Emerald AI Conductor

Emerald AI is a US company with offices in Washington DC, Boston and San Francisco, founded and led by CEO Varun Sivaram. It makes the power demand of AI data centers flexible so that they can act as grid partners, which the company says could unlock up to 100 GW of capacity on existing grids while protecting reliability and affordability.

Its Conductor platform connects grid operators and data centers, orchestrating compute workloads and onsite energy. Temporal flexibility briefly slows or pauses batchable AI jobs during grid stress within SLA guardrails, spatial flexibility shifts workloads across regions, and resource flexibility dispatches batteries and onsite generation. Conductor optimises across data center fleets, simulates AI power demand with a digital twin, and produces auditable measurement and verification data. Partners include NVIDIA, Oracle Cloud Infrastructure, National Grid, Digital Realty and Salt River Project.

• Temporal flexibility: slows or pauses batchable AI workloads during grid stress within SLA limits.
• Spatial flexibility: shifts workloads across regions to where power is available.
• Resource flexibility: dispatches batteries and onsite generation with compute orchestration.
• Fleet optimisation: coordinates flexibility across multiple data center sites.
• Digital twin: simulates AI power demand for planning and interconnection studies.
• Measurement and verification: auditable performance data and automated reporting.

Data center developers in Vietnam could propose flexible load commitments to secure earlier grid connection. Cloud and AI service providers could shift training jobs away from evening peaks to cut costs and support grid stability. Power corporations could pilot demand response contracts with large data centers.

Flexible data centers can connect to the grid faster because they reduce draw during the few stress hours that limit capacity. In a 2025 Arizona trial an Oracle data center cut power use by 25% over three hours during peak demand, and a UK demonstration with National Grid achieved 100% compliance across 200 power targets. Auditable measurement and verification supports utility programmes and reporting.

AI data center demand is growing faster than grid capacity, and long interconnection queues are pushing utilities toward flexible connection agreements. Vietnam and Southeast Asia are attracting new hyperscale and AI data center investment while grids in Hanoi, Ho Chi Minh City, Johor and Batam face constraints, making load flexibility a practical tool for faster connection.

Emerald AI raised a 150 million USD Series A at a 1.05 billion USD valuation in August 2026. Investors include Energy Impact Partners, Lowercarbon Capital, NVIDIA, National Grid, GE Vernova, Siemens, Samsung Ventures, JERA Ventures and Aramco Ventures. It was named a 2026 World Economic Forum Technology Pioneer, and its first commercial flexible AI factory is a 96 MW facility in Manassas, Virginia.