For years, software companies have adhered to the “seats” model, charging fees based on the number of users. However, artificial intelligence is radically changing this system. As AI tools become more advanced, the costs of using them are becoming increasingly expensive. Now, a new era is dawning that could fundamentally change the software industry: pay-as-you-go.
Why is AI becoming expensive?
At the heart of this change lies a new type of AI – powerful models that don’t just respond quickly, but think step-by-step, check their work, and perform multiple calculations before responding. This requires enormous computing power in a process called “inference-time compute”.
For example, OpenAI’s latest o3-high model uses 1,000 times more processing, i.e., “tokens,” to answer a single question compared to the previous o1 model. Barclays analysts estimate that the cost of preparing one complex response per user query could be approximately $3,500.
Now imagine the costs that would arise when this number is multiplied by millions of users.
Why traditional pricing models are no longer effective
Until recently, SaaS companies such as Salesforce, Microsoft, or Zoom traditionally used a stable payment model for each user. This method was beneficial when the cost of providing the service to the user was practically zero. However, with artificial intelligence, this cost can change dramatically for each user – the more the user utilizes the system, the higher the cost becomes.
This puts the flat-rate system at risk. If one customer uses significantly more AI power than another, the company may incur losses.
The consulting company AlixPartners clearly explained this problem: “AI agents may drive a much higher cost of revenue compared to traditional SaaS offerings, forcing companies to rethink how they manage those costs.”
“AI agents may drive a much higher cost of revenue compared to traditional SaaS offerings, forcing companies to rethink how they manage those costs.”
To address this issue, many companies are switching to an activity-based pricing model. In this model, the customer pays based on the number of tokens processed, questions asked, features used, or automations triggered.
This approach is not new in other areas of technology. For example, the developer platform Vercel charges more as the number of site visits increases. Recently, OpenAI CEO Sam Altman proposed a credit-based system – the user pays for a basic subscription and uses credits to access various AI functions, purchasing additional credits if needed.
This approach gives customers more control and helps software companies manage the growing costs of AI infrastructure.
What does this mean for the future of software?
We are entering a new stage of the AI economy – now each interaction has specific costs. Software companies that want to survive will have to adapt their pricing systems to this reality. Fixed tariffs may gradually disappear, replaced by flexible and usage-based plans.
For users, this means greater transparency – but also the need to monitor their usage volume. In the future, using powerful AI tools may become similar to managing your phone’s internet data.
Prepared by Navruzakhon Burieva
