Commercial open source firms make money by selling something that they don’t give away for free. If you’ve been following my writing or even attended my **open source business workshop **you know that I’ve been calling what companies sell the closed complement. Closed, because customers don’t get it for free, and complement, because it somehow sits next to the open-source software so that charging a price doesn’t upset the user community.
For most of the new commercial open source firms since the financial crisis, the closed complement has been the provision and operation of the open-source software as a cloud service. It is closed, because you don’t make the service available for free, and it is a good complement, …
Commercial open source firms make money by selling something that they don’t give away for free. If you’ve been following my writing or even attended my **open source business workshop **you know that I’ve been calling what companies sell the closed complement. Closed, because customers don’t get it for free, and complement, because it somehow sits next to the open-source software so that charging a price doesn’t upset the user community.
For most of the new commercial open source firms since the financial crisis, the closed complement has been the provision and operation of the open-source software as a cloud service. It is closed, because you don’t make the service available for free, and it is a good complement, because it is obvious that service provision is somehow separate from software and costs money.
But times are changing. All around me companies are vocally complaining about cloud service costs. Also, there is a clear trend back to on-premise and digital sovereignty. Software-as-a-service as a revenue source, whether for open-source or closed-source software, isn’t going away, but others are rising and might well take over.
Primarily, in this day and age, I think the new closed complement will be based on data and compute that massages the data into a particular form. Right now this is preparing LLMs for AIs. This then I expect will be a strong new closed complement:
**The update service to an ever evolving and improving LLM. **
You can make the software that gathers and prepares the data and creates or updates the RAG and LLM open source. For 99% of customers, operating this pipeline themselves will be too much. They’d rather subscribe to a commercially run pipeline that delivers them the updates onto their premises.
One important aspect of commercial open source is how to keep the competition at bay. Here, nobody stops you from having not only a particularly efficient pipeline, but lots of secret sauce in it as well. A competitor, who tries to clone your update service using your open-source software will be hard pressed to deliver even on-par results.
Voilá, the times are a-changing.