Duolingo lost its streak
58.7 million people open the app every day. Bookings grew 8%.
Duolingo closed at $146.13 on 21 August. That is simultaneously a good number and a dreadful one. It is up roughly 29% from where the stock sat in the days around its second-quarter results on 5 August. It is also about 73% below the $540.68 it closed at on 14 May 2025.
What makes $DUOL worth an afternoon is not the drawdown. Plenty of 2021-vintage software names have done worse. It is that the business did not break while the stock did. Revenue in Q2 2026 grew 18% to $298.5 million. Daily active users grew 23% to 58.7 million. The company passed a billion dollars of annual revenue last year. The app is still the top-grossing title in the Education category on both app stores.
The whole argument sits in one comparison, and it is right there on page two of the shareholder letter. Users grew 23%. Total bookings grew 8%. Everything below is an attempt to work out which of those two numbers is telling the truth about the next three years.
ℹ
Where these numbers come from
Every figure in this piece is taken from Duolingo's Q2 2026 shareholder letter and the unaudited condensed consolidated financial statements filed with it on 5 August 2026, or from dated market data. Anything we have calculated ourselves is labelled as derived.
1. The shape of the fall
Duolingo's peak was 14 May 2025, at a closing price of $540.68. The break came on 26 February 2026, when the company reported a strong fourth quarter — revenue of $282.9 million, up 35% and ahead of estimates — and then guided 2026 bookings growth to 10–12%, against the 24% it had just posted. The stock fell hard the following day and kept going. Its 52-week low is $87.89.
Since then it has clawed back a lot of ground. The Q2 shareholder letter records the share price at 30 June as $115.02. By 7 August it closed at $130.90, by 18 August $139.64, and by 21 August $146.13.
Where the stock has actually traded
| Date | Closing price | Note |
|---|---|---|
| 14 May 2025 | $540.68 | All-time closing high |
| 27 Feb 2026 | ~$101 | Day after FY26 guidance |
| 30 June 2026 | $115.02 | As disclosed in the Q2 letter |
| 7 Aug 2026 | $130.90 | Two days after Q2 results |
| 18 Aug 2026 | $139.64 | |
| 21 Aug 2026 | $146.13 | Most recent close at time of writing |
| 52-week low | $87.89 | As reported 20 Aug 2026 |
$DUOL — two years
The full round trip: the May 2025 peak, the February 2026 guidance break, and the rebound since the Q2 print.
So this is not a story about a stock that is still falling. It is a story about a stock the market has already repriced twice — once violently downward on a guidance change, and once upward on a quarter that beat. The question is what it has repriced to.
2. Users up. Money up rather less.
Here is the quarter as the company reported it.
Duolingo Q2 2026 versus Q2 2025
| Metric | Q2 2025 | Q2 2026 | Change |
|---|---|---|---|
| Daily active users | 47.7M | 58.7M | +23% |
| Monthly active users | 128.3M | 140.6M | +10% |
| Paid subscribers | 10.9M | 12.7M | +17% |
| Subscription bookings | $227.3M | $250.3M | +10% |
| Total bookings | $268.0M | $289.1M | +8% |
| Revenue | $252.3M | $298.5M | +18% |
| Gross margin | 72.4% | 72.6% | +20bps |
| Net income | $44.8M | $33.2M | -26% |
| Adjusted EBITDA | $78.7M | $77.3M | -2% |
| Adjusted EBITDA margin | 31.2% | 25.9% | -530bps |
| Free cash flow | $86.3M | $78.6M | -9% |
Read down that column and the fan opens up in front of you. Engagement is the strongest line. Every step closer to cash is weaker than the one before it.
Q2 2026 year-on-year growth, by metric
The further down the funnel you go, the smaller the number gets.
Two things in that chart deserve more attention than they usually get.
First, revenue growth of 18% is flattered by accounting, not performance. Duolingo recognises subscription revenue rateably over the life of a subscription, most of which run twelve months. Revenue in Q2 2026 is therefore substantially the cash collected during 2025, when bookings were growing much faster. Bookings are the forward-looking number. Bookings grew 8%.
Second, the gap between DAU growth and MAU growth is 13 percentage points. Daily users grew 23%; monthly users grew 10%. Duolingo is getting a lot better at making the people it already has open the app every day. It is growing the total pool of people who touch the app at all far more slowly.
The per-user arithmetic
Divide the money by the users and the picture sharpens. These four lines are our arithmetic on the company's disclosed figures, not company-reported metrics.
Derived per-user economics (our calculation from disclosed figures)
| Derived metric | Q2 2025 | Q2 2026 | Change |
|---|---|---|---|
| Revenue per DAU | $5.29 | $5.08 | -4% |
| Total bookings per DAU | $5.62 | $4.92 | -12% |
| Paid subscribers as a share of DAUs | 22.9% | 21.6% | -130bps |
| Non-subscription bookings | $40.7M | $38.8M | -5% |
Bookings per daily active user fell 12%. The share of daily users who pay fell from 22.9% to 21.6%. And the part of bookings that is not subscriptions — advertising, the English Test, in-app purchases — actually shrank, from $40.7 million to $38.8 million.
Duolingo added eleven million daily users and got less money out of each of them. Whether that is a bargain depends entirely on what those users do next year.
3. The 26% profit fall is mostly a tax story
Net income fell from $44.8 million to $33.2 million, a 26% drop, and that headline did a lot of work in the coverage. It is also close to meaningless as a read on the operating business. Here is the income statement, line by line.
Where the net income decline actually came from
| Line item | Q2 2025 | Q2 2026 | Change |
|---|---|---|---|
| Income from operations | $33.363M | $33.945M | +1.7% |
| Interest income | $11.427M | $11.831M | +3.5% |
| Income before income taxes | $46.450M | $45.366M | -2.3% |
| Provision for income taxes | $1.669M | $12.208M | +631% |
| Net income | $44.781M | $33.158M | -26.0% |
| Effective tax rate (derived) | 3.6% | 26.9% | +2330bps |
Pre-tax income fell 2.3%. Operating income rose. The entire visible collapse in net income is the tax line going from $1.7 million to $12.2 million — an effective rate moving from 3.6% to 26.9%, which is roughly the 23–25% the company has told investors to expect for the year.
This matters both ways. It means the bear case cannot lean on "profits are falling", because on an operating basis they are not. It also means the flattering comparison is gone for good: Duolingo now pays a normal tax rate, and every future year-on-year net income comparison will be made against a properly taxed base.
4. All of the growth is subscriptions
Duolingo reports revenue in five buckets. Only one of them is working.
Revenue by product type
| Product | Q2 2025 | Q2 2026 | Change |
|---|---|---|---|
| Subscription | $210.7M | $258.0M | +22% |
| Advertising | $20.6M | $21.1M | +2% |
| Duolingo English Test | $10.1M | $10.1M | 0% |
| In-app purchases | $10.4M | $8.0M | -23% |
| Other | $0.5M | $1.3M | +148% |
| Total | $252.3M | $298.5M | +18% |
Revenue by product type
Subscription revenue carries the quarter. Everything else is flat or shrinking.
Advertising grew 2%. The Duolingo English Test — the product with the clearest institutional moat and the highest price point — was flat to the dollar. In-app purchases fell 23%. Subscription is 86% of revenue and rising, which means the diversification story that used to accompany this stock has quietly stopped being true.
Some of this is by design. Management has said explicitly that it is trying to increase revenue without adding friction for free users, and the advertising line is the most obvious casualty of that. But a company that depends on one product line for all of its growth has a narrower set of ways to surprise you.
5. Where the margin went
Adjusted EBITDA margin fell 530 basis points. It did not go into cost of revenue — gross margin actually improved slightly, to 72.6%, which the company attributes to a measured pace of AI feature rollout and AI cost efficiencies. It went into operating expenses.
Operating expenses
| Operating expense | Q2 2025 | Q2 2026 | Change |
|---|---|---|---|
| Research and development | $73.7M | $92.2M | +25% |
| Sales and marketing | $29.6M | $40.0M | +35% |
| General and administrative | $46.0M | $50.6M | +10% |
| Total operating expenses | $149.2M | $182.8M | +23% |
| Revenue (for comparison) | $252.3M | $298.5M | +18% |
Operating expenses, Q2 2025 vs Q2 2026
Sales and marketing grew twice as fast as revenue.
Total operating expenses grew 23% against revenue growth of 18%. Sales and marketing grew 35% — nearly twice the rate of revenue — which is consistent with what management has said about building out performance marketing, and is the clearest evidence that the user growth is not purely organic any more.
⚠
Stock-based compensation exceeds net income
Duolingo booked $38.6 million of stock-based compensation related to equity awards in Q2 2026, against GAAP net income of $33.2 million. The company guides SBC to around 15% of revenue for the full year and expects 3.5-4.0% dilution from equity grants before buybacks. Adjusted EBITDA adds all of that back.
The company is buying some of that dilution back. It repurchased roughly $44.4 million of stock in Q2 and about $71.9 million in total through 1 August under a $400 million authorisation, which it says covers nearly all of the dilution from 2024 and 2025 combined.
6. How much of the user growth repeats?
This is the part of the quarter that most deserves scepticism, and to its credit the company volunteers it rather than burying it.
Duolingo attributes the DAU acceleration to three factors: product improvements, marketing, and a one-time event in June called Streak Revival, which let eligible learners restore their longest-ever streak by opting in and completing three lessons. The response was enormous — 15.4 million learners revived a streak, including nearly 8 million who had no active streak at all when it began.
Management says two of the three drivers are permanent and expects DAU growth above 20% for the rest of the year. That may well be right. But roughly 8 million reactivated dormant users is a material number against a base of 58.7 million, and the honest position is that we will not know how many of them stayed until the Q3 print.
The genuinely encouraging number underneath all this is retention. Current User Retention Rate — the share of engaged daily users who come back the next day — is at an all-time high of 84%, up about a point year on year. That compounds, and it is the strongest single argument the bulls have.
7. What management has actually promised
Duolingo reaffirmed its topline guidance and raised its profitability outlook. Here it is in full.
Q3 and FY 2026 guidance as issued 5 August 2026
| Guidance | Q3 2026 | FY 2026 |
|---|---|---|
| Bookings | $307M | $1285M |
| Bookings growth YoY | 8.9% | 10.9% |
| Revenue | $302M | $1207M |
| Revenue growth YoY | 11.1% | 16.3% |
| Adjusted EBITDA | $76M | $320M |
| Adjusted EBITDA margin | 25.2% | 26.5% |
| Gross margin | ~71.0% | ~71.6% |
Note what happens to revenue growth. The full year is guided at 16.3%; the third quarter alone is guided at 11.1%. The rateable-recognition tailwind from the fast-growing 2025 bookings is running out, and reported revenue growth is converging down towards the bookings growth rate. Anyone anchoring on the 18% headline should look at the 11.1% instead.
ℹ
The currency wrinkle
On a constant-currency basis Q2 bookings grew 6%, not 8% - foreign exchange helped the reported figure. Over half of Duolingo's bookings come from outside the US, and the company estimates every 1% move in the dollar against its currency basket is worth about $4 million to second-half bookings.
8. What you are paying for it
Here is the arithmetic at the most recent close, using the share count and balance sheet from the 30 June filing.
Valuation arithmetic at the 21 August close
| Input | Value | Source |
|---|---|---|
| Closing price 21 Aug 2026 | $146.13 | Market data |
| Fully diluted shares at 30 June 2026 | 50.7M | Q2 shareholder letter |
| Implied fully diluted market value | $7.41B | Derived |
| Cash and short-term investments at 30 June 2026 | $1.31B | Q2 balance sheet |
| Implied enterprise value | $6.09B | Derived |
| FY2026 adjusted EBITDA guidance | $320M | Company guidance |
| Implied EV / guided adjusted EBITDA | 19.0x | Derived |
Nineteen times guided adjusted EBITDA, for a business guiding to 10.9% bookings growth, with net cash of $1.31 billion, 72.6% gross margins and a stock-based compensation charge that exceeds its net income. Whether that is cheap is a matter of what you think bookings do in 2027, and nothing in the Q2 filing answers that.
What the sell side thinks
The interesting thing about the analyst community here is not the direction — targets have been rising since the print — but that the stock has run past the average of them.
Published price targets and ratings around the Q2 print
| Firm | Rating | Price target | Date |
|---|---|---|---|
| DA Davidson | Buy - upgraded from Neutral | $160 - raised from $130 | 18 Aug 2026 |
| Wedbush | Not stated | $150 - raised from $139 | 20 Aug 2026 |
| UBS | Buy | $150 | Reported 18 Aug 2026 |
| Citi | Neutral | $140 | Reported 18 Aug 2026 |
| Truist | Hold | $120 | Reported 18 Aug 2026 |
| Barclays | Not stated | $115 | Reported 18 Aug 2026 |
| Consensus of 24 analysts | Hold | $126.46 | Reported Aug 2026 |
A consensus target of $126.46 against a price of $146.13 means the average analyst covering this stock currently sees it as overvalued by about 13%. That is an unusual place to be three weeks after a beat, and it tells you the rebound has been driven by something other than the sell side.
9. The two cases, stated fairly
The bull case
- Retention is at an all-time high of 84% and improving, and small gains there compound into large DAU gains over time.
- The bookings slowdown is a deliberate, disclosed choice. The company removed monetisation friction and pushed the Video Call AI feature down from the premium Max tier into the cheaper Super subscription, trading near-term cash for a bigger engaged base.
- Gross margin came in at 72.6% against a company expectation of about 71.0%, specifically because AI is getting cheaper to serve. The same technology cast as the threat is currently expanding the margin.
- Operating income grew year on year. The profit decline is a tax artefact.
- $1.31 billion of net cash, positive free cash flow every quarter, and an active buyback covering nearly all recent dilution.
The bear case
- Bookings per daily active user fell 12% and the paying share of daily users fell from 22.9% to 21.6%. More users, less money each.
- Every revenue line except subscriptions is flat or shrinking. Advertising grew 2%, the English Test was flat, in-app purchases fell 23%.
- Reported revenue growth is converging down on bookings growth. Q3 revenue is guided at 11.1% against the 18% just reported.
- Operating expenses grew 23% against revenue growth of 18%, with sales and marketing up 35%, suggesting growth is becoming more purchased and less organic.
- Some of the 23% DAU acceleration came from a one-time June event that reactivated roughly 8 million dormant users, and their retention is untested.
- At $146.13 the stock trades above the consensus analyst target of $126.46.
10. What would actually settle it
Both cases above are constructed from the same filing, which is what makes this a real argument rather than a narrative fight. The disagreement is not about the facts. It is about whether an engaged free user is an asset that converts later or a cost that never does.
There are three numbers in the third-quarter report that will move the needle, and they are worth writing down now.
- Whether DAU growth holds above 20% without another one-off event. That tests how many of the 8 million reactivated dormant users actually stayed.
- Whether bookings per DAU stops falling. The company does not report this; you have to divide total bookings by DAUs yourself. It fell 12% in Q2.
- Whether paid subscribers as a share of DAUs stabilises around 21.6% or keeps sliding. This is the single cleanest measure of whether the free-user strategy converts.
Duolingo has done something unusual and, on its own terms, honest: it told the market in February exactly what it was going to sacrifice and why, then did it. The market marked it down 73% from the peak and has since marked a good chunk of that back up. Neither reaction was obviously wrong on the information available at the time.
The streak, in the end, is the perfect metaphor for the whole company. It is the most powerful engagement mechanic anyone has built in consumer software — 15.4 million people completed three lessons in a week just to get a number back. The open question, which the Q2 filing does not answer, is how much that number is worth in cash.
Key takeaways
- Users grew 23% and bookings grew 8% in Q2 2026. That single gap is the entire investment argument, and the filing supports both readings of it.
- The 26% net income decline is almost entirely a tax effect. Pre-tax income fell 2.3% and operating income rose. The effective tax rate went from 3.6% to 26.9%.
- Bookings per daily active user fell about 12% and paying users fell from 22.9% to 21.6% of DAUs. More users, less money from each.
- All of the growth is subscriptions, up 22%. Advertising grew 2%, the English Test was flat, in-app purchases fell 23%.
- Some of the DAU acceleration came from Streak Revival, a one-time June event that reactivated nearly 8 million users with no active streak. Q3 tests whether they stayed.
- At the 21 August close of $146.13 the stock trades roughly 19x guided FY26 adjusted EBITDA, and about 13% above the $126.46 consensus analyst target.
This article is analysis, not investment advice. All financial figures are taken from Duolingo's Q2 2026 shareholder letter and accompanying unaudited financial statements filed 5 August 2026. Market prices and analyst targets are as of the dates stated. Figures marked as derived are our own arithmetic on the company's disclosed numbers.