Why do all Top 500 supercomputers run Linux?
Because Linux can be stripped down, modified, and scaled freely with no per-node license cost — the exact combination a supercomputer needs. By the late 2010s, all 500 machines on the Top 500 list ran it.
* Tux, the Linux kernel mascot — the kernel that runs every machine on the Top 500 list. — Larry Ewing, CC0 / attribution, via Wikimedia Commons. *
A supercomputer is thousands of nodes wired into one machine, and that scale punishes every weakness Linux happens not to have. The reasons compound:
| Factor | Why it matters at supercomputer scale |
|---|---|
| Customisable | Strip the OS to bare essentials so the hardware spends cycles on computation, not overhead |
| Modifiable (open source) | Patch the kernel for exotic interconnects or schedulers |
| Scalable | The same kernel design spans one CPU to millions |
| Stable | Long jobs run for weeks; a crash wastes enormous compute time |
| No license fees | Per-node licensing across thousands of nodes would be ruinous |
The same traits explain Linux's reach far beyond supercomputers — as of the mid-2020s it runs the large majority of web servers, the major clouds (AWS, Google, Azure), every Android phone's kernel, and countless routers and IoT devices.
Irony: Torvalds' 1991 post said Linux would "never be big and professional." It became the default platform for the world's most demanding computing.