Besides... Let's assume everyone got used to it... But on the other hand, open models that compete with Claude and OpenAI are starting to be really good (this might feel unrelated, but it is), and they are manageable to run locally, and a lot of people want to do that. Now, the winners would be hardware companies, but that would increase the demand for RAM and GPUs even more, which might lead a company like Oracle to raise the prices related to AI features. Now, people will try to connect the system to their local models, or they might push their users to reduce their usage of AI features. Oracle might lose a lot of money in this case.
UNLESS Oracle AI is able to do some things that open models won't, because they are trained on different data, maybe... + what companies would pay for it is higher than the cost of the seats they would get rid of. Unlikely? (Also add to that... if this was the case, the increase in profits should cover what Oracle already invested in all of this)