Emerald AI, a company leveraging software to prevent runaway electricity demand from computing facilities, is now valued at $1.05 billion.

Silicon Valley Power, a municipal utility, is significantly smaller than California's utility giant PG&E. However, its grid connects to roughly 55 data centers within a mere 20-square-mile area, making it one of the most energy-intensive regions in the state.

This concentration puts the utility under immense pressure from the ongoing wave of AI infrastructure construction. Even with efforts to double its capacity, Silicon Valley Power cannot keep pace with the immediate power needs of clients such as Nvidia and Intel.

"All the spare capacity we might have had in the past is now spoken for," said Nicolas Procos, utility director at Silicon Valley Power. "So now you're talking about expanding the system, or coming up with creative solutions."

As one such solution, Silicon Valley Power recently brought on board Emerald AI, a startup founded just two years ago. Using its proprietary software, Emerald helps data centers manage their power consumption more flexibly based on input from the utility.

Building and running AI models requires immense computational power, along with the grid energy needed to fuel these workloads. This resource strain has sparked a nationwide backlash against data center construction, potentially hindering industry growth.

Emerald is a startup attempting to ease this problem. On Tuesday, the company announced it has raised $150 million in funding at a valuation of $1.05 billion. Venture capital firms DCVC and Energize Capital co-led the round. Nvidia, Samsung Ventures, GE Vernova, and Salesforce Ventures also participated. Early investors include former Secretary of State John Kerry and Google veteran Jeff Dean, who recently launched his own startup.

Nvidia's involvement underscores how the chip giant is financing the growth of the AI economy.

"You can't deliver compute without energy," said John Tough, managing partner at Energize Capital.

However, the power required by AI systems is variable, and compute-intensive tasks can overload the grid during peak periods. Through Emerald's software, data centers can reduce energy consumption when utilities request it and utilize excess capacity when available.

Emerald was founded by Varun Sivaram, a 37-year-old renewable energy industry veteran who now serves as CEO and has also held climate policy roles in government. The company's chief scientist is Ayse Coskun, a Boston University computer science professor who has pioneered research in data center flexibility.

Emerald generates revenue through several channels. By optimizing efficiency, it helps utilities supply more power to data centers and takes a share of the additional sales. The company also assists data center operators in working with their customers, such as AI labs or hyperscale cloud providers, to identify which AI workloads can be reduced and where excess compute capacity can be utilized. Emerald takes a cut of the resulting additional revenue.

Emerald's software is powered by AI agents that autonomously decide how data center operators and their clients should best reduce power loads. The company has tested its product across five data centers, with partners including Nvidia and Oracle. It currently holds contracts in California and Virginia.

Sivaram stated his goal is to make data centers more flexible for the grid, and in doing so, "turn them from villains into heroes."