Investment Thesis
Duolingo (NASDAQ:DUOL), while a top language learning app, may soon feel the pressure from OpenAI. OpenAI’s ChatGPT-4o in my opinion, is in a league of its own, particularly in language translation, and that could pose a challenge to language learning apps. Described as omni-capable of integrating multiple modalities such as text, audio, and visual inputs, it can perform tasks such as real-time conversation translations across multiple languages that can rival language learning platforms. Adding on to this, with the launch of OpenAI’s new platform, Chat-GPT can now handle 50 languages.
Another advantage that sets ChatGPT-4o apart is, that it’s easily available on smartphones, computers, and other personal devices, so users can access their tool for language practice compared to dedicated language learning apps. Users might choose to practice a new language while being engaged in another activity, rather than setting aside specific times to use language learning software. On top of this, the model’s ability to adapt to analyze and respond to a user’s inputs in real time could make the learning experience more personal.
Last year, Duolingo announced that it integrated GPT-4 into their platform. The partnership with OpenAI has enabled features within the Duolingo Max service, which utilizes AI to provide detailed feedback and simulated conversation practice with virtual characters to adjust in real-time to the user’s strengths and weaknesses.
Luis von Ahn, CEO of Duolingo, emphasized that AI can learn from user interactions to improve teaching methods through real-time feedback to refine the AI’s performance.
This also means GPT 4o can do this too (since OpenAI’s system underpins their apps).
Von Ahn has also articulated that AI’s role in their platform extends beyond mere functionality, since it can create a more engaging and responsive educational environment.
Generative AI has become more sophisticated at complex language tasks—including real-time conversation translation—that other language learning platforms, including Duolingo, are not likely to replicate at this moment (when compared to the capabilities of ChatGPT-4o). I predict that users will turn to multifunctional AI tools that integrate translation into broader platforms.
I think that if AI models like ChatGPT-4o can provide equal or superior language learning experiences, users might prioritize these multifunctional AI tools over traditional language-specific platforms. Duolingo, in my opinion, may need to innovate rapidly to rival ChatGPT-4o and redefine the value they offer to language learners in an AI-augmented industry. Even if they integrate GPT 4o into their tech stack, I believe users may opt for new AI first language services (vs. Duolingo’s traditional interface). For this, the stock valuation feels steep and thus, I recommend a strong sell.
Background
Duolingo has pioneered gamified language learning on an educational platform, but the rapid development of GPT-4o could potentially disrupt Duolingo’s business model by offering a more flexible, accessible, and integrated learning tool. GPT-4o is the latest iteration of OpenAI’s generative pre-trained transformer models, which can handle multiple input and output modalities, including text, audio, image, and video. It has been optimized for enhanced multilingual capabilities.
OpenAI announced that they have improved tokenization across a wide array of languages, so they can process non-English languages more effectively. This is crucial for language learning applications, because the accuracy of understanding and generating text across different languages is what many learning apps are struggling with.
GPT-4o can also respond to audio inputs in milliseconds, so human response times during conversation are sustained. This is important in real-time translation and conversational practice, which is what a language learning must feature.
This is the same reason why the model is a threat to the future of Duolingo, despite the latter having integrated the earlier multimodal large language model version. GPT-4o can understand and generate speech in over 50 languages and can be integrated across different digital interfaces and platforms, so users can engage in language learning contextually than the more structured environments provided by apps like Duolingo. For instance, users might prefer an AI model that can instantly translate languages and provide contextual information within any application or website, rather than switching to a dedicated language learning app. Many users use Duolingo as an app to practice a language on, before taking an international trip. Users may now use GPT-4o directly to build custom study courses or directly help with translation in real-time situational instances.
Another benefit is that GPT-4o can process different types of inputs simultaneously. For example, a user can converse in a foreign language, get real-time corrections, and receive contextual information about cultural nuances or pronunciation tips directly through the AI.
Duolingo’s AI integration last year is a strategy to prevent OpenAI from overshadowing the traditional, gamified learning module they offer. CEO von Ahn believes that AI can improve the personalization and efficacy of language learning, and argues that AI-enhanced platforms can potentially match the effectiveness of human tutors.
Of its 31 million daily active users, new Duolingo Max accounts comprise 10% of the total. Despite these, Duolingo’s stock recently fell by 12% after reporting profitability and revenue growth. This is attributed to the slowing rate of new subscriber additions, marking the slowest record since 2022. In a quote from the earnings call from their CEO:
…the way we see it, we had like ten quarters of accelerating DAU growth and we kept saying almost every quarter like this can’t happen forever -Q1 Call.
Investors are concerned about the company’s growth trajectory, especially as the initial surge in user engagement during the pandemic has now begun to plateau. I believe it’s clear that their AI strategy is not working, given the slowing growth in subscribers. This was before GPT 4o came out.
New Startups Threaten Duolingo
Even outside of a standalone GPT-4o, smaller, more nimble startups threaten growth. New AI-driven startups in the language learning and translation industry have compelled Duolingo to use AI and address their limitations (I believe this was a reactive response and not proactive, which is concerning). These startups leverage advanced generative AI technologies to enhance the speed, accuracy, and efficiency of language translation and learning processes.
For example, Praktika has raised $35.5 million in a Series A funding round to enhance their AI-powered avatars that make learning languages feel more natural. This contrasts with traditional language-learning apps that often rely on more mechanical methods like multiple-choice questions or drag-and-drop exercises. The platform, which claims to have 1.2 million active monthly users across 100 countries and nearly $20 million in revenue since May of 2023, will simulate real-life interactions with AI avatars that can adopt different accents and personalize the learning experience based on user interaction.
The AI avatars can mimic human-like interactions by using varying tones of voice and emotional expressions, and is expected to improve language retention and user engagement through more relatable and less structured sessions.
Another startup that Duolingo faces is IYO ONE. In April 2024, IYO ONE CEO Jason Rugolo delivered a speech at TED to introduce the “audio computer” that can augment the sounds around a user, translate conversations in real time, and naturally respond to one’s voice. He demonstrated the capabilities of the technology through a simulated conversation in a challenging, noisy environment. He explained the use of various software technologies like beamforming, machine-learning denoising, and AI-powered transcription and translation, which help in enhancing and manipulating sounds.
German startup DeepL offers services that include writing, editing, and translation across multiple languages and has secured funding to scale their operations.
Google is also in the race. The company is integrating AI into tools that directly compete with Duolingo by offering real-time conversational practice, a feature also found in Duolingo. Google’s advantage over Duolingo is obviously the massive data and advanced AI capabilities to enable it to offer nuanced language learning tools.
AI’s application can also be found in specialized areas like Manga translation by Japanese firms. The startup, Orange, will launch a tool that will translate Manga into English and other languages faster and cheaper than traditional methods.
Valuation
For me, the biggest issue with Duolingo is that as growth slows, the valuation feels extremely high (especially in light of new disruption risk).
For reference, their forward non-GAAP P/E is 50.14 which is 215.71% higher than the sector median of 15.88.
While right now, paying a 215.71% premium means that you access revenue growth that is 904.99% higher than the sector median (36.40% vs. 3.62%), as management mentioned growth is clearly slowing. And this was before the OpenAI event with GPT-4o.
In essence, I think as growth significantly slows, the P/E multiple will need to come down.
Already we are seeing EPS estimates for the FY 2026 year start to decline. I think this will accelerate as we see the market starts to price in slower growth rates due to reaching critical mass and due to the OpenAI GPT-4o translation directly (or indirectly through AI-enabled startups).
If we see the stock’s P/E come down closer to 25 times forward earnings (still a strong premium above the sector median forward P/E) this would represent over 50% downside in shares. I think this is really possible. It’s clear growth is slowing before GPT-4o. I believe the market still needs to price this in.
Bull Thesis
Duolingo plans to retain their position as the top language learning app in terms of revenue, through improving their offerings and pricing. The company recently expanded their services to include math and music lessons. Users can learn musical concepts like pitch, meter, and rhythm, as well as basic math skills such as calculating tips and hourly wages.
The company is also experimenting with their pricing structures to enhance affordability and accessibility with various pricing points for their Duolingo Max plan, which is currently priced at $29.99 per month or $167.99 annually in the U.S. In order to reach a larger global audience, Duolingo adjusts their prices based on regional economic conditions.
Despite this, there are concerns that will limit their financial performance in the coming months. The company has reduced their contract workforce by 10% to accommodate new AI roles and capabilities.
The company also recently faced a serious data security issue, when hackers exposed the personal information of their 2.6 million users.
Duolingo Must Evolve
I believe that the company must push for more aggressive development of AI-driven features beyond basic gamification, such as enhancing the natural language processing capabilities beyond GPT4. More realistic conversational scenarios that can fulfill the demands of a hybrid model to blend AI with human instructions are definitely a good start. We should consider, however, that developing and implementing advanced AI technologies is resource-intensive and requires hiring top talent, which can be expensive and affect the company’s short-term profitability.
They should also prioritize improving their cybersecurity measures to sustain user trust and comply with the increasingly stringent global data privacy regulations. However, investment in cybersecurity is another unavoidable, yet an expensive cost.
These necessary expenditures come at a time when Duolingo’s shares might already be considered overvalued by some investors (including me). For investors, the cost of Duolingo’s evolution could dampen profits and make shares even less appealing.
Takeaway
I believe that Duolingo is at a point where it must evolve extensively to remain competitive in the educational technology and language learning market (they don’t want to suffer the same fate as Chegg (CHGG)). OpenAI’s ChatGPT-4o has proven to be more intuitive and engaging based on demonstrations, real user feedback, and can redefine the efficacy and appeal of Duolingo’s offerings.
New product developments indicate a shift towards more dynamic and integrated language learning and translation services, where startups create more value propositions through advanced AI capabilities, that could redefine user expectations and market standards. Duolingo may need to innovate and rethink their model to maintain their lead in the market.
Adding on to this, if AI translation becomes even more sufficiently advanced in the coming years, it might reduce the demand for learning new languages through apps and potentially shrink the market overall.
Given these factors, Duolingo’s stock currently presents a less than attractive investment opportunity in my opinion. I think it is a strong sell.
Read the full article here