The AI Energy Management Alliance launched Wednesday with an ambitious proposition: the power-hungry data center need not be a grid burden. Google, Nvidia and Emerald AI, joined by Anthropic, utilities and power producers, want regulators to reward facilities that can temporarily cut electricity use when supplies tighten.¹

The mechanics are familiar to utilities but newly tailored to AI. Instead of running at full tilt around the clock, data centers could pause noncritical computing jobs, shift workloads to sites with spare capacity, draw on batteries or use nearby generation. In return for reliable demand reductions, they could receive quicker grid connections and potentially avoid some expensive upgrades. Google says it has committed 1 gigawatt of demand reduction through utility agreements nationwide.¹

Emerald AI’s software is central to that promise, linking utility requests directly to data centers so they can respond quickly, much like batteries. The alliance says a broader use of load shifting and paused computing could open space for as much as 100 gigawatts of additional data-center capacity. It is building on decades-old demand-response programs, which traditionally paid large industrial users to curb consumption during peaks.²

But the coalition arrives after a sharp turn in public sentiment. Eighty-four percent of Americans say they are concerned about data centers’ effect on local power prices, while strong majorities in both parties support making developers pay for the grid upgrades they require.¹ Nvidia sustainability chief Josh Parker called the plan a “win-win-win solution,”¹ but its credibility rests on whether operators actually deliver when the grid is stressed.

That is the key fault line. Emerald AI chief executive Varun Sivaram says flexibility must be “verifiable and enforceable” before data centers earn faster or larger connections.¹ Even advocates acknowledge the limits: smarter demand response may reduce the need for new generation, but it “won’t eliminate it entirely,” according to Emerald AI chief scientist Ayse Coskun.² Regulators now face the test of whether to create the fast lane—and communities the harder question of whether it can be trusted.