News Isambard 2, the world's first Arm-based supercomputer, retires after six years of service — Isambard 3 will tap Nvidia's Grace CPU Superchips

Wow, no mention of Fugaku? I figured that's what this was going to be about, because it was the first ARM-based supercomputer that I was aware of (and the first to be notable for something other than the mere use of ARM CPUs):

One thing that's interesting about it is that each core featured 2x 512-bit SVE pipelines. In the case of Nvidia's Grace, each of its V2 cores has only 4x 128-bit SVE2. So, the clock-normalized performance for vector FP operations is likely worse for Grace. SVE2 improved somewhat over the original SVE, but not by that much.
 
What is the typical lifespan of a super computer?
I don't know, but some of the longer-lived ones seem to have a mix of node types. At some point, they start adding nodes that have better specs and more up-to-date hardware than the others.

So, perhaps the answer to your question partly depends on how you define a supercomputer. At the end of the day, it's really just a network of machines, running specific protocols, and services. They're initially sold as one big contract, but they can often be upgraded in a piecemeal approach, like what happened here:

"Alpine was part of ORNL's storage solution for Summit and its peripheral systems, holding scratch data from the supercomputer and its external nodes, which pre- and post-process Summit's calculations. The Alpine storage system held 250 petabytes of capacity inside 32,494 10TB NL-SAS drives. Made up of 77 IBM Elastic Storage Server (ESS) nodes, the system could have 2.2 TB/s random read and write speeds at its peak. Still, in recent years, drive failure rates have hit unacceptable levels, necessitating installing a stopgap replacement storage system, Alpine2."

https://www.tomshardware.com/tech-i...paration-for-the-worlds-fastest-supercomputer