UII UPDATE 553 | NOVEMBER 2026
Intelligence Update

Cheaper tokens, but higher bills

The price of AI inference is falling. For a given level of capability, a token now costs a fraction of what it cost a year ago. But enterprise AI bills are rising, not falling. Organizations are using the savings to consume more AI: larger models, longer reasoning chains, autonomous agents and wider adoption across the lines of business are all increasing the number of tokens used. Unless consumption is managed, cheaper tokens will not mean cheaper AI, either in the cloud or on-premises.

Tokens are getting cheaper

Estimates of how quickly inference prices are falling vary, but all point the same way. A study by MIT researchers found that the cost of completing a fixed benchmark task with frontier models fell by 80-90% annually between April 2024 and November 2025. Earlier analyses by Epoch AI found even steeper falls on some tasks.

Request an evaluation to view this report

Apply for a four-week evaluation of Uptime Intelligence; the leading source of research, insight and data-driven analysis focused on digital infrastructure.

Posting comments is not available for Network Guests