Every large AI campus now begins with the same unglamorous question: where does the electricity come from? Billions of dollars flow into concrete, chips and cooling, and many projects still stall at the utility’s front door. San Francisco-based Rune Energy has decided to reverse the sequence. The startup builds solar-powered data centers beside power plants that already produce more than the grid can use. Compute travel to the energy, rather than energy chasing compute. That small change in geography carries a large change in strategy. This is a strategic argument as much as an engineering one. Power access has become a major constraint in parts of the compute economy, and some operators struggle to secure new grid connections. Rune treats that scarcity as a placement problem. If solar farms have electricity they cannot deliver to the grid, a suitably configured data center could use some of that power, depending on the site’s electrical design and the availability of generation. Investors tracking AI infrastructure should watch where that reframing leads.

Inside the RELIC Module Behind Solar-Powered Data Centers

Rune calls its unit Renewable Energy Linked Intelligent Compute, or RELIC. Each module plugs directly into the direct current that a solar array generates. Network equipment, a cooling system, power electronics and GPUs all live inside one container. The reported container dimensions are about eight feet long, two feet wide and five feet high, according to IEEE Spectrum’s account of Rune’s product. A module can draw up to 100 kilowatts and weighs about 2,000 pounds. Picture a heavy-duty equipment crate that happens to hold a small supercomputer. Capacity scales by the stack. According to industry reports recently cited, one RELIC module can hold anywhere from 128 Nvidia B300 GPUs to over 2,000, and operators can interconnect several modules into a large GPU network. Palivela describes the difference from a conventional site in plain terms. “The only difference [between Rune and a conventional data center] is that it’s not in a centralized building connected to transformers and switchgear,” Palivela says. He also identifies the customer segments Rune is targeting. “We sell this to inference customers and those who are doing small training runs or fine-tuning training runs on state-of-the-art models.”

Where the Wasted Energy Hides

Layden puts a number on the opportunity. “It’s our estimate that there’s 50,000 gigawatt hours of energy wasted by utility-scale solar every year,” says William Layden, Rune’s co-founder and CEO. A solar farm can generate more electricity than the grid can absorb at certain moments, especially when output runs high and demand runs low. Operators then curtail production, which means they throttle panels that could otherwise deliver power. Curtailment is one distinct source of unused solar generation, separate from equipment outages, conversion losses and inverter clipping. Equipment outages, conversion losses and inverter clipping drain value as well. Inverter clipping deserves a closer look. Inverters typically convert the DC electricity from solar panels into alternating current for the grid.

When an array produces more DC power than its inverter can handle, the inverter clips the excess. Rune connects before that conversion, which lets it reach some of that power. The company’s cofounder frames the whole situation as a mismatch rather than a shortage. Palivela’s diagnosis is blunt. “We’re already in an energy abundant future,” says Varun Palivela, Rune’s co-founder and CTO. “But the energy abundance is stranded in time and space.” The power constraints facing large data centers, he adds, are “not a real bottleneck, but a resource allocation issue.” That reading frames at least part of the power challenge as a resource-allocation problem rather than a shortage of electricity everywhere. Logistics, unlike physics, yields to better business models.

The Direct Current Advantage, Tested

Tapping DC power early lets Rune skip a conversion step. As per the reports, the company signs power-purchase agreements with solar-farm owners and buys electricity straight from them. Those deals give the solar facility income for power that would otherwise go unused. Utility-scale solar systems can use a nominal DC voltage of 1,500 volts, although the actual voltage depends on the system’s design and operating conditions. Rune’s technology needs no inverter, and it does use DC to DC converters to step the voltage down to a lower range. “You need to get the power in a conditioned state to allow the GPUs to consume it,” Layden says.

Outside experts urge caution on the efficiency story. Constance Crozier, an assistant professor at Georgia Tech’s School of Industrial Systems Engineering, has studied data center energy demand. She argues that the main advantage may not come from cutting out the conversion. “DC to AC conversion is typically quite efficient—95 to 97 percent—so I don’t expect the efficiency savings to be exceptional,” she says. “I would imagine the main advantage is reduced component cost.” However, lower component costs could still matter in a market where data center hardware investments can reach billions of dollars, although Rune’s actual savings have not been quantified in the article.

Speed, Supply Chains and a Product Mindset

Layden insists Rune is a product company, not a construction company. Rune manufactures RELIC modules at its Mountain View, California, facility and designs them for ultra-fast, simple deployment once they reach a solar farm. “You literally take two wires, plug it into the DC link, and you’re done,” says Layden. Layden says Rune recently fulfilled a request to bring a data center online within a week, a company-reported deployment timeline that the article does not independently document. “We were able to do that,” Layden says. “We are the fastest time to compute.” The modular design sidesteps a stubborn supply-chain problem. A traditional facility can require large transformers, switchgear, cooling systems and other specialized electrical equipment. Procurement timelines for some specialized electrical equipment can extend for months or longer, depending on the equipment, supplier and project requirements. Rune’s containers carry their own power electronics, which lets them avoid much of that queue. Speed to power could become a competitive advantage if Rune can sustain faster deployments than competing providers without compromising reliability or cost.Buyers who need inference capacity quickly may value faster deployment alongside electricity prices, reliability and the total cost of computing.

Cloudy Days and the Battery Question

Intermittency remains solar’s obvious weak spot. What happens when clouds linger for several days? “We have batteries integrated at the Relic level to make sure we always have compute 24/7,” says Layden. Rune modules carry no backup grid connection, which raises the stakes on that design. Crozier says sizing the batteries correctly is key. “They need to size the batteries to store enough energy for nighttime and cloudy day operations, and generate enough solar to fill these batteries on ‘good solar’ days,” she says. That arrangement carries a trade-off. Some of the available power at any given moment goes to battery charging rather than directly to compute. Rune responds with extensive modeling of optimal data center and battery sizing. The team uses historical data to estimate how much power each site is likely to offer under different conditions. Current sites sit in California, Texas and Massachusetts. For example, a Texas site and a Massachusetts site need different battery profiles.

Behind the Meter and What Comes Next

Rune’s strategy belongs to a wider push to go behind the meter. Data-center operators connect directly to power generation to avoid relying on the grid. The approach has gained attention as operators struggle to secure new grid connections. Direct connections to nuclear plants, natural-gas generators and other power sources are among the approaches being explored to supply large data center loads. Solar plus batteries adds a renewable option to that list. Crozier sees broader value in the pairing. “Far more significant flexibility can be obtained by co-locating batteries with a data center, especially if they’re paired with a generation source,” Crozier says. “This way the data center can remove part of the load from the grid without changing its operation.” That flexibility matters to utilities as much as to tenants. A data center that can reduce its grid demand without slowing workloads could ease pressure on a constrained electricity system, provided its on-site generation and storage can reliably supply the displaced load. Rune plans to scale its operations to 100 megawatts of compute over the next 12 months. Meanwhile, the real test will be whether the economics hold at that size.

The Compute Forecast Takeaway

The larger lesson sits in how the industry measures scarcity. Operators have long treated electricity as a bill to negotiate; Rune treats it as a place to go. If curtailed solar really totals 50,000 gigawatt hours a year, that stranded supply represents a large pool of untapped compute fuel. Execution risks remain, from battery sizing and reliable power delivery to whether demand for inference and fine-tuning workloads can support the model at scale. Anyone modeling future capacity should track whether modular, solar-fed compute can demonstrate reliable operation and viable economics as deployments expand beyond initial installations. The answer will show whether wasted energy becomes valuable or stays a footnote.