• Google and Nvidia are just two of the tech giants behind the new AI Energy Management Alliance
• The group is focused on building future data centers that coordinate power usage with the grid
• Moving training to off-peak hours and balancing on-site energy are some of the solutions

Emerald AI, Google and Nvidia have come together to form the AI Energy Management Alliance (AEMA), and it could be exactly what we need to tackle rising energy consumption within the data center field.

Together, the group hopes to deploy flexible AI data centers that can dynamically adjust how much electricity they consume from the grid based on a number of different conditions.

This important development comes in response to rising power availability pressures, which are leading to years-long delays for many new projects.

Future data centers could adapt how much power they use

In an announcement, Nvidia Head of Sustainability Josh Parker explained how these future campuses could shift compute workloads, discharge storage, use paired generation or respond to other system contingencies to dynamically change how much electricity they’re drawing from the grid at any given time.

The concept involves shifting high-compute tasks like training to off-peak times, using on-site batteries or generation in conjunction with the grid and exporting excess stored energy back to the grid during times of stress.

Crucially, the AEMA must now set out technical specifications, performance metrics and open standards to allow this to happen at scale.

And while only Emerald AI, Google and Nvidia are behind the alliance from launch, a number of additional partners are set to join the scheme. SiliconANGLE reports that Anthropic will be one of them.

The concept itself isn’t new – Google also announced earlier this year that it had over 1GW of demand-response capacity – but doing so at scale would likely make a considerable impact across the board.

Both Nvidia and Emerald AI have already developed software to handle power consumption in relation to the grid.

“By creating a common framework for performance, reliability and collaboration, AEMA aims to help the US build the infrastructure of intelligence at the speed and sustainability the moment demands,” Parker wrote.

All of this, of course, comes against a backdrop of mounting energy pressures. Recent Bloomberg data suggests that US data centers alone could use as much natural gas as the entire of Germany and Japan combined by 2035, with a similar Moody’s study projecting that total US data center energy consumption could top 426 TWh by 2030.