This is a really good point, and I hope they come up with a another proof of work algorithm that does something useful like protein folding, or switch to "proof of stake" which is not compute intensive alternative.