Emerald AI tests Conductor software for flexible data centers in Virginia

Emerald AI, a Virginia-based startup backed by NVIDIA, has developed Conductor software designed to reduce data center power consumption by 25-40% during peak grid stress without compromising AI performance. The system dynamically adjusts workloads and energy usage in real time, responding to grid signals to avoid overloading the electrical network. Testing in May 2025 showed Conductor cutting power use by 25% over three hours while running 256 NVIDIA GPUs during peak demand in Phoenix. A UK demonstration achieved over one-third power reduction in under 60 seconds, aligning with sudden demand spikes like those caused by mass kettle use during soccer matches. Additional pilots with Silicon Valley Power in Santa Clara further validate the technology’s adaptability. Northern Virginia, home to the world’s densest concentration of data centers, faces grid constraints that delay new projects. Emerald AI’s 96 MW Aurora AI Factory in Manassas, built with NVIDIA and Digital Realty, is designed for real-time power modulation and aims to begin commercial operations in early 2026. The company raised $24.5 million in seed funding in July 2025, with investors including NVIDIA’s NVentures, Jeff Dean, and Fei-Fei Li. Founded in 2024 by CEO Varun Sivaram, Emerald AI is headquartered in Clarendon, Virginia. While not directly involved in crypto, Emerald AI’s technology could influence energy-intensive sectors like Bitcoin mining by automating demand-response strategies. This approach contrasts with traditional mining operations, offering a scalable solution for grid-friendly energy management.