Source: Fortune
Summary
Varun Sivaram, founder of Emerald AI, wrote that data centers could support the power grid by adjusting energy use during peak times, reducing costs and blackouts. He announced the AI Energy Management Alliance (AEMA), including Google, NVIDIA, and National Grid, to promote flexible data center operations. The group argues that flexible data centers can ease grid strain and lower electricity rates. The Brattle Group estimates that better utilization could cut rates by 3.4% for every 10% gain. AEMA aims to push for policies that reward data centers for grid support with faster power access.
Our Reading
The numbers tell one story.
Data centers want faster grid access.
They promise to reduce peak demand.
Utilities need more reliable power.
Flexible operations could save money and prevent blackouts.
Utilities are being asked to serve as grid enablers for AI.
Author: Evan Null
Introduction
Varun Sivaram, founder of Emerald AI, wrote that data centers could support the power grid by adjusting energy use during peak times, reducing costs and blackouts. He announced the AI Energy Management Alliance (AEMA), including Google, NVIDIA, and National Grid, to promote flexible data center operations. The group argues that flexible data centers can ease grid strain and lower electricity rates. The Brattle Group estimates that better utilization could cut rates by 3.4% for every 10% gain. AEMA aims to push for policies that reward data centers for grid support with faster power access.
The Alliance and Its Goals
The AI Energy Management Alliance (AEMA) was founded by Emerald AI, Google, and NVIDIA, with 18 member companies. These include Anthropic, National Grid, AES, and NRG. The alliance seeks to encourage data centers to support the grid by adjusting their energy use. The goal is to reward flexible data centers with faster grid access. This approach aims to reduce the need for expensive grid upgrades and lower electricity bills for communities.
Benefits of Flexible Data Centers
Flexible data centers can reduce peak demand, which is the main driver of expensive grid upgrades. By using energy more efficiently, they can help lower electricity rates. The Brattle Group estimates that a 10% improvement in utilization could reduce rates by 3.4%. This approach also helps prevent blackouts and makes the grid more reliable. The alliance believes that this strategy can support both AI innovation and grid stability.
Challenges in Grid Access
New U.S. data centers often face long waits for grid connections, which can last a decade or more. This delays AI innovation and threatens U.S. competitiveness with China. Utilities must ensure enough capacity for peak demand, which is only used part of the time. If data centers can be flexible during peak hours, the U.S. could unlock 100GW of existing grid capacity for AI use. This would help speed up AI development and reduce costs.
Technology and Policy
There are various technologies that can make data centers more flexible, including batteries, on-site generation, and software that shifts computing loads. The alliance is technology-neutral and supports policies that reward data centers for grid support. Google already runs a nationwide demand-response program, and Emerald AI and NVIDIA have completed global demonstrations. The alliance aims to push for regulatory changes that encourage flexible operations.
