GW is already a measure of rate of energy use. You can talk about GWh per day (which is really just another way of saying 0.041 GW), but GW per day is only sensible in the context of a ramp in power consumption.
Assuming the original source of that number didn't incorrectly conflate GWh into GW, that puts your comparison as underestimating the energy usage 24-fold.
Not if you put your data-center right next to the power station.
Watts are Joules per second, you'd only measure "watts per day" as a rate of change of power supply, not absolute power supply.
Also, 5GW / 170000 homes ≈ 29kW/home, even the USA isn't that heavily supplied with electricity.
(My home in Germany is very efficient, 0.5 kW average, including heating and cooling as well as all devices).
Just to give perspective 5GW would power about 10 Million H100s or about 5 million G200. Completely farcical, won't happen.