The Rebundling

A Tale of 2 Exit (types)🚪

My plan for this week was to write about the modern data stack (MDS) rebundling, and that’s what I will do, but man, the acquisition news just keeps coming.

Generally, these acquisitions represent either extensions of larger products into new data services or smaller growth-stage startups expanding their offering to help reduce the number of tools their customers are paying for. Overall, it seems to reflect what many expected to happen with the MDS. There were too many categories, too many products, so some of them had to consolidate or disappear.

Let’s start on the smaller end and work our way up.

Coalesce acquired Castordoc back in March. Castordoc is a data catalog, of the enterprise variety, not one of the other 4 varieties. Coalesce is data transformation platform, a closed-source, enterprise competitor of dbt. Adding a catalog makes sense to me. These are 2 product categories that are adjacent and have a lot of potential for synergy.

Coalesce raised a $50M series B last year and has been growing headcount steadily, 117% in the past 2 years, so it seems they are doing well. Castordoc hasn’t raised since 2022, and headcount had been declining, so it may be that they were actively looking for a home, and Coalesce was a good fit.

Hex acquired Hashboard. Hex has been a darling of the notebook category in the MDS, but was it ever going to be big enough to justify a venture-scale exit on its own? To be honest, I was expecting to write about how Hex was another case of a data startup that was going to have to live up to some sky-high COVID/ZIRP era valuations they hadn’t grown into yet but it turns out they raised a round back in 2023, well into the post-COVID doldrums, so it seems like they must be doing something right. Anyway, Hashboard, which raised a $7M seed in ‘22, gives them a play to go after more traditional BI. Makes sense! Not too earth-shattering and probably not an amazing exit for the Hashboard, as it seems they were probably in a similar position to Castordoc.

Speaking of sky-high ZIRP era valuations, let’s get on to Fivetran’s acquisition of reverse ETL platform Census. Census raised a total of $80M on a $630M valuation in the heat of the most feverish froth for data startups. At its peak, reverse ETL seemed to be the next big category, then that didn’t really pan out. That said, this one again does feel like a natural fit, at least conceptually. Fivetran does integrations to get data out of APIs into the data warehouse, why shouldn’t they also do integrations to get data back into those APIs from the data warehouse? Census had been slowly declining in headcount the past couple of years and hadn’t raised since 2021. As such, it seems unlikely the acquisition got close to the $630M valuation. Perhaps it will work better as a combined product. Others have been skeptical, Ethan Aaron wrote a thoughtful post on this when the acquisition was announced.

So those were the “finding a good home for the technology and team” type acquisitions that happen when things aren’t quite working out.

What’s going on in the world of things that are working out 👀?

Datadog made two big acquisitions to foray deeper into the data space.

First, they announced the acquisition of Metaplane, a data observability platform that raised a $14M seed round. Datadog’s business is observability, so this feels like a natural extension of their product. Given the recent funding, I don’t think Metaplane was looking for an exit, but if a great company like Datadog with natural synergies for your product comes knocking, I can see why they would decide to partner up.

The second Datadog acquisition was Eppo, a feature flagging and experimentation platform. For this one, we have an announced deal amount of $220M. Eppo had raised a series B last year and was growing revenue, so this also looks like a case of a growing company that Datadog wanted to snap up. Presumably to extend its observability stack, so customers can see an incident, roll back or iterate with flags, and prove the business impact, all inside one console.

That was all pretty big news, and I could end here, but there are some additional huge rumors swirling this week.

Databricks is said to be in talks to buy Neon for $1B. Neon is a serverless Postgres offering. I think the heart of the story here is that we’ve been seeing charts like the below coming out the past few months:

Because Neon makes it extremely easy and cheap to spin up new databases. It makes the perfect complement to AI coding agents like Bolt, Lovable, etc. Those services are creating tons of new apps powered by Neon databases. If you’re Databricks, you have a great data lakehouse offering but no OLTP offering. If you buy Neon, customers could run their app’s production database on your platform and you can naturally capture their analytical and AI workloads. Now this is just a rumor so it’s very possible that the deal doesn’t go through, as it seems the Redpanda/Snowflake deal I mentioned in my first newsletter did, but we will see.

Oh, by the way, in keeping with my apparent commitment to mention Cursor every week, they turned down a massive $3B+ offer from OpenAI. OpenAI then turned around and bought Windsurf for $3B. Cursor then announced a $900M round at a $9B valuation. While it seems wild for Cursor to turn down what must have been a meaningfully bigger offer than Windsurf got, I have to say I get it. Cursor has gotten to $300M ARR faster than any company in history, and $ 3 B would only be a 10x on current ARR. Given their incredible growth rate, the $ 9 B valuation from their funding round is actually a pretty reasonable, 30X current ARR (subject to the caveats expressed in my early post on this topic). I mean, how do you value a company that is doing numbers like this:

$1-$100M ARR - 12 months
$100-200MARR - 3 months
$200-300M ARR - 1 month

Interesting times!

BTW, I ran out of time to write in depth but data.world was acquired by ServiceNow this week as well.

See you next week!

Warmly,
Paul Dudley