The archive · Climate & Energy · Strategic decision · 2024–2026
Emerald AI's $150M bet: data centers become flexible grid assets
Emerald Conductor software shifts AI data-center power use around grid stress — five live demos done, commercial at full scale, $150M Series A at $1.05B
Emerald AI
What the business is
Emerald AI's Emerald Conductor platform dynamically shifts and reduces AI data-center power consumption in response to grid conditions, so AI facilities connect faster, act as flexible grid assets, and keep promised compute performance.
How it started
Founded in November 2024 in Washington, D.C., by Dr. Varun Sivaram, a Rhodes Scholar and solar physicist who had been chief strategy and innovation officer at Orsted, chief technology officer at ReNew Power, and a Biden-administration clean-energy official. He concluded that power, not chips or capital, is AI's binding constraint, and that software is the fastest way through it.
What happened
With partners including NVIDIA, EPRI, Oracle, Nebius, and National Grid, Emerald completed five demonstrations at commercial data centers. A field test published in Nature Energy cut a 256-GPU cluster's power consumption by 25% over three hours during grid strain in Phoenix without breaking performance promises. The company then deployed across an entire data center in California, launched the Silicon Valley Power Flexible Load Interconnection Program, and began work on the roughly 100-megawatt Vera Rubin AI Research Factory in Manassas, Virginia, with Digital Realty and NVIDIA. An oversubscribed $150M Series A at a $1.05B valuation, co-led by Energize Capital and DCVC, closed in August 2026, bringing total funding above $220M.
How it ended up
Commercial scaling: deployments at full data-center scale are live in California, with expansion worldwide planned for later 2026; the outcome is still unfolding.
Background
Emerald AI, a Washington, D.C.-based startup, announced on August 25, 2026 that it raised $150 million in an oversubscribed Series A at a $1.05 billion valuation, co-led by Energize Capital and DCVC with participation from NVIDIA, Samsung Ventures, Siemens, Aramco Ventures, Salesforce Ventures, GE Vernova, RWE, and others. Total funding passed $220 million, and 12 Fortune Global 500 companies are now investors on the company's Strategic Advisory Board.
The company's bet is that AI data centers can be turned from inflexible power consumers into responsive grid assets. Its Emerald Conductor software orchestrates AI compute workloads and on-site energy resources to cut a facility's power draw when the grid is stressed, while protecting critical workloads. In a field test published in Nature Energy, the software cut a 256-GPU commercial cluster's consumption by 25% over three hours during grid strain in Phoenix without compromising performance; Emerald says the approach can unlock 100+ GW of untapped US grid capacity.
Founder and CEO Dr. Varun Sivaram — formerly chief strategy and innovation officer at Orsted, CTO of ReNew Power, and a Biden-administration clean-energy official — founded the company in November 2024 on the conviction that power, not chips or capital, is AI's binding constraint. Emerald completed five demonstrations at commercial data centers in Arizona, Illinois, Virginia, Oregon, and London with NVIDIA, EPRI, Oracle, Nebius, and National Grid, then moved to commercial deployment across a full California data center.
Emerald also launched what it calls the first-in-the-nation Flexible Load Interconnection Program with Silicon Valley Power — expanded grid access in exchange for verified, dispatchable flexibility — and is working with Digital Realty and NVIDIA on the roughly 100-megawatt Vera Rubin AI Research Factory in Manassas, Virginia. It was named a 2026 TIME100 Most Influential Company and a World Economic Forum 2026 Technology Pioneer. The round's thesis: software, not new plants, is the fastest way through the AI power crunch.
What has to be true
- A $150M Series A at $1.05B roughly 21 months after founding, with 12 Fortune Global 500 investors, shows demand-side grid software became investable.
- The Nature Energy field test — 25% power reduction over three hours on a 256-GPU cluster without breaking performance promises — is a rare public proof point for AI demand flexibility.
- Working with utilities and grid operators (Silicon Valley Power, EPRI, Dominion, PJM) turned a technology demo into interconnection policy.
- Betting on demand-side software rather than supply builds like nuclear or batteries differentiates Emerald in a crowded AI-power market.
What can be applied
When the constraint is a decade-long build like grid power, the wedge is demand-side software: make customer load flexible instead of waiting for new supply, and investors pay for it.
Aftermath
As of September 2, 2026, Emerald AI says its software is deployed commercially at an entire California data center, where it sustained grid-responsive flexibility during peak strain, and additional large-scale deployments are planned later this year. The company serves AI firms, data center operators, and electric utilities; its next milestone is bringing online the roughly 100-megawatt Vera Rubin AI Research Factory in Manassas, Virginia, with Digital Realty and NVIDIA. Series A proceeds are earmarked for worldwide expansion.
Sources
- Emerald AI Raises $150 Million Series A at $1.05 Billion Valuation to Scale Power-Flexible AI Data Centers
- Emerald AI raises $150M at $1.05B to turn AI data centres into grid allies
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card