RICHLAND, Wash.-As more companies invest in data centers to support their shift to AI-driven systems, demand for electricity will rise significantly.
To help developers, engineers, utilities and policymakers understand the complexities and best practices of building a data center, the Department of Energy's Pacific Northwest National Laboratory has codeveloped a comprehensive guide to siting, building and maintaining the facilities. The guide was written by experts from PNNL, the National Electrical Manufacturers Association and ASHRAE.
The new guide acts as a one-stop shop for anyone involved in the development of a data center, said Kieren McCord, a systems engineer at PNNL and lead author of the guide.
"Policymakers, utilities and builders of data centers can find research-backed best practices in this guide for building data centers that keep energy costs low while safely maintaining operations," McCord said.
Data centers have been storing our data for decades, but in the last 20 years, large tech companies have been building what's known as "hyperscaler" data centers, which house thousands of individual servers in buildings up to 1 million square feet. These servers store, process and analyze terabytes of data from all kinds of industries-from tech to manufacturing to research-as well as help to train AI models used in those same fields.
"All customers benefit from these collaborative efforts to ensure that new AI data center customers are good grid citizens," said Steve Rosenstock, senior manager of customer technology solutions for the Edison Electric Institute. "The broad participation in this project shows how committed we all are to customer efficiency, reliability and affordability."
Working with the grid
As of 2023, data centers in the United States use about 4.4% of the country's total electricity consumption, and that could rise to 12% by 2028. Because AI-driven data centers use so much power, researchers are looking into ways that the facilities can support grid stability.
"A grid-interactive design enables a data center to actively communicate and coordinate with the electrical grid. That way, they can evolve from passive energy consumers to active participants in managing electricity usage from the electrical grid," said Srinivas Katipamula, a mechanical engineer and Laboratory Fellow at PNNL who wrote the guide's section on grid-interactive data center operation.
The new guide describes several strategies for building a grid-interactive data center. For instance, developers could use microgrids or on-site battery systems to store and use energy when demand on the grid is high, Katipamula said.
During times when electricity demand is low, a data center's batteries would charge. When demand for electricity starts to rise throughout the day or at different times of year (such as in the summer, when more people use air conditioning), data centers could pull power from their battery systems. A microgrid would allow a data center to isolate itself from the larger grid while continuing to operate.
Data centers can also employ smart controls that automatically decrease or increase operations in response to demand on the grid.
"PNNL brings world-class expertise on grid integration, energy efficiency, and resilience to this collaboration. Their engineering- and research-driven approach is exactly what the data center industry needs as these facilities become increasingly grid-interactive. Modern data centers are participating in demand response, integrating on-site generation and storage, and managing real-time power quality, which dramatically increases technical complexity," said NEMA President and CEO Debra Phillips.
A major contributor to a data center's power usage is cooling the servers-another issue addressed by the new guide.
The "thermal management" problem
Data centers store thousands of individual servers, each about the size of a laptop. These units are running all the time, creating heat.
"It's hard to estimate how hot the data center interior can become, but think about when your laptop feels warm on your lap. Now multiply that by an entire room of server racks," McCord said.
Data centers spend about 20-40 percent of their energy simply on cooling. The new guide describes several ways data center operators reduce their electricity loads while cooling, especially during times when demand on the grid is already high.
Some operators will raise the set temperature and let the space get a bit warmer, similar to when utilities ask homeowners to set their air conditioning temperatures a little higher to relieve grid pressure during a heatwave.
Some operators will pre-cool the data center in anticipation for a rise in electricity demand.
"That way, during the demand time, the room and servers are still warming up from the colder temperatures so that electricity isn't needed for cooling," McCord said.
Some tech companies are even working on improving the ways that individual servers can themselves cool down to the chip level, McCord added.
Industry collaborators
The new guide "translates complex technical challenges into clear, actionable strategies that help operators enhance performance, control costs and make more effective use of energy, while strengthening reliability at both the facility and grid level," said 2025-26 ASHRAE President Bill McQuade.
Besides grid-interactive design and cooling guidance, the new guide also covers planning and siting, operations and maintenance and integrated design, among other topics. The guide is available here.
To create the new guide, PNNL, NEMA and ASHRAE teamed up with more than 50 industry partners, including tech companies like NVIDIA and IBM, as well as heating, ventilation and air conditioning manufacturers like Carrier and Siemens.
"This guide brings together the most comprehensive industry expertise on data centers in a single resource," said Bing Liu, director of buildings and industrial programs at PNNL, who launched this industry-lab partnership a year ago. "Rather than being frozen in time, it's a dynamic online resource that can be updated, remain relevant and stay accessible to anyone involved in developing a data center."
