Sometimes you don't know up front. Perhaps you learn a lot about gaussian splatting and push forward that tech by five years (presuming they started on this five years ago).

Or maybe you learn this helps you build a world model that would have accelerated Waymo's progression, or you sell it to robotics companies.

That's the nice thing about having a money printer and a ton of smart, curious, and driven people. You can afford to do things without a strict eye towards profit, that's reserved for tiny companies living on a razor's edge. Or consultants I guess.