The way I've come to think about Oracle's prices for Solaris support is that it is effectively increasing the price of the machine by 8% (or 12% if you want hardware support) per year that you keep it. Whether this matters depends on the sort of environment that you have.

In some organizations, the hardware costs are a small part of the total cost of operating a machine, and if not they're a small part of the value that the machine contributes to your bottom line (this includes a lot of the sort of environments that Oracle traditionally sells to). Even if this is not so, I believe that many organizations turn over production machines relatively frequently for various reasons; at a three year lifespan a Solaris support contract only increases the machine's hardware cost by a bit less than a quarter, which may be considered cheap.

(This relatively short lifespan is true even of some machines here in the department; as you might expect, research compute clusters age quite fast. This year's cutting edge server is not even really worth the heat from keeping it powered on in a few years.)

Universities are in many ways the worst case for this pricing model. We tend to run machines for a very long time (five years is on the low end) and the hardware cost is usually the only cost of the machine, since we generally don't have expensive licensed software and the like. And of course universities have no return on investment; these machines don't make us any money, they're a pure cost sink. With a many year lifespan and essentially no other costs, the extra cost added by Solaris support adds up rapidly. If we run a Solaris server for six years on the same hardware (which is perfectly normal around here), the server costs us almost half again as much as its sticker price. Or to put it the other way, we get two servers instead of three.

(Of course we're being thrifty by not replacing hardware very often. But once you start thinking that way, you wind up having to ask if we could be even more thrifty by not running Solaris at all.)

In however many years from now when our current hardware starts being either flaky or inadequate we may wind up concluding that Solaris really is worth it and that it lets us save one machine out of every three (or more). But I honestly think that it's going to be a hard sell, because it's hard to see how Solaris is that good.

(Or maybe we'll have a real hardware budget at that distant future time. In a university environment, one can always dream.)