Revenue & profit estimatorBeta

Estimated revenue, compute cost, model license fee and operator profit for each chip from measured throughput at a target interactivity, per GPU-hour or per gigawatt-year

Benchmark config

Model

Target P90 interactivity (tok/s/user)

050100

Inference profit estimator

The inference profit estimator computes revenue per chip-hour for each chip at a chosen interactivity and utilization, stacked into compute expense (TCO $/chip/hr), the model lab license fee and operator profit.