> the AV1 video codec was deployed when no consumer had HW decoding capabilities

This was a bug. An improved software decoder was deployed for Android and for buggy reasons the YouTube app used it instead of a hardware accelerated implementation. It was fixed.

Having worked on a similar space (compression formats for app downloads) I can assure you that all factors are accounted for with decisions like this, we were profiling device thermals for different compression formats. Setting aside bugs, the teams behind things like this are taking wide-reaching views of the ecosystem when making these decisions, and at scale, client concerns almost always outweigh server concerns.

Well I will say if your running servers hit billions of times per day. Offloading processing to the client when safe to do so starts make sense financially. Google does not have to pay for your CPU or storage usage ect...

Also I will say if said overhead is not too much it's not that bad of a thing.

This is indeed an issue but it's widespread and everyone does it, including Google. Things like servers no longer generating actual dynamic HTML, replaced with servers simply serving pure data like JSON and expecting the client to render it into the DOM. It's not just Google that doesn't care, but the majority of web developers also don't care.