The energy used in PoW must be "wasted".

Because of the difficulty adjustment and the economic incentives, the equilibrium cost of PoW approaches the value of the block reward. Even if the result of PoW is useful, then the cost of the "wasted" PoW will still approach the value of the block reward. So, nothing is accomplished by making the PoW generate something useful.

Let me demonstrate by example:

Suppose a hash rate of 1 exahash per block costs 1 BTC in energy costs and earns 1 BTC per block on average. In this case, 1 BTC worth of energy is "wasted".

Now suppose that a use for hashes is found and 1 exahash can be sold for 0.5 BTC. We can now say that only 0.5 BTC worth of energy is "wasted". Not so fast. Mining costs 1 BTC per block, but now earns 1.5 BTC so miners will certainly increase their capacity in order to make more money.

In this scenario, miners will double their hash rate on average. Note that the difficulty will also double because the total hash rate has doubled.

Now, 2 exahash per block costs 2 BTC, but it earns 1 BTC per block plus the hashes sell for 1 BTC. As you can see, we are back where we started where 1 BTC worth of energy is "wasted", even though mining also generates something with value.

In case you were wondering, the case where hashes initially sell for more than the value of the block reward also gives the same results, assuming a normal demand curve.