US-based firms InfraPartners and Emerald AI are partnering on a new data center design intended to reduce energy costs for operators.

The partnership’s ‘Flex-Ready Data Centers’ aim to make power infrastructure inside such facilities more responsive to fluctuations in power supply across the grid.

“Many technology companies are pledging to pay for new power infrastructure for AI data centers, but building new generation and transmission still takes years of planning and permitting,” Ayse Coskun, Emerald AI’s chief scientist, told DCD.

“By enabling data centers to dynamically adjust their power consumption in response to grid conditions, these facilities can operate as controllable and predictable loads. This flexibility creates the possibility for faster grid interconnection, as operators can account for flexible demand when evaluating new large loads. The result is a mutually beneficial model where AI infrastructure can scale while supporting a more reliable, affordable, and sustainable power system.”

The new design comes amid rising concerns in the US and Europe that the growth of the data center industry is outpacing the capacity of national grids, forcing governments to invest millions in network upgrades and power plants.

InfraPartners and Emerald AI’s new data center design affords individual operators greater flexibility to these needs, and reduces their own power costs, by linking the former’s Building Management System (BMS) with the latter’s Emerald Conductor AI platform.

Data on power and cooling needs, and on the current state of the data center’s storage and energy systems, is fed via the BMS to an optimization engine run by Emerald AI. The facility can then dynamically adjust its load to conditions on the grid in real-time, either increasing power consumption when surplus renewable energy is available or cutting its consumption during more expensive peak hours.

The feasibility of Emerald Conductor has already been demonstrated as part of the Electric Power Research Institute’s DCFlex Initiative in Phoenix, Arizona, wherein Emerald AI successfully proved that the power usage of an AI compute cluster could be reduced by a quarter for three hours during periods of high grid demand.

The firm also participated in a live demonstration of the technology to the UK’s National Grid in late 2025.

InfraPartners has also investigated means by which the sector’s power needs can be adapted, partnering last month with Nvidia Prologis and EPRI to study the feasibility of smaller, containerized data centers located in close proximity to substation sites.