Dai Wang

Smoothing the Ramp, Not the Peak: Scheduling-Induced Power Dynamics of LLM Inference and Their Grid-Scale Consequences

Large language model (LLM) inference serving is a fast-growing electricity load whose power dynamics remain uncharacterized from a grid-planning perspective. Using real, measured GPU power traces, we show that chunked prefill scheduling, a latency-motivated technique already deployed by default in production LLM servin...