I wish I could run models locally. I just don't have the hardware for it. So I rent access to one that still works well enough and hasn't annoyed me yet. But "yet" is doing a lot of work there. The article is right that the trend is worrying. You're at the mercy of someone else's cost cutting and someone else's political compass. I just can't afford to opt out right now. That's the part that stings the most.
I believe local AI is the future of most daily workflows