Skip to content
AI Atlas
PaperActive

Characterizing Job Power Elasticity for Power-Flexible AI Training

arxiv.org/abs/2609.11542

quality89

Updated 58 min ago · first seen 12 Sept 2026

paper_01M29X34PACMAZXTTJFBCQH7EM

Published
12 Sept 2026
T1 · 58 min ago
arXiv
2609.11542
T1 · 58 min ago
Category
cs.AI
T1 · 58 min ago

Abstract

Large language model (LLM) training is among the fastest-growing sources of electricity demand in modern data centers, and power availability is a primary bottleneck to continued AI infrastructure growth. Making the power consumption of these workloads flexible could unlock additional power for AI growth, limit increases in electricity prices, and improve the utilization of existing grid infrastructure. However, to realize this flexibility, we must first understand how the performance of training workloads changes when GPU power is reduced. This paper presents the first systematic characterization of \emph{job power elasticity} (the sensitivity of throughput to power reductions) in LLM training. To quantify elasticity, we introduce the \emph{Power Flexibility Index (PFI)}, a normalized metric that quantifies the performance cost of power reductions and provides a control primitive for SLA-aware power flexibility. We collect data from 131 LLM training runs on H200 (plus 24 H200 validation runs and 34 matched H100 runs), including both dense and mixture-of-experts models, pretraining and fine-tuning tasks, and up to 32 GPUs. We find that LLM training jobs exhibit substantial but variable power elasticity, and we identify telemetry signals that predict PFI at runtime. Finally, we demonstrate that PFI-aware power allocation maximizes total tokens/second throughput under power constraints. Under a 30\% power reduction, PFI-aware power allocation recovers ~1.5k tokens/s per job, 63\% of the performance gap between an equal-weight allocation and an oracle with perfect information. Our results establish power elasticity as a measurable property of training jobs and provide a foundation for power-aware, grid-responsive AI infrastructure.

Authors 5

Ayse Coskun, Charles Dawson, Philip Colangelo, Shayan Sengupta, Varun Sivaram

Specification

Official page

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

Arxiv announce type
new

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

arXiv id
2609.11542

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

Categories
cs.AI

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

PDF

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

Primary category
cs.AI

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

Published
12 Sept 2026

Source:arXiv (Atom API + RSS)T1observed 58 min agohigh

Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →

Provenance

Attributed facts

9

Source tiers

T19

Freshest observation

58 min ago

Conflicts

None