Rune Raises $40M to Turn Stranded Solar Power Into AI Compute

Rune has raised $40 million in Series A funding as the energy infrastructure startup looks to address one of artificial intelligence’s increasingly significant constraints: finding enough electricity to power rapidly expanding compute infrastructure.

The round was led by Spark Capital, with participation from Union Square Ventures, Lowercarbon Capital, Activate Capital, Committed Capital, Timeless Partners, and Logos Fund. The financing brings Rune’s total funding to $53.5 million.

Alongside the funding, Rune unveiled RELIC, or Renewable Energy Linked Intelligent Compute, a modular data-center system designed to install directly at renewable-energy generation sites. Rather than waiting for new utility connections and conventional data centers to be constructed, Rune’s approach places GPUs alongside existing solar infrastructure and draws on electricity that might otherwise be curtailed, clipped, or economically unattractive to sell to the grid.

The company says RELIC systems can bring compute customers online in as little as six weeks, potentially creating another route to AI infrastructure at a time when access to electricity is becoming as important as access to GPUs.

Turning Solar Plants Into Compute Infrastructure

Rune is effectively trying to rethink where a data center needs to exist.

Traditional data centers are generally built around access to the electrical grid. Developers secure land, obtain utility capacity, build power and cooling infrastructure, and connect facilities to transmission and distribution networks. As demand for AI infrastructure has accelerated, that process has increasingly run into power availability and interconnection constraints.

Rune instead moves the compute to the electricity.

Its RELIC units are installed at operating renewable-generation facilities, where they can consume electricity directly from the plant. Rune says this allows it to capture several categories of underutilized generation, including solar power that is curtailed because the grid cannot accept it, electricity lost through clipping when solar output exceeds an inverter’s capacity, and power that becomes uneconomic to sell during periods of low or negative pricing.

The architecture is notably different from simply placing a conventional data center beside a solar farm. Rune’s technical description of RELIC says the systems are installed behind the inverter on a renewable plant’s DC bus bar, allowing the compute equipment to consume electricity before it is converted into alternating current for the grid.

That direct-current architecture can eliminate some equipment normally required in data-center power infrastructure, including additional transformers and AC conversion stages.

A Modular Data Center Built Around Variable Power

One of the more interesting aspects of RELIC is that it is designed around the intermittency of renewable energy rather than attempting to eliminate it.

Rune describes RELIC as a responsive computational load that can automatically determine when surplus power is available and adjust consumption accordingly. Its systems can ramp or curtail compute loads in milliseconds, with the renewable plant retaining control over its primary electricity operations.

The company has designed the platform around 100-kilowatt building blocks that can be combined into deployments exceeding 100 megawatts. Rather than constructing a large permanent data-center building, the modular units can be transported to an existing generation site and connected with relatively limited site preparation.

Rune’s current platform supports bare-metal GPU clusters ranging from eight to 1,024 GPUs, allowing deployments to be sized around individual workloads. The company says it has more than 80 MW of contracted power and a development pipeline exceeding 1 GW.

That architecture also creates an unusual relationship between software and energy availability. Instead of expecting the power supply to continuously accommodate the compute workload, Rune can make certain computational workloads responsive to the available supply.

Not every AI workload is naturally suited to interruption or fluctuating power availability, making orchestration an important part of the model. Rune has built software controls that monitor the generation facility and determine how much otherwise-unused electricity can safely be directed toward computing without interfering with the plant’s existing grid commitments.

AI’s Growing Electricity Problem

The timing of Rune’s expansion reflects a broader change in the economics of AI infrastructure.

GPUs remain expensive and difficult to deploy at scale, but electricity availability has emerged as another limiting factor. The U.S. Department of Energy estimates that American data centers consumed approximately 192 terawatt-hours of electricity in 2024, representing about 4.7% of total U.S. electricity consumption. Its reference forecast puts data-center electricity consumption at roughly 464 TWh by 2028.

At the same time, renewable generation continues to expand. Utility-scale solar generated roughly 296,000 GWh of electricity in the United States during 2025, an increase of 34% from the previous year, while wind and utility-scale solar together supplied approximately 17% of U.S. electricity.

Those two trends create the opportunity Rune is targeting: rapidly expanding demand for computational power alongside a growing pool of renewable generation that cannot always be efficiently absorbed by the electrical grid.

Rather than building more generation specifically for data centers, Rune is attempting to locate flexible computing capacity where excess generation already exists.

RELIC Goes Live at a 200 MW Texas Solar Facility

Rune is also revealing a new deployment at an existing 200 MW solar facility in Texas, where RELIC has been installed without requiring new grid infrastructure or major modifications to the underlying renewable asset.

Texas is a particularly relevant testing ground. Solar capacity has expanded rapidly across the Electric Reliability Council of Texas (ERCOT) market, with the U.S. Energy Information Administration forecasting that utility-scale solar generation could reach 78 billion kWh in the region during 2026, potentially exceeding coal generation for the first time.

As renewable penetration increases, periods in which electricity production exceeds transmission capacity or market demand can become more frequent. Flexible compute infrastructure could potentially act as a buyer during some of those periods rather than forcing generators to reduce output.

Rune says its systems do not require a grid connection for the compute infrastructure itself and consume no water, another potentially significant distinction as data-center developers increasingly face scrutiny over both electricity and water requirements.

Building a New Layer Between Energy and AI

Rune was founded by CEO William Layden and CTO Varun Palivela, bringing together experience from both renewable energy and semiconductor engineering.

Layden previously worked with hydroelectric facilities using otherwise-unused electricity for computing-intensive cryptocurrency mining and later evaluated technologies across SoftBank Energy’s 1.7 GW solar portfolio. Palivela previously worked on high-performance processor architecture at NUVIA before its $1.4 billion acquisition by Qualcomm, as well as holding engineering roles at Qualcomm, Arm, and Marvell.

That combination is reflected in Rune’s strategy. The company is not developing another AI model or competing directly with GPU manufacturers. Instead, it is focusing on the physical infrastructure connecting processors to electricity.

The $40 million Series A will help Rune expand deployments with renewable-energy operators while adding compute customers seeking inference capacity.

If the model works at scale, RELIC could turn an inefficiency in renewable generation into infrastructure for another rapidly growing industry. Instead of waiting years for entirely new data centers and grid connections, Rune is betting that at least part of AI’s next wave of computing capacity can be placed where unused electricity already exists.