[zeteo]

Using robots for this task is more a question of fixed cost than anything else. Rather than build a conveyor belt to each cubbyhole in the warehouse - some of which may not be used for weeks on end - it's much cheaper to have a few robots that go to each spot as needed. And it's also more flexible, since it's easy to add or remove robots as needed.

In computing terms, conveyor belts everywhere cost O(warehouse space) to build, whereas mobile robots cost O(activity volume). Profits are also, most likely, O(activity volume).

[onemoreact]

I think maintenance costs / downtime is also a big factor. With robots most failures simply reduce throughput but with conveyor belts you risk shutting down sections of the warehouse when something breaks or needs to be repaired. On the other hand you can order a few extra robots and swap then out for maintenance.

As to energy costs, moving a shelf a at most 1/2 mile at low speeds on a level floor using electric motors probably uses less energy than you might think. Don't forget human pickers are say 150lb people walking up and down the same isles and they use about 150-200W to do so. The shelves weigh more (up to 1,000 lb), but you can a turn off most of the lights and electric motors are more efficient than walking. Assuming it's 400W a robot that's around 4 cents an hour per robot or something like 300$ a year.