My guess: it's the price at which they make more money than if they rent the TPUs to other companies.
The Gemini team has had trouble securing enough TPUs for their user's needs. They struggle with load and their rate limits are really bad. Maybe at a higher price, they have a better chance at getting more TPUs assigned?
You're probably joking, but this is interesting. If we throw more RAM at AI, it can help us optimize programs to reduce our RAM needs, I haven't thought about it like that