Grid-Interactive AI Data Centres for Renewable Smart Grids: A Systematic Review

Abstract

With the increasing number of artificial intelligence workloads being run in hyperscale and cloud data centres, there is a growing demand for electricity in power systems with high shares of variable renewable energy. This systematic review analyzes how grid-interactive AI data centres can help renewable smart grids with flexible computing and energy-management strategies. There has been a search of studies published between the years 2022 and August 2026 on Scopus, Web of Science, IEEE Xplore, ACM Digital Library, ScienceDirect, SpringerLink, Google Scholar. A search was done with various terms to get better relevancy. Out of the total number of studies, 43 studies were finally included. There are four evidentiary themes relating to architectures (enabling), coordination (renewable-energy), performance (multi-dimensional) and barriers (implementation). Insights reveal that the shifting workload, relocation, battery-backed UPS, thermal control enables an increase in renewable-energy use, peak demand decrease, operating costs decline, and supports grid services. Nevertheless, the outcomes are heavily dependent upon the forecast accuracy, workload flexibility, battery degradation, converter inertia, service level requirements and electricity market conditions. There are times when maximising economic savings does not lead to maximising carbon reductions and therefore we need to value costs, carbon intensity, grid reliability and computing joint policy evaluation. Simulation-based studies and limited field verification are the dominant evidence at present. It is therefore essential to develop greater standardisation, lifecycle carbon assessment, uncertainty-aware control and operational demonstrations for reliable deployment.