How Google and Nvidia Want Data Centers to Share Power
Emerald AI, Google, Nvidia and Anthropic have started a new group to make room on the power grid for more data centers TechCrunch. It is called the AI Energy Management Alliance (AEMA). It was announced on Sept. 16, with details reported on Sept. 17.
The members include tech companies and power companies. Google, Nvidia, Anthropic and Emerald AI are members alongside utilities AES, Constellation, National Grid and NRG Energy.
Nvidia shares the launch announcement on its official blog NVIDIA, where the group is called a first-of-its-kind coalition. The launch release is titled "Global Technology Pioneers Emerald AI, Google, and NVIDIA Launch the AI Energy Management Alliance to Advance Flexible AI Data Centers" Business Wire.
The idea centers on software from Emerald AI. The software lets a power company ask a data center to use less power for a short time. The data center can pause less urgent jobs or move some work to another data center where the grid has spare room. Like a neighbor who turns off a dryer during a crunch to keep the lights on for everyone. The stated aim is to make data centers flexible helpers for the power grid to unlock power for AI work.
The group says this approach could allow an extra 100 gigawatts of data centers to connect to the grid. Emerald AI will turn on the world's first power-flexible AI factory at nearly 100 megawatts Fortune.
Emerald AI raised $150 million in a Series A funding round led by Energize Capital and DCVC. That money supports rollout of its software across participating sites.
Other signs point to more building. Nvidia said it expects AI spending to stay strong for years. Under a separate agreement, Anthropic will commit up to 1 gigawatt of computing powered by Nvidia's Grace Blackwell and Vera Rubin hardware, which are Nvidia's current and next AI chips.
The broader context here is timing of power use, not just amount. Grid planners have usually treated big users as always-on, so they save extra space for rare peaks. A data center that can cut or move its use when asked changes that rule.
In my view, the idea makes sense but will be hard to run day to day. Pausing less urgent jobs sounds easy until operators must sort work across training, fine-tuning, batch inference and storage, apply those rules for different customers, and prove to power companies they will respond on time, every time. Moving work also needs spare wires and spare computers free somewhere else at that moment.
Worth flagging here is what success would allow. If it works at nearly 100 megawatts and then spreads, grid operators gain a new way to connect new users without waiting for new power plants and wires everywhere. That does not remove the need for more power. It changes when flexible users can connect. Over time, that kind of coordination is how basic systems usually grow to meet demand.
