Building More Efficient and Resilient AI Data Centers
Unleashing U.S. global dominance in AI through more efficient data centers
(Graphic by Kelly Machart | Pacific Northwest National Laboratory)
While data centers have been around for decades, those designed to support computing infrastructure for artificial intelligence (AI) require significantly more advanced hardware, high-performance computing capabilities, and robust data management. These energy-intensive AI data centers must handle increased processing demands, including AI training spikes, which demand much more power from the grid or micro-grid they are attached to.
From 2017 to 2023, data centers significantly boosted their annual contribution to the U.S. gross domestic product, from $355 billion to $727 billion, fueled by electricity-intensive computing equipment and vital cooling systems that drive operational efficiency. From 2019 to 2023, new data centers, especially computationally intensive generative AI data centers, led to a 10 percent growth in electricity demand across the 10 states with the highest demand growth. By 2028, data centers in the U.S. are anticipated to use about 12% of the country's total electricity consumption.
AI data centers that are built with energy management in mind can help alleviate this growing power demand in a variety of ways that leverage effective use of resources and smart integration with the grid. Pacific Northwest National Laboratory (PNNL) is committed to helping unleash U.S. global dominance in AI by improving the energy performance of the nation’s data centers.
PNNL is using its decades of experience in building technologies and controls research, development, and validation to address the growing demands of AI data centers on the nation’s electricity infrastructure.
PNNL collaborated with ASHRAE, the National Electrical Manufacturers Association (NEMA), and over 50 industry partners to develop a best practices framework for the efficient planning, siting, design, commissioning, retrofit, and operation of energy-intensive AI data centers.
The framework is a one-stop shop for AI data center planning and development that addresses all aspects of energy source, energy use, and water use in data center facilities, including electricity generation and opportunities for waste heat reuse. It also includes recommendations based on facility climate zone and load density, and will act as a dynamic, living resource accessible online to anyone involved in data center design and management.
The information in the framework will enable the country to grow its AI data center capacity while lowering impacts on the grid and consumers by improving facility reliability, safety, and performance. The framework synthesizes expert knowledge across multiple domains impacting data center performance by leveraging PNNL building systems expertise alongside leading industry experts in data center development.
Stay tuned for the framework launch on June 10, 2026!
Bing Liu, Director of Energy Innovation
bing.liu@pnnl.gov