A household cancels three streaming subscriptions and consolidates two productivity tools into one. This looks, from the customer side, like a clear trend toward simplification. From the vendor side, the response has been to add new pricing dimensions rather than compete on the old ones — OpenAI's API bills per token, Adobe's Firefly features run on a monthly generative-credit allowance — so that even a consolidated toolset can produce a bill that varies unpredictably month to month.
The mechanism is that AI features introduced a genuinely new cost structure into software: variable compute cost per use, rather than the largely fixed cost of serving a static application. A vendor offering an AI writing assistant faces real marginal costs that scale with usage, which creates a legitimate business reason to meter differently. It also creates an opportunity to introduce pricing complexity that happens to obscure how much a customer will actually pay in a given month.
Customers experience this as a kind of pricing opacity the flat-fee era mostly avoided. A subscription used to mean a predictable monthly number. An Adobe credit allowance or an OpenAI token bill means a base fee plus a variable component most customers have no intuitive sense of how to estimate in advance, and no easy way to compare across competing products that meter the same underlying capability differently.
This asymmetry favors vendors in the short term, since unpredictable variable costs are harder for customers to benchmark and shop against than a flat fee. It carries a longer-term risk too: customers who cannot predict their own Adobe or OpenAI bill tend to develop a generalized distrust of the pricing relationship, which shows up eventually as churn even among customers who were never, on any single bill, actually overcharged.
Some vendors are experimenting with usage caps and transparent cost calculators specifically to counter that trust erosion, treating pricing clarity as a competitive feature rather than an afterthought — a bet that predictability is worth more in retention than the extra revenue captured by more aggressive, less legible metering.
The net effect for now is a strange coexistence: fewer total subscriptions per household, but less confidence about what an Adobe or OpenAI bill will actually cost next month. Consolidation solved the problem of too many bills. It did not solve the newer problem of a bill that no longer behaves the way a bill used to.