Designing Datacenter Power Delivery Hierarchies for the AI Era

• URL: http://arxiv.org/abs/2605.16255v1
• Date: Fri, 15 May 2026 17:58:58 GMT
• Title: Designing Datacenter Power Delivery Hierarchies for the AI Era
• Abstract summary: Demand for AI accelerators is rapidly increasing rack power density, with projections approaching 1MW per deployment by 2027.<n>As power densities increase, a datacenter designed for a different target density may strand power, i.e., may be unable to use all the power that its delivery hierarchy has provisioned.<n>We develop a framework for evaluating datacenter power delivery designs using throughput, power, and cost metrics over realistic arrival, oversubscription, and decommissioning sequences.
• Score: 4.15652461187777
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Demand for AI accelerators is rapidly increasing rack power density, with projections approaching 1MW per deployment by 2027. This poses a major challenge for datacenter power delivery designers. As power densities increase, a datacenter designed for a different target density may strand power, i.e., may be unable to use all the power that its delivery hierarchy has provisioned. Designs must remain efficient over long datacenter lifetimes and multiple hardware generations. Power utilization is particularly important as grid power capacity is a scarce resource in the AI era. Designing an efficient power delivery hierarchy for the long run is difficult because rack placement feasibility, workload impact, and cost depend jointly on electrical topology, deployment granularity, placement policy, power oversubscription, and workload mix. Moreover, each of these factors evolve over time, have inter-dependencies across multiple resource dimensions, and generally do not lend themselves to closed-form analysis. To address this challenge, we develop a framework for evaluating datacenter power delivery designs using throughput, power, and cost metrics over realistic arrival, oversubscription, and decommissioning sequences. The framework combines projection models for GPU, compute, and storage deployments with operational factors grounded in production data from Microsoft Azure. Our results show that multi-resource stranding materially changes deployable capacity, effective capital expenditure, and delivered performance, and quantify how rising density from rack- and pod-scale AI systems shapes these outcomes. For AI datacenter design, the relevant planning objective is not installed megawatts, but deployable capacity over time.

Related papers

• PowerAtlas: Towards Electricity-Computing Co-Scheduling for Power Systems [42.32975870695121]
  PowerAtlas is a framework for electricity-computing co-scheduling.<n>It integrates historical instances, domain knowledge, and physical constraints.<n>It produces joint decisions satisfying grid operational rules and service-level agreements.
  arXiv  Detail & Related papers  (2026-07-29T09:53:50Z)
• A Phased Development Framework Enabling Islanded Operation of Sustainable AI Data Centers With Onsite Grid-Following and Grid-Forming Energy Architectures [1.7313417696295232]
  Current projections estimate that approximately 50 GW of AI data center capacity will require grid connectivity in the United States by 2030.<n>This manuscript proposes a phased development framework for AI data center expansion.<n>A modular construction architecture is presented, along with a detailed analysis of integrated energy systems.
  arXiv  Detail & Related papers  (2026-07-19T19:43:35Z)
• A Survey on the Green Development of Large Models: From Resource-Efficient Architectures to Hardware-Software Co-Design [71.32556809242666]
  The rapid expansion of large-scale AI models has led to significant performance breakthroughs across diverse domains.<n>It has also raised critical concerns regarding computational costs, energy consumption, and environmental sustainability.<n>This survey aims to offer a holistic roadmap toward sustainable, scalable, and socially responsible development of large models.
  arXiv  Detail & Related papers  (2026-07-10T04:02:22Z)
• Power-Flexible AI Data Centers: A New Paradigm for Grid-Responsive Compute [0.7399816551817894]
  Traditional power-system planning treats large computing facilities as inflexible peak loads, leading to costly infrastructure upgrades and long delays in grid interconnection.<n>Recent work has shown that AI clusters can reduce electricity consumption during peak demand through software-based workload orchestration.<n>We describe an architecture integrating grid signals, workload scheduling, and power telemetry for fine-grained cluster power control.<n> Experimental results from a real-world deployment on a 130 kW GPU cluster demonstrate multiple forms of flexibility, including rapid load reduction, sustained curtailment, and carbon-aware operation while preserving service levels for priority jobs.
  arXiv  Detail & Related papers  (2026-06-23T19:13:19Z)
• Urban Power Grid Topology and Hierarchy Identification from Open Data [53.468443050795344]
  This paper develops an open-data-driven framework for identification of urban power grid topology.<n>We fuse public infrastructure data (power-lines, substations, transformers, poles) to map the high and medium-voltage skeleton.<n>We then leverage geospatial machine learning on OpenStreetMap to group power demand clusters.<n>We reconstruct the complete grid topology that connects 7,330 buildings and all major electricity infrastructure assets.
  arXiv  Detail & Related papers  (2026-06-19T11:51:38Z)
• An Agentic AI Framework with Large Language Models and Chain-of-Thought for UAV-Assisted Logistics Scheduling with Mobile Edge Computing [47.615329461292504]
  In cloud manufacturing, unmanned aerial vehicles (UAVs) can support both product collection and mobile edge computing (MEC)<n>In this paper, we propose an agentic-AI-assisted optimization framework with two components.<n>First, we develop an agentic AI that combines large language models, retrieval-augmented generation, and chain-of-thought reasoning to translate user input into an interpretable mathematical formulation for the hybrid scheduling problem.
  arXiv  Detail & Related papers  (2026-05-13T09:13:19Z)
• From Barrier to Bridge: The Case for AI Data Center/Power Grid Co-Design [6.39525307115435]
  A single hyperscale training campus can draw power comparable to a mid-sized city.<n>This paper argues that the resulting entanglement of compute and power infrastructure requires a shift from implicit coexistence to explicit co-development.
  arXiv  Detail & Related papers  (2026-05-04T19:08:43Z)
• Measurement of Generative AI Workload Power Profiles for Whole-Facility Data Center Infrastructure Planning [1.2396819107017578]
  We present a methodology that bridges the gap by linking high-resolution workload power measurements to whole-facility energy demand.<n>We measure power consumption of AI workloads at 0.1-second resolution for AI training, fine-tuning and inference jobs.<n>The resulting whole-facility energy profiles capture realistic temporal fluctuations driven by AI workloads and user-behavior.
  arXiv  Detail & Related papers  (2026-04-08T17:56:41Z)
• DCcluster-Opt: Benchmarking Dynamic Multi-Objective Optimization for Geo-Distributed Data Center Workloads [14.834687262955585]
  We present DCcluster-Opt: an open-source, high-fidelity simulation benchmark for sustainable, geo-temporal task scheduling.<n>It combines curated real-world datasets, including AI workload traces, grid carbon intensity, electricity markets, weather across 20 global regions, cloud transmission costs, and empirical network delay parameters.<n>A modular reward system enables an explicit study of trade-offs among carbon emissions, energy costs, service level agreements, and water use.
  arXiv  Detail & Related papers  (2025-10-31T03:07:12Z)
• Improving AI Efficiency in Data Centres by Power Dynamic Response [74.12165648170894]
  The steady growth of artificial intelligence (AI) has accelerated in the recent years, facilitated by the development of sophisticated models.<n> Ensuring robust and reliable power infrastructures is fundamental to take advantage of the full potential of AI.<n>However, AI data centres are extremely hungry for power, putting the problem of their power management in the spotlight.
  arXiv  Detail & Related papers  (2025-10-13T08:08:21Z)
• PowerGrow: Feasible Co-Growth of Structures and Dynamics for Power Grid Synthesis [75.14189839277928]
  We present PowerGrow, a co-generative framework that significantly reduces computational overhead while maintaining operational validity.<n> Experiments across benchmark settings show that PowerGrow outperforms prior diffusion models in fidelity and diversity.<n>This demonstrates its ability to generate operationally valid and realistic power grid scenarios.
  arXiv  Detail & Related papers  (2025-08-29T01:47:27Z)
• Joint Resource Management for Energy-efficient UAV-assisted SWIPT-MEC: A Deep Reinforcement Learning Approach [50.52139512096988]
  6G Internet of Things (IoT) networks face challenges in remote areas and disaster scenarios where ground infrastructure is unavailable.<n>This paper proposes a novel aerial unmanned vehicle (UAV)-assisted computing (MEC) system enhanced by directional antennas to provide both computational and energy support for ground edge terminals.
  arXiv  Detail & Related papers  (2025-05-06T06:46:19Z)
• Power Modeling for Effective Datacenter Planning and Compute Management [53.41102502425513]
  We discuss two classes of statistical power models designed and validated to be accurate, simple, interpretable and applicable to all hardware configurations and workloads. We demonstrate that the proposed statistical modeling techniques, while simple and scalable, predict power with less than 5% Mean Absolute Percent Error (MAPE) for more than 95% diverse Power Distribution Units (more than 2000) using only 4 features.
  arXiv  Detail & Related papers  (2021-03-22T21:22:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.