Speak: The $1 Billion Startup building DuoLingo on Steroids

In partnership with

(5 minutes)

Hi 👋, this is the Today in AI Newsletter: The weekly newsletter bringing you one step closer to building your own startup.

We analyze a cool, industry-shaping AI startup every week, with a full breakdown of what they do, how they make money, how much they’ve raised, and the opportunity ahead. 

Let’s get to the good stuff in this email: 

• This startup is helping helping people speak a new language using AI

• They’re doing tens of millions in ARR, and are on track to profitability

• They’ve just raised a $78 Million Series C at a $1B valuation led by Accel

So what’s the startup and who are the founders behind it? Here’s the story of Speak 📈

Speak was founded in September 2016 by Connor Zwick and Andrew Hsu with the mission to build a truly superhuman AI tutor. 🦸

Speak is a full fluency solution for learning another language.🗣️ In other words, it helps people learn another language through actually speaking it.

It’s a mobile app helping people learn how to have real world conversations and communicate with other people using a new language. 

Speak has a whole pedagogy behind the way learning works. 📚

What they do is teach people chunks of high-frequency words that always come together in everyday speech. 

They then teach this to people and have them say it a ton, until it becomes a habit.

People then develop these patterns and chunks of words that they can pull out and practice in simulated conversations. 💬

These conversations are based on a real goal that’s tied to your motivation to learn the language in the first place.

This makes it extremely tailored to each user based on their interests, motivations and current speaking level. 💫

And clearly, this is working. It works better than other learning methods that are more grammar-focused.

Connor and Andrew knew that over the next 5 to 10 years, with more data and compute, models will get so good that they’ll be able to fully replace humans in the learning process… 📇

2025 Prediction: A Surge of Self-Serve CTV Buyers

Roku predicts that 2025 will be a breakthrough year for self-serve CTV advertising. Roku Ads Manager makes it easy to integrate CTV into your 2025 marketing mix. Easily segment your target audience, optimize campaigns in real-time, and drive conversions with interactive ad formats and shoppable ads with a Shopify integration. Roku Ads Manager makes CTV advertising accessible and impactful for businesses of any size.

Discover CTV performance on Roku

Backstory 👀

Ever since he was in school Connor was always interested in computers, how they worked and he was super entrepreneurial. 🧱

When he got his first iPhone, he realized that it was the same size as an index card. 📱

He had an idea… Replicate flash card studying on the iPhone instead of having physical index cards.

In high school, he went on to start FlashCards+. By the time he got to Harvard, the app had over 5 million users. 👥

He then sold the company at age 21, and dropped out of Harvard one year in for the Thiel Fellowship. 🏆

This is where he met Andrew Hsu, a super smart kid who had 3 Bachelors degrees by the age of 16. 📜

The public school system wasn’t for him, so he got homeschooled and sped through the curriculum.

At age 12, he attended the University of Washington to study Biochemistry, Neurobiology and Chemistry.  🧬

He then went to Stanford at age 16 for his PhD 🧑‍🔬 and 3 years in, decided to drop out for the Thiel Fellowship.

The 2 hustlers were both interested in AI so grew very close. 🤖 They both took a year to do a deep dive into Machine Learning.

Initially they thought Computer Vision was the best route to go down, but they then looked into Speech Recognition.

This was a form of tech you could actually build a relationship with and there was a persona behind it. 🧑‍🍼

Connor and Andrew then built an entire speech recognition system that not only understood what people were saying, but also their accents. 🔊 They did this only using random data on Youtube…

They then thought “Can we build a language learning experience that people will use at all?” 🤔

The Hustle 🤑

The youngsters then raised a seed off their MVP with zero revenue. 💪

For the first few years it’d been a struggle to find PMF, so they decided to niche down to one market: South Korea. 🇰🇷

At one point 1% of South Korea’s GDP was spent on learning English - it just made sense to try it.

Their first day on the App Store, 3 people paid and they made $18 in revenue. 😢

Their value proposition then switched to help South Koreans “speak more English in the first 20 minutes of using Speak than they have their entire lives.”

PMF clicked and every KPI skyrocketed. 🚀

Fast forward to 2022, Speak was the #1 education app in South Korea and was being used by 6% of the population. 🏆Over 3 million users…

Since then, they’ve started expanding to over 40 other countries including the US and Japan. 🌎

According to Connor, newer markets are doing even better than Korea. Japan, for example, is now giving them more new ARR than Korea.

They’re also now expanding B2B through selling to Children’s Education and Universities. 🏫

Stats 📊

📈Speak has now had over 10 million users, a user base that’s at least doubled every year for the last 5 years.

⏱️This year alone, Speak has created over 25 million personalized lessons, with an average daily usage of 10-20 minutes.

🗺️The startup now has over 75 employees with offices in SF, Seoul, Tokyo and Ljubljana.

🧑‍💼 They also now have over 200 enterprise customers with an 85%+ adoption rate across employees. This includes 8 of the top 10 largest employers in Korea…

💸According to Connor, Speak is now doing tens of millions in ARR with hundreds of thousands of paying subscribers, and is approaching profitability.

💰On December 10th 2024, Speak raised a $78M Series C at a $1 billion valuation led by Accel with participation from the OpenAI startup fund, Khosla Ventures and Y Combinator bringing their total raised to $152 million.

Speak vs DuoLingo 🦜

Speak sounds awfully similar to DuoLingo. So what does it do different?

DuoLingo’s main audience are native english speakers: people in America, UK, Australia etc. 🗣️ The majority of their subscribers were also not learning a language before using DuoLingo.

DuoLingo’s a lot more casual…

Speak’s main audience at the beginning were foreigners trying to learn English; they’re now trying to expand so that anyone can learn any language using their platform. 🤓

Speak’s a little more serious where they’re solving the bigger problem of connecting humans around the world both culturally and socially. 🎉

Before you go 👋

Run CTV Ads on Roku This Q5

Peak shopping season isn’t over yet! Two thirds of consumers plan to shop the week after Christmas, and “Q5” – the period between Christmas and mid-January – has become a significant shopping window.* Roku Ads Manager makes it easy to run self-serve CTV ads and reach shoppers where they’re streaming post-holidays. Get started for as little as $500 and find your next customers on the big screen. (*National Retail Federation, 2023)

Discover CTV performance on Roku

Thank you for reading.

On a side note, Happy New Years everyone and enjoy the holidays!! 🥳 

Know someone who'd love this newsletter? Share the love.🌟