I had some thoughts about Canonical as a company. Here they are—

Enterprise Support

Canonical barely even “owns” any software: it’s all open source. Any company could run Ubuntu, or OpenStack on their own hardware, and never pay a cent in support.

Yet, with $205M gross revenue at a 82% margin, Canonical is clearly doing something right.

The success of Canonical reveals a deep truth about what I’ll call the stack—the infrastructure code that powers the world’s applications and business requirements.

Nobody wants to worry about the stack. Nobody cares that they run on Linux vs FreeBSD, or musl vs glibc. Nobody cares what container tech they use, as long as it works without them even thinking about it.

Any time spent away from the business’s core product is a serious and risky decision. Worrying about the stack, as many hobbyist Linux users know deep down, is a terrible waste of time. So, companies with serious technology requirements let Canonical figure out the deployment problem for them, and happily pay the fee.

Ubuntu isn’t perfect. Canonical doesn’t promise this either. But it doesn’t have to be: it just has to be good enough. It turns out paid support is an acceptable substitute for perfect, bug-free software.

Mission

Canonical’s mission focuses on the standard FSF tenets of free software: freedom to download, run, copy, distribute, study, share, change and improve the software free of licensing restrictions. I would argue that this is at the core of Canonical’s strategy to dominate.

You can find hobbyist Linux distributions like Debian that live and die by these principles, even more strongly so. However, Ubuntu is unique in that it can be used license free, along with enterprise support.

This has allowed Ubuntu to proliferate the server world, and has also allowed Ubuntu to reap the network effects of being the #1 server operating system in the world. Canonical doesn’t want to make money on software licensing—that would limit Ubuntu’s reach. Besides, recurring support revenue is a more enticing business model.