Emerald AI to Deploy Conductor at 96-MW Virginia AI Factory as Grid Flexibility Test Goes Live
Updated
Updated · MIT Technology Review · Jun 16
Emerald AI to Deploy Conductor at 96-MW Virginia AI Factory as Grid Flexibility Test Goes Live
1 articles · Updated · MIT Technology Review · Jun 16
Summary
A 96-megawatt AI factory in Manassas, Virginia, is set to become Emerald AI’s first live-grid deployment of Conductor later this year, letting the software cut data-center power use during demand spikes while preserving priority workloads.
The project aims to ease a key bottleneck for new data centers: power access. In PJM, bringing new generation online can take 8 years, while studies suggest flexible loads could tap existing grid capacity much faster.
Emerald’s earlier trials scaled from a 130-kilowatt UK simulation to 256 Nvidia A100 GPUs in Phoenix, where Conductor trimmed chip power 25% for 3 hours while keeping computing performance acceptable.
Northern Virginia is the highest-stakes proving ground: about 500 data centers there account for 13% of global capacity, and Virginia says electricity demand could rise 183% by 2040 if proposed projects proceed.
Supporters including Nvidia and PJM see flexible demand as a near-term bridge for AI growth, while critics say it cannot replace the new generation and transmission still needed to avoid future shortages.
Will making AI data centers 'power-flexible' lower public electricity bills, or just help tech giants expand faster?
With AI's energy needs fueling a fossil fuel surge, can smart software alone truly avert a national power crisis?
Power-Flexible AI at Scale: How Aurora’s 96MW Manassas Data Center Is Shaping the Future of Grid-Responsive AI Infrastructure
Overview
The Aurora AI Factory, which opened in Manassas, Virginia in early 2026, is the world’s first commercial-scale, power-flexible AI data center. As AI workloads grow rapidly, they demand huge amounts of electricity, causing a surge in energy use and putting pressure on local electrical grids, especially in regions like Northern Virginia. This strain leads to bottlenecks for new data centers and raises concerns about energy reliability and costs for communities. Aurora directly addresses these challenges by using advanced power management to adjust its energy use in real time, helping stabilize the grid and setting a new standard for sustainable AI infrastructure.