I personally aim for 15 years as the time to start shopping for a new SEM or probe, the actual usable lifespan being closer to 20. In academia I find it typically takes about 5 years from first mentioning the possibility of replacement to an appropriate successful funding opportunity coming up. About 20 years is enough time for a control computer to become well and truly archaic, which I'd say is the limiting factor with most modern machines.
In the SEM world this also depends on capabilities. A 15-20 year old SEM with a 5 axis motorized stage has infinitely more utility remaining than a system with a manual stage.