This seems like a lot of money? I like Streamlit fine but I don't see $800m of value in it or even its potential.
What am I missing?
On the other hand, definitely encouraging me to go start a side hustle building some ML-tinged BI app on top of Dremio or Fugue or Materialized or Datadog or some other contender for data platform du jour and just hunt down Fortune 500 users to give it away to in exchange for (9 figure!) clout.
It's par for the course of snowflake really not understanding data visualization or how their end users need a UI to be (no find replace still for example!)
We changed the URL above to what seems to be the best third party article (for reasons I was explaining in a different thread earlier: https://news.ycombinator.com/item?id=30531133).
I've been using Streamlit for roughly half a year to build various dashboards and GUIs for custom analysis scripts and I am honestly neither surprised that they were acquired nor by the price tag.
Our data is quite complicated and needs a lot of custom code & filtering. Even relatively simple analytical questions need some custom code, if people don't want to do almost everything manually (i.e.: Origin, Export to Excel, crop and move around data...).
Over the years I built up libraries and functions for some of these tasks and with Streamlit I can just slap a sufficiently nice interactive GUI on top in 1-2 hours and iterate the analysis live during a call (give them the URL and ask what additional filters, visualizations they would need, adapt the code in the background and they see the results instantly).
Overall, congratulations - well earned.
Also, I considered applying for a position there and regret not doing so.
Congrats to the Streamlit team. My every interaction with them has been positive. Add to that the great product and community they’ve built, and you get a company that’s deserving of this outcome.
I'm knee deep in the data tooling space myself and it's quirky at first but not too surprising.
This resembles the Google Cloud acquisition of Kaggle, or Microsoft's acquisition of Github. For better or for worse, these larger players / platforms are buying a large community around the hip tool. Snowflake wants to be a giant data platform company, not just "another data warehouse".
No stranger than Confluence acquiring Chartio. Snowflake has been making some interesting ventures and investments (eg Sigma Computing).
As a user of data tooling, bundling data pipelining, cloud warehousing, visualization, and MLOPs into a consistent environment is helpful and lowers tooling cost. Streamlit is frankly fantastic and one of my personal favorite tools.
Source: Myself, and I have family that run Datateer, an up and coming pipelining/analytics ops player in the space.
Yeah it definitely would. Historically, to OP's hesitation, databases & data warehouses haven't purchased BI tools, application layers, etc. But this is starting to change (e.g. Databricks bought Redash, an open source BI platform)
I think it's a key area snowflake have been lacking. Built in nice dashboards that are as easy to throw together as a streamlit app would be a killer feature. Their lack in this area was a key issue for me.
True, I’d say that BI is saturated with a number of well built products, but perhaps the purchasers of those products aren’t fully aware of what’s available?
Yeah the market is super fragmented but enormous. Even huge players like Tableau have a small slice of the market. It's really wild-west feeling and no one is confident they know what they're doing when it comes to implementation/ops.
A few months ago I tried building a configuration manager for robusta.dev using streamlit.
Basically I wanted to autogenerate a frontend for arbitrary Pydantic models (in my case configurations of Kubernetes automations). The tooling was still lacking so I abandoned the project, but it was really cool just how far I could get with writing almost no code. A way better experience than all the no code tools IMO
Snowflake is basically GCP BigQuery or AWS Athena, but they charge $40/TB of (highly compressed) storage and have a decent margin on compute too. It's also completely proprietary, so you're not exiting their platform anytime soon. The other big player in this space is DataBricks, which is a reasonable product, but their pricing is based on both instance type and volume of data consumed by a query, which can get pretty insane quickly when dealing with large amounts. Again, they rely on lock-in to a significant extent.
Snowflake employee here, speaking on my own behalf.
It's no harder to get out of Snowflake than it is to get out of any other SQL based RDBMS. If we were relying on the fact that we manage the files in S3 instead of the customer to keep our customers on board, we wouldn't have a business.
I will grant that there's non-ANSI features in Snowflake that our competition don't offer, but it would be weird to describe "have awesome features" as being lock-in.
They are doubling revenue every X months (currently about 12). And it's a big market and getting bigger, so they might be able to keep X pretty low for a pretty long time. They are adding customers + adding revenue at existing customers. Double revenue every 12 months and you get to 32x in 5 years. And it's pretty sticky.
Yup, it's interesting stuff. The efficiency claims (ie for customers) seems believable (like maybe 70% true?) so it will be interesting to see where those growth numbers go from here... up/down/sideways, it all seems very possible
What am I missing?
On the other hand, definitely encouraging me to go start a side hustle building some ML-tinged BI app on top of Dremio or Fugue or Materialized or Datadog or some other contender for data platform du jour and just hunt down Fortune 500 users to give it away to in exchange for (9 figure!) clout.