AI just collapsed the cost of building software
In 2026, nearly a trillion dollars evaporated from software stocks. Hedge funds made billions shorting companies like Salesforce, HubSpot, and Atlassian. Meanwhile, a CNBC reporter with no engineering background opened Claude Code, described a project management tool in plain English, and had a working prototype integrated with Gmail, Stripe and a vector database in 45 minutes.
Traders are calling it the SaaS apocalypse. They are only seeing half the story. It’s not that software is dying. The economics of building software have fundamentally changed, and the consequences of that shift will play out very differently depending on which side of the proprietary divide you stand.
If you build proprietary, closed-source SaaS, this is an existential crisis. If you build OpenSource, this is the biggest opportunity in a generation. But it will require adaptation, speed, and a willingness to let AI into the workshop. When software creation costs collapse, proprietary software stops being a defensible product and becomes a commodity.
The creation cost collapse
The Mindset AI team recently published an excellent analysis drawing parallels between what is happening in SaaS and what already happened in music, publishing, and photography. The pattern is always the same: when the cost of creation collapses toward zero, the volume of what gets created explodes, per-unit economics disintegrate, and the industry restructures around whoever controls distribution and whatever cannot be commoditized.
In 1972, the Rolling Stones spent millions recording an album. Today, an entire album can be produced on a laptop. The music industry did not die, but it was completely rebuilt, and many of the old gatekeepers did not survive the transition.
Software is now entering that same phase. The difference is that the timeline is compressed. What took the music industry two decades to go through may play out in software over a handful of years. For most of the history of software, the hardest part was writing the code. That may no longer be true.
The underlying mechanism is simple. When the cost of creating something collapses, the competitive advantage shifts away from creation and toward distribution, ecosystems, and network effects. The same thing happened with music, photography, and publishing. Software is now entering that phase, except the creative agent is no longer just a human developer. It is a human developer paired with an AI system that can generate, refactor, and extend software at unprecedented speed.
In that world, proprietary licensing stops being a moat and starts looking like friction.
The DocuSign test: how fast can you be disrupted?
I have a friend who built Holosign.co, a fully functional e-signature platform. Drag-and-drop field placement, multi-party signing with controlled order, SHA-256 audit trails, legally binding under both the ESIGN Act and eIDAS, team management with roles and permissions. Basically the core of what DocuSign sells.
Holosign charges $19 per month. Not per user. Per organization. Unlimited users, unlimited documents. DocuSign charges $15 per user per month for their starter tier, and $45 per user per month for anything resembling team features. For a ten-person team, that is $450 per month versus $19.
And here is the thing that should keep every SaaS CEO awake at night: someone else could undercut even that $19. The tools to build this kind of application are becoming so accessible that it is only a matter of time before someone spins up an OpenSource variation, slaps an AGPL license on it, and gives it away. In a world where software can be recreated in days, the AGPL may turn out to be the most economically disruptive software license ever written. Not because they are altruistic, but because the cost of building it has dropped to the point where giving it away and monetizing through services, hosting, or support makes more economic sense than trying to defend a proprietary moat.
That is the trajectory for the entire SaaS industry. Not for one vertical, not for one niche. All of it. When the cost of building software approaches zero, the idea of charging per-seat licensing begins to look absurd.
And this is not just happening in lightweight SaaS tooling. In enterprise infrastructure, Proxmox is rapidly replacing VMware after Broadcom’s acquisition turned licensing into a hostage situation. Proxmox downloads have exploded since Broadcom’s VMware acquisition, with enterprises increasingly migrating production workloads to an OpenSource virtualization stack. Choosing Proxmox is now the pragmatic choice.
OpenSource is the endgame, but the road is rough
Here is my prediction, and I will be blunt about it: the vast majority of what we currently know as commercial SaaS will become OpenSource software within a relatively short time interval. Not because of ideology. Because of economics.
When creation costs collapse, the proprietary model loses its primary advantage. You cannot charge premium prices for something a motivated developer (or a motivated developer with an AI pair programmer) can replicate in days. The value moves from the code itself to the ecosystem around it: hosting, integration, support, customization, and compliance. That is exactly the model OpenSource has been refining for decades.
And increasingly, the AI tools driving this shift are themselves OpenSource.
But this transition will not be smooth. Not at all. In fact, the pain has already started.
FOSDEM 2026: the canary in the coal mine
But the OpenSource world is not entering this transition unscathed. If you want to understand the tension the OpenSource world is already feeling, you only needed to be in Brussels at the end of January. FOSDEM 2026, the annual gathering of over 8,000 free and OpenSource software developers, was dominated by a mood that mixed excitement with genuine anxiety. The keynotes leaned heavily into sustainability, funding crises, and what multiple speakers described as “adversarial AI.”
The most visceral example came from Daniel Stenberg, creator and sole lead maintainer of cURL, a tool that runs on over 10 billion devices. The cURL project had been running a bug bounty program through HackerOne since 2019, paying out around $86,000 for legitimate security findings. Then generative AI tools went mainstream, and the project was flooded with what Stenberg calls “AI slop”: long, confident, technically detailed vulnerability reports that are completely fabricated. One report described a critical HTTP/3 stream dependency cycle exploit that, if real, would have been catastrophic. It referenced functions that do not exist in cURL. The AI had hallucinated the entire thing.
By late 2025, only about one in twenty or thirty reports was accurate. Stenberg described the situation as a denial-of-service attack on his volunteer security team. In January 2026, cURL shut down the HackerOne bounty entirely. Stenberg was blunt: the project needed to ensure its survival and the mental health of its maintainers.
This is not an isolated case. The FOSDEM hallways were full of similar stories. A talk titled “The Synthetic Senior” addressed the growing dilemma of welcoming new contributors while navigating an influx of high-volume, low-context pull requests that demand deep review time. Traditional mentorship does not scale to this volume. The proposed solutions were telling: “helpful friction” like intentional barriers to surface genuine contributors, self-serve onboarding with automated pre-flight checks, and healthy boundaries around closing contributions that lack necessary context.
The mood was not defeatist, but it was sober. FOSDEM 2026 made it clear that the OpenSource community is already living through the early, painful stages of the AI transition. And the maintainers are bearing the brunt.
Ironically, Stenberg himself pointed to the other side of the equation in the very same talk. He noted that AI-powered code analyzers, when wielded by experienced engineers, are uncovering deep bugs in cURL that no previous tool ever found. He runs three different AI review bots on his pull requests. They catch missing tests, flawed assumptions, even violations of obscure protocol specs that no human had read in over a decade. AI is simultaneously attacking and defending his project. The question is who is using it, and how.
OpenSource maintainers: adapt or be forked
That is the other side of that same coin. The uncomfortable truth is that OpenSource maintainers are facing their own reckoning, and the answer to the FOSDEM anxiety is not to reject AI but to wield it. The same AI tools that are flooding projects with slop can also transform how those projects are maintained, and maintainers who refuse to embrace these tools will simply be outpaced.
Think about what an OpenSource maintainer deals with daily: an avalanche of bug reports, pull requests of wildly varying quality, feature requests that need triage, documentation that is perpetually outdated, CI pipelines that need babysitting. Most of this is done by a handful of exhausted volunteers who are already burning out.
Now imagine a maintainer who uses AI tools, whether OpenSource LLMs like Llama or Mistral, or commercial ones like Claude, to handle the bulk of that work. Automated triage of bug reports. AI-assisted code review that catches the obvious issues before a human ever looks at them. Documentation that updates itself when the API changes. Test generation that keeps pace with new contributions.
The maintainers who embrace this will be able to handle ten times the volume with the same effort. The ones who don’t? Their projects will be forked by people who do. Not out of malice, but because the forkers will simply be able to deliver better quality assurance, faster response times, and more consistent maintenance. When someone can fork your project and run it better than you can because they are leveraging AI and you are not, the community will follow the fork.
That’s how it is supposed to be. If you remember the inspiring essay from the early 2000s: “The Cathedral and The Bazaar” by Eric S. Raymond
New business models, or automation
The OpenSource world desperately needs new business models that ensure maintainers can sustain their work. The current situation, where critical infrastructure depends on volunteers who are one bad month away from abandoning their projects, was already fragile before AI entered the picture.
With AI dramatically increasing the volume of OpenSource software being created, the need for reliable, well-maintained projects will only grow. Companies that depend on these projects need to fund them. Foundations need to rethink their grant structures. New models, perhaps AI-assisted maintenance services, or platforms that connect corporate users with maintainer teams, need to emerge.
Because if they don’t, the alternative is clear: the maintainers themselves will be automated away. An AI system that can triage issues, review pull requests, write tests, update documentation, and cut releases does not need a salary, does not burn out, and does not take weekends off. That is not a future anyone should want for the soul of OpenSource, but it is the future we will get if we do not build sustainable alternatives.
The Free Lantern: proof that creation costs have collapsed
The same force that may soon produce an AGPL alternative to every SaaS product is already reshaping creative industries. The Free Lantern is a complete music project, an actual artist with an LP in production, releasing tracks monthly, built with the help of AI tools. What makes it remarkable is not just that it exists, but what it is about. The Free Lantern creates music specifically for the Linux and OpenSource community. Funk, disco-pop, tracks that celebrate distro culture, the joy of a clean terminal, the community spirit of free software. Before 2025, this project would not have been financially possible. The niche is too small to justify traditional music production costs. Studio time, session musicians, mixing engineers, mastering. The economics simply would not work for a relatively small audience of Linux enthusiasts.
But with AI tools, the creation cost dropped to nearly zero. The creative vision, the songwriting, the curation, the community connection, that is all human. The production barrier that would have killed the project before it started simply no longer exists. Now apply that logic to software.
A Linux-themed funk track and a niche ERP for organic goat farmers follow the same economic rule. Both were too expensive to justify building in 2023. Both become inevitable in 2026. Every underserved niche, every industry vertical too small for a VC-funded SaaS startup to pursue, is about to get its own tailored OpenSource tooling.
Not because someone is being generous, but because the cost of building it has dropped below the threshold where it makes sense to just do it.
This is the Jevons Paradox in action. When something becomes drastically cheaper, we do not use less of it. We use wildly more, in ways nobody predicted. The long tail of software is about to explode, and OpenSource is the only distribution model that can sustain it.
The proprietary world will be hit hardest
Let me be clear about where the pain will concentrate. The OpenSource world will go through a difficult adjustment. Maintainers will need new tools, new workflows, new funding models. Some projects will be forked, some communities will fracture, and there will be growing pains.
But this is nothing, absolutely nothing, compared to what the proprietary software world is about to experience.
Proprietary SaaS has been built on a simple premise: building software is expensive, so customers will pay recurring fees for access to it. When that premise collapses, the entire business model collapses with it. Per-seat pricing becomes a tax on productivity. Complex UIs become a liability rather than a moat. Switching costs evaporate when an AI agent can replicate the functionality of your product in an afternoon. And releasing it OpenSource ensures peer review, strong security, collaboration and all the other benefits we already know for decades.
As Sam Altman put it, every company is now an API company whether they want to be or not. Interfaces stop mattering. Outcomes matter. And when AI agents are choosing tools based on cost, speed, and functionality rather than brand loyalty and switching cost friction, the competitive dynamics of software change completely.
OpenSource, by contrast, has always been about the code, the community, the shared infrastructure. Those things do not become less valuable when creation costs drop. They become more valuable, because they are the foundation everything else gets built on.
What happens next
The SaaS apocalypse is real, but calling it an apocalypse frames it wrong. It is a restructuring. The total amount of software in the world will increase dramatically. The total revenue generated by software will likely grow. But the distribution of that revenue will shift, violently, away from the companies that have been extracting rent from proprietary code and toward the ecosystems that can adapt.
OpenSource is positioned to win this transition. Not automatically, not painlessly, but structurally. The question is whether the OpenSource community will move fast enough to seize the moment: embracing AI tools, building sustainable maintainer models, and scaling up to absorb the tidal wave of software that is about to be created.
Proprietary SaaS was built for a world where software was hard to create. AI has changed that permanently. We are entering an era where more people can build software, improve it, and share it than ever before. And that has always been the promise of Free and OpenSource Software.
Jasper Nuyens is the founder of Linux Belgium BV, providing enterprise Linux consulting and OpenSource infrastructure services, and co-founder of OpenSource-Enterprise.com. He is also the creative force behind The Free Lantern, a music project for the OpenSource community built with the assistance of AI tools.
