If you can capture a significant fraction of the gains made by automating lots of white-collar work, then those billions/trillions look more reasonable.

"Correct" investment volume towards early industrialization might have looked just as unreasonable without the benefit of hindsight.