Energy is the most intractable bottleneck to the advancement of AI. Over the next three years, the AI industry seeks to bring nearly 50GW of data centers online in the United States, but S&P Global forecasts that just half of that can be plugged into the power grid.

Emerald AI was founded to help address this critical challenge, and I am thrilled to announce today that we have raised $25 million in a Strategic Expansion Round, bringing our total funding to $68 million in just 16 months since our founding. This round provides us with the capital to scale solutions aimed at addressing this power crunch.

This round was led by Energy Impact Partners (EIP), a financial investor whose unique model is backed by leading electric power utilities from around the world, as well as top industrial firms. EIP was joined by existing Emerald AI financial investors as well as key strategic partners who sit at the epicenter of the AI data center and energy grid ecosystem. The round included participation by Amplo, Eaton, GE Vernova, IQT (In-Q-Tel), Lowercarbon Capital, NVentures (NVIDIA’s venture capital arm), Radical Ventures, Salesforce Ventures, Samsung Ventures, Siemens, and other remarkable investors, giving Emerald an unmatched foundation to deploy our software to transform AI factories into grid allies around the world.

Alongside this funding, we are officially launching the Emerald AI Strategic Advisory Board, a group that includes seven Fortune 500 companies. This board is designed to bridge the AI and energy divide, enabling us to collaborate with partners across the entire value chain to transform data centers from grid liabilities into flexible assets.

AI Factories Belong on the Grid—and Can Strengthen It

As AI companies face decade-long interconnection queues and complex regulatory structures, the industry is understandably seeking to go off-grid—up to and including into space—potentially making AI even more expensive and technically complex to power. The best, most scalable, and most affordable way to maximize AI deployment is to use every last MWh of energy from the terrestrial power grid, alongside other innovative solutions. The key to that is power flexibility.

Emerald AI’s Conductor platform is built to make that possible by orchestrating AI workloads, coordinating them with onsite energy resources, and interfacing with the power system to deliver precise, grid-responsive power flexibility while ensuring AI quality of service. At scale, power-flexible AI factories could unlock up to 100GW of grid capacity on the existing U.S. power system while helping protect energy affordability and reliability for local communities.

This month, we announced two major milestones that show how quickly this vision is moving from concept to deployment.

• At NVIDIA GTC, we announced Emerald AI’s integration with NVIDIA DSX Flex to help turn AI factories into power-flexible grid assets.
  

• And at CERAWeek in Houston, NVIDIA and Emerald AI announced a collaboration with leading energy producers AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, and Vistra to pioneer a new class of flexible, hybrid AI factories built on the NVIDIA Vera Rubin DSX AI Factory reference design with DSX Flex—using Emerald Conductor to orchestrate onsite energy resources alongside computational flexibility so projects can connect faster and support the power grid and local communities.
  

Our announcements at the back-to-back Super Bowls of AI and energy were fitting—Emerald AI is positioned at the intersection of those two foundational sectors. And we were delighted to see front-page coverage on the Wall Street Journal of our collaborations.

Launching the Emerald AI Strategic Advisory Board

To realize our vision of integrating the fast-moving tech sector with the highly regulated energy industry requires convening the vanguard of both ecosystems. That is precisely what Emerald AI is doing today through both technical solutions and by helping forge the coalitions we need.

The Strategic Advisory Board will unite existing investors—such as NVIDIA, Salesforce Ventures, and National Grid—alongside a newly expanded group of strategic partners who have invested in Emerald AI:

• Eaton
  
• GE Vernova
  
• IQT (the not-for-profit strategic investor for the U.S. national security community and America’s allies)
  
• Samsung Ventures
  
• Siemens
  

AI must evolve from being the grid’s fastest-growing power consumer into one of its strongest allies. If AI factories can be power-flexible, ramping down consumption during rare periods of peak grid strain, they can help defer costly, time-consuming upgrades. By aligning their energy use with grid capacity—and, when needed, coordinating with bridge power and other onsite resources—power-flexible AI factories can connect faster, make better use of existing infrastructure, and generate more value for the system—reducing costs for local communities rather than driving new upgrade expenses.

Proving the Model: From Concept to Commercial Scale

Thought leadership and strategic capital mean little without execution. From day one, Emerald AI has been focused on proving that power-flexible AI is a tangible reality.

I am incredibly proud of our team for successfully completing five live demonstrations at commercial data centers over the last year across Arizona, Illinois, Virginia, Oregon, and London. Conducted in collaboration with EPRI and harnessing NVIDIA accelerated computing—including Blackwell Ultra GPUs running production-grade AI workloads—these deployments proved that we can flex power usage to relieve grid strain without compromising the stringent performance and latency requirements of critical customer workloads.

That progression began with the first peer-reviewed evidence of AI power flexibility, published in Nature Energy from our Arizona demonstration, and expanded to a spatial-flexibility demonstration that shifted inference workloads from Virginia to Chicago while maintaining quality of service. This month, we reached two further milestones: the UK’s first live demonstration of grid-responsive AI infrastructure in London, where we reduced electricity demand by more than a third in under a minute while high-priority workloads continued to run; and our most advanced joint technical demonstration with NVIDIA to date in Hillsboro, Oregon, integrated with the full NVIDIA DSX software stack.

Across our demonstrations, we have now successfully proven:

• Temporal Flexibility: Slowing or pausing AI workloads such as fine-tuning runs that have some customer-designated flexibility on completion time. Harnessing onsite energy resources can also provide temporal flexibility.
  

• Spatial Flexibility: Seamlessly rerouting latency-sensitive loads such as serving inference queries from a data center in a power market facing constraints to another facility halfway across the country where power capacity is available.
  

These milestones are the foundation for our commercial-scale deployment later this year in 2026. Together with NVIDIA, Digital Realty, EPRI, and PJM Interconnection, we are advancing toward the deployment of the world’s first power-flexible AI factory at commercial scale: the 96MW NVIDIA Aurora facility in Manassas, Virginia.

We are building a future where breakneck AI innovation doesn’t break the grid—it helps strengthen it. With this new capital, the backing of our Strategic Advisory Board, and a growing ecosystem of partners across AI, energy, and infrastructure, we have the resources we need to unlock the energy bottleneck and power the next era of human progress.

‍