Demand response (DR)—the intentional modification of electricity consumption in response to grid conditions or price signals—has expanded from emergency curtailment into a portfolio of market-based services that can support reliability, integrate variable renewable energy, and reduce infrastructure costs (U.S. Department of Energy, 2006; International Energy Agency, 2023). At the same time, data centers have become a fast-growing and highly concentrated source of electricity demand, with U.S. data center electricity use rising to an estimated 176 TWh in 2023 (4.4% of U.S. electricity consumption) and projected to reach roughly 325–580 TWh by 2028 under scenario assumptions driven in part by accelerated AI server adoption (Shehabi et al., 2024). This co-evolution—rapid load growth and increasing need for flexibility—makes data center participation in DR both attractive and complex. This paper synthesizes peer-reviewed research, national laboratory field studies, standards documents, and selected utility/market operator materials to explain (a) what "data center DR" means in operational terms, (b) how data centers technically deliver flexibility, (c) how DR is communicated, controlled, and verified, (d) how participation is monetized across tariffs and capacity/ancillary markets, (e) what empirical case studies indicate about feasibility and outcomes, and (f) what regulatory, reliability, cybersecurity, and environmental considerations constrain or enable scaling. Key findings supported by the reviewed evidence include: (1) measurable, automation-friendly load flexibility exists in both cooling/infrastructure systems and IT workloads, but the magnitude and response time vary significantly by mechanism and facility type (Ghatikar et al., 2012; Wierman et al., 2014). (5) Recent deployments illustrate two emerging archetypes: "compute-aware DR" (shifting non-urgent tasks across time and sometimes geography) and "grid-interactive UPS/battery services" (fast frequency response and reserve-type services), with high-profile examples reported by large operators and aggregators in multiple regions (Google Cloud, 2023; Microsoft, 2022; Baringa Partners LLP, 2023).
Samuel et al. (2026) studied this question.