Marginal cost is the reason enterprise software economics have been changed by artificial intelligence.
Simply put, the additional cost of supplying traditional software seat number 5,000 is very close to zero, once the software has been written and deployed commercially. That is why profit margins for software were traditionally so high.
The same impact happens based on usage. The cost of supporting a very-active license user was not very different from supporting a light-usage customer or user.
AI breaks that model as each inquiry involves additional real cost. So marginal-cost pricing doesn't perform as in the past. Instead, costs have to account for usage, since AI inference imposes real new costs.
Basically, infrastructure cost has become variable (at scale) for the first time in software-as-a-service history, since the cost of the next operation is not insignificant.
So AI inference costs have to be recovered in some way other than the licensed, per-seat mechanism. One way or the other, variable usage has to be accounted for and priced into the retail models.
No comments:
Post a Comment