I overheard a fascinating conversation between two engineers at a coffee station here at KubeCon this week, where one of them said that whilst there were plenty of vendors on the show floor with interesting solutions, he would only ever consider software that is 100% open source; before he spends any time understanding what a product actually does, he wants to know if it is OSS. He even said that he is perfectly happy to pay for software, as long as it is open source.
It is a really interesting perspective (and I suspect a very common one at a CNCF event), but it left me with a question I could not shake on the walk back to our booth. If the software is open source, and AI is driving the cost of developing software toward zero, why would anyone pay for the original when they can take an OSS project, throw the code into AI, and have it add whatever additional features they want? It feels a lot like KFC publishing the eleven herbs and spices, and then wondering why thousands of KFC clones have popped up and nobody buys their chicken anymore.
For most of the history of commercial open source, the code being free did not matter all that much, because the real barrier was labor. Understanding a large unfamiliar codebase, adding enterprise features (SSO, RBAC, audit logging) and then keeping all of it running took a team of engineers that cost far more than the subscription; open core vendors (like Portainer) built entire businesses in that gap.
AI is narrowing that gap very quickly. An engineer with a capable coding agent can now read their way into an unfamiliar codebase and bolt on an SSO integration or an RBAC layer far faster than an engineering TEAM could two years ago. So if the paid features are the only thing separating the open source edition from the commercial one, and those features can be generated in a week, what exactly is the buyer paying for?
The traditional defense is that creating a fork is cheap and owning one is not; every upstream CVE is yours to merge, every upstream release is yours to rebase your custom features against, and when something breaks, it is your engineer responsible for the fix. The catch is that the same AI that writes the features can also triage the CVEs, rebase the patches and explain the code back to the on-call engineer; if the cost of ownership collapses the way the cost of creation already has, the gap that commercial open source lives in closes from both sides. I genuinely do not know where that lands, and I am skeptical of anyone at this conference who claims they do.
What does our coffee station engineer actually want? I think he wants insurance. If the vendor gets acquired, triples the price, or changes the license, he is not stranded; he holds the recipe, and he can walk away. That is a completely rational position, and I understand why it is so popular in this community.
So what is the moral of this story?
I think that the value of both open source and commercial software is shifting away from the code itself, toward innovation. Are you driving innovation into the market, or are you merely delivering what is needed today? If you are driving innovation, that train cannot stop, ever… Customers need a vendor that deeply understands their market, their problems, and has built a perspective across hundreds or thousands of customer engagements. Nobody asked for the iPhone before it existed, and the same is true of most software that changed a market. Does this turn the software into a “feature factory”? Only if the people designing the software are disconnected from the market, as if building features is near free, whats most important is knowing WHAT to build, and only experience can tell you that.
Open source is going to have a pretty significant challenge too. I wonder how many PRs get opened against the original code, with features contributed by a 3rd party, vs a fork being created and the fork exclusively contains the changes. Should we be prepared to see hundreds or even thousands of variants of OSS products? I think that’s likely.
