@Does anyone want it?
Damn Good Question!
I can see an immediate ROI for self hosted conference bridging options compared to using a 3rd party. I can also see an immediate ROI for self hosted automated telemarketing campaigns that tie into the internal development stack and client database, both things usually only seen in SMEs not SMBs.
For businesses that have a vested infrastructure in an existing KSU or traditional PBX, there is no immediate ROI, and what ROI there is usually isn't realized for years.
SMBs usually have a pretty flat network topology, and now I must introduce a routed layer four aware network infrastructure, POE switches, battery backups on the Servers, Gateways, and switches as well as emergency services over POTS or PRI if I'm using an ITSP. The price of those alone is usually worth the maintenance contract on the existing KSU for quite a few more years.
On top of that, slapping a new card into the PBX is dirt simple. Trying to figure out why Oracle isn't replicating properly to the fail over server usually requires a bit more skills than (un)rescrewing in two bolts!
A couple of years ago, maybe the argument still works now, many people were upgrading they're network from one type of topology to another, and maybe they were doing a rip and replace on the wired infrastructure too. In that particular scenario, usually VOIP plays out as a good investment, but that's a pretty atypical case.
On top of that, if the SMB is already using DS1 for the PSTN handoff instead of POTS, they're probably already saving bucket loads of cash anyways. This makes VOIP from the ITSP provider pretty unattractive.
I'm with you, most people that actually sit down and crunch the numbers, and yes all of them, usually the ROI is greater than three years. Most COTS VOIP systems and the network infrastructure they reside on, don't maintain that kind of warranty.