Well yes sure, but remember AV1 decoding only became standard like 1-2 GPU generations ago. Encoding only this generation. iPhones only got support with the 15 Pro so it will be another generation before it trickles down to the base models. And what about the hundreds of millions of Android phones in Asia and the likes with dirt cheap SoCs. Pretty sure they don't have dedicated AV1 decoding hardware for a long time.

So that's a TON of hardware being made slow and inefficient if everything were to be AVIF tomorrow. Not saying AVIF decoding will be a big hurdle in the future but how long until all this hardware browsing the web has been replaced? That's why I think somethings that's efficient and fast on CPUs without any specialised hardware is more suited for a replacement.

Servers often come without GPU, and they’re usually the ones encoding image formats.