Is inference spend per employee a good AI metric?

Answered by Falk Gottlob · Reviewed · 1 min read

THE SHORT ANSWER

Not on its own. It shows where money went, not what came back. Mercor, which raised a $750M Series D at a $20B valuation on November 10, 2026 as The Next Web reported, says it spends 3x as much on inference as on employee salaries and that the spend pays for itself. The decision behind that, keep hiring and buy compute per person, is worth copying, and swyx's AINews reports the run rate went from $1B in February to $3B now. But the ratio divides one cost by another, and RuntimeWire reported that expert payouts above $4M a day sit outside it. My own small version: I built 39 agents in 80 days, paid inference on all of them, and 13 had zero references anywhere else in the system. Report three lines under the ratio: the share of runs something used, reviewed hours, and tokens per finished unit.

SOURCES

THE LONG VERSION

RELATED ANSWERS