Pivoting development to address a temporary shortage isn't necessarily good business strategy.
Someone told me that the best thing is to assume that in a couple of years when the AI bubble bursts will have a deluge of RAM and GPUs with lots of RAM, and it would be best to design for that.
I think that's too optimistic, but seriously, investing into optimisation as a response to a market shortage doesn't sound to me like the best use of development time unless it's simple to do. If it's simple to do, it would often already be done.