I'm going to post here instead of there since I don't want to get bogged down in the investment stuff I know nothing about
And you'd be attacked by consanguineous Intel bulls which brain has suffered far too long from an impoverished genetic pool, so the discussion would be void.
I agree that Google is not in any place to make a custom ARM CPU core any time soon. That isn't necessarily relevant. If Google is looking to make some custom chip where most of the heavy lifting is done by custom logic it makes a lot more sense to affix a standard ARM chip in there than it does to use anything off chip. I have no idea what's even feasible for Google to make dedicated accelerator hardware for.
IBM PowerEN might be a hint of what can be accelerated.
I don't buy for a minute that it'd be just as easy for Google to throw an Intel CPU on this custom ASIC and have it manufactured by Intel. Just because Intel says they're willing to negotiate custom SoCs with user provided IP blocks doesn't mean that they're automatically the best choice, they're in a totally different position vs ARM and the big foundries like TSMC when it comes to license model, support, IP libraries, etc.
Intel isn't even able to build its own modems on its own process in a timely manner, so there's little doubt they're still not in a position to be a foundry for mass production of lower cost chips coming from others. And when you start adding HW accelerators you don't care what the CPU is, might be Intel, ARM, POWER, MIPS, and out of the three Intel is the less likely to easily license its core.
That being said, I'm not sure I buy this whole story of Google building its own chips. At least not until there's some more evidence.