I did not know your Nest thermostat has a Zigbee radio. Even more so, what a shame this was never used to enable local connectivity.
There are many reasons why cloud-only products exist, and you have correctly described two of those reasons. However, what I meant with my comment is that Matter could potentially make such products less attractive, because it could provide an alternative.
I’ll mention a few examples…
Many years ago Miele introduced connected appliances through a dedicated slot in the appliances where you could insert a communication module which could then communicate with a gateway over powerline. This gateway then had a (primitive) local API. They might have sold a few hundreds of those modules and gateway, I don’t know. Then later they introduced a Zigbee gateway and new Zigbee modules you could retrofit into the communication module slots. To me it seemed that they were moving in the right direction, although the functionality was very limited. The communication module slots seems like a good idea for appliances that could last 20 years. But, of course, not many people wanted to buy a 300 € gateway just to have a primitive API to mostly monitor your appliances.
So they migrated to Wi-Fi, and at the same time abandoned the slots for communication modules (at least to my knowledge). My dishwasher came with built-in Wi-Fi. And at the same time they dropped the local API and introduced Miele Cloud. For me, as a consumer and home automation enthusiast, this certainly went into the wrong direction. Now I have a slow-performing solution completely out of my control and at the mercy of Miele. I can use it as long as it’s not down for maintenance, and as long as it will be able to communicate with my Wi-Fi access point without compromising security or speed.
Next example: Velux Active. I will make this one short: Provides a cloud integration (Velux with Netatmo), closed, to be used with their app. Integrations not supported. And also the gateway supports HomeKit, which then provides a local integration possibility (not yet supported by openHAB though).
Next, Danfoss Link. A proprietary Z-Wave solution with thermostats and a controller. Local communication between thermostats and gateway, but no local API. Cloud solution for getting online (via app). The cloud solution is anonymous, i.e. no account needed. This solution was later dropped in favour of a new Zigbee-based solution, Danfoss Ally.
To be honest, I believe these companies really tried to reach as many “average” users as possible and just “app-enable” their solutions.
The Miele Zigbee solution was of course terrible, since it required users to buy an expensive gateway just to get a washing machine online without many reasons to do so. It could have been made cheap, but I think even then, it wouldn’t appeal to many, as it needs to be connected to the router with cable and turned on all the time. Migrating to Wi-Fi and cloud seems like a logical move, because all users could then instantly connect with their phones. Unfortunately the communication module slots and local API died on the way.
With Velux I think they also just wanted a simple solution that could appeal to as many users as possible. I believe they even also provided a “Google Assistant” logo in the box. In this case I’m not so sure data collection or subscriptions was even a priority.
The same with Danfoss. Like with Velux, simplicity (for the user) and taking care of security in their cloud backends, as you mentioned.
So if we assume that they were not ill-intended, I think these kind of companies would happily strive for Matter compatibility if they could put that on the box label and they believed it would help them sell more products (by making them more attractive than the competitors’ products). This could also save them some money, as they probably would have less integrations to support, and in some cases wouldn’t need to provide a cloud infrastructure. But then again, this is based on the assumption that they actually have good intentions and just never saw a feasible way to actually provide something which would not be cloud-based, have an open API and still attract average users. So far this has been almost impossible to accomplish.
It will be the chicken and egg. Only when Matter is successful and adopted, it will be attractive to provide Matter compatibility. So, I’m not as positive as I was maybe one year ago, but I do still hope that both Matter and Thread will grow and reach some kind of critical mass that could get things rolling. And then we would again be able to buy products which would work until they break, which we could integrate with, and which would be fast.
EDIT: Google is of course a completely different type of company than those I just mentioned. They have a completely different business model, so you are spot on with your data collection and subscription points. It would have been interesting to see where Nest would be today without the Google takeover.