Emerald AI makes AI’s power demand flexible, transforming data centers into valuable grid partners, protecting energy affordability and reliability for local communities, and unlocking massive capacity on existing grid infrastructure to power the AI revolution today.
Emerald AI is tackling AI's power crunch by transforming AI data centers into flexible grid assets that dynamically adjust power consumption in response to power grid conditions, an approach that can unlock up to 100 gigawatts of capacity on the existing U.S. grid for AI while protecting grid reliability and energy affordability for communities
Emerald AI is among a global cohort of companies tackling the world's most pressing challenges. For us, it's a clear signal that power flexibility, not simply more power, is becoming central to scaling AI innovation all while building a reliable grid.
Silicon Valley Power and Emerald AI today announced a pilot program to demonstrate how flexible data centers can help support grid reliability, affordability, and more efficient system planning as advanced digital infrastructure continues to grow in Silicon Valley.
Participating power companies include AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, and Vistra
1
2
3
AI’s explosive growth is colliding with grid limits, delaying AI infrastructure by years and raising energy costs for everyone.
Intelligent flexibility unlocks 100GW of capacity for AI data centers — speeding innovation, lowering rates, and avoiding blackouts.
Emerald AI enables data centers to provide relief when the grid is stressed while assuring critical AI performance.
/ 3
Emerald AI's customers span the largest AI firms, global data center operators, and leading power utilities, and we partner with companies across the AI and energy value chain defining the intelligence era.
“AI factories are the engines of the intelligence era, and like any great engine, every system must be designed together — energy, compute, networking and cooling as one architecture. NVIDIA and Emerald AI are working together to enable a future for AI where performance, efficiency and grid responsiveness can be tapped into immediately.”
Jensen Huang
Founder and CEO, NVIDIA
“NextEra looks forward to working with NVIDIA and Emerald AI to help design efficient energy campuses and flexible AI factories that economically support rising demand while further strengthening America’s energy infrastructure.”
John Ketchum
Chairman, President and CEO, NextEra Energy
“Compute infrastructure that is responsive to real-time grid conditions while meeting the performance demands unlocks a new model for scaling AI.”
Pradeep Vincent
Chief Technical Architect and Senior Vice President, Oracle Cloud Infrastructure
“Emerald AI demonstrates how innovative technologies can help us optimise the grid, enable increased investment in advanced computing, and deliver real benefits to the wider UK economy.”
Steve Smith
Chief Strategy and Regulation Officer, National Grid UK
“Emerald AI’s innovative solutions provide a valuable opportunity to explore how flexible power management can support both our customers’ growth and the broader energy ecosystem.”
Chris Sharp
Chief Technology Officer, Digital Realty
AI flexibility can unlock 100GW to power the AI revolution today
Briefly slow or pause batchable AI workloads when the grid is stressed and resume post‑event, all within SLA guardrails.
Shift workloads across regions on fiber to where power is abundant — within latency bounds.
Dispatch batteries and other onsite energy alongside compute orchestration for maximum, reliable flexibility.
Briefly slow or pause batchable AI workloads when the grid is stressed and resume post‑event, all within SLA guardrails.
Shift workloads across regions on fiber to where power is abundant — within latency bounds.
Dispatch batteries and other onsite energy alongside compute orchestration for maximum, reliable flexibility.
Emerald Conductor is the flexibility management platform for AI infrastructure. We connect utilities and data centers in real time, simulate a digital twin model of AI power, orchestrate compute alongside onsite energy, and verify performance—so AI can grow in harmony with stronger, more affordable grids.
Provide operators with AI demand visibility and enable data centers to respond to grid signals.
Flexibly control power use, meeting grid needs while preserving AI performance.
Orchestrate spatiotemporal workload flexibility to support grids nationwide.
Auditable M&V data and automated reporting, establishing data centers as credible grid resources.
Emerald AI has deployed power flexibility at commercial data centers around the world, validating the results with partners, customers, and top scientific journals.
Arizona · 2025
During peak grid demand on a hot Phoenix afternoon, Emerald Conductor flexes down AI power demand at an Oracle data center by 25% over three hours to meet utility needs.
Platform: NVIDIA A100 · Partners: EPRI, Oracle, NVIDIA, SRP, Databricks
United Kingdom · 2025
When football halftime hits, a gigawatt of tea kettles switches on. Emerald Conductor shaved power to ease national grid kettle spikes and met 200 power targets overall with 100% compliance.
Platform: NVIDIA B300 · Partners: EPRI, Nebius, NVIDIA, National Grid
Virginia ↔ Chicago · 2026
During Virginia's winter peak, Emerald geo-shifted inference traffic within milliseconds from a constrained Virginia cluster to a Chicago cluster with available capacity — ensuring inference was served within latency constraints.
Platform: NVIDIA H100 · Partners: EPRI, Oracle, NVIDIA, Commonwealth Edison, PJM
Oregon · 2026
Deploying on NVIDIA Grace Blackwell Ultra clusters, Emerald Conductor shaved power against volatile market pricing signals and sustained multi-day load reductions through heat dome and freezing rain stress events.
Platform: NVIDIA GB300 · Partners: EPRI, Portland General Electric, NVIDIA
California · 2026
Emerald's first multi-MW, full data center deployment cut power use during peak California CAISO demand on a hot summer day, reducing load after sunset when solar generation goes dark.
Platform: NVIDIA H100
Virginia · 2026
Emerald AI partners with NVIDIA and Digital Realty to launch a 96MW Vera Rubin AI Factory in Manassas, Virginia, that sets a new reference architecture for power-flexible AI infrastructure built on the NVIDIA DSX Platform.
Platform: NVIDIA Vera Rubin · Partners: NVIDIA, Digital Realty, EPRI, PJM, Dominion
01
Get to 'Power-On' in Months, Not Years. Accelerate your interconnection queue, upsize an existing connection, and turn your energy consumption into a monetizable asset. We make your facility a preferred partner for any utility.
02
Integrate the Future, Safely. Transform the largest new source of demand into your most valuable flexible resource. Gain unprecedented visibility, accurate forecasting, and dispatchable load to enhance grid stability and defer costly infrastructure upgrades.
03
Scale Without Constraint. Never let power availability dictate the pace of your innovation. Our platform ensures you have swift access to the energy you need to train and deploy the world's most advanced models, while assuring stringent AI performance standards.
04
Enable Innovation, Ensure Reliability, Protect Affordability. Accelerate the AI economic boom while strengthening grid resilience and protecting ratepayers. Our platform provides a powerful, market-based tool to manage new demand efficiently, achieving policy goals without relying solely on costly and slow-to-build infrastructure.
Publication
Emerald AI, EPRI, National Grid, Nebius, and NVIDIA recently conducted a live, UK-first demonstration at Nebius’ new AI Factory in London. Demonstrating that high-performance AI infrastructure could operate as a power-flexible, grid-responsive asset without disrupting mission-critical workloads.
Publication
Our Arizona demonstration results were peer-reviewed and accepted in one of the world's top scientific journals, demonstrating the pioneering advances of Emerald's Conductor platform to make AI data centers flexible, reliable partners to the grid.
%20Jack%20Megrue.avif){kind=logo}
