Not an engineer so maybe someone else will have to chime in but i doubt the radial heat from a GPU is enough to heat an element which in turn can flash water to steam.
Surely you cant get the density needed before the GPU's themselves burst in to flames?
Again, probably need a physicist or engineer to figure out if that's enough to get water to the 500-600 degrees typically used in steam driven turbines.
I do not believe this would be possible. If we stick with 70 degrees being the max temperature of the GPUs^, as we increase the number of GPUs then the coolant will approach 70 degrees but not go over, hovering from 69-70.
To break it down, if you have a comically large stove top heater set to 70 degrees and a regular sized one set to 70 degrees, if you put a pot on either the max temperature of the water in the pot is still 70 degrees. The rate at which it heated to 70 would change but max temp wouldn't.
^Max safe temp of GPUs is around 100c. What will actually happen is that the GPUs will heat the coolant to this temperature, then either the coolant will lose heat through normal means (when water isn't heated it cools), or if the coolant stays hot eventually a GPU will turn itself off through overheat safety mechanisms. Once this GPU is off the coolant loop will drop temperature for a while then slowly approach max temp again, until either another GPU gets turned off or the system finds its equilibrium where the coolant is at a stable temperature and no more GPUs are turning themselves off.
Surely you cant get the density needed before the GPU's themselves burst in to flames?