Google, Nvidia y Emerald AI they support an alliance that seeks to adapt the electricity consumption of artificial intelligence data centers to network conditions.
An alliance seeks to make data center consumption more flexible
Google, Nvidia and the startup Emerald AI joined a new stage of the AI Energy Management Alliance (AEMA), a coalition that seeks to promote data centers capable of adjusting their electricity demand when the system needs it.
The organization brings together 20 companies and entities from the technology and energy sectors, its members also include Anthropic, National Grid, AES, Constellation, NRG, and RWE.
The initiative builds upon the structure of the Advanced Energy Management Alliance, created in 2014 around companies specializing in demand response. Now, the focus will be on the rapid growth of artificial intelligence infrastructure and the pressure it can exert on electricity grids.
Data centers that adjust their electricity demand
The data centers since they typically maintain a high and constant consumption 24 hours a day, AEMA’s proposal aims to introduce greater flexibility so that some installations can temporarily reduce their load when the network is under pressure.
This adjustment could be achieved by shifting or pausing certain computing loads related to artificial intelligence. It could also be supported by battery storage systems or electricity generation located near the facilities.
Google has already committed 1 gigawatt of electricity demand that could be reduced when necessary through agreements with utility companies in the United States.
With this model, the coalition seeks to enable data centers capable of demonstrating reliable reductions in energy consumption to accelerate their connection to the network and, in some cases, avoid infrastructure expansions.
Flexibility must be verifiable
The success of the model, however, will depend on whether the facilities can actually deliver the promised demand reductions. Varun Sivaram, CEO of Emerald AI, argues that access to faster or higher-capacity connections should be conditional on that flexibility being measurable, verifiable, and enforceable.
The issue becomes more relevant as the growth of artificial intelligence the need for new electrical capacity to power data centers is increasing. At the same time, various communities have expressed concern about the potential impact of these facilities on electricity prices and local infrastructure.
A survey by AP-NORC and the Energy Policy Institute at the University of Chicago indicates that 84% of Americans are concerned about the effect of data centers on local electricity rates.
In addition, 79% of Democrats and 76% of Republicans surveyed support data center developers paying for the network upgrades needed to connect their facilities.
Google is committed to flexible energy management
According to AEMA, reducing consumption during times of grid stress could decrease the need to build some of the additional electrical infrastructure and facilitate the incorporation of new loads.
The approach also seeks to change the relationship between large data centers and the communities where they are located. The proposal is that these facilities should be able to respond to the needs of the electrical grid instead of maintaining a completely rigid demand.
The coalition will take this proposal to federal and state regulators, regional network operators, and utility companies. The regulatory framework will be crucial, in June, US federal regulators asked several regional operators to analyze new alternatives for connecting large consumers capable of flexibly meeting their demand.
The evolution of these standards will determine the extent to which data centers with flexible power consumption can obtain faster connections and become an additional tool for managing the increased demand associated with artificial intelligence.
Source: Axios
Photo: Shutterstock
