Duolingo is My Next Big Bet.

Q1 2025 ER Digest

I initiated a position in Duolingo in March, which is now up over 35%. This is an update of my original Duolingo deep dive and my Duolingo Q4 2024 ER digest.

Duolingo’s business gets bigger and stronger as they increase their ability to produce content that optimises student outcomes per dollar spent.

AI is enabling Duolingo to increase this ability exponentially and I believe non-linearly better financial results will emerge as a result, in a relatively short period of time.

As has been the case with Meta since 2022, AI has greatly accelerated Duolingo’s value creation process. Duolingo has added 148 new courses over the past year, when the previous 100 courses took them twelve years. In the graph below you may notice that Duolingo’s MAUs have ticked up QoQ, from 117M to 130M - the largest absolute increment in the entire time sequence depicted. In my view, this growth is driven by AI exponentiating Duolingo’s ability to produce content.

See Duolingo CEO Luis von Ahn’s remarks about this, during the Q1 2025 earnings call:

Teaching requires quite a bit of content.

It used to be the case that a lot of this content was made by hand with some automation, but a lot of it by hand. By now, we are able to make the content for all of our subjects mostly, close to 100%, automatically.

So that allows us to just have a lot more content that is made a lot faster and much cheaper. So that’s one place where it’s helping us.

In the abstract, Duolingo’s business is about producing content and tweaking it to optimise user lifetime value. This is true all the way from the content they produce to acquire new customers (such as the recent market campaign, in which Duolingo’s owl faked his own death), to the content they produce to engage users across tiers. AI fundamentally permeates this stack and makes its iteration exponentially more efficient. As time goes by, the stack generates additional data which then speeds up the iteration even further.

For instance, improving Duolingo’s monetisation is about figuring out what content to put into what tier in order to maximise the value to price ratio. That’s essentially a function of running many A/B tests and finally implementing the modifications that optimise outcomes. Enhancing user engagement throughout various courses and across educational verticals (from languages, to math, music and chess currently) is fundamentally the same.

For example, see Duolingo CFO Matt Skarupa’s remarks on why they decided to launch Duolingo Max:

We are optimizing for platform LTV. And so when we run our A/B tests, we have to take into account, are these Max subs incremental or not in figuring out the overall impact on platform LTV.

And that’s why we launched Max is because we ran enough of the experiment to figure out that they were incremental.

There [was] enough incrementality to it that it was platform LTV positive. And we still think that’s true. And in fact, we think it will become more true over time.

Hence, I believe that Duolingo’s earning power will broadly track the evolution of AI models going forward. The business will get bigger and stronger as AI models get better. As I explain in this post, Duolingo is what I call a Singularity Scaler.

While some analysts speculate that ChatGPT will disrupt Duolingo, they fail to understand a basic premise of the digital economy: a marginal advantage at the user experience level tends to equate to an exponential advantage in terms of adoption, which ultimately yields an advantage in terms of model performance. In other words, so long as Duolingo remains laser focused on enhancing student outcomes per dollar spent, it will be increasingly harder for generic AI models to disrupt it.

I believe Duolingo will maintain a lead in user experience via novel features that pique user interest in a way that other platforms without an exclusive focus on education cannot. For example, see Luis’s remarks about how he sees the Video Call with Lily feature evolving:

So, there is that type of stuff we got to get better at. And then we are going to have a whole world for her where she is going to be able to change her background immediately and start pretending that she is like your taxi driver or something.

So, we are really – the idea is to practice as much language as possible and to increase engagement. And the more we increase the engagement, the more people will tell their friends not only to use Duolingo, but if they are a Duolingo user, they should get Duolingo Max to be able to practice with Lily.

Going back to the idea of AI permeating and enhancing Duolingo’s entire value creation process, it seems that AI now enables non-technical folks at Duolingo to create prototypes. The prototype of the new chess vertical, which will allegedly launch in the next few weeks, was created by people with no technical background and no chess background. This means we are likely to see an explosion of new verticals and interesting features over the coming years.

Luis’s remarks during the Q1 2025 earnings call about the development of the incoming chess vertical were highly insightful:

I should mention something amazing about chess is that it really started with a team of two people, neither of whom knew how to program, so they were not programmers.

And they basically made prototypes and did the whole curriculum of chess by just using AI.

Also, neither of them knew how to play chess. And we started that for several months.

And eventually, when they had a really good prototype, we had a whole team to professionalize it and put it in the app. So we’re very happy with it. This is, at the moment, mainly going to be to increase users.

Duolingo’s new math and music verticals continue to progress well, but are still much smaller than the language vertical. Nonetheless, Luis says Duolingo will produce much more content for these verticals driven by AI. Since these verticals are already being monetised, it likely won’t be long until they are meaningfully accretive to the top and bottom line. See Luis’s remarks about this during the Q1 2025 earnings call:

We are, of course, also investing in math and music and chess. We’re spending less effort on that because they are much smaller businesses than language learning. But you’ll see us – for example, for math, you’ll see us vastly increase the content because of AI within the next few months, the content is probably going to quadruple that we have in math because of AI in, call it, I don’t know, a quarter or so.

Further, there is no limit to the number of verticals that Duolingo can deploy over time. My long term thesis remains that Duolingo is building the underlying infrastructure for a more effective and cheaper educational system. While gross margins (blue line, below) are expected to continue declining going into Q2, as Duolingo performs more AI inferences, management is still guiding for margins to recover going into H2 2025. Meanwhile, both revenue (purple line) free cash flow per share (orange line, below) continues growing spectacularly and it’s highly likely that the growth of this metric will continue to accelerate.

In general, the optimisation of video call and Max is on trend with what we expected on the last call.

And that’s good news for us because it does lead to margin expansion in the back half of the year like we thought.

And we also saw evidence in the first quarter that rates were coming down for those type of API calls. So just in general, we feel confident with the current course and speed of those optimisations.

-Duolingo CFO Matt Skaruppa doing the Q1 2025 earnings call.

Until next time!

---

⚡ If you enjoyed the post, please feel free to share with friends, drop a like and leave me a comment.

You can also reach me at:

Twitter: @alc2022

LinkedIn: antoniolinaresc

Disclosurđ

These are opinions only of the individual author. The contents of this piece do not contain investment advice and the information protfrdvided is for educational purposes only and no discussions constitute an offer to sell or the solicitation of an offer to buy any securities of any company. All content is purely subjective and you should do your own due diligence.

Antonio Linares makes no representation, warranty or undertaking, express or implied, as to the accuracy, reliability, completeness or reasonableness of the information contained in the piece. Any assumptions, opinions and estimates expressed in the piece constitute judgments of the author as of the date thereof and are subject to change without notice. Any projections contained in the Information are based on a number of assumptions as to market conditions and there can be no guarantee that any projected outcomes will be achieved. Antonio Linares does not accept any liability for any direct, consequential or other loss arising from reliance on the contents of this presentation. Antonio Linares is not acting as your financial, legal, accounting, tax or other adviser or in any fiduciary capacity.