Nvidia-backed AI orchestration platform Emerald AI has released the results of its UK demonstration project in partnership with National Grid.

The project saw Emerald AI utilize its Emerald Conductor platform to manage a cluster of 96 Nvidia Blackwell Ultra GPUs at a Nebius data center in London to validate its ability to dynamically adjust power consumption in response to real-time signals within AI data centers, without disrupting workloads.

The project took place over five days in December of last year. It simulated more than 200 real-time “grid events” to test the Conductor's ability to dynamically adjust the data center's power consumption. According to the company, the platform successfully adjusted power usage to the requested level, cutting power demand by up to 40 percent without impacting critical workloads.

“With our partners, we’ve shown they can be connected and managed without major new network capacity, flexing their power up or down in real time to support the whole system. This approach will enable us to connect significant new demand more quickly and help to lower network charges for customers over time,” said Steve Smith, president of National Grid.

The partners claim that the platform was able to successfully react to spikes in demand during halftime of major football matches, successfully followed load reduction requests for up to 10 hours, and was able to successfully cope with a simulated system stress event, shedding 30 percent of its load in approximately 30 seconds to ensure grid resilience.

“This trial demonstrates that AI infrastructure can be a dynamic force for the grid. With dozens of realistic AI workloads running simultaneously, we delivered fast emergency curtailment and sustained, precise peak reduction. The same approach we validated here can be applied to much larger AI factories, as the industry scales,” said Dr. Varun Sivaram, founder and CEO of Emerald AI.

The trial was the third completed by Emerald AI following similar demonstration projects in Phoenix, Arizona, in partnership with EPRI’s DCFlex Initiative and Oracle, and Chicago, Illinois. In Phoenix, the company partnered with Oracle on a proof-of-concept demonstration. During the test, the conductor was able to modulate real AI workloads, achieving 25 percent power reduction over three hours while maintaining workload performance.

In Chicago, the company raised the stakes. Unlike the Phoenix demonstration, where the team was aware of the workload profiles before the demo, in Chicago, the conductor handled unknown, random workloads. This demonstration presented a significantly more challenging environment and, according to Sivaram, the platform proved resilient.

According to Emerald AI, the latest trial in London is expected to serve as an operational blueprint for its first commercial project in Virginia. The company announced plans for the project in October, with the orchestration software set to be deployed at Nvidia’s under-construction 96MW Aurora data center in Manassas, Virginia.

Emerald AI originally launched last year following a $25.5 million seed funding round. Since then, it has raised around $67.5m, following a further seed extension round last month.