An iPhone, Macbook, etc all cost in the $1000 range.

There was a post about the new iphone using A19, which includes a feature that makes local inference much easier.

If that makes it to M5, I think the local inference case continues to grow with each M processor.