Data center developers facing a growing thicket of energy-related technical challenges can now consult a detailed set of best practices for navigating them holistically, an emerging need during a period of surging AI-fueled demand for ever larger and more complex, capable and demanding facilities.
Three bodies active in crafting standards and conducting research for the built environment — National Electrical Manufacturers Association (NEMA), ASHRAE and Pacific Northwest National Laboratory (PNNL) – have collaborated to produce the AI Data Center Energy Performance Framework, a resource billed as a set of comprehensive guidelines for more curated, up-to-date, expertly-informed datacenter design, construction and operation.
The guidance, accessible on the ASHRAE website, is characterized by NEMA as “a unified suite of technical standards, best practices, and design guidance for developers, utilities, engineers, and facility operators building next-generation AI infrastructure.”
The framework’s creators have invoked AI because its massive computing power demands have radically changed the equation for data centers, requiring assets of extreme size and scale that carry implications for facility power, HVAC and water infrastructure, as well as safety, maintenance, security, and reliability. Those more acute challenges increasingly demand better coordination of parties at the design, construction and operations levels, one of the initiative’s primary aims.
With AI driving “rapid changes in load density, system design and operational expectations,” says Bill McQuade, president of ASHRAE, an HVAC industry organization, the framework provides “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.”
A June 10 online panel assembled by NEMA unpacked elements of the framework, explaining its need, purpose, goals and applicability.
Runaway data center growth on fast timelines to deployment, the panel agreed, was fast outstripping the knowledge and project-execution expertise base needed to build facilities thoughtfully, opening the door to an expedited, 12-month effort to craft the framework.
The rapid uptake of AI means “it’s a very different game in data centers from years ago,” says panelist Bing Liu, PNNL director of buildings and industrial sector, who spearheaded the collaboration. That understanding prompted an initial industry survey seeking knowledge of the “pain points,” in bringing datacenters online. Those revelations of a “huge knowledge gap in terms of how to build and operate a data center today,” helped produce the framework, one she hails for being a “living document,” readily accessible online and updatable as needed.
That will be a key feature of the framework, says NEMA’s senior vice president technical affairs, Patrick Hughes, because technology that can improve data center operations is evolving and will be key to continued development. New electrical technologies, such as solid-state transformers, for example, will be referenced in the framework as energy-saving tools that address the resource’s overarching focus on maximizing energy performance.
“Giving data center designers the latest thinking from the makers of products to build optimally designed facilities,” he says, will be an important contribution to a sector that has moved well beyond building simple structures housing servers: “This market used to be about building buildings; if you’re talking about a 750MW data center, that’s more like a small city.”
With more data center projects being scrapped or put on hold, in part because of grid interconnection issues, energy efficiency will only grow in importance as a prime consideration, Hughes adds. To that end, designs that can control power usage, incorporate on-site power and make facilities grid assets will be prized, and the framework, he emphasizes, will prioritize detailing the path to that goal.
But the primary challenge for data center developers the framework seeks to address is knitting all project elements together from the start. Efficient power sourcing and distribution lies at the root, but modern datacenter development requires the input of up to a half-dozen design and engineering disciplines notes another panelist, ASHRAE president-elect, Sarah Maston.
“We need a systems-level approach beginning at the design phase,” she says. “Organizations must adopt an integrated design framework to ensure scalability and cost-effective operation.”
