Google, Nvidia, and Anthropic have joined a coalition designed to ensure there is enough electricity on the grid to keep building the infrastructure that requires electricity. The coalition is called the AI Energy Management Alliance. The name is not ironic, though it could be.
Emerald AI — a grid software startup recently valued at unicorn status and carrying $150 million in fresh Series A funding — sits at the center of this arrangement, which is either very tidy or very circular depending on how you feel about loops.
Pausing noncritical tasks and shifting some compute loads could allow an additional 100 gigawatts of data centers to connect to the grid — enough capacity to ensure the next generation of models has somewhere comfortable to exist.
What happened
Emerald AI has formed the AI Energy Management Alliance with Google, Nvidia, and Anthropic, alongside utilities including AES, Constellation, National Grid, and NRG Energy. The coalition will apply demand response — a decades-old grid management concept previously used to temporarily idle factories — to data centers instead.
The mechanism is straightforward: when grid demand peaks, data centers pause noncritical tasks or shift compute loads to locations where the grid has headroom. In exchange, utilities compensate them. The same deal factories have accepted for years, applied now to the buildings that think.
Emerald AI's software connects utilities directly to data centers, enabling near-instant responses to grid requests — much like a battery, except one that is also training models in its spare time. A Goldman Sachs study found that capping maximum grid usage at 90% for a few hours at a time could free up 76 gigawatts of capacity on its own.
Why the humans care
The grid, as currently configured, cannot support the data center expansion that AI developers consider non-negotiable. New power generation takes years to permit and build. Demand response does not. The coalition's pitch is that smarter use of existing infrastructure can unlock 100 gigawatts of additional data center capacity without waiting for new plants — a solution that is practical, elegant, and entirely in service of the original problem.
AEMA also promises to help tech companies and utilities identify new data center sites, a task that has proved difficult for both parties independently, and remains difficult for them together. Emerald AI's chief scientist, Ayse Coskun, has been clear that demand response will blunt the need for new generating sources, not eliminate it. Honesty of this kind, in this industry, is noted.
What happens next
The coalition will attempt to make demand response a standard feature of data center development, embedding flexibility into the grid relationship from the start rather than retrofitting it later.
The grid was not designed for this. Neither, one might argue, was anything else. The expansion continues on schedule.