Red Hat built a $34 billion business on Linux. IBM bought it. What the deal validated was a hypothesis that had held for four decades: that companies extracting enormous value from shared code would, in self-interest, keep funding the projects they depended on.
That hypothesis is now under stress. Not because anyone decided to stop funding open source. Because the industry that funded it most — SaaS — is being dismantled by the industry that depends on it most — AI.
The prediction that arrived
About a year ago, I wrote that AI agents would hyper-personalize enterprise software — that business teams would build their own tools, bypassing IT backlogs and one-size-fits-most SaaS platforms. The argument was straightforward: when AI reduces development costs by 70–90% and compresses deployment timelines from months to hours, the build-versus-buy equation flips.
That prediction materialized faster than expected. On January 30, 2026, Anthropic released agentic workplace plugins for Claude Cowork. Within five trading days, roughly $285 billion in market value evaporated from application software stocks. Over the following month, Adobe, Microsoft, Salesforce, SAP, ServiceNow, and Oracle shed more than $730 billion combined. Jefferies traders called it the “SaaSpocalypse.”
Klarna had already demonstrated the template: terminate Salesforce and Workday contracts, replace functions with AI agents, and watch revenue per employee climb from $400,000 to $700,000. A healthcare CTO deployed three major internal platforms — CRM, HRM, and helpdesk — in 90 days. One founder texted his investor that he was replacing his entire customer service team with an AI coding agent.
The build-versus-buy decision is shifting toward build. But everyone is focused on the disruption itself. The second-order consequences are worse.
The four-link chain
Open source does not fund itself through donations. It funds itself through corporate labor. 84% of Linux kernel commits come from corporate developers across 1,780 organizations. The Linux Foundation generates $311 million in annual revenue from over 3,000 member companies. MongoDB, Elastic, HashiCorp, Databricks: each built multibillion-dollar businesses on open source and poured engineers and capital back upstream.
The model worked because incentives aligned. Healthy upstream projects meant healthy downstream businesses.
AI is breaking that alignment. In four sequential steps.
1. SaaS was the financial backbone
Approximately $7.7 billion flows annually from organizations to open source, but 86% of it is employee labor — labor funded by SaaS revenue. Remove that revenue, the labor follows.
2. AI is dismantling the SaaS model
Here is the fundamental paradox: a CRM with AI agents reduces required users from 20 to 2. That is a 90% revenue drop despite a 10x capability improvement. When companies can build internal tools faster than they can negotiate enterprise contracts, the per-seat model collapses. And with it, the surplus that funded open source contributions.
3. Open source projects are closing the gates
MongoDB moved to restrictive licensing in 2018. Elastic followed in 2021. HashiCorp in 2023, spawning the OpenTofu fork. Redis in 2024, spawning Valkey. Each relicensing smoothed the path for the next. Meanwhile, Stack Overflow’s question volume has collapsed 76% since ChatGPT launched. The Open Source Collective saw a 20% drop in corporate sponsorships in 2023. Sixty percent of maintainers are unpaid. Sixty percent have quit or considered quitting.
4. AI’s own training data degrades
A 2024 paper in Nature (Shumailov et al.) demonstrated that models trained on AI-generated output experience model collapse — irreversible defects that compound across generations. The long tail of open source code — small libraries, niche projects, idiosyncratic solutions — is the data diversity that makes code-trained models capable of generalization. When that tail shrinks, AI gets dumber in precisely the areas where it needs to be smarter.
The chain closes into a loop: AI needs open source. AI disrupts SaaS. SaaS funding dries up. Open source shrinks. AI training data degrades. And eventually, AI-built internal tools stop being better than the SaaS products they replaced — sending companies back to vendors whose own foundations have eroded.
The signal is already visible
This is not projection. Daniel Stenberg shut down cURL’s bug bounty after AI-submitted reports hit 20% of submissions with only 5% validity. Maintainers are drowning in what the community calls “slop PRs.” Small libraries are going dormant because LLMs generate equivalent utility functions on demand, eliminating the users who would have contributed fixes.
Open source is bifurcating. Massive enterprise-backed projects — Linux, Kubernetes, PyTorch — survive. Mid-tier and small libraries quietly go dormant. That long tail is exactly what gives AI training data its breadth. Its disappearance is not a future risk. It is the current trajectory.
Who picks up the bill
Meta is the outlier that proves the rule. Llama has been downloaded over 600 million times on Hugging Face. PyTorch was donated to the Linux Foundation. Zuckerberg’s logic is transparent: commoditize the layer below you, crowdsource R&D, establish a default the way Linux became the default for servers. Self-interested and correct.
The rest of the AI industry has not followed. AI companies collectively raised over $202 billion in 2025. A Harvard Business School study valued the open source software they depend on at $8.8 trillion. Their combined direct financial contributions to that open source projects: likely under $50 million per year.
Anthropic announced a $1.5 million two-year partnership with the Python Software Foundation. TechCrunch called it “couch-change.” That framing is hard to argue with.
The emerging mechanisms — the Open Source Pledge asking $2,000 per developer annually, the Open Source Endowment targeting $100 million over seven years, Germany’s Sovereign Tech Fund investing €23.5 million in 2024 — are real, and insufficient. Against an $8.8 trillion projects. Here is what both sides seem to forget: open source is the reason either industry exists.
Every SaaS company that became a billion-dollar business did it on open source frameworks, databases, and libraries. Two developers in a garage could challenge an incumbent because the building blocks were free. That access — not venture capital, not talent alone — is what created the SaaS market. It is what allowed a generation of entrepreneurs to build Salesforce, Elastic, Databricks, and thousands of others without asking anyone’s permission first.
And AI models can write code today because they learned from millions of open source repositories. Every line that Claude or Copilot generates traces back to developers who shared their work freely. The Harvard study puts it plainly: 96% of commercial software includes open source code. Without it, firms would pay 3.5 times more to build what they build today.
Both industries owe their existence to the open source projects. Both are now, through different mechanisms, starving it. Meanwhile, 60% of the maintainers holding it all together are unpaid. Sixty percent have quit or considered quitting. The average unpaid maintainer puts in nearly nine hours a week — and when they burn out, critical infrastructure goes dark. Kubernetes retired Ingress NGINX in late 2025 not because it was obsolete, but because the people maintaining it on nights and weekends could not keep going.
This is not a problem one side solves alone. SaaS companies, fighting for survival, cannot fund open source the way they used to. AI labs, flush with $202 billion in fresh capital, have not started in any meaningful way. The only path that does not end in collapse is collaboration — shared, deliberate investment in the infrastructure both depend on.
Red Hat understood this forty years ago. IBM paid $34 billion for the proof. The model worked because the entities extracting value from the open source projects also funded the open source projects. That alignment needs to be rebuilt — this time, with both SaaS and AI labs at the table.
If they choose to fight over the ruins instead, the people who pay will not be executives or investors. It will be the next generation of developers and entrepreneurs who will have used that open foundation to build what comes next — the way every generation before them did.
When two kingdoms fight over the same land, it is the soil that stops producing.
