Emerald AI has formed an alliance with Google, Nvidia, Anthropic, and a number of electric utility companies to accelerate the connection of new data centers to power grids. The alliance, called AI Energy Management Alliance (AEMA), says that demand management could make up to 100 gigawatts of additional grid capacity available to data centers.
The idea is to make data centers more flexible in their electricity consumption. When grid loads rise, noncritical tasks can be temporarily paused or some computing operations can be shifted to other data centers with more available capacity. This means a facility would not have to reduce its operations entirely or run backup generators every time the grid reaches peak demand.
From Factory Management to AI Workloads
Demand management is a practice that has been used for decades, with large industrial companies agreeing with utilities to reduce their consumption during peak periods in exchange for financial compensation. Data centers, meanwhile, have often participated in these programs by running backup generators, a mechanism with environmental impacts, particularly when it relies on diesel.
Emerald AI is attempting to offer a software-based alternative that directly connects utilities’ requests with data center systems. According to the article, the platform can coordinate the shutdown of nonessential tasks or rapidly redistribute computing workloads, making a data center more comparable in its response to a battery that adjusts its consumption when needed.
A Multi-Party Alliance
In addition to Emerald AI, Google, and Nvidia, the alliance includes Anthropic and utilities including AES, Constellation, National Grid, and NRG Energy. One of its goals is also to help technology companies and utilities identify new locations for data centers, as finding sites with sufficient electrical capacity has become an obstacle to expansion.
This approach does not start from scratch; Google is developing its own load-management tools, while Enel X allows data centers to use uninterruptible power supply systems to reduce consumption peaks. A study published by Goldman Sachs last year indicates that limiting peak grid usage to 90% for a few hours could free up 76 gigawatts of capacity.
What Changes in Practice?
The alliance’s importance lies not in adding a new source of electricity, but in turning the flexibility of AI workloads into a resource that grid companies can rely on when planning new connections. Emerald AI’s funding, which reached $150 million in a Series A round led by Energize Capital and DCVC, also gives the company capital to expand its technology’s deployment.
However, the 100-gigawatt figure represents the capacity the alliance hopes to make available, not capacity that has already been definitively provided. Ayse Coskun, Emerald AI’s chief scientist, acknowledges that the technology could reduce the sector’s need for new generation sources, but it will not eliminate that need. The approach’s effectiveness also remains tied to data centers’ ability to identify tasks that can be deferred or shifted, and to the willingness of grid operators to coordinate these responses quickly and safely.