I have found that we swap out servers faster than we need to. A five-year cycle is reasonable. An eight-year cycle is pushing it, but still doable depending on business needs.

Kicking the tires on three-year-old servers is the first place I go when buying new hardware. It may not be a good strategy at this time, as they may also suffer from significant price inflation. I also recognize that budgetary politics may require you to buy new before the old is worn out.

However, if it's possible to keep the same hardware, I've had good luck saving money by moving clients off VMware and onto XCP-ng/Xen Orchestra using its VMware migration tool. There are other options that may better suit your needs.

For me, the next step is to find ways to make the current implementation of business rules through the VMs more efficient and to make better use of the available resources. I.e., are our VMs over-provisioned or underutilized? Then you move into incremental upgrades, for example, adding more memory to the servers.

Lastly, you can stretch resources by identifying friction points and shifting them to areas where they don't matter as much. This advice comes from the perspective of an environment where the business uses custom solutions for most of the production cycle, rather than just Lego blocking commercial packages.