The idea has always been that by the time the reward shrinks to nothing, transaction fees will more than make up the difference as a result of increased usage. Of course, back when I was introduced to this idea, the notion that in 2018 Bitcoin blocks would still be capped at 1MB would have been considered absurd. Everyone understood that decentralization didn't come from everyone running their own full node, but from the incentive structures that govern the behavior of the miners.

There is also the possibility that eventually it could move to Proof of Stake for block generation and consesus, which would also solve the problem of unsustainable energy consumption.