Note that the clients being FOSS is of little relevance because all they do is forward a recording to a blackbox proprietary service run by a for-profit company.
The code that has access to your audio and does the actual task at hand is not FOSS in the slightest.
That doesn’t seem right; that’s only ~18W. Each one of those systems alone will exceed that at idle running 24/7. I’d expect 1-2 orders of magnitude more.