Fauna

» see fauna.com as it lived, on the wayback machine

~ the obit ~

Fauna was a database with no server you could point at. You signed up, got an endpoint and a key, and from then on your database was an address on the internet that answered questions. There was no cluster to size and no version to upgrade on a Sunday night. It reached the public as FaunaDB in 2017, and the company had been building toward it since 2012, when Evan Weaver and Matt Freels left Twitter. Weaver had been employee number fifteen there and ran infrastructure through the years when the site falling over was a running joke, which gave the pitch a certain authority. These were people who had learned at three in the morning what a database actually does.

The technical claim was about transactions. Most of the hosted databases of that era asked you to give up consistency in exchange for working across regions, so your app in Frankfurt might read something your app in Virginia had already changed. Fauna said no, you can have both, and shipped its own query language to prove it. FQL was not SQL and it was not Mongo either. You wrote queries as nested function calls, which people either liked very much or bounced off entirely, and that split turned out to matter more than anyone expected.

The money arrived in the shape the pitch deserved. A $25 million Series A in 2017, led by Point72 Ventures and advertised at the time as the largest round ever raised by an operational database vendor. Another $27 million in 2020, led by Madrona, with GV and GitHub founder Tom Preston-Werner on the list and Bob Muglia taking the chairman's seat. Call it $60 million across the life of the company, against a market where the incumbents were free and forty years old.

I never put anything in Fauna. My work is computer vision on aircraft, where the machine has to answer before the next frame lands and frequently has no network at all, so a database that exists only at someone else's endpoint was never on the table for me. There is still something uncomfortable in reading the shutdown notice. The whole promise of a managed database is that you stop thinking about the machine, and then at noon Pacific on a Friday the machine is not there.

~ the rap sheet ~

~ who it was for ~

Fauna's natural customer was a small team shipping a web app with no operations person on staff. That was a real population in 2019 and 2020, when the fashionable way to build a site was a static front end on Vercel or Netlify plus a handful of functions, and the one piece nobody had a clean answer for was where the data went. Fauna was the answer in a lot of those tutorials, and its own homepage carried Vercel's wordmark up in the integrations row. That is the crowd it was courting, and that crowd is also where the problem lived. Teams small enough to want a database with no operations are usually too small to pay very much for one, and teams big enough to pay are usually big enough to hire somebody who likes running Postgres. Fauna needed to cross that gap. It never did.

The numbers make the shape of it visible. Fauna's marketing claimed more than 80,000 development teams. Its own goodbye letter, written by the same company ten weeks before the lights went out, said thousands of development teams and hundreds of paying customers. Both of those are probably true, and the distance between them is the entire business.

~ the cracks ~

Postgres came back. That is the shortest version of what happened to the whole category between 2021 and 2024. A run of companies worked out how to sell managed Postgres with the serverless billing model and none of the relearning, so a developer could get the convenience without adopting a new query language, and FQL went from being Fauna's differentiator to being the tax you paid for choosing Fauna. Every proprietary language is a bet that the product will outlive the cost of learning it.

The end came on 19 March 2025 with a post titled "The Future of Fauna". Hacker News found it the same day. The company's own end-of-life document gives the reasoning without decoration: driving broad based adoption of a new operational database that runs as a service globally is very capital intensive, and in the current market the board and investors determined it was not possible to raise the capital needed to achieve that goal independently. No acquisition, no quiet absorption into a larger vendor. The investors simply stopped.

~ what is left of the code ~

Fauna promised to open source the core database along with the drivers and the command line tools, and it did. What it could not open source was the service, because the service was the product: the global replication, the endpoint that was always up, the thing you were actually buying. You can read the code that ran your data today. You cannot ask it anything at the address your application was written against. The service was turned off on Friday 30 May 2025 at twelve noon Pacific time, and accounts were deleted. Anyone who had not finished migrating by then watched their application start returning errors on a schedule published two months in advance.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »