DCFlex Deliverables

Interconnection queue data can provide valuable early insight into the potential scale, timing, and location of data center development; however, requested capacity alone is not a reliable indicator of future load growth due to uncertainties related to project realization, development timelines, load realization, and phased energization. This Quick Insight summarizes an EPRI framework that transforms interconnection queue data into planning-ready regional load growth forecasts.

Cooling systems account for 30–40% of data center load yet remain one of the least-utilized resources for grid flexibility today. This report evaluates how modern cooling architectures can provide dispatchable, demand flexibility without impacting data center information technology equipment’s reliability and operation. Findings show that ride-through can be extended from seconds to tens of minutes or more through thermal mass, precooling, setpoint relaxation, and thermal energy storage (TES).

This report surveys the landscape of energy storage technologies applicable to data centers and evaluates their fit against EPRI’s Flex MOSAIC™ framework for characterizing large-load flexibility. It reviews electrochemical, mechanical, and thermal energy storage options, covering process descriptions, technology readiness, and key performance metrics. The report then maps these technologies against Flex MOSAIC’s five flexibility classes, identifying which storage types are best suited to each class’s notification, duration, and utilization requirements.

Rapid growth in artificial intelligence and high-density digital workloads is driving unprecedented demand for new data center capacity, but grid interconnection delays and transmission constraints increasingly limit the ability to secure reliable power. As a result, onsite generation, ranging from uninterruptible power supply systems and emergency backup generators to peaking units, primary power, and bridging power, has become central to data center design.

EPRI collected and analyzed responses from 1,126 qualified and verified data center professionals from twelve countries with high data center concentrations and projected growth. This report presents key findings from the survey results, including

This white paper introduces a set of metrics for assessing portfolio readiness in the context of data center load growth. It provides an overview of the challenges system resource planners encounter when integrating data center loads and introduces metrics that can be incorporated into the resource planning process to address these challenges. The paper also explores how these metrics fit within the broader portfolio planning framework. The developed framework and metrics were informed by an extensive review of existing frameworks and literature along with a series of expert interviews.

As part of EPRI’s DCFlex initiative, a field demonstration was conducted to evaluate hydrotreated vegetable oil (HVO) as an alternative fuel for backup diesel generators used in data centers. The project compared HVO and conventional fossil diesel under real-world operating conditions at a data center facility in Dallas, Texas, using a 2 MW backup generator.

This report examines what constitutes a distribution-connected data center (DxDC), emphasizing that classification depends on more than voltage level or facility size; it is shaped by interconnection configuration, planning practices, jurisdictional treatment, operational visibility, and administrative considerations. The paper explores the diverse drivers behind distribution-level interconnection, with time-to-power emerging as the primary motivation, while highlighting that distribution connection does not eliminate underlying grid constraints or transmission impacts.

This report examines how data center load behavior may affect synchronous generator reliability. Large data centers, especially those supporting AI, can create cyclical and oscillatory demand patterns that differ from traditional loads and introduce new operational challenges. The report describes how these load oscillations can interact torsionally with synchronous turbine-generator shafts, amplifying mechanical stress, accelerating fatigue, and reducing reliability.

The DCFlex Load Shape Library provides access to standardized hourly data center load shapes derived from measured power consumption data from multiple facilities. The library enables users to explore, visualize, and compare demand characteristics across different facility sizes, ownership models, and workload types.