If your AI assistant has started feeling stingier lately, you are not imagining it. Google has begun rolling out compute-specific usage limits on Gemini Notebook, making it the third major AI vendor in a fortnight to restructure how much you actually get for your subscription. The era of “pay your monthly fee and use it as much as you like” is ending, replaced by a more metered reality where the expensive stuff is rationed.
Why this is happening now
The reason is unglamorous and unavoidable: running frontier AI is staggeringly expensive, and the flat-rate subscriptions were always a loss-leader landgrab. For a while the labs ate the cost to win users. Now, with growth captured and investors asking about margins, the maths is reasserting itself. Compute-specific caps are how vendors stop their heaviest users from torching the economics: instead of one blanket limit, they meter the costly operations (the big reasoning runs, the long context, the heavy tools) separately from the cheap ones.
Three vendors moving the same way in two weeks is not a coincidence, it is the industry collectively deciding the all-you-can-eat phase is over. Expect more of it, dressed up in the friendly language of “fair usage” and “optimising your experience,” which mostly means you will hit a wall sooner on exactly the powerful features you signed up for.
The takeaway
For anyone who relies on these tools for real work, the practical advice is to read the new limits before they bite, and to notice which of your workflows lean on the expensive operations now being rationed. The bigger picture: AI pricing is maturing from land-grab to real-business economics, and “unlimited” is going the way of the unlimited mobile plan, which is to say, it was never really unlimited, and now they are telling you so. Budget your heavy usage accordingly, because the meter is running whether you can see it or not.
Related: Suno’s move to cap what its users can do.