Last week, Reuters reported that AI specialists are raising concerns about Nvidia’s acquisition of SchedMD, the company behind Slurm, the scheduling software running roughly 60% of the world’s supercomputers.
I keep a close eye on big/quiet acquisitions like this. But this story isn’t really about Nvidia.
It’s about a question that keeps coming back: what happens to open source software when it gets bought?
Not all open source projects face this. Linux isn’t going anywhere. Neither is PostgreSQL. The projects with broad, distributed governance and no single commercial entity behind them tend to stay genuinely free. But the ones that become deeply embedded critical infrastructure while being maintained by a small, acquirable company? Those follow a pattern so predictable it should have its own name. You think you own the place because you’ve been living there rent-free. Then someone buys the building and you realise you were always a tenant.
The pattern
An open source project solves a real problem. It gets adopted widely, often because it’s the only tool that works without a sales call. A corporation notices the project has become critical infrastructure. They acquire the company that maintains it. They promise to keep it open. They mean it, at first.
Then the economics kick in. Enterprise features appear. The paid tier gets patched faster. The open source version doesn’t get killed. It just quietly falls behind until the gap between free and paid becomes the whole product. It’s not the first time it has happened:
- MySQL: Oracle acquired it in 2010. Promised continuity. The community edition slowly lost transparency and external contributors. The original creator forked it into MariaDB before the ink was dry. Today, MySQL’s external contributor base has effectively collapsed.
- Terraform/HashiCorp: Open source for nine years. HashiCorp switched the licence to restrict commercial use in 2023. No community vote. No warning. IBM then acquired HashiCorp for $6.4 billion. The community forked it into OpenTofu.
- Redis: Changed its licence in March 2024 to restrict cloud providers. Within a month, AWS, Google, and Oracle had backed a fork called Valkey. Within a year, Redis had lost a lot of its external contributors.
Three projects. Three acquirers. Same playbook. And now Slurm.
Why it keeps happening
It’s not malicious. It’s rational.
An open source project with wide adoption has market penetration no sales team could achieve, trust no marketing campaign could buy, and switching costs that grow every year the software is embedded deeper into production. The company that controls its maintenance controls the on-ramp to an entire ecosystem.
The acquisition price isn’t paying for code. Code is freely available. It’s paying for the keys to the building: the ability to set the roadmap, hire the contributors, and decide what gets fixed first. The tenants didn’t change. The ownership did. This is the bit the “just fork it” crowd underestimates. Yes, you can fork GPL-licensed software. But when the best engineers now work for the acquirer, a community fork is technically possible and practically starved of the people who know where the bodies are buried.
Final thoughts
Slurm is GPL licensed. Nvidia says it’ll stay open source and vendor neutral. The pattern doesn’t care about press releases. It cares about incentives.
If your infrastructure depends on an open source project maintained by a single company, you don’t own your infrastructure. You’re a tenant. And the rent is whatever the new landlord decides to charge.
Know the cost of migration before it becomes a crisis. And stop treating “open source” as a synonym for “independent.” It’s only as free as the governance behind it.
Sources: Reuters · Nvidia Blog