ASHRAE, Pacific Northwest National Laboratory and the National Electrical Manufacturers Association have introduced an AI Data Center Energy Performance Framework to help data center owners, engineers and operators manage the growing technical demands of high-density computing.
The framework,
hosted by ASHRAE, is designed as a practical resource for new construction, retrofits and ongoing facility operations. It focuses on the building systems behind AI infrastructure, including cooling, electrical distribution, energy use, water use and reliability.
The release comes as artificial intelligence and high-performance computing continue to reshape the data center market.
AI workloads often require higher rack densities and can create more variable power and cooling demands than traditional computing environments. That shift is putting more pressure on mechanical and electrical systems, as well as local utility infrastructure.
ASHRAE cited
Pew Research Center data showing that the U.S. has more than 3,000 operational data centers, with another 1,500 in development. As more facilities come online, operators are facing a familiar but more urgent question: how to support growth without creating avoidable energy, cooling or uptime risks.
The new framework does not focus on one part of the facility. Instead, it looks across the full lifecycle, from early planning and design to commissioning, retrofit work and daily operation. That broader approach matters because AI-focused facilities are increasingly shaped by the interaction between power delivery, thermal management and system resilience.
ASHRAE’s contribution is tied to HVAC, thermal management and facility performance, including its work through Technical Committee 9.9 on mission critical facilities, technology spaces and electronic equipment. NEMA brings electrical equipment and safety expertise, while PNNL contributed federal energy systems research and helped coordinate the working group.
Power and Cooling Planning Can No Longer Happen in Silos
For developers and operators, one of the clearest messages from the framework is that mechanical and electrical planning need to be more closely aligned. In AI data centers, cooling design, power distribution, equipment protection and uptime strategy are connected decisions, not separate workstreams.
A facility may improve cooling performance but still face reliability concerns if electrical systems are not sized, coordinated or protected for the operating profile. At the same time, power upgrades can increase heat loads and shift cooling requirements. The framework encourages teams to evaluate those tradeoffs earlier in the process.
The guidance is not positioned as a universal mandate. Data center operators face different constraints depending on climate, utility capacity, water availability, financing, customer requirements and local infrastructure. A site in a water-stressed region may prioritize water efficiency, while another facility may focus on grid resilience, cooling energy or electrical equipment performance.
That flexibility may be useful in a market where AI hardware and deployment models continue to change quickly. PNNL has framed the framework as a living online resource, meaning it can be updated as technologies and operating practices evolve.
ASHRAE is also planning to continue the discussion at its 2027 Data Center and AI Integration Conference, scheduled for March 3-5, 2027, in Dallas. The event is expected to focus on artificial intelligence, infrastructure performance and system integration.
For businesses building, financing or operating data centers, the framework points to where the market is moving. AI is not only changing compute demand, it is also changing how facilities are designed, powered, cooled and maintained. Operators that address those building system challenges early may be better prepared for higher densities, tighter energy constraints and more complex reliability requirements.
