An AI agent uses compute differently than a traditional application. Over a single session, it writes and runs code, calls a model, and waits on a tool or a person before picking the work back up. As a result, it needs real compute in short, intensive bursts.