I’ve tried to post three times now on open source, but I keep rambling on about too many details as it’s something I’m passionate about. But with people talking about Confluent and Hashicorp, I feel I need to post something.
Essentially my thesis is that open source companies can’t be great companies. Don’t get me wrong, I’m not against open source. I’ve worked for multiple open source companies. Open source it is a great model from a technology perspective. And from a customer perspective. And you can have a good open source business model. But there are some fundamental flaws that make me think no open source company can be a great Saul-level company. Basically it’s just impossible to have a big enough moat.
There are essentially three surviving open source business models.
-
“Pure” subscriptions. Give the software away completely for free and just charge for support, services, and conveniences. This is very difficult because it’s hard to prevent “free riders”. Essentially Red Hat is the only large scale example of this and there are some unique factors in play for them. Arguably they are the exception that proves the rule.
-
“Open Core”. Give the core of the software away, but keep certain features proprietary so that you can upsell. This is still common, but is almost impossible to succeed with. Firstly, because it’s a tight rope. If you keep too much proprietary then you don’t get the ubiquity and benefits of open source. But if you keep too little then you have the problem mandeville describes: no one pays. Plus, since the base code is open, the “community” can always step in and write their own competing versions to your proprietary features. (This happened to one of the open source companies I worked for, someone wrote an open source alternative to one of our commercial add-on features. It had a huge impact on us.)
-
“Hosted Service”. Give the software away, but sell a hosted version of the software. Hailed as being the savior of the open source business model, due to the success of companies like Databricks it has actually proven to be full of challenges as well. Yes, products like Atlas and Confluent Cloud have had some great initial growth numbers, but the nature of open source means that it will be hard to have a moat. The hyperscalers (and others) have jumped in to compete. And since they just want to sell compute, hyperscalers can set prices ridiculously low. There have been some attempts to prevent this with restrictive licenses, but that approach is fraught with challenges as well.
But fundamentally, my big issue with open source is that you end up competing with your biggest customers. Fundamentally the open source business model is about selling expertise. Confluent says “Why bother hosting your own Kafka when you can have the experts at Confluent host it for you?” Elastic says “Why not pay for subscription to Elastic so that you have Elastic’s expertise when you have an issue?”
The answer to “why not?” is cost. If I have one set of Kafka brokers, it makes sense to pay for help from Confluent. But if I have a huge number of brokers, and Confluent wants to charge me millions of dollars, it starts to make sense to just hire my own experts and use Kafka for “free”. I believe this is why we’ve seen fast growth for total customers for Confluent, but slow growth in the numbers of large customers. Similarly, if I am Microsoft, and I have only a few customers interested in a product, it makes sense to partner with someone like Mongo or Elastic. But once a threshold has been passed, I’m not going to partner with DataStax, I’m going to create my own version: Cosmos. I’m not going to partner with Confluent, I’m going to create my own version: Event Hub.
Open source is (almost by definition) about guaranteeing that no company can gain a moat. Which is why, although I might like open source as a buyer, as an investor I prefer companies like Crowdstrike, Monday and Datadog that can gain a lasting proprietary differentiation via their “secret sauce”. If the secret sauce is open, it isn’t very secret and it’s too easy to copy once the profits start rolling in. I am aware that the numbers for Atlas and Confluent Cloud look good. But I think they will hit a ceiling.
–CH