Emerald AI raises $24.5M to cut data center power use by 25%

Emerald AI has raised $24.5 million in funding from high-profile investors including Nvidia, former climate envoy John Kerry, and Kleiner Perkins chair John Doerr to develop technology that makes data centers more flexible power consumers. The startup’s platform allows grid managers to remotely reduce AI data center power consumption by up to 25% during peak demand periods without affecting AI performance, potentially transforming data centers from grid burdens into “grid allies.”

The big picture: Energy demand has emerged as the primary constraint limiting AI growth, with the Department of Energy predicting that U.S. data center electricity consumption could nearly triple over three years, reaching 12% of total national output by 2028.

Why this matters: The AI boom is driving up greenhouse gas emissions and knocking tech companies off their net-zero climate goals, while communities fear rising electricity rates and grid instability from massive new data centers.

How it works: Emerald’s platform applies demand-side management principles specifically to data centers, giving power system managers unprecedented flexibility during peak usage periods.

• The technology allows grid operators to assume that large data centers won’t demand their full power allocation during critical moments like hot summer days when air conditioning usage peaks.
• During a recent test in Phoenix on a 96-degree day, Emerald reduced power consumption by AI chip clusters by 25% for three hours while maintaining full AI performance.
• The approach eliminates the need for expensive “peaker plants” that only operate during peak demand and are often the most polluting sources in regional power fleets.

In plain English: Think of the electricity grid like a highway system that must be built to handle rush hour traffic. Normally, data centers are like massive trucks that always demand full lanes, even during the busiest times. Emerald’s technology turns these data centers into smart vehicles that can temporarily use less space when the highway gets crowded, without slowing down their actual work.

What they’re saying: Political and business leaders are framing AI energy needs as a national security priority.

• “This is a competition we must win,” said Pennsylvania Republican Senator Dave McCormick at an AI and energy summit at Carnegie Mellon University.
• “Energy is the limiting factor,” explained Jonathan Gray, president and CEO of Blackstone Group, an asset management firm. “Unless we get the energy side right, we can’t do this.”
• “Our goal is to make these data centers flexible in their power consumption,” Emerald AI CEO Varun Sivaram told Newsweek.

The broader impact: Sivaram believes this technology can address three major community concerns about data centers: rising electricity rates, grid crashes, and reliance on dirty diesel generators for backup power.