The growing power demand of data centers driven by AI highlights the urgent need for sustainable energy solutions, with Microgrids and Distributed Energy Resources (DER) offering a more resilient, cost-effective alternative to traditional grid upgrades, as demonstrated through Xendee’s multi-year optimization approach.
By: Giovanni Maronati, Michael Stadler, Timothy Grunloh, Reynaldo Guerrero, Nishaant Sinha
The Data Center Power Surge: AI and Beyond
In the United States alone the power demand of data centers is expected to triple by 2030, rising from 3% to 11–12% of total electricity consumption[1],[2]. Hyperscale data centers, capable of rapid scaling to meet the growing computational needs of AI and cloud services, are at the forefront of this surge. These facilities can require upwards of 1 GW of power, equivalent to the output of a nuclear power plant[3]. Although improvements in GPUs and AI server efficiencies have been notable - such as a recent 47% improvement in performance per Watt achieved by one leading hardware provider - the sheer scale of AI workloads continues to drive net increases in energy consumption[4]. At the same time, historical efficiency gains observed between 2015 and 2019, averaging 15% annually, have begun to decelerate to 1-2% per year, underscoring the urgency for alternative energy solutions[5],[4].
Grid Challenges and Energy Sustainability
The rise of data centers has placed immense stress on regional power grids, particularly in areas of concentrated activity, such as Northern Virginia’s “Data Center Alley”[1],[6]. Utilities and regulators face complex trade-offs: rapid energy infrastructure development to meet immediate demand risks delaying decarbonization goals, while renewable generation projects require longer timelines to scale effectively[7]. For example, Duke Energy forecasts a significant increase in gas capacity to meet an additional 2 GW of demand by 2030[8]. As indicated by data from the U.S. Energy Information Administration (EIA), the peak hourly U.S. electricity demand stayed constant for the last decade[9]. Thus, investments in transmission lines have also stalled. Data shows that the added new high voltage transmission lines dropped by 62% in the late 2010s compared to the early 2010s[10]. Thus, the U.S. electric grid is not prepared for significant load growth as expected due to AI.
The Role of Microgrids: A Resilient, Sustainable Solution
A Microgrid connects loads and distributed energy systems (e.g., PV, battery systems, wind turbines, but also thermal resources such as Combined Heat and Power running on hydrogen, among others) within clearly defined electrical boundaries that acts as a single controllable entity with respect to the grid. A Microgrid can operate in either grid-connected or in island mode, including entirely off-grid applications. By integrating generation, energy storage, and flexible loads, Microgrids can:
|
Different Approaches to Powering Data Centers with Microgrids
A single-step approach, where utilities upgrade their distributions system and power plants to meet rising loads, has been used multiple times. However, this method often fails to account for sustained demand growth or future energy technologies like SMRs. Also, the rising power demand requires expensive upgrades to the cables and transformers which cannot be done quickly. New transmission projects take about 7 to 10 years to develop, approve, and construct, according to estimates from the California Independent System Operator (CAISO)[11]. As a result, costly grid expansions risk becoming stranded investments when more advanced, low-carbon DER solutions come online that would reduce the need for grid upgrades to the full extent.
In contrast, a multi-year approach using DERs and Microgrids offers a more strategic and cost-effective solution by leveraging Xendee’s Adaptive Multi-Year Algorithm. This innovative method begins by prioritizing near-term optimization, accounting for projected changes such as increasing utility rates, technology degradation, and shifting energy demands. The process involves an iterative annual analysis where each year’s design and operational decisions are optimized based on updated inputs and the results from previous years. This ensures that investments align with current conditions and anticipated future needs.
The algorithm evaluates whether investments should be made immediately or delayed until conditions are more favorable. For instance, a project might avoid renewable installations in the initial years due to high costs but integrate solar PV and storage as installation prices drop and tariffs increase. By the end of the analysis period, cumulative capacity accounts for system degradation and future requirements, ensuring no oversizing while maintaining operational efficiency. Additionally, detailed year-by-year cash flow tables are generated, providing advanced financial projections to guide planners and instill investor confidence.
This multi-year method addresses energy needs in two phases:
However, with such a multi-year approach it is important to design the DERs in an effective way to avoid sunk costs later on. In other words, the DERs installed now, need to augment the SMRs of the future to create a very cost-effective solution for data centers.
By phasing investments in this manner, Microgrids achieve near-term reliability and sustainability while preparing for long-term growth with advanced energy solutions. In Par 2 of this blog series, we will demonstrate this approach by using Xendee’s data center and multi-year modeling capabilities.
Xendee is a software platform for designing optimized Microgrid and EV infrastructure, and operating them efficiently in real-time. This allows users to create reliable, bankable DER systems that reduce engineering costs, energy prices, and CO2 emissions while also improving energy security and resilience to power outages
To dive into the comparative analysis, and understand this powerful multi-stage blueprint for meeting the energy demand of data centers with DER and SMR, read Part 2 of this series.
References:
[1] McKinsey & Co., “How data centers and the energy sector can sate AI’s hunger for power,” 2024. [Online]. Available: https://www.mckinsey.com
[2] International Energy Agency, “Electricity 2024 – Analysis and Forecast,” 2024.
[3] W. Fork, An Overview of Nuclear Energy Demand and AI, U.S. Nuclear Capital Summit, 2024.
[4] Schneider Electric, The AI Disruption: Challenges and Guidance for Data Center Design, 2023.
[5] Goldman Sachs Global Investment Research, AI, Data Centers and the Coming US Power Demand Surge, Apr. 2024.
[6] PJM Interconnection, Data Center Load Growth Forecast, 2023.
[7] S&P Global Commodity Insights, “Rising datacenter demand forces reckoning with US utility decarbonization goals,” 2024.
[8] Duke Energy, 2023 Integrated Resource Plan Update, Jan. 2024.
[9] U.S. Energy Information Administration, Hourly Electric Grid Monitor, October 2023
[10] GridStrategies, The Era of Flat Power Demand is Over, December 2023
[11] CAISO, The CAISO’s 20-year Transmission Outlook, May 2022. http://www.caiso.com/InitiativeDocuments/20- YearTransmissionOutlook-May2022.pdf