Google, Nvidia and Emerald AI announce alliance for flexible AI data-center power management

Google, Nvidia and Emerald AI have formed the AI Energy Management Alliance (AEMA) to advance flexible load-management capabilities for AI data centers, including demand response.

The alliance is intended to support digital infrastructure that can contribute to power-system efficiency, affordability and reliability. Its founding partners include Anthropic, AES, Constellation, National Grid, NRG Energy, RWE, Fluence and Voltus, alongside other companies from the data-center, energy, software and storage sectors. Tyler Norris, Google’s lead for energy-market innovation for AI and infrastructure, has been named AEMA’s first chair.

For data-center operators and power providers, the initiative centres on whether portions of AI compute demand can be shifted, reduced or otherwise managed during periods of system stress. Such flexibility could position large AI loads as a grid resource, coordinated with energy storage, local generation and contractual demand-response mechanisms.

Google has also published supporting work from The Brattle Group on practical implementation of FERC approaches, and from Aurora Energy Research on flexible data centers and onsite resources in ERCOT. The material cited by the source notes that traditional grid upgrades can take five to seven years or longer, underscoring the potential value of demand-side measures where network capacity is constrained.

Separately, Google said in March 2026 that it had incorporated an aggregate 1 GW of demand-response capacity into long-term power contracts with several US utilities. This is announced contractual capacity, not evidence of commissioned or delivered demand-response capability.

The source also cites an assessment that reducing electricity consumption for fewer than 100 hours annually could potentially release tens of gigawatts of new capacity through flexible computing, storage or generation. That estimate should be treated as indicative: the underlying methodology, geographic scope and precise capacity figure were not disclosed.

AEMA could become a forum for developing operating and commercial practices under which AI data centers connect to grids with demand-management commitments. This could increase the importance of load-management software, energy storage and demand-response contracting in markets facing constrained grid capacity, although the alliance has not disclosed participant obligations, target timelines, funding, commercial arrangements or the volume of load it expects to manage.