Emerald AI Raises $150M for Grid Software

Emerald's core product, Emerald Conductor, coordinates AI computing workloads with onsite batteries and generation to modify a facility's grid power draw during periods of stress, according to SiliconANGLE. FinSMEs describes the platform as turning data centers into flexible grid assets, serving AI companies, data center operators, and electric utilities.

SiliconANGLE reports that Emerald spent the previous year conducting five demonstrations on live commercial loads in Arizona, Illinois, Virginia, Oregon, and London, with partners including Nvidia, Oracle, National Grid, regional utilities, and grid operators. Sivaram told SiliconANGLE that the demonstrations showed data centers can "adjust their power use precisely when the grid needs relief, without compromising critical computing workloads."

The New York Times reports that Silicon Valley Power, the municipal utility serving Santa Clara, California, recently engaged Emerald to help data centers vary their power consumption in response to utility input. Nicolas Procos, Silicon Valley Power's utility director, told the Times that its previously spare capacity had been "spoken for" amid demand growth, leaving system expansion and flexible-load approaches as the available options.

According to SiliconANGLE, Emerald has conducted a commercial deployment in California in which a data center flexed its full load during peak demand. The outlet also reports that Silicon Valley Power created a Flexible Load Interconnection Program with Emerald, under which data centers committing verified dispatchable flexibility can receive faster and larger interconnection service.

SiliconANGLE further reports that Emerald is working with Digital Realty and Nvidia on tests at the nearly 100-megawatt Vera Rubin AI Research Factory in Manassas, Virginia. Dominion Energy, PJM Interconnection, and the Electric Power Research Institute are participating in those tests, which the outlet reports are tied to a facility due online later this year.

FinSMEs reports that Emerald intends to use the financing to expand operations and development efforts, while SiliconANGLE reports that it will support commercial deployments globally.

For infrastructure teams, the reported work centers on a practical constraint in AI capacity planning: grid interconnection and transmission construction can take substantially longer than new compute deployment. Companies pursuing comparable flexible-load arrangements typically need workload schedulers that can distinguish delay-tolerant training or batch jobs from latency-sensitive inference, alongside telemetry and controls that can demonstrate a committed response to utilities.