AI companies shift from flat fees to pay as you use pricing

In short: AI services are increasingly moving from monthly subscriptions to pay as you use pricing, and that is changing who benefits and who pays.

What's going on

For the past few years, many popular AI chatbots and AI tools were priced below what they likely cost to run. The gap was often covered by investors, which helped AI spread quickly.

A new Financial Times column says that is starting to change. More AI services are switching from flat fees to usage-based pricing, meaning customers pay more when they use more. One example mentioned is Microsoft-owned GitHub, which began moving in this direction in April.

This shift matters because running AI requires a lot of “compute,” which is basically rented computer power in large data centers (like paying for electricity and machines in a giant server warehouse). The column argues that, so far, a lot of the money in the AI boom has flowed to chip makers, since AI needs many expensive chips to run.

As companies face higher and less predictable AI bills, some are cutting back on usage. The column also notes a reported increase in the use of Chinese AI models, which are often open source and cheaper. Open source means the underlying code is shared publicly, so others can use and modify it (like a recipe anyone can copy).

What to watch

Investors are watching whether cheaper AI leads to much more total usage, or whether AI becomes more like a commodity, where providers struggle to charge premium prices. Another pressure point is the rising cost of building AI infrastructure, with estimates that spending could reach $1.2tn next year across five large companies, according to Morgan Stanley.

Source: Financial Times