Research has a large asymmetry to it. Once someone shows something can be done. Others can follow quite easily.

And more substantial is that when someone shows something can be done, its orders of mangnitude easier for $ENGINEER at $CORP to get $CSUITEs attention to get a budget / justify the risk.

Your self driving car example is the best example of this. Since waymo and tesla got popular NVIDIA really started pushing their self driving cars for everyone tech.

https://youtu.be/EzAVW1VgzcI?si=gfnOJoaO9hfIFXX1

And NVIDIA isn't the only one.