Author: Corey Olson, Senior Principal, Systems Engineering, Motivair by Schneider Electric
A data center design that may not be commissioned for another three to five years creates a difficult target for teams specifying their infrastructure. The cooling system must support tomorrow’s compute densities while meeting current and evolving efficiency requirements.
NVIDIA’s Vera Rubin can support AI workloads reaching up to 227 kW per rack. At densities like these, facility water temperatures, flow rates, economization, heat rejection, controls, and the interaction between liquid- and air-cooled equipment all become part of the thermal architecture.
At the same time, data center cooling efficiency standards are becoming more specific about how facilities use energy and water. California is developing its 2028 Title 24 requirements, including proposed provisions specifically for liquid-cooled IT. At the federal level, H.R. 9372 points toward greater standardization of how data center energy and water performance is measured.
Cooling compliance cannot be left until the final stages of design. The thermal architecture specified now needs to anticipate the efficiency environment the facility is likely to operate under years from now.
Data center cooling efficiency standards are moving toward system-level performance
Efficiency codes have traditionally shaped familiar aspects of data center mechanical design, from economization and fan power to controls and monitoring. However, the level of attention paid to the performance of the cooling system as an integrated system, including liquid-cooled infrastructure, is changing.
California provides one of the clearest indications of this direction.
California’s Title 24 trajectory signals what’s next for liquid cooling design
California updates its Building Energy Efficiency Standards on a three-year cycle. The 2025 standards took effect January 1, 2026, while development of the 2028 standards is already underway. Those standards are scheduled to be adopted in 2027 and take effect January 1, 2029.
For a data center being designed now, 2029 is not a distant planning horizon. It may be the year the facility begins operating.
The Statewide Codes and Standards Enhancement (CASE) initiative’s working draft for the 2028 cycle is particularly relevant to design teams because it proposes requirements specifically addressing liquid-cooled IT. For facilities with more than 2 MW of liquid-cooled IT load, the draft proposes new economizer thresholds at 75°F wet bulb and 75°F dry bulb. It also proposes minimum water and air temperature differentials and variable-flow requirements intended to support economizer performance under actual operating conditions.
While these data center cooling efficiency standards are proposals, not current requirements, they suggest that specifying liquid cooling will increasingly require design teams to think beyond whether a Coolant Distribution Unit (CDU), cold plate, or other component meets an individual performance requirement. Facility water temperatures, flow control, economizer operation, and heat rejection need to work together to deliver efficient performance across changing IT loads and environmental conditions.
Federal policy also puts more emphasis on how performance is measured
H.R. 9372, the Data Infrastructure Energy Measurement and Standards Act, would direct the National Institute of Standards and Technology (NIST), in consultation with the Department of Energy, to develop standardized definitions, methodologies, and best practices for measuring data center energy and water use.
For specifiers, the move toward standardized measurement is significant because the methodology accounts for differences in IT, power-chain, and cooling configurations, as well as climate-driven variation. That means considering from the outset how temperatures, flow, cooling energy, water use, controls, and part-load performance will be measured and validated. Those decisions can be much harder to retrofit.
Three specification decisions that can reduce future efficiency and compliance risk
No design team can predict exactly how every state or federal data center cooling efficiency standard will develop over the facility’s operating life. Nor should a project be engineered around regulations that remain proposals. Specifiers can, however, make architectural decisions that preserve options. Three areas deserve particular attention.
1. Design for measurable performance
For liquid-cooled infrastructure, specifications can establish requirements to measure and validate factors such as supply and return temperatures, flow, cooling energy, water use, controls behavior, and part-load performance.
The immediate implication of H.R. 9372 is that data center performance is receiving greater scrutiny at the system level. Building measurement into the original design gives operators a stronger foundation to demonstrate and optimize performance later, regardless of how specific reporting or efficiency requirements evolve.
2. Preserve flexibility between liquid and air
Liquid cooling does not mean eliminating air cooling from the facility. Colocation providers in particular may need to support different customer hardware, rack densities, and deployment schedules within the same building. Neocloud and hyperscale environments can also change as successive compute platforms are deployed.
A flexible design can combine direct-to-chip cooling and CDUs for the highest-density racks with rear-door heat exchangers (RDHx) or conventional air cooling where those approaches remain appropriate.
3. Treat heat rejection and water strategy as first-order design decisions
Efficiency discussions can become overly focused on the equipment closest to the IT load, but removing heat from a chip is only one part of removing it from the facility. Chiller configuration, dry or evaporative heat rejection, economization, climate, water availability, facility-water temperatures, and potential heat recovery all affect the performance of the complete cooling system.
Data center cooling efficiency standards are also expanding beyond energy alone. There are emerging water-efficiency requirements and proposals at the state and municipal level, like the Virginia measure addressing cooling technologies in designated water-scarcity areas.
Specifiers should focus on creating a thermal system that can be measured, adjusted, and operated efficiently as rack densities, environmental conditions, customer requirements, and codes change; this adaptability is becoming one of the most important design criteria.
Higher rack densities put the entire thermal architecture to the test
As rack densities increase, cooling performance becomes more dependent on the interfaces between systems. A cold plate may efficiently capture heat, but that heat still has to move through the coolant loop, CDU, and facility-water system before ultimately reaching the heat-rejection equipment. At every stage, temperatures, flow rates, controls, equipment selection, and system design affect what happens next.
Coordination across Direct-to-Chip accessories, CDUs, Heat Dissipation Units (HDUs), rear-door heat exchangers, Chillers and Heat Rejection, and the supporting facility infrastructure is critical. For phased capacity expansion, prefabricated pods can also form part of the broader infrastructure strategy.
Schneider Electric’s comprehensive liquid cooling portfolio provides a natural system-level perspective here. With Motivair by Schneider Electric, the company’s liquid cooling capabilities extend from direct-to-chip thermal management and coolant distribution through facility cooling and heat rejection. This supports the chip-to-chiller approach: considering the complete thermal ecosystem rather than optimizing individual pieces in isolation.
For specifiers, the value of that approach is less about sourcing more equipment from one portfolio and more about reducing the design gaps between systems. The temperatures expected by the IT loop need to work with the CDU. CDU conditions need to align with the facility-water strategy. The facility-water strategy needs to support the intended economizer and heat-rejection performance. Controls and measurement need visibility across those boundaries.
Design for the data center cooling efficiency standards your facility will operate under
California’s proposed 2028 Title 24 provisions bring liquid-cooled IT directly into efficiency codes and put greater emphasis on operating temperatures, flow, economization, and system performance. H.R. 9372 points in a related direction at the federal level by focusing on more consistent measurement of data center energy and water use.
Neither means specifiers should attempt to design against requirements that have not yet been finalized. But designing only to the minimum requirements in force when a project breaks ground carries its own risk when commissioning may be three to five years away.
The more durable approach is to build adaptability into the thermal architecture from the beginning: establish operating temperatures deliberately, preserve economizer opportunities, account for water and heat rejection, design for measurable performance, and create enough flexibility to support changing rack densities and cooling technologies. Schneider Electric’s Liquid Cooling Hub provides more detail on comprehensive direct-to-chip architectures as well as retrofit and greenfield planning guidance. For a closer look at how to select and design the right system for your facility, download the Liquid Cooling System Architecture Guide.
Add a comment