ChatGPT Has a Serious Problem and It's Not Its Executive Team...

A lot has happened to everyone's favorite AI over the last couple of weeks. The big executive/board battle is what most talk about these days, but there's something a bit more interesting - in my opinion - going on that's gone unnoticed.

Right after OpenAI's explosive demo day, and before Sam Altman became famous for getting fired over a Meet, OpenAI stopped accepting new paying subscribers for ChatGPT.

That's right, they simply said no to new money.

According to a tweet by Sam, they had to pause because they couldn't sustain growth - a great problem to have.

But my immediate thought is: "How much money are we talking here?"

And as I started looking at the numbers I realized that we can learn something very important from this. We can kind of reverse engineer ChatGPT's revenue churn.

That's pretty interesting!

Disclaimer: Before I dive into the numbers, I should make it clear that I'm looking at ChatGPT's mobile revenue here and not all revenue. Just what the mobile apps make from in-app purchases on the App Store and Google Play.

Losing Hundreds of Thousands Every Day!

We analyzed ChatGPT's revenue growth over the last few months, including the steep increase it saw before demo day, and estimate that since the shutdown, OpenAI has lost roughly $127K in net revenue (or roughly $181K in consumer spending).

In the two weeks since the shutdown, that amounts to nearly $2M. And this estimate looks at net revenue, which is what OopenAI would have kept after Apple and Google take their fees.

The loss comes from not being able to get new paying users and also current paying users churning (canceling their subscriptions).

That's A Lot of Churn!

For a long time now I've been saying that while ChatGPT is amazing, many simply don't know how to get value out of it. OpenAI stopping new paying subscriptions gives us a rare opportunity to estimate ChatGPT's churn.

If there was no churn, ChatGPT's revenue would simply look like the previous month as subscriptions are renewed.

That, however, isn't the case. The difference between the two is the churn. Roughly.

Looking at roughly two weeks of data, we calculate ChatGPT's revenue churn to be 14%. There's no absolute number of "good" churn, but by many counts, 14% isn't good at all. It's pretty bad - and that aligns with the theory that people don't know what to do with ChatGPT.

The fact that while subscriptions are on revenue continues to grow at an amazing speed is a testament to how exciting ChatGPT looks from the outside, but without more, retention is hard.

A Big Opportunity for Developers

OpenAI's GPT is an amazing product. As-is, it's a hard sell for the mass market, but only because it can do too many things. Focused products that use AI are already making a lot, and that will continue to grow.

Now, I'm not talking about basic ChatGPT wrappers that don't add value - those are slowly disappearing. I'm talking about actual tools that do something with AI.

If you're a developer and you're not looking for ways to harness AI, you're missing out.