There was a really neat article here a while back comparing AI to previous tech/industry revolutions.

The article compared market conditions to argue that the current revolution is similar to the shipping container revolution of the '60s-'80s -- with a lot of market interest and no real moat or market leader, high competition drove prices down, kept margins low, and most of the benefits went to the consumer.

This seems like the best case scenario here. AI ends up expensive to train but cheap to copy and run, so we end up with lots of competitors or even just running it in-house. It's cheap to use, so there'll be lots of businesses trying similar AI-driven ideas and pushing prices of their products down for the consumer.