Pragmatic Engineer: AI Token Spend Up 10x in 6 Months, Inside 15 Companies' Reckoning

Gergely Orosz, the Pragmatic Engineer, just published the most comprehensive look at enterprise AI token spending in the wild. He interviewed developers and engineering leaders at 15 companies across all sizes. The findings are stark: AI token spend has increased approximately 10x in the last six months at most companies, with no signs of slowing down.

The interviews reveal a pattern that's consistent across the industry: companies are shocked by how much token consumption is accelerating, they're unsure how to manage it, and they're split between two strategies: "let it rip and start measuring" or "curb spending." Most are choosing the first strategy, which means token costs are about to get worse before they get better.

Orosz's research is based on first-hand accounts from leaders at 15 companies. The data points are striking.

At one SaaS company of about 2,000 people, the Dev Productivity Lead described their approach: "Model routing helped keep our costs growing less dramatically. For example, changing the default model reduced cost by 30%. This is our strategy with AI spend, summarised: Short term: spend, spend, spend! Experiment and use whatever models make sense. Measure the impact. Measure key outcomes and report on spend monthly. When spend vs results diverge: adjust."

The pattern is consistent: 10x increases, no forecasting, and leadership generally unconcerned for now, provided they can measure the impact eventually.