Pomegra
Emerald AI raised $150M at a $1.05B valuation on Aug. 25, co-led by Energize Capital and DCVC, selling software that makes AI data centers shed load on command as a power-grid flexibility tool.
Key Takeaways
- Energize Capital and DCVC co-led the oversubscribed round; Nvidia, Samsung Ventures, GE Vernova, and Salesforce Ventures also participated.
- Emerald's platform cut data center power draw by 25% on demand and held that reduction for three consecutive hours in a 2025 EPRI pilot with Nvidia and Oracle.
- Total lifetime funding now exceeds $220M; 12 Fortune Global 500 companies are among the company's investors.
Lead
Emerald AI closed a $150 million Series A on August 25, reaching a $1.05 billion valuation and unicorn status in the same announcement. The round was oversubscribed and co-led by Energize Capital and DCVC, with Nvidia, Samsung Ventures, GE Vernova, and Salesforce Ventures also joining. The Manassas, Virginia-based company sells software it says can convert any AI data center into a flexible grid asset - one that reduces power consumption on command - and argues the approach could unlock more than 100 gigawatts of latent US grid capacity without laying new transmission lines.
What Does Emerald AI Actually Do?
The company's core product, the Emerald Conductor platform, coordinates the power draw of AI computing clusters in real time, enabling a data center to ramp down consumption rapidly when a utility or grid operator signals the need. The approach does not shut off servers outright; it throttles workloads that are not latency-sensitive while protecting jobs that are. In a June 2025 pilot run under EPRI's DCFlex initiative alongside Nvidia, Oracle Cloud Infrastructure, Salt River Project, and Arizona Public Service, a cluster running the Conductor cut energy consumption by 25% and sustained that reduction for three consecutive hours.
Emerald AI says it has now completed five commercial demonstrations globally and has moved into full commercial deployment at multi-megawatt, data center scale. Its most prominent current project is the Vera Rubin AI Research Factory in Manassas - nearly 100 megawatts built with Digital Realty and Nvidia, tested with EPRI, Dominion Energy, and PJM Interconnection, and expected to come online later in 2026.
Why Are Grid Operators and AI Firms Both at the Table?
The data center buildout driven by AI demand has run into a US power grid with little room left in many regions. New transmission capacity takes years and billions of dollars; new generation faces similar delays. Flexible load - the ability to dial demand up or down quickly - is among the few near-term tools grid operators can use to maintain reliability as power-hungry AI infrastructure comes online at scale. Emerald AI's pitch is that its software makes data centers behave more like the industrial interruptible-load programs utilities have used for decades, but at the speed and granularity that GPU clusters require.
For AI operators, the economics matter separately. Data centers that can offer grid services may capture revenue from demand-response programs, potentially offsetting power costs, and may face fewer constraints when pursuing new interconnection agreements - which have become a serious bottleneck for new capacity across multiple US regions.
Strategic Investor Mix
The participation of Nvidia, GE Vernova, and 12 undisclosed Fortune Global 500 companies on the cap table spans three distinct constituencies: AI compute hardware, power generation equipment, and presumably large data center operators or utilities. Strategic investors at this stage typically reflect commercial relationships already underway, not speculative alignment. Total funding exceeds $220 million, which means Emerald AI had already deployed significant capital before reaching commercial scale - a figure worth noting as the company enters what it calls its scaling phase.
The $1.05 billion valuation implies a meaningful step up from its prior financing, though the prior round's terms were not disclosed. At that valuation, the round implies roughly a 7x revenue multiple if the company is near $150 million in contracted or recurring value - a reasonable assumption given the strategic investor base, though the company has not disclosed revenue figures.
Outlook
Emerald AI enters commercialization with credible validation but in a sector where hardware economics, grid policy, and utility contracting cycles all move slowly. The 100-GW market-size claim represents long-run potential, not near-term addressable volume. The more immediate question is whether flexible-load software can sustain pricing power, or whether utilities and large data center operators will eventually build equivalent coordination tools internally. The Vera Rubin factory coming online later this year will be the first high-profile test of the Conductor platform at close to 100 megawatts of real operating load.
