Pulse Report: AI Cost of Compute

Senior operators are spending real money on AI while most still cannot see it, track it, or forecast it.

The headline is a confidence gap. Leaders are deploying AI, but the money moves faster than the visibility. Only 17 of 99 see AI spend as a clean line item. 56 say it is buried or invisible. A third, 36 of 99, do not track whether costs beat forecast, and 31 who do look have already been burned. Confidence in 12-month forecasts is thin. 47 of 99 score it 1 or 2 out of 5. Just 4 pick the top mark. Unit economics are rarer still. Only 5 of 99 know their cost per AI task precisely, while 46 say no. Controls lag too. 30 have none yet, and those who act lean on budget alerts and quotas, not routing or caching. Pricing is now shaping strategy. 46 of 99 report that consumption-based pricing changed a decision, including 8 projects killed outright. A few are ahead. One director in tech described a proxy on top of a cloud model that routes traffic and throttles spend. That remains the exception. For most, the next step is simple and overdue: make AI spend a named line item, track overruns monthly, and measure cost per outcome. Visibility comes first. Control and forecasting follow.