It’s not a waste of resources if you learn something. Think of this as research rather than product development. You can try many things (from VR, to miniaturised computers, to cloud gaming, controllers with wonky form factors…) to see what results in a good experience. You don’t need to get anywhere near a full fledged product to understand those things, so the waste of resources isn’t massive anyway.
I’d bet at the moment people decided “this is useful, I even want this for me, so let’s turn it into a product” the steam deck looked more like a screen, a gamepad and a raspberry pi all taped together or jammed into a 3d printed prototype chassis.
If people have spare capacity to work on these projects, the material cost at such a point can be under <5k which is peanuts for a company like Valve.
What? Elections in Japan?