All of this is true, but from what I read from experts there's a lot of work to be done on the software side to be able to do it across a wide range of hardware targets.
The current DXR 1/1.1 brute force approach (mostly written by Nvidia) that is only feasibly playable on $2000 / 500W GPUs from Nvidia is awfully inefficient and is never going to get wide adoption.
The RTX ON stuff served Nvidia well for the world's Battaglias to claim you should only buy GPUs from one specific vendor so you can turn on some clotheless emperor effects, but if it's ever going to be widespread then the API and engine devs need to spend many man-months researching and implementing much more efficient approaches.
It's only after all consoles and most mid-end iGPUs are able to run this adequately that devs can forget about making shadowmaps, cubemaps, tricks to hide the deficiencies in screen-space effects, etc. and in turn shorten development times (and save money).
