InFactPlanner: Planning Sustainable Geo-Distributed LLM Data Centers

2026-08-13

AI Summary

This work addresses the sustainability challenges—encompassing energy consumption, carbon emissions, water usage, and latency—posed by large-scale deployment of large language models (LLMs), for which existing evaluation methods lack efficiency in assessing the holistic impact of diverse infrastructure configurations. The paper introduces the first trajectory-driven decision-support framework that integrates real query traces, hardware–model specifications, site-specific parameters, time-varying grid carbon intensity, and renewable energy models to enable multidimensional what-if analyses for both single-site and geographically distributed LLM data centers. The framework supports rapid, scalable comparisons with energy estimation errors below 10% and reveals significant trade-offs between sustainability and latency objectives: the lowest-carbon deployment strategies are often not those minimizing latency, and carbon benefits are highly contingent on local grid composition.

Abstract

The rapid growth of LLM inference is shifting sustainability concerns from one-time training to continuous serving, where infrastructure decisions shape energy use, carbon emissions, water consumption, and service quality. Yet operators often need to compare deployment alternatives before large-scale infrastructure is built, making direct measurement costly, slow, and sometimes infeasible. We present InFactPlanner, a trace-driven decision-support framework for what-if analysis of sustainable AI data center deployment for LLM inference across single and geo-distributed sites. InFactPlanner combines query traces, hardware-model profiles, candidate site configurations, PUE/WUE parameters, renewable generation models, and time-varying grid carbon intensity to estimate power, energy, carbon emissions, water use, latency, and server utilization. The framework abstracts low-level serving effects into configurable hardware-model profiles, enabling rapid comparison of site selection, capacity placement, hardware, model, renewable integration, and routing choices. We validate the energy accounting pipeline by reproducing reference LLM inference energy estimates with less than 10% deviation, evaluate scalability across multiple data centers and server counts, and demonstrate scenario-driven decision analyses for hardware selection, renewable placement, geographic deployment, and carbon-aware routing. Our results show that sustainability-optimal choices can differ from latency-optimal ones, and that the carbon value of deployment depends strongly on the local grid mix.

Problem

• sustainable AI
• LLM inference
• geo-distributed data centers
• carbon emissions
• infrastructure planning

Innovation

• sustainable AI
• geo-distributed data centers
• LLM inference
• carbon-aware routing
• trace-driven simulation