Do we really need remote code execution on the bios level? Is this a case of 'we can, but should we?'
Remotely managing a corporate desktop or laptop, e.g. fixing an OS-level problem remotely, may also be important.
OTOH I'd prefer this functionality clearly delineated, usinf strong encryption, and with an explicit reliable "off" switch (preferably physical).
The proper solution is an optional management chip on the motherboard, not the CPU. PCI bus mastering NICs with wake-on-lan and other management features have existed for decades; it wouldn't be particularly difficult add the rest of the ME features.
There is no universe where what we have now is a good solution. There is no universe where there aren't superior alternatives (like custom bios for datacenter users, like just spending a little bit more money and having two versions)
If this is a problem, configure your OS better, or get a better OS. Otherwise you'll soon get need another ME on top of your ME to "fix an ME-level problem remotely".
How many layers of machines do we really need???! most datacenters already run in VMware with dockers inside running Java VM inside executing Javascript inside.
I have yet to see or even hear someone defend the ME on the grounds that they use it daily. With a few SPARC servers as the exception, we have a 100+ Intel based servers and we don't feel the need for using Intels Management Engine.