The effect of vibe coding tools on BaaS adoption
Supabase recently tweeted its growth numbers: it took Supabase 5 years to get its first million users, 4 months (!) to get its second million and 3 months to get its third million.
That sort of non-linear growth is unusual in BaaS, and in a follow-up tweet Supabase CEO Paul Copplestone clarified that much of the growth comes from AI Builder/Platforms.
In an interesting discussion Firebaser Michael Bleigh pointed out that Firebase is seeing renewed growth in usage too since they added a vibe coding layer to Firebase Studio.
Disclosure: I worked on the Firebase team from 2015 til 2024. Like many teammates, I always rooted for Supabase (and other BaaS providers) to be successful too. More choice for developers is always a good thing. But if there’s a bias in this article, my employment history probably explains it. 😳
I work in a different segment of the developer tooling space now, building tools that make it easier for developers (both traditional coders and others) to build beautiful cross-platform apps - and our 4-year old FlutterFlow product provides connectors for both Supabase and Firebase1.
Such connectors deliver sort-of reseller flow to the backend systems. Most apps built on these tools will need to store their data somewhere in the cloud, and if the tool provides a pre-built integration with a certain backend - the builders are more likely to pick that than roll their own.
The AI tools/platforms that Paul mentioned and the vibe coding connectors that Michael mentions are also tools that make it easier to build apps, but target an even broader audience than FlutterFlow does.
No matter your opinion on how sustainable these vibe-coded apps are, it’s clear that there’s a boost in the number of apps that are being built that we haven’t seen since… I don’t know, maybe when VB6 was released? And many of these apps will use a backend, typically the one that is most conveniently integrated into the tool.
What I am most curious about is how “sticky” these new apps are going to be.
It’s no big secret that most of the apps that are started on Firebase are never shipped, they’re abandoned (but not deleted). Even of those apps that ship, the vast majority of them never become successful. That’s also why Google can offer a (still reasonably generous) free tier for Firebase: apps that don’t see significant usage also consume (almost) no backend resources (yay NoSQL)!
This assumes incredibly optimistic 10% completion and success ratios
I have no data on what percentage of apps that are started on Firebase or Supabase ever ship any version, nor on what percentage of shipped apps become successful. I just know that the numbers are low, and multiplying them leads to a very small success ratio - which is where the revenue for the BaaS providers comes from.
If you then look at the resource consumption and assume that each created app consumes resource unit (e.g. database usage) for just the developer/team, each published app uses 100 resources, and each successfull app uses 10,000 resources, we see this pattern:
This assumes 100x multipliers for resource consumption across the tiers
There’s a lot still missing in this simple chart and the ratios are probably not accurate, but it illustrates the point: the revenue (and cost) of a BaaS comes from a small fraction of the apps, all the others combined are just a small percentage.
The charts above use a linear growth rate of apps built: after all, the number of developers in the world is slowly growing and at best linear. Or at least… it was. As the charts for Supabase growth above show, the number of apps being built is now growing exponentially - because there is a vastly expanded pool of people building apps with the new tools.
Even though I no longer work on BaaS myself, I am extremely curious to see how these success ratios are going to be in this age of vibe coding. I expect that we’ll see more completions, so apps that ship to production in some form. But will these apps be successful? Will they see enough usage to justify the costs of the BaaS provider? Will they be sticky enough to keep the users coming back?