Actually they are cheap - the reason why a significant amount of on chip resources can be assembled into the IME on all of Intel's chipsets is because IC process technology has advanced by many orders of magnitude. The Pentium 4 represented the edge of consumer grade CPUs 10 years ago, and it was around that time we reached practical limits on frequency. As hardware manufacturing shifted towards more cores and caches, SoC ICs became the standard. Slapping a cheap, low power system of CPU and memory on every chipset is easy, cheap and natural.
With extra hardware being cheap and easy, Intel put 2 and 2 together and tried to monetize this by offering businesses extra capabilities as a standard. Obviously this introduces new attack vectors, and the opaque nature of this new "standard" feature attracts attention from all sides.