AI Absorption Ledger / DUOL

Duolingo

DUOL · Q2 2026 · reported 2026-08-05 · revenue $298mn

Assessment

Q2 2026 brings the first level management has given for the AI bill, and it describes a cost that is falling. The CFO put AI expenses in cost of goods sold in the tens of millions of dollars and internal AI use closer to , both without a period. The CEO said the cost of a Video Call fell from at launch to under , mainly by moving to open source models. Gross margin came in at , ahead of the expected , and the 10-Q again attributes the improvement to lower per-unit third-party AI costs. The spend channels are sized by the ledger at , the CFO’s phrase through the fixed undated band, and .

The cost fall is what moves AI out of the priced tier. Video Call, put behind Max because it was expensive, now reaches most new Super subscribers, and the CEO says Max may keep unlimited calls or be retired, aiming to lose little revenue. Neither tier has a reported figure; the ledger reads Max as an existing tier under an AI name, product-defensive in both quarters, with ballparks of for the Max premium and for Super.

Content production and AI in engineering went unmentioned: the prior quarter's output figures were not repeated, and the only per-head remark answered a question about the bonus plan without naming AI. Research and development grew , faster than revenue, on headcount; no contractor, headcount reduction or AI-attributed saving appears in the call, the letter or the 10-Q.

Sized channels against the income statement, Q2 2026

4 of 6 channels sized

Each blue mark is one channel's dollars for the quarter; a bar is the range of an estimate. Grey marks are the company's reported lines. The distance between them is the point: how large the AI channel is next to the line it sits in.

New money and old money, Q2 2026

2 new3 expanded1 relabelled

Each channel is tagged once for whether its money existed before language models, from the company's annual report and call at the start of the period. A bar splits one flow's sized dollars by that tag. The incremental total is the part that would not be there without the models: a new channel counts in full, an expanded one only for what AI added, a relabelled one at zero.

Paid for AI$1.7mn to $27mn sized

new $1.7mn to $27mn

Incremental total $1.7mn to $27mnpoint $5.0mn
Revenue arriving through AI$0 to $38mn sized

expanded $0 to $4.6mnrelabelled $0 to $34mn

Incremental total $0 to $4.6mnpoint $619k

The sized total counts every channel the company credits to AI, including relabelled money that existed before language models and the ledger's own estimates for it. The incremental total counts relabelled channels at zero. A flow is split by layer where its dollars sit at more than one: end use, compute sold to builders, and hardware. The same dollar can be a buyer's spend, a cloud's revenue and a chipmaker's revenue, so the layers are never added together.

Paid for AI2 channels · $1.7mn to $27mn sized · $1.7mn to $27mn incremental

cost of-revenue

Model and inference bill for AI features in the product

0.28% to 8.4% of the quarter’s revenue

Incremental total: counts in full.

Matched line moved : the whole line, not this channel.

our inferencedisclosure: bounded· motive: product-defensive· before LLMs: new

The CFO put AI expenses in cost of goods sold in the tens of millions of dollars and, with hosting, among the largest items, with no period. The CEO said the cost per Video Call fell from at launch to under , at least lower, mainly by moving to open source models, and that all AI use in China runs on local models. Gross margin of beat the expected and the year-end expectation rose to from . The size shown is the ledger's own estimate, : the phrase is in the words-to-numbers table, and with no period it converts to the quarter through the fixed undated band. The former estimate, a judgment share of cost of revenues, is no longer used. The counterparty is mixed: third-party model providers, and the hosting and compute providers that serve the open source and local models the company is moving to.

Evidence: 18 quotes, 11 figures, 2 confounds, 5 from before coverage

Third-party AI model costs carried in cost of revenues: what the company pays to serve Video Call, the Max features and the AI-generated content in the learning product. The filing names them beside payment processing and hosting fees and does not split them out.

Why this motive

Inference cost in cost of revenues for features moving into the standard paid tier: the cost per Video Call fell far enough to give it to Super subscribers (claim c2), and the CFO keeps the AI cost savings in margin while putting Video Call out to all Super users (claim c9). The product-defensive tell, as in the prior quarter.

Before LLMs: new

At the anchor the 10-K already listed generative AI costs inside cost of revenues beside payment processing and hosting fees, with no amount; cost of revenues was for FY2024. The CFO said on the Q1 2024 call that every use of the Max AI features carried an incremental cost. The bill exists only because of LLMs, so the tag is new whatever the anchor shows.

“Cost of revenues predominantly consists of third-party payment processing fees charged by various distribution channels in addition to hosting fees and generative AI costs.”
Filing, mdna, 10-K periodic report, 2025-02-28
“Other business purchase commitments consist of hosting costs, web services and generative AI costs.”
Filing, mdna, 10-K periodic report, 2025-02-28
“We talked about the fact that for MAX, there's incremental cost for every use case, for every usage of it. And that does lower the gross margin of the MAX product.”
CFO, qa, earnings call, 2024-05-08
“Further, our ability to continue to develop, maintain or use such technologies may be dependent on access to specific third-party software and infrastructure, such as processing infrastructure for the training of our own machine-learning models or the use of third-party AI models.”
Filing, risk factors, 10-K periodic report, 2025-02-28
“Now, that said, we expect that the cost of generative AI, the cost of large language models, will go down. So over time, we'll probably be able to roll out everywhere.”
CEO, qa, earnings call, 2024-05-08

Figures

  • Gross margin (three months; the same 10-Q row also carries the six-month figures) · 2026-CQ2
  • Gross margin, prior year (three months) · 2025-CQ2
  • Gross margin expected for Q2 2026 when guided · 2026-CQ2
  • Gross margin above the expectation, percentage points · 2026-CQ2
  • Gross margin guided for Q3 2026 · 2026-CQ3
  • Gross margin guided for fiscal 2026, raised on AI cost trends · FY2026
  • Gross margin expected at the end of 2026, per the CFO · 2026-CQ4
  • Gross margin expected at the end of 2026 on the Q1 call, per the CFO · 2026-CQ4
  • Cost per Video Call when the feature was first put on the platform, per the CEO · FY2024
  • Cost per Video Call now, upper bound, per the CEO · 2026-CQ2
  • Fall in the cost per Video Call since launch, at least · 2026-CQ2

Reported line it is matched to

Cost of revenues was of revenue against a year earlier, against the prior-year share; gross margin rose from to . The 10-Q attributes the increase primarily to lower per-unit third-party AI costs.

2026-CQ2: 2025-CQ2: 2026-CQ2: 2025-CQ2:

What else could explain it

  • line composition: Cost of revenues is mostly payment processing fees and hosting; the AI part is not split.
  • mix shift: Subscription revenue grew faster than the rest, which changes the line mix without any AI cost moving.

Quotes

“When we first started adding Video Call to the platform, the first time that we put a Video Call on the platform, I remember the team that was working on it told me, "Okay, we can give this to users, but it's going to cost, like, $0.30 per call to give to users." That was expensive, and this is why we decided to put it behind our most expensive plan, which is Duolingo Max.”
c1 · CEO, qa, earnings call, 2026-08-05
“The good news is that, through a lot of really hard work, we've been able to bring down the cost of Video Call. It is now under $0.01 per Video Call. The reason for that is mainly a move towards open source models. It's just a lot cheaper to do that, and we don't see a loss in quality in there.”
c2 · CEO, qa, earnings call, 2026-08-05
“We're developing a lot of features, and there may be some that, because of costs of them or something, we end up putting them behind Duolingo Max. That is not the goal. The goal really is to try to give, particularly speaking features, which are the most expensive ones to provide.”
c5 · CEO, qa, earnings call, 2026-08-05
“I do expect that what'll happen is that the cost per usage of AI, whatever it is, the cost per token, if you want to call it, will come down for us because we will continue moving more and more to open source.”
c7 · CEO, qa, earnings call, 2026-08-05
“Internally, of course, we're still using many models from, say, OpenAI, Anthropic, and we'll continue doing that.”
c8 · CEO, qa, earnings call, 2026-08-05
“What we're basically saying is by going up about a point and a half on that adjusted EBITDA margin, we are seeing AI cost savings that will give us a bit structurally a better margin, even with our goal to put voice call out to all super users over the course of the year.”
c9 · CFO, qa, earnings call, 2026-08-05
“For gross margin, we now expect to end the year closer to 70% as compared to the 69% we initially expected, as we drive more AI content into our products, offset by AI cost savings.”
c10 · CFO, prepared remarks, earnings call, 2026-08-05
“Yeah, our expenses on AI in the cost of goods sold are in the tens of millions of dollars, so they're significant to the cost of goods sold. Hosting is another big cost for us as well, but those are the two biggies going through the cost of goods sold.”
c11 · CFO, qa, earnings call, 2026-08-05
“The reality is that in China, we simply cannot use the AI models, the kind of U.S. AI models. We have to use local models. That's by law.”
c13 · CEO, qa, earnings call, 2026-08-05
“At the highest level, our view inside the company is that it is in our best interest as a company to use open-weight models as much as possible. If I had a magic wand, I would try to move everything to an open-weight model. It's not always possible because sometimes the frontier models are more advanced. From a company standpoint, it is just significantly better because it's way cheaper to use open-weight models.”
c14 · CEO, qa, earnings call, 2026-08-05
“Gross margin was 72.6%, a small improvement over the prior year, and was ahead of our expectation of approximately 71.0%. This reflects our measured pace of AI-powered feature expansion (such as Video Call), as well as AI cost efficiencies.”
c16 · Filing, press release, 8-K earnings release, 2026-08-05
“We expect a gross margin of approximately 71.0% in Q3 and approximately 71.6% for the full year, better than the trajectory we outlined on our Q1 call based on AI cost trends.”
c17 · Filing, press release, 8-K earnings release, 2026-08-05
“The increase was primarily attributable to an increase in subscription gross margin, reflecting continued reductions in per-unit third-party AI costs.”
c18 · Filing, mdna, 10-Q periodic report, 2026-08-06
“Cost of revenues predominantly consists of third-party payment processing fees charged by various distribution channels in addition to hosting fees and third-party AI costs.”
c19 · Filing, mdna, 10-Q periodic report, 2026-08-06
“It just turns out that for a lot of applications, you don't need the absolute smartest model. The reality is the quality's indistinguishable for many applications that we use.”
c21 · CEO, qa, earnings call, 2026-08-05
“We do that in China. All of our usage of AI uses Chinese models.”
c23 · CEO, qa, earnings call, 2026-08-05
“Cost of Revenues and Gross Margin. Total gross margin increased to 72.6% from 72.4% during the three months ended June 30, 2026 and 2025, and total gross margin increased to 72.8% from 71.8% during the six months ended June 30, 2026 and 2025.”
c24 · Filing, mdna, 10-Q periodic report, 2026-08-06
“That's certainly what we have. What I would say is that our expectation is that over the next some amount of time, that portfolio will be weighted a little more towards open-weight models than it is today.”
c27 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026direction only · our inference · product-defensive · $3.9mn to $28mn
  • Q2 2026bounded · our inference · product-defensive · $833k to $25mn

vendor bill

AI used inside the business

0.28% to 0.84% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: quantified· motive: exploratory· before LLMs: new

The CFO put internal AI use closer to with no period; the CEO said the company still uses frontier lab models internally and wants as much as possible on open-weight models. The 10-Q puts increased software and third-party AI costs at of the year-over-year rise in Research and development, for the half. The size shown is the ledger's own estimate, , the CFO figure converted to the quarter through the fixed band for an amount with no period.

Evidence: 6 quotes, 6 figures, 2 confounds, 4 from before coverage

AI models and tools the company pays for in its own work rather than in the product: engineering, content production and operations. The 10-Q carries the increase inside Research and development as software and third-party AI costs, combined.

Why this motive

Carried from the prior quarter: the CFO gives a level and the CEO a model mix (claims c8, c12 and c14), and no displaced line or measured return is attached to the internal spend. The exploratory reading stands.

Before LLMs: new

At the anchor the 10-K said AI and machine learning helped produce content and that the company depended on third-party AI models, and listed generative AI costs among purchase commitments; Research and development was for FY2024, with web services and technology costs up and no AI share named.

“Other business purchase commitments consist of hosting costs, web services and generative AI costs.”
Filing, mdna, 10-K periodic report, 2025-02-28
“For example, we use AI and machine learning technologies to help us produce content for both the learning application and the Duolingo English Test.”
Filing, risk factors, 10-K periodic report, 2025-02-28
“Further, our ability to continue to develop, maintain or use such technologies may be dependent on access to specific third-party software and infrastructure, such as processing infrastructure for the training of our own machine-learning models or the use of third-party AI models.”
Filing, risk factors, 10-K periodic report, 2025-02-28
“Increased web services and technology costs of $3.3 million”
Filing, mdna, 10-K periodic report, 2025-02-28

Figures

  • AI used inside the business, per the CFO (period not stated) · TTM as-of 2026-06-30
  • Increase in software and third-party AI costs within Research and development, year over year (combined) · 2026-CQ2
  • Increase in software and third-party AI costs within Research and development, six months, year over year · 2026-H1
  • Research and development · 2026-CQ2
  • Research and development, prior year · 2025-CQ2
  • Research and development growth, year over year · 2026-CQ2

What else could explain it

  • line composition: The 10-Q increase combines software and third-party AI costs.
  • other: The CFO figure carries no period; read as annual it is about a quarter of the high end of the range.

Quotes

“Internally, of course, we're still using many models from, say, OpenAI, Anthropic, and we'll continue doing that.”
c8 · CEO, qa, earnings call, 2026-08-05
“Inside the business, we also use AI, and that is more closer to the $10 million range internally.”
c12 · CFO, qa, earnings call, 2026-08-05
“At the highest level, our view inside the company is that it is in our best interest as a company to use open-weight models as much as possible. If I had a magic wand, I would try to move everything to an open-weight model. It's not always possible because sometimes the frontier models are more advanced. From a company standpoint, it is just significantly better because it's way cheaper to use open-weight models.”
c14 · CEO, qa, earnings call, 2026-08-05
“The increase was primarily due to higher net personnel costs of $13.8 million, driven primarily by growth in headcount, including increased stock-based compensation expenses related to equity awards of $5.0 million, and increased software and third-party AI costs of $3.3 million.”
c20 · Filing, mdna, 10-Q periodic report, 2026-08-06
“The increase was primarily due to higher net personnel costs of $23.6 million, driven primarily by the growth in headcount, including increased stock-based compensation expenses related to equity awards of $8.9 million, and increased software and third-party AI costs of $5.5 million.”
c26 · Filing, mdna, 10-Q periodic report, 2026-08-06
“That's certainly what we have. What I would say is that our expectation is that over the next some amount of time, that portfolio will be weighted a little more towards open-weight models than it is today.”
c27 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026direction only · our inference · exploratory · $1.7mn to $5.0mn
  • Q2 2026quantified · our inference · exploratory · $833k to $2.5mn

Cost displaced by AI2 channels · 2 not sized

other · cheap to verify

Course content produced with AI

Not sized

The call, the letter and the 10-Q say nothing this quarter about content output or AI in content creation; the prior quarter's ballpark rested on output figures that were not repeated.

inscrutabledisclosure: not mentioned· motive: efficiency· before LLMs: expanded

Neither the call nor the letter repeated the course-unit figures or mentioned AI in content production. The channel is carried with no reading of its own. The counterparty is mixed: the company’s own content staff and outside contractors, whose cost both sits in Research and development.

Evidence: 0 quotes, 4 from before coverage

Learning content (course units, new subjects) generated with AI tools in place of the handmade process, by employees and contractors whose cost sits in Research and development. Management gives output volumes, not the cost of producing them.

Why this motive

Carried from the prior quarter; the quarter's sources are silent on AI in content production.

Before LLMs: expanded

Course content was already being generated in large part with large language models at the anchor: on the Q1 2024 call the CEO said the first big batch of English content was quicker and cheaper than the mostly handmade process before, with human involvement much less. The FY2024 10-K recorded a decrease in net contractor costs within Research and development of without naming a cause. What moves in coverage is the volume published per quarter. The size is the change AI made, not the whole line.

“What was amazing about it is this is really the first big batch of content that we were able to generate in large part because of generative AI. I mean, we use large language models for this. It was a lot quicker and a lot cheaper to generate than what we had been doing in the past, which was mostly handmade.”
CEO, qa, earnings call, 2024-05-08
“Now, this is not to say that humans are not involved anymore. They're still involved, but the involvement is much less because we've been using generative AI.”
CEO, qa, earnings call, 2024-05-08
“For example, we use AI and machine learning technologies to help us produce content for both the learning application and the Duolingo English Test.”
Filing, risk factors, 10-K periodic report, 2025-02-28
“The above increases were partially offset by a decrease in net contractor costs of $3.9 million.”
Filing, mdna, 10-K periodic report, 2025-02-28

By quarter

  • Q1 2026direction only · shape match · efficiency
  • Q2 2026not mentioned · inscrutable · efficiency

engineering · cheap to verify

Experiments per head from AI in engineering and product

Not sized

No source this quarter attributes engineering or product output to AI. The CEO mentioned experiments per week growing faster than headcount only in answer to a question about the user-growth bonus plan, with no AI term, and the 10-Q attributes the Research and development increase to personnel costs on headcount growth and to software and third-party AI costs.

inscrutabledisclosure: not mentioned· motive: exploratory· before LLMs: expanded

The quarter is silent on AI in engineering: the only per-head remark answered a question about the bonus plan and did not name AI. Research and development grew , faster than revenue, on headcount. The channel is carried with no reading of its own.

Evidence: 0 quotes, 4 figures, 2 from before coverage

Engineering and product payroll avoided because AI tools raise the number of A/B tests and product changes each person ships. The line it would show in is Research and development, which keeps growing on headcount.

Why this motive

Carried from the prior quarter; the quarter's sources do not tie engineering or product output to AI.

Before LLMs: expanded

At the anchor the product organization already ran hundreds of A/B tests a quarter, and the team was employees including more than engineers, paid inside Research and development ( for FY2024). AI tools change tests per head, not the existence of the line. The size is the change AI made, not the whole line.

“The foundation of our product strategy is our relentless focus on improving learner engagement through A/B testing, and we run hundreds of A/B tests on new product features each quarter.”
Filing, business, 10-K periodic report, 2025-02-28
“Our team, which as of December 31, 2024, consisted of approximately 830 passionate employees, including more than 380 engineers, aims to build the most sophisticated education platform in the world.”
Filing, business, 10-K periodic report, 2025-02-28

Figures

  • Research and development · 2026-CQ2
  • Research and development, prior year · 2025-CQ2
  • Research and development growth, year over year · 2026-CQ2
  • Increase in net personnel costs within Research and development, year over year · 2026-CQ2

By quarter

  • Q1 2026direction only · our inference · exploratory · $0 to $3.4mn
  • Q2 2026not mentioned · inscrutable · exploratory

Revenue arriving through AI2 channels · $0 to $38mn sized · $0 to $4.6mn incremental

pricing packaging · cheap to verify

Price premium of the Max tier for AI features

0% to 11.3% of the quarter’s revenue

Incremental total: counts at zero.

our inferencedisclosure: described· motive: product-defensive· before LLMs: relabelled

The CEO said Video Call was put behind Max because it cost a call at launch, and that with the cost now far lower the company may keep unlimited Video Call for Max only or retire Max, aiming to lose little revenue. No Max subscriber or revenue figure was given; the channel stays relabelled. The size shown is the ledger's own estimate, , with a low end of zero.

Evidence: 6 quotes, 2 figures, 2 confounds, 10 from before coverage

Subscription revenue the company earns above Super Duolingo from Max, the top tier, which it presents as the tier for generative AI features (Video Call, explanations, role play) and which also carries non-AI packaging. The company does not report Max subscribers or revenue; in 2026 management moves Video Call into Super and says Max may be repriced or retired.

Why this motive

Carried from the prior quarter and reinforced: the AI feature that defined Max now reaches most new Super subscribers at no stated extra price (claims c15 and c9), and the CEO says Max may be retired while trying to keep its revenue (claims c3 and c4). Included at no additional cost is the product-defensive tell.

Before LLMs: relabelled

Max is a third subscription tier the company had wanted before LLMs: on the Q1 2024 call the CEO called the AI features a good excuse to start it, said the top package need not hold only AI features, and described non-AI packaging (unlimited hearts) being tested in it. The 10-K describes it as priced above Super Duolingo, with no price, subscriber count or revenue given; the CFO called Max subscribers relatively small. Subscription revenue was for FY2024, and up to of daily active users could see Max in May 2024. No measure of an AI-attributed premium appears at the anchor or in coverage, so the money is existing subscription revenue under an AI name.

“We talked about the fact that for MAX, there's incremental cost for every use case, for every usage of it. And that does lower the gross margin of the MAX product.”
CFO, qa, earnings call, 2024-05-08
“Duolingo Max is a premium subscription tier launched in 2023 and offered to a portion of our user base and priced higher than Super Duolingo. It gives learners access to the existing features of Super Duolingo in addition to incremental features and exercises powered by generative AI technology.”
Filing, business, 10-K periodic report, 2025-02-28
“About a year ago, we started experimenting with a third tier, which we call Duolingo Max, where the idea was that this coincided with the large language models and generative AI coming out. We decided to use the AI features as a good kind of excuse to start the third tier, which is something we've wanted to do for a while.”
CEO, qa, earnings call, 2024-05-08
“The other thing that you'll see us do is you'll see us start shifting features around to see what is the best packaging. I mean, there's no real reason for the highest package to be just AI features.”
CEO, qa, earnings call, 2024-05-08
“We're seeing a bunch of evidence that people are willing to pay a substantially higher price for it.”
CFO, qa, earnings call, 2024-05-08
“Furthermore, in the third quarter of 2024 we announced an AI-powered product enhancement, Video Call, within our highest subscription tier that allows users to practice conversation skills through video calls with one of our characters.”
Filing, risk factors, 10-K periodic report, 2025-02-28
“Now, that said, we expect that the cost of generative AI, the cost of large language models, will go down. So over time, we'll probably be able to roll out everywhere.”
CEO, qa, earnings call, 2024-05-08
“We're running an experiment to put unlimited hearts in Max.”
CEO, qa, earnings call, 2024-05-08
“Yeah, Sean, the mix question is going to be one that we already mentioned, which is right now, the number of folks who are subscribing to Max is relatively small.”
CFO, qa, earnings call, 2024-05-08
“And certainly in 2024, the impact of Max is in our guide, but it's commensurate with the fact that it's only a 5%-10% rollout now. So it's relatively modest in the 2024 guide.”
CFO, qa, earnings call, 2024-05-08

Figures

  • Subscription revenue · 2026-CQ2
  • Cost per Video Call when the feature was first put on the platform, per the CEO · FY2024

What else could explain it

  • bundling: Max also carries non-AI packaging, so its premium is not all paid for AI.
  • mix shift: Subscribers may move from Max to Super now that Super carries Video Call; revenue can shift between the channels.

Quotes

“When we first started adding Video Call to the platform, the first time that we put a Video Call on the platform, I remember the team that was working on it told me, "Okay, we can give this to users, but it's going to cost, like, $0.30 per call to give to users." That was expensive, and this is why we decided to put it behind our most expensive plan, which is Duolingo Max.”
c1 · CEO, qa, earnings call, 2026-08-05
“That calls into question, what are we doing with Duolingo Max? My answer is, I don't know yet. There's a few possibilities, and one possibility could be that while Super subscribers get a limited version of Video Call, like limited number of Video Calls, Duolingo Max subscribers get unlimited. That is a possibility. Another possibility, truthfully, is that we may actually sunset Duolingo Max.”
c3 · CEO, qa, earnings call, 2026-08-05
“What I will tell you is that we're going to have an answer to this in the next couple of quarters, in addition to that, we're going to try to do this without a loss of revenue. That's partly why we're not going super fast here, because we're trying to figure out how to do this without losing much revenue.”
c4 · CEO, qa, earnings call, 2026-08-05
“We're developing a lot of features, and there may be some that, because of costs of them or something, we end up putting them behind Duolingo Max. That is not the goal. The goal really is to try to give, particularly speaking features, which are the most expensive ones to provide.”
c5 · CEO, qa, earnings call, 2026-08-05
“Additionally, most new Super Duolingo subscribers now have access to Video Call, and we expect to extend that access to existing Super subscribers later this year. This means that more learners can now use our flagship conversational practice feature, demonstrating how AI is helping us deliver higher-quality learning at scale.”
c15 · Filing, press release, 8-K earnings release, 2026-08-05
“The first one is that if you have access to Video Call, the engagement is very good. In fact, that feature has just gotten significantly better.”
c25 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026described · our inference · product-defensive · $0 to $33mn
  • Q2 2026described · our inference · product-defensive · $0 to $34mn

product revenue · cheap to verify

Video Call added to Super Duolingo

0% to 1.6% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: product-defensive· before LLMs: expanded

Most new Super subscribers now have Video Call and existing ones are to get it later in the year; the CEO said it is one of the main reasons people subscribe and that cost, now under a call, made the move possible. No price or conversion figure was given. The size shown is the ledger's own estimate, .

Evidence: 6 quotes, 2 figures, 2 confounds, 3 from before coverage

Subscription revenue gained in the middle tier because its subscribers now get Video Call, the AI conversation feature that was in Max only: conversion of new subscribers and any price increase for Super with Video Call. Separate from the Max premium, which this move may erode.

Why this motive

The AI feature is included in the existing tier with no price change stated (claim c15), and the CEO calls it one of the main reasons people subscribe (claim c6). Included at no additional cost is the product-defensive tell; carried.

Before LLMs: expanded

At the anchor Video Call sat only in the highest tier, announced in the third quarter of 2024, and Super Duolingo carried the non-AI paid features; subscription revenue was for FY2024. Moving Video Call into Super changes what an existing tier holds, not whether the revenue exists. The size is the change AI made, not the whole line.

“Duolingo Max is a premium subscription tier launched in 2023 and offered to a portion of our user base and priced higher than Super Duolingo. It gives learners access to the existing features of Super Duolingo in addition to incremental features and exercises powered by generative AI technology.”
Filing, business, 10-K periodic report, 2025-02-28
“Furthermore, in the third quarter of 2024 we announced an AI-powered product enhancement, Video Call, within our highest subscription tier that allows users to practice conversation skills through video calls with one of our characters.”
Filing, risk factors, 10-K periodic report, 2025-02-28
“Super Duolingo offers learners additional features to enhance their learning experience.”
Filing, business, 10-K periodic report, 2025-02-28

Figures

  • Subscription revenue · 2026-CQ2
  • Cost per Video Call now, upper bound, per the CEO · 2026-CQ2

What else could explain it

  • bundling: Video Call enters Super alongside longer free trials and other tier changes; any lift is inside subscription revenue.
  • mix shift: Revenue gained in Super may be revenue lost from Max.

Quotes

“What I will tell you is that we're going to have an answer to this in the next couple of quarters, in addition to that, we're going to try to do this without a loss of revenue. That's partly why we're not going super fast here, because we're trying to figure out how to do this without losing much revenue.”
c4 · CEO, qa, earnings call, 2026-08-05
“If suddenly all users have it, there's less incentive to buy, and then we would have to find other reasons to get people to buy. There's a little bit of a trade-off here. It's both cost, and this is one of the main things that gets people to buy.”
c6 · CEO, qa, earnings call, 2026-08-05
“What we're basically saying is by going up about a point and a half on that adjusted EBITDA margin, we are seeing AI cost savings that will give us a bit structurally a better margin, even with our goal to put voice call out to all super users over the course of the year.”
c9 · CFO, qa, earnings call, 2026-08-05
“Additionally, most new Super Duolingo subscribers now have access to Video Call, and we expect to extend that access to existing Super subscribers later this year. This means that more learners can now use our flagship conversational practice feature, demonstrating how AI is helping us deliver higher-quality learning at scale.”
c15 · Filing, press release, 8-K earnings release, 2026-08-05
“Look, I would love to make it available to all Duolingo users. I don't think we can do that right now.”
c22 · CEO, qa, earnings call, 2026-08-05
“The first one is that if you have access to Video Call, the engagement is very good. In fact, that feature has just gotten significantly better.”
c25 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026described · our inference · product-defensive · $0 to $2.3mn
  • Q2 2026described · our inference · product-defensive · $0 to $4.6mn

Reported lines, year-over-year growth

Revenue +18.3%

Q2 2026. Growing slower than revenue: cost of revenue (+17.3%), general and administrative (+10.0%). A displaced cost shows up as a line that stays under the dashed revenue line. These are the audited lines, as first reported; nothing here is attributed to AI by the filing. Not drawn: advertising expense, which one-time items move by more than 60% in a quarter; the values are in the table below.

Cost of revenueResearch and developmentSales and marketingGeneral and administrativeRevenue
0%10%20%30%40%50%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026Sales and marketingResearch and developmentRevenueCost of revenueGeneral and administrative
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Revenue$231mn$252mn$272mn$283mn$292mn$298mn
Cost of revenue$67mn$70mn$75mn$77mn$79mn$82mn
Research and development$70mn$74mn$83mn$80mn$83mn$92mn
Sales and marketing$27mn$30mn$35mn$34mn$39mn$40mn
General and administrative$43mn$46mn$44mn$48mn$46mn$51mn
Advertising expense$18mn$20mn$25mn$24mn$28mn$28mn