Emerald AI Raises $150M at $1.05B to Make Data Centers Flexible Grid Loads

Emerald AI has raised $150 million in a Series A round valuing the data-center power startup at $1.05 billion. It plans to expand global commercial deployments of software that cuts or shifts a facility’s electricity use during periods of grid stress.

Energize Capital and DCVC co-led the oversubscribed round. Emerald AI launched publicly in July 2025 and says investors have put more than $220 million into the company to date.

The product changes when computing draws power

Emerald Conductor schedules AI workloads against onsite batteries and generation. When the grid is stressed, the software reduces the data center’s draw; Emerald says critical training and inference jobs can keep running through that adjustment.

The approach targets a timing problem. New transmission can take a decade of permitting and construction, while the International Energy Agency expects data centers to drive nearly half of U.S. electricity-demand growth through 2030. Emerald estimates flexible operation could unlock more than 100 gigawatts on the existing U.S. system, but that is a company estimate rather than a demonstrated nationwide result.

Demonstrations are becoming operating programs

Nvidia, Oracle and National Grid were among the partners in Emerald’s five live commercial-load demonstrations. Emerald says the technology now runs commercially at full data-center scale, a company-reported milestone that the new funding is intended to extend.

Two disclosed routes to commercial use

• Emerald said a California data center flexed its entire load during a period of peak demand.
• Silicon Valley Power created a Flexible Load Interconnection Program with Emerald, offering faster and larger interconnection to data centers that commit to verified, dispatchable flexibility.

The upcoming test is reliability, not just flexibility

The nearly 100-megawatt Manassas project is a concrete upcoming test of the model. Dominion Energy, PJM Interconnection and the Electric Power Research Institute are helping run tests there. The central unresolved question is whether data centers can repeatedly deliver promised reductions when utilities need them while preserving priority computing workloads.