Emerald AI raised $150 million in an oversubscribed Series A at a $1.05 billion valuation, the Washington, D.C. company said on August 25, 2026. Energize Capital and DCVC co-led the round. Founder and CEO Dr. Varun Sivaram said the company will use the money to take commercial deployments worldwide with AI labs, data-center operators, and utilities.

Emerald Conductor is the product. The software is meant to shift AI workloads and onsite energy so a site can drop its draw when the grid is tight without stopping the jobs the operator cares about most. Emerald says that approach, applied across the U.S. build-out, could unlock more than 100 gigawatts on the existing grid, capacity that would otherwise wait on new transmission. The company cites the International Energy Agency’s view that data centers could account for nearly half of U.S. electricity-demand growth through 2030.

From five demos to a commercial site

Over the past year Emerald completed demonstrations at commercial data centers in Arizona, Illinois, Virginia, Oregon, and London. Partners named in the announcement include NVIDIA, the Electric Power Research Institute, Oracle, Nebius, and National Grid, plus regional utilities. The company says it is now running at a full data center in California that flexed during peak strain.

In Santa Clara, Emerald partnered with Silicon Valley Power on a Flexible Load Interconnection Program that trades verified, dispatchable flexibility for larger grid access. In Manassas, Virginia, it is working with Digital Realty and NVIDIA on what it calls the first power-flexible AI factory, a nearly 100-megawatt Vera Rubin research site tested with EPRI, Dominion, and PJM Interconnection and due later this year. Emerald says 12 Fortune Global 500 companies are now investors and sit on a strategic advisory board. The published participant list also includes Samsung Ventures, Siemens, Aramco Ventures, Salesforce Ventures, GE Vernova, RWE, In-Q-Tel, and others.

“We founded Emerald AI on the conviction that the intelligence driving the AI revolution could solve its own greatest bottleneck: power.” Dr. Varun Sivaram, founder and CEO, Emerald AI

Decoded Take

Chip supply is no longer the only gate on new AI campuses. Interconnection queues and community power bills are. Emerald is selling a software answer to a permitting problem: prove you can shed load, and the utility may let you plug in sooner. That only works if operators will pause or move training jobs when the grid asks, and if utilities treat those promises as real capacity. The California commercial site and the Manassas factory are the tests. Watch whether those facilities publish verified megawatts of flex during peak hours, and whether other utilities copy the Santa Clara interconnection bargain. If they do not, this stays a well-funded demonstration with a unicorn price.