Infrastructure (servers, bandwidth, etc) costs money.
For better or worse, most devices don't have local interfaces. Some matter devices exist in the past year or two, but not all data/functions are available via matter. Older devices likely won't ever be updated to support matter.
Also lump in the fact that HA is a locally run/focused application, it isn't super compatible with a cloud based API/data delivery system without some additional development from the OEM end.
Last I did the math for an internal system.. in 24 hours HA traffic was ~20% of total, for less than 1% of users. That's wild. Mostly because each instance is pinging the API directly every X or less minutes.
If an executive here heard that math, they'd likely ask to block it as well. Right or wrong that's how people react.