250-$300 prints from 2025.

Peer Financial Comparison

Duolingo is, by quite a margin, the most profitable of the consumer-engagement set on both operating margin and FCF margin, with ROIC that is the highest in the table. Yet on EV/Sales it sits between Stride (a slower-growth education services company) and Spotify (a comparable consumer-subscription brand). Roblox, with similar revenue growth and worse operating margin, trades at 11.4x EV/Sales versus DUOL's 3.8x — a gap explained by Roblox's larger TAM narrative and its much heavier SBC (23% of revenue dilutes future EPS far more aggressively than DUOL's 13%). Chegg is the cautionary tale: revenue down 39% year-over-year as generative AI cannibalized its homework-help product, and the stock at 0.1x EV/Sales. The market is currently pricing DUOL somewhere between "the next Spotify" and "the next Chegg," with weight shifting toward the latter after the May 2026 bookings guide.

What to Watch in the Financials

What the financials confirm: Duolingo is now a genuinely profitable, cash-generative consumer-subscription business with an unleveraged balance sheet, sector-leading FCF margins, and returns on capital that put it in the top decile of US software. The compounder thesis is no longer aspirational — FY2025 proved it on three out of four lines (gross margin held, operating margin expanded, cash flow scaled).

What the financials contradict: The narrative that growth is durable in the 30-40% range is breaking down. Q1 FY2026 revenue growth of 27% and bookings guide of 10.5% for the year are not consistent with the prior trajectory. Either bookings re-accelerate in 2H 2026 as the AI-tutoring features land, or the long-term steady-state growth rate is materially lower than recently assumed — in which case further multiple compression is the path of least resistance.

The first financial metric to watch is bookings growth — specifically, whether the FY2026 print exceeds the 10.5% management guided in May, and whether quarterly bookings re-accelerate in the back half of 2026 as new AI-tutoring features roll out. Revenue follows bookings by one to two quarters; analyst price targets follow bookings by about ten minutes.

Web Research — What the Internet Knows

The Bottom Line from the Web

The filings show Duolingo's KPIs; the web shows that the market lost faith in management's narrative over a 90-day window. Between November 5, 2025 (Q3 print) and February 27, 2026 (Q4 print), DUOL fell from a high of ~$540 to $101, the consensus 12-month price target collapsed from $300+ to $95, and at least six law firms opened securities-fraud investigations alleging undisclosed DAU deceleration. The deferred-tax valuation-allowance release booked $256.7M of one-time GAAP income into FY 2025 — without it, "net income" would have been about $157M, not $414M.

What Matters Most

1. Six law firms are investigating securities-fraud claims; the alleged class period centers on undisclosed Q2 2025 DAU deceleration

Levi & Korsinsky, Pomerantz LLP, Portnoy Law Firm, Faruqi & Faruqi, Bronstein Gewirtz & Grossman, and Johnson Fistel have all opened investigations. The Levi & Korsinsky trigger was a JMP Securities note (July 28, 2025) showing third-party DAU growth decelerated from ~51% YoY in Q1 2025 to ~39% in Q2 2025, after which DUOL fell over 6% in a session. The Faruqi/Pomerantz/Portnoy investigations all anchor on the Feb 26, 2026 disclosure that triggered a $16.45 (14.01%) one-day decline. As of the most recent web evidence, all six probes are at the "pre-litigation investigation" stage — no consolidated complaint has been confirmed in the extracted sources.

Sources: zlk.com/cases/duolingo, Pomerantz via Morningstar/PRNewswire 2026-03-05, Faruqi 2026-04-02 GlobeNewswire, Reddit megathread.

2. The $256.7M deferred-tax valuation-allowance release inflated FY 2025 GAAP net income by ~2.6x

FY 2025 reported net income was $414.1M, of which $256.7M was a one-time, non-cash income-tax benefit from releasing the valuation allowance against federal and state deferred tax assets ($222.7M was booked in Q3 2025 alone). Excluding the release, normalized FY 2025 net income would have been roughly $157M. The disclosure is technically standard — the cumulative three-year income test was satisfied for the first time — but the optical lift in headline GAAP earnings sits awkwardly alongside the FY 2026 guidance for a margin step-down from 29.5% Adj EBITDA margin to ~25%.

Sources: Q3 2025 shareholder letter, Q4/FY 2025 shareholder letter.

3. The Audit Committee Chair became CFO three days before the worst guidance miss in company history

On January 12, 2026, Duolingo announced that Matt Skaruppa (CFO since 2020) would step down. The replacement: Gillian Munson, the sitting Audit Committee Chair and a board member since September 2019. Effective date: February 23, 2026. Q4 2025 results were reported February 26, 2026 — meaning Munson was CFO for three calendar days before announcing 2026 guidance that triggered a 22% intraday decline. The CFO-from-Audit-Chair pathway is unusual: it leaves the board without a designated financial-expert director until the seat is refilled, and proxy advisors typically scrutinize such transitions for whether they were planned succession or pressure-driven.

Sources: Duolingo IR press release, Reuters 2026-01-12, CFO Dive.

4. The Feb 27, 2026 analyst purge — 12-month targets cut by 50–70% across the bulge bracket

Within 24 hours of the Q4 2025 print, the major banks slashed price targets in lockstep: JPMorgan Overweight→Neutral, $200→$95; Morgan Stanley Overweight→Equalweight, $245→$100; Evercore ISI (Mark Mahaney) Outperform→In Line, $330→$114; Truist Buy→Hold to $100; Scotiabank Sector Outperform→Sector Perform, $300→$100; UBS held Buy but cut to $125 from $285. Current consensus price target sits near $95, implying ~16% downside to the prevailing $113 share price.

Source: marketscreener consensus log, Benzinga ratings.

5. Short interest spiked to 20.2% of float (March 13, 2026) — bears positioned aggressively before Q1

Per ad-hoc-news (March 25, 2026): "Short interest has climbed to 20.21% of the public float as of March 13, 2026, up 9.28% from the prior month… 7.63 million shares, days-to-cover at 2.2." This is an exceptionally high short ratio for a $5B large-cap consumer-internet name and explains part of why the Q1 2026 beat (May 4) did not produce a relief rally — the position was already crowded.

Source: ad-hoc-news.de 2026-03-25.

6. The April 2025 "AI-first" memo backlash is now formally cited in the 10-K as a contributor to user-growth deceleration

The chronology, per multiple sources: late-2023 "AI-first" strategy → Jan 2024 Washington Post report on ~10% contractor workforce reduction → April 2025 internal memo and external press attention (htxt.co.za, PCMag, Entrepreneur) → social-media backlash → May 25, 2025 PCMag "Amid Backlash, Duolingo Backtracks on Plans for AI Pivot" → Aug 6, 2025 Q2 call: CEO Luis von Ahn explicitly attributes lower-end DAU growth to the memo: "I said some stuff about AI, and I didn't give enough context. Because of that, we got some backlash on social media." → Sept 17, 2025 CNBC "How AI makes my employees more productive without layoffs" (walkback campaign) → Feb 26, 2026 Q4 call: "we've probably gotten in trouble for adopting AI fast, for example, by writing stupid memos that I wrote one time."

Sources: Wikipedia, PCMag 2025-05-25, CNBC 2025-09-17, Q2 2025 transcript via fool.com.

7. Q1 2026 beat the numbers but the stock still fell ~14%

Q1 2026 (reported May 4, 2026): Revenue $292.0M (+27% YoY), bookings $308.5M (+14%), DAUs 56.5M (+21%), paid subs 12.5M (+21%), Adj EBITDA $83.4M (28.6% margin), FCF $147.8M, cash $1.139B. GAAP EPS $0.89 beat by $0.14; revenue beat by $3.4M. Despite the beat, the stock fell ~14% the next session as the reaffirmed (rather than raised) full-year 2026 guidance forced the market to price the slowdown as structural rather than transitory. DAU growth at 21% is the slowest since the 65% print in Q4 2023 and the 30% print in Q4 2025.

Source: Q1 2026 8-K, Zacks 2026-05-13.

8. The 85% global DAU share claim is the moat — but it relies on category definition

From the Q4 2025 earnings call (Feb 26, 2026): "Duolingo holds about 85% share of the global daily active user base in language learning apps, minimizing near-term competitive threats." Sensor Tower data corroborates the dominance (Duolingo 28% churn vs Babbel 58%, Simply Piano 64%, Brainly 35%). But the framing — "language learning apps" — excludes ChatGPT-style free LLM tutors, which is the bear case repeatedly flagged by Motley Fool, Seeking Alpha (Ian Bezek, Jan 12, 2026), and the originating Chegg-substitution analogy.

Sources: Q4 2025 transcript, Sensor Tower, Motley Fool 2026-03-31.

9. Insider activity skews to selling; one director bought near the lows

Pre-pivot (July–August 2025): Severin Hacker (co-founder, CTO) sold 10,000+ shares at $339–$370; Matt Skaruppa sold ~2,522 shares; Natalie Glance sold ~2,300 shares. Post-pivot: Seeking Alpha reported on March 5, 2026 that a director bought 5,000 shares worth ~$499K — at least one inside buyer near the bottom. Munson received a 133,753-share option grant on March 25, 2026 as her sign-on award.

Sources: secform4.com, TIKR, Seeking Alpha 2026-03-05.

10. The Duolingo English Test is an under-appreciated optionality leg — and the UK Home Office contract is live

DET went from ~500 accepting institutions in 2019 to over 6,000 institutions worldwide by 2025 per third-party analyst commentary. The Telegraph (April 21, 2026) reports Duolingo is in the running to test migrant English skills for the UK Home Office — a sovereign-government testing contract would be a category-changing reference customer. CEO commentary in Q2 2025 acknowledged DET revenue was below expectations due to fewer international university applications.

Source: The Telegraph 2026-04-21, alphatalon Substack.

Recent News Timeline

What the Specialists Asked

Governance and People Signals

CFO transition: the most material people event

The CFO transition is the single highest-conviction governance signal in the web record. The board lost its sitting Audit Committee Chair three days before the worst guidance event in company history, replaced its CFO of nearly six years on a six-week timeline, and has not (per extracted sources) publicly named a new Audit Committee Chair. Proxy-advisor reaction will likely surface in the 2026 proxy season.

Insider activity skews to selling pre-crash

Dual-class voting power

Class B shares (20 votes each) are held only by founders Luis von Ahn and Severin Hacker. As of September 30, 2024, affiliated stockholders held 78.8% of total voting power. No sunset trigger was surfaced in the extracted sources — this is a permanent founder-control structure.

Say-on-pay (2025 AGM)

98.8% of votes cast were in favor of Say-on-Pay (154.3M for / 1.9M against). Board re-elections passed with small withhold rates. ISS/Glass Lewis-specific recommendations not surfaced in the web record.

Industry Context

The most material industry shifts visible in the web record beyond what the filings cover:

App-store economics are slowly improving in the consumer's favor. Apple cut China App Store commissions after government pressure (March 13, 2026). The EU DMA and Epic ruling created precedents that have already affected pricing in some regions. Duolingo has not quantified the web-vs-IAP billing mix, leaving 200–400 bps of potential gross-margin upside unmodeled by sell-side.

AI-tutor substitution remains the open bear case. Despite management's repeated assertion that user churn is driven by "personal busyness" rather than ChatGPT, neither management nor any third-party analyst has published a retention-vs-LLM cross-tabulation. This is the single biggest unresolved variant on the thesis.

The Duolingo English Test institutional adoption curve (500 → 6,000+ institutions, 2019–2025) is among the most underdiscussed bull catalysts. A UK Home Office migrant English-skills contract win (tender live as of April 2026) would re-rate DET as a separate franchise. The CEO has publicly noted DET revenue is below internal expectations due to weak international student application volumes — i.e., the franchise is growing through a macro headwind, not because of one.

Category definition matters. Duolingo's "85% global DAU share" claim defines the market as language-learning apps, excluding free LLM tutors. The market's de-rating of the stock implies investors are using a broader definition.

Web Watch in One Page

The Duolingo thesis turns on whether the behavioural moat survives the generative-AI substitution cycle while the bookings reset proves transitional. Five active watches cover the load-bearing variables outside the quarterly calendar. The first follows paid-conversion durability — the single leading indicator that distinguishes a freemium compounder from a Chegg-style monetisation erosion. The second tracks a live sovereign-government tender (UK Home Office migrant English testing) that would re-rate the Duolingo English Test as a standalone franchise. The third watches whether free large-language-model providers expand directly into language tutoring, the structural risk the 10-K self-incrimination already named. The fourth follows Apple and Google in-app-purchase fee policy and DMA-style enforcement that directly drives ~28% of cost of revenue. The fifth tracks whether the six open pre-litigation probes into 2025 user-growth disclosure consolidate into a federal class action that would force reserve disclosure and re-anchor the management-trust discount. Together they cover the long-duration thesis test (paid conversion), the asymmetric optionality (DET), the structural failure mode (LLM substitution), the margin lever (platform fees), and the overhang (legal/governance).

Active Monitors

Rank	Watch item	Cadence	Why it matters	What would be detected
1	Paid % of MAU and subscriber-vs-DAU growth gap	Daily	Single leading indicator of whether the freemium funnel survives free-LLM substitution; the durable thesis breaker is two consecutive quarters of paid % flat-or-down YoY	Quarterly KPI disclosures, sell-side panel notes (JMP, Cleveland, AppFigures), Sensor Tower / data.ai updates, management commentary on paid attach and renewal cohorts
2	UK Home Office migrant English-test tender decision	Daily	First sovereign-government adoption of the Duolingo English Test would re-rate DET as a separately-disclosed franchise; £816M live tender with zero embedded value in consensus	UK Home Office award announcements, shortlist updates, contract scope, competitor bids, security-review outcomes
3	Free large-language-model substitution in language learning	Daily	The structural failure mode the FY25 10-K already named — ChatGPT, Gemini, or Claude expanding into dedicated language tutoring is the canonical Chegg-substitution vector	New language-learning features from frontier LLM providers, AI-native tutoring apps gaining material traction, ChatGPT category crossover in mobile education rankings, comparative engagement studies
4	Apple and Google in-app-purchase fee policy and DMA enforcement	Daily	~28% of FY25 cost of revenue is platform fees; every 5pp shift = ~3-4pp gross margin; the FY26 margin guide is one regulatory decision away from re-rating either way	New Apple/Google commission policies, EU Digital Markets Act enforcement orders, US Open App Markets bill progress, web-billing carve-outs that would let Duolingo bypass IAP
5	Securities class-action consolidation and governance escalations	Daily	Six pre-litigation probes are anchored on undisclosed Q2 2025 DAU deceleration; consolidation into a federal complaint would force reserve disclosure and re-anchor the management-trust discount	Federal-court complaints with class periods covering 2025 statements, lead-plaintiff motions, settlement disclosures, 8-K Item 8.01 filings, proxy-advisor commentary on the CFO-from-Audit-Chair transition

Why These Five

The report identifies one load-bearing thesis variable (paid % of MAU), one live binary catalyst (UK DET tender), one structural failure mode (LLM substitution), one margin-and-cash lever (platform fees), and one durable overhang (litigation consolidation plus governance signals). Each monitor maps to a specific 5-to-10-year underwriting driver or failure mode rather than to the next earnings date. The set is deliberately built around evidence that arrives between the quarterly prints, where the multiple actually moves before consensus updates — paid-conversion panel data lands weeks ahead of earnings, the UK tender will be decided outside any earnings window, platform-fee rulings come from regulators on their own schedule, and class-action complaints land when the plaintiffs' bar is ready, not when Duolingo reports. The two report-rated highest-impact items — paid conversion durability and the UK sovereign tender — anchor the top of the ranking; the structural risks (LLM substitution and platform-fee shifts) fill the middle; the legal/governance overhang sits last because it widens the discount rather than re-rating the underlying business.

Where We Disagree With the Market

The market is pricing Duolingo as the Chegg analog — but Duolingo's own paid-conversion data is moving in the opposite direction. Consensus has compressed EV/Sales from 18.4x at the FY2024 close to 3.8x today, anchored a $95 12-month price target against a $113 print, and built a 20.2%-of-float short position around the assumption that the FY25-to-FY26 bookings cut (33% → 10.5%) is the start of a structural deceleration driven by free-LLM substitution. The operational fingerprint of that substitution thesis is paid % of MAU falling while DAU holds — exactly what happened at Chegg. Duolingo's paid % of MAU has done the opposite, rising from 8.8% to 9.3% over four quarters while paid subscribers grew +28% YoY. Three additional gaps compound the mis-framing: the headline FY26 bookings number is mechanically distorted by deferred-revenue base effects and an amortizing April-2025 brand event, the founder-PSU-plus-$400M-buyback structure is uniquely shareholder-friendly in a way the SBC-discount narrative is not crediting, and a live UK Home Office DET tender carries near-zero embedded option value. None of these views requires the company to be cheap on owner earnings; the variant is that the direction of the leading indicator is wrong in consensus.

Variant Perception Scorecard

The scorecard is deliberately uneven. Consensus clarity is high because every market signal points the same way — bulge-bracket price targets clustered $94-100, FY26 consensus revenue exactly at the guide ($1.21B), forward P/E 17x, short interest 20.2%, six law-firm investigations, and a 14% drop on a 27% Q1 revenue beat. Evidence strength is materially lower because the central variant claim rests on a one-quarter trajectory in paid % of MAU and a forward-looking read on bookings reacceleration that requires Q2/Q3 FY26 prints to confirm. Variant strength sits in the middle: the disagreement is specific and observable, but the highest-conviction signal that would settle it is two quarters away.

Consensus Map

Five debates explain virtually all of the current $113 price. The market's signal cluster is unusually tight — bulge-bracket targets within $94-100, consensus revenue exactly at the guide, and short-interest positioning that requires the substitution thesis to validate. The disagreement is not that any single market view is wrong on direction; it is that the leading indicators inside the company's own KPIs already contradict views 1 and 2, and views 3 and 5 are quantifiably mis-weighted.

The Disagreement Ledger

Disagreement #1 — the Chegg framing fails its own diagnostic test. A consensus analyst defending the substitution thesis would say: the 78% drawdown reflects the right read of generative-AI risk, the 23-point bookings cut is the same vector that hollowed Chegg, and management's "stupid memos" walkback confirms the brand is mood-sensitive. Our disagreement is that the operational signature of Chegg substitution is paid % of MAU falling while DAU holds — and Duolingo's paid % is moving in the opposite direction over the same four-quarter window the market has compressed the multiple. If we are right, consensus has to concede that the substitution thesis as priced does not survive the company's own quarterly KPI page. The cleanest disconfirming signal is a single Q2 FY26 print of paid % of MAU flat or down YoY — that converts our view to wrong inside one quarter.

Disagreement #2 — bookings is the wrong denominator at the wrong horizon. The consensus analyst's strongest counter is that bookings leads revenue by 6-12 months on a 12-month subscription book, so the FY26 cut is the FY27 revenue. Our disagreement is narrower: the FY26 bookings figure is mechanically depressed by three identifiable distortions (deferred-revenue base effect, FX, brand-event echo), and paid subscriber count (+28% YoY) is the cleaner read of the recurring monetization engine. If we are right, FY27 bookings will print materially above the implied 8-10% extrapolation, forcing sell-side to re-underwrite the structural-deceleration thesis. The disconfirming signal is a Q3 FY26 bookings print below Q2 — two consecutive quarters at the trough.

Disagreement #3 — the buyback mechanic is uniquely shareholder-friendly and not in the multiple. A consensus analyst would argue the $400M buyback is just SBC offset and that diluted share count will continue creeping above 50M as SBC scales to 15% of revenue. Our disagreement is that the founder PSU framework mandates zero new CEO/CTO equity grants through 2031, so the buyback shrinks share count permanently rather than refilling the comp pool — an unusual structure for a software business. Combined with $1.05B net cash (22% of market cap) and CEO total comp of $768K, the per-share compounding trajectory is materially better than peer SBC math implies. The disconfirming signal is cumulative buyback below $150M by Q3 FY26 with diluted shares creeping above 49.7M — that would tell us management is preserving optionality (M&A) rather than returning capital.

Disagreement #4 — DET option value is currently zero in the price. The consensus analyst would point out that DET has run below management's internal expectations on weak international university applications, and that TOEFL/IELTS dominate volume by ~3x. Our disagreement is bounded: we are not arguing DET re-rates the whole business on probability-weighted basis; we are arguing the UK Home Office tender is a live binary with no consensus attribution, and the asymmetry favors a positive surprise. The disconfirming signal is an award to TOEFL/IELTS or a UK-domiciled bidder — that closes the binary at zero, which is already in the price, so the downside is bounded.

Evidence That Changes the Odds

The strongest two evidence items (paid % of MAU trajectory and the paid-sub-vs-DAU growth gap) refute the consensus framing directly inside the company's own quarterly disclosure. Items 3-5 establish the moat substrate that makes the Chegg analog wrong; items 6-7 bound the downside legs of the bear thesis; item 8 disciplines the bull case by recognizing valuation is fair-not-cheap on owner earnings. The fragility column is where the variant view can break — every claim depends on one or two quarterly prints continuing the current trajectory.

How This Gets Resolved

Seven signals together cover every leg of the variant view, but only signals 1, 3, and 5 are decisive for the long-term underwriting case the variant rests on. Signal 1 is observable in 78 days (Q2 FY26 on August 5); signal 3 in 170 days; signal 5 within the next two quarters. Signal 1 is the single most important — it is the leading indicator the consensus is not weighting against the lagging bookings figure, and one print of paid % of MAU flat-or-down YoY would force the variant view to retreat. Signals 2, 4, 6, and 7 are near-term confirmations rather than long-term thesis updates; treat them as confirming evidence rather than thesis breakers.

What Would Make Us Wrong

The strongest argument against the variant view is that paid % of MAU at 9.3% is the trough, not the trend. The Chegg substitution mechanic does not play out in two quarters — it took FY22-FY25 for Chegg to lose two-thirds of its paid base, and the operational signature was visible in renewal-cohort retention well before it showed up in net-add growth. Duolingo does not disclose renewal-cohort retention data. The paid-sub +28% YoY figure is gross of churn, not a clean read of cohort durability. If the trade we are pushing back on is right about the substitution direction, a Q2 FY26 paid % of MAU print holding at 9.3% would be the highest-conviction near-term refutation of our variant — but it would not invalidate the Chegg analog over the longer 24-month horizon the bear case requires. The honest read is that one print is necessary but not sufficient.

The second-strongest objection is that management credibility has been spent in a way the variant view discounts too cheaply. The CFO swap from Audit Chair in 7 weeks, the "stupid memos" walkback, the MAU demotion concurrent with deceleration, the twice-revised Adj EBITDA/FCF definitions, and the six pre-litigation probes are not separately decisive but cumulatively raise the underwriting bar. If the variant requires a market re-rating, that re-rating depends on the same management team to deliver paid-conversion and bookings prints that sell-side will trust — and there is a reasonable read in which sell-side has decided the management discount is structural rather than cyclical. A class-action consolidation event in the next 6-12 months would lock that discount in regardless of the operating prints.

The third objection is that the variant rests on a base effect we cannot independently verify in real time. The disagreement #2 claim that FY26 bookings growth is mechanically distorted by deferred-revenue amortization, FX drag, and brand-event echo is structurally defensible, but the size of each distortion is not separately disclosed. Management has not published a "normalized" bookings figure stripping the brand event. The bear can fairly argue the variant view is post-hoc rationalization of an unfavorable headline — and that the right way to underwrite a 23-point bookings cut is to take the headline at face value until two prints disprove it.

The fourth objection is that the founder PSU framework expires in July 2031, and the variant attributes structural value to a window that is itself finite. After 2031, a new pay package has to be designed, and any framework that re-introduces ordinary-course equity grants would unwind the "buyback shrinks share count" mechanic the variant relies on. We are recognizing a current advantage with a defined expiration; the bear is right that the post-2031 structure is undefined.

The first thing to watch is the Q2 FY26 paid % of MAU print on August 5, 2026 — if it lands at 9.3% or higher, the Chegg framing has failed its diagnostic test inside the company's own disclosure, and consensus has 60-90 days to start re-underwriting before the Q3 FY26 print on or around November 5 either confirms or kills the variant view.

Liquidity & Technical

Duolingo trades a deep, liquid tape: roughly $215M changes hands every day, so a 5% position is implementable for funds up to about $4.6B inside a five-day window — liquidity is not the constraint here. The constraint is technical: the stock is in a structural downtrend (price sits 42% below its 200-day average), but a violent relief rally is forming, with RSI recovering from a sub-25 capitulation low and MACD freshly back above zero.

1. Portfolio implementation verdict

2. Price snapshot

3. Price history with 50- and 200-day moving averages

Price is below the 200-day moving average by 42% — a deep, mature downtrend. A death cross fired on 2025-08-26 after the prior golden cross from 2024-09-27 expired. The stock ran from roughly $170 in mid-2024 to a $530 peak in May 2025, then unwound the entire move in nine months. The most recent two months show price reclaiming the 50-day from below — the first technical positive on the chart since the August death cross.

4. Relative strength

5. Momentum — RSI and MACD

Near-term momentum has flipped from extreme oversold into a fragile bullish reading. RSI bottomed at 20.4 on 2025-11-17 (peak capitulation around the November earnings reaction) and printed a second sub-25 low in February 2026 — a higher low in price but a lower-low echo in RSI, which is the classic positive divergence pattern. The MACD histogram has been positive for thirteen consecutive weekly samples, with the current print at +0.42. This is the kind of base that often precedes a multi-month relief rally, but the absolute readings are still modest — momentum is no longer falling, not yet thrusting.

6. Volume, sponsorship and volatility regime

50-day average daily volume has tripled from roughly 0.9M shares in May 2025 to peaks above 2.9M in early 2026 — that is the signature of institutional repositioning, not retail. Average daily traded value is now near $215M, which is unusually deep for a $5.5B-cap name.

The three highest-conviction days in the last twelve months are all earnings-window reactions: two crushing down-days (a −25.5% print on 2025-11-06 and a −14.0% print on 2026-02-27) and one explosive up-day (+13.8% on 2025-08-07 — a high-water mark just before the death-cross unwind began). The fact that the two biggest down-days came on 7-plus times average volume is unambiguous: institutions distributed into both releases, not retail. Catalyst attribution is left blank because primary-source news is not staged for this run.

Volatility bands (5-year history): calm below 43%, normal 43–79%, stressed above 79%. The current 30-day realized at 56.7% sits in the upper half of the "normal" range — sharply down from the 94% November spike but still elevated. The November 2025 print is the third-highest 30-day reading in the stock's listed history. Volatility is normalizing, not normalized.

7. Institutional liquidity panel

This is the case where the tape is the easy half of the decision. ADV runs near $215M, market cap is $5.47B, and annual turnover is 979% — i.e., the entire float changes hands nearly ten times a year. That kind of churn is rare and reflects how much the name has been traded around macro and earnings inflection points.

A. Average daily volume and turnover

B. Fund-capacity table

For each ADV participation rate, how big a five-day position is feasible and what fund AUM that supports at common portfolio weights:

At a comfortable 10% ADV participation rate, this name supports a 5% position for funds up to $2.3B. At a more aggressive 20% participation, the same 5% position is implementable up to $4.6B AUM. A 2% portfolio weight is feasible all the way out to $11.4B AUM — institutional in the truest sense.

C. Liquidation runway

D. Execution friction

Median daily price range over the trailing 60 sessions is 2.49% — meaningfully above the 2% threshold, so a single-day market order on a large block will eat noticeable impact cost. Combine with the elevated realized vol and the practical rule for this name is: VWAP across the day, not market on close, and never on earnings windows.

Bottom line on liquidity: the largest size that clears 5 days at 20% ADV participation is 2.0% of market cap ($109M), and at the more conservative 10% participation it is also 2.0% but takes a full week. Liquidity is decisively not the constraint — a fund can act in this stock at any normal institutional position size.

8. Technical scorecard and stance

Stance: neutral with a bearish skew on a 3- to 6-month horizon. The structural picture is broken — a primary downtrend with price 42% below the 200-day moving average does not normally resolve in a clean V-bottom — but the near-term tape says capitulation is most likely behind us. Position-sizing matters more than direction here.

The two levels that decide it:

Bull confirmation: a sustained weekly close above $196 (the 200-day SMA). That would force the death cross to mechanically reverse and reestablish the multi-year uptrend; combined with the standing positive divergence in RSI, the relief rally would convert into a genuine trend reversal.

Bear reaffirmation: a daily close below $89.71 (the 52-week low). Loss of the November-to-February double bottom would expose the IPO-vintage $61 all-time-low zone and confirm that the bounce off February's $101 capitulation was a bear-market rally, not a base.

Liquidity is not the constraint — it is the cleanest part of the setup. A fund considering an entry can build a 2–5% position over a week without market impact, which means the right action is to watchlist and stage: do nothing above $130 (still vulnerable to a re-test of $90), begin scaling on either a clean break above $130 with above-50d volume or a re-test that holds the $90 area, and reserve full size for confirmation above the 200-day at $196.