It all depends on which reports you've been looking at. There's been a great deal of variety in the predictions and that's because there's a great deal of variety within the broad spefctrum of factors goiing into the microserver market. There's enough to make you think that this is going to be a booming market segment soon--and there's enough to write it off as a small piece of a muc larger, much more significant pie.
Micro-servers are where the growth is. Calxeda was a great attempt and did demonstrate use in a niche segment of that market. Their problem was that they only had a 32b offering while the likes of AppliedMicro, AMD, Broadcom have all announced 64b designs with Applied having silicon for an year now. I can see why investors hesitated in putting in more money.
Tier #1 server OEMs are fading away, who/what is replacing them. Google, Amazon, FB, Taiwan ODMs and data centers. The area really in flux now is the hyperscale segment which is focused on data centers. That's another laugh about server market reports, they are not seeing the real market which is measured in server sockets and not servers. Lastly don't discount OCP, it might be the catalyst to standardize most hyperscale architectures.
It may be that traditional server OEMs are fading, or it may just be that a new market driven by service providers is springing up in parallel. Either way, the wave of change is pretty much undeniable.