AI consumption pricing turned software cost into a meter. Here is how to read yours before the bill does.

Software used to be priced at the signature. Now it accrues at the meter. Ramp's payments data shows AI token spend up twenty point seven times in thirteen months, and the FinOps Foundation reports ninety-eight percent of practitioners now manage AI spend, up from thirty-one percent two years ago. A seat is a price you approved once. A meter is a price you discover every month.

Bill walks into the first room of the Grid, the meter room, and breaks down the four parts of every consumption agreement: the unit, the rate, the commitment, and the true-up. Along the way: why the median business pays two thousand dollars a month while the average pays one hundred forty thousand, how Coinbase cut its AI bill nearly in half while usage grew, the controller who found ten thousand dollars a month in a term he had never heard, and the arithmetic that turns any invoice into a meter reading.

Plus, a listener question from Elena in Cincinnati: finance wants next year's AI budget as one fixed number, and consumption pricing makes that feel impossible. What do you actually give them?

IN THIS EPISODE

• The anatomy of a meter: the unit, the rate, the commitment, and the true-up, and why you need all four named for every consumption agreement you own
• Ramp's April benchmarks: a sixty-times gap between median and average AI spend, and the twenty-times price difference decided by a model default
• How Coinbase halved AI spend while usage grew: better defaults, routing, and caching, not friction and spend alerts
• The AngelList controller who learned the phrase "prompt caching" from a spend briefing and recovered ten thousand dollars a month the same day
• The three numbers to hand finance instead of one fixed budget, and the one-hour homework that produces your first meter reading

CHAPTERS

• (00:15) - Meter Room Opens
• (01:16) - Reading the New Bill
• (04:12) - What a Meter Is
• (07:32) - Real-World Savings
• (09:31) - How to Read Usage
• (12:55) - Budgeting the Unknown
• (14:21) - Utility Lessons
• (15:48) - Measurement Discipline
• (16:38) - Homework and Next Steps

TRANSCRIPT

Click here to view the episode transcript.

SOURCES & FURTHER READING

• Ramp, AI token spend controls launch (July 2026): https://www.prnewswire.com/news-releases/ramp-launches-ai-token-spend-controls-302827389.html
• Ramp, AI token cost benchmarks (April 2026 data): https://ramp.com/blog/ai-token-cost-for-businesses
• State of FinOps 2026, Linux Foundation / FinOps Foundation: https://www.linuxfoundation.org/press/state-of-finops-survey-ai-value-and-skills-top-priorities-as-finops-matures-across-technology-value-98-manage-ai-90-saas-64-licensing-48-data-center-1
• Coinbase CEO Brian Armstrong on cutting AI cost: https://finance.yahoo.com/markets/crypto/articles/coinbase-ceo-halved-ai-costs-130000536.html
• FOCUS 1.4, FinOps Foundation: https://www.finops.org/insights/introducing-focus-1-4/

ABOUT THE SHOW

Operational ITAM is a podcast about the unglamorous machinery of enterprise technology — hardware and software asset management, licensing, audit defense, SaaS governance, and the money quietly leaking out of all of them.

Bill Van Nort has led IT asset management, end-user computing, IT operations, and workplace technology at large organizations across banking, mortgage, and automotive. He has reclaimed millions in software spend and survived audits from the biggest publishers on the planet.

Consulting enquiries and listener case files: operationalitam.com

The Operational ITAM Podcast is an EpicB Media LLC production.