Skip to lesson
Exit
GPU Economics & Capacity Planning1 / 3

2 min lesson

Reserved vs. on-demand vs. serverless

Answer this: "A team reserved a GPU cluster at a 40% per-hour discount but runs it at ~25% sustained utilization. Why might this be more expensive than on-demand and what would you check?"

Step 1 of 3

Reserved vs. on-demand vs. serverlessthe buy decision is a bet on demand certainty

The cheapest hour comes from committing ahead of time and the most flexible hour comes from buying it the moment you need it. Serverless is a third mode - priciest per unit, but the cleanest way to absorb pure burst and experiments. Choosing the mix is a bet on how confident you are in the forecast.

RESERVED VS. ON-DEMAND

Interactive diagram. Tab through its regions; each focused region shows its detail in the panel below.

diagram: compare

The buy decision is a bet on demand certainty - pick the column that matches your confidence in the forecast.