CleanAI, an energy and data center development platform, thinks it has an answer to the grid delay problem facing data center developers: co-located, “hybridized” solar, storage, and natural gas microgrids.
In a landscape where speed to power now trumps just about everything — including emissions profile and price — for data center developers, CleanAI is pitching its solar-gas microgrids as both faster and more scalable than pure gas microgrid setups (like that deployed by Elon Musk’s xAI in Memphis).
Founded last year by a team of veteran energy developers, the startup is promising to deliver data center developers clean power “at the speed of tech” and bring their microgrids online in as little as 18 months. And according to a pitch deck viewed by Latitude Media, CleanAI claims that its solar-gas setup can achieve eight cents per kilowatt-hour, which would be significantly cheaper than most utility power. That price assumes a grid connection, meaning that the gas would serve as pure backup.
The company’s primary offering to data centers isn’t the service of designing and building custom hybrid microgrids; instead, it’s offering what CEO and co-founder Luke Hansen describes as “powered land as a service.” With this model, CleanAI acquires and develops land, builds, owns, and operates the on-site generation assets, and then sells dispatchable power directly to data centers to build there.
It’s a distinct approach from that taken by traditional powered land developers, Hansen explained, in that CleanAI is selling energy as an operating expense, rather than an up-front capital expense, explained Hansen, who previously developed energy storage projects at General Electric and then Avantus.
For data center developers, the set-up is designed to alleviate both the need for diesel backup and the cost and risk of joining the load queue for interconnection; as the AI boom takes off, the latter process is notoriously plagued with delays.
But there’s a caveat: CleanAI’s model only really works in Texas at the moment, thanks to the state’s regulatory framework allowing power developers to sell behind-the-meter power directly to a data center, without an utility acting as an intermediary.
Elsewhere in the country, including in California, Arizona, and Nevada, selling power directly behind the meter is either not allowed or only possible via complicated experimental tariffs where utilities act as intermediaries, Hansen explained. Those states are working on new rate cases or policies, he added, but for now their policy environments aren’t as conducive to CleanAI’s model.
The “clean transition tariff,” for example — a special rate case designed in part by Google and geothermal developer Fervo, approved in Nevada in May — wouldn’t apply to a CleanAI microgrid. The CTT only applies to 100% clean power, therefore excluding a hybrid microgrid that might have only 98% clean power, Hansen said.
Texas policy is also conducive to another key selling point for CleanAI’s model: the ability to skirt a grid connection, and therefore get power much sooner than is typically possible. ERCOT uses a unique system for adding new generation to the grid, known as “connect-and-manage.” It’s a faster system than the norm, because it allows generation to connect as soon as local upgrades have been identified and completed, without waiting for broader network upgrades to be planned and built. Instead of tying new power generators up in long interconnection studies, ERCOT lets them come online and then manages grid constraints or bottlenecks in real time.
Finding patient capital
Of course, CleanAI’s business model is a capex heavy one: they’re buying land and energy infrastructure up front. That means the work can’t be executed without some serious capital, so the company is eschewing traditional Silicon Valley-style venture rounds in favor of institutional funding sources.
As Hansen put it: “You don’t develop a gigawatt project off of $1 million VC early-stage money.” In other words, VC funding is really too small and too expensive for large infrastructure projects like those CleanAI is building.
While they did pitch a few VCs who were willing to invest, in the end CleanAI opted for a better “match for the asset class,” meaning mostly family offices, which write bigger checks and accept more reasonable returns.
A software backbone
Underpinning CleanAI’s microgrid business model is its proprietary software platform, which both simulates and controls the hybrid microgrid systems, enabling custom design and real-time management for each data center.
That platform involves a digital twin of a hybrid system, which tests how a microgrid’s particular hybrid generation mix responds to weather and load scenarios. It leverages decades of hourly weather data to validate system sizing and reliability, even under rare or extreme conditions like Texas heat waves. The simulations, Hansen explained, encompass not just energy production capabilities, but also the performance of grid-forming inverter controls, needed to stabilize power and smooth AI workload spikes.
Embedded in the software are intelligent control and optimization algorithms that manage energy dispatch for thousands of scenarios in real time, balancing solar generation, battery charge cycles, and gas backup to meet reliability targets set by the customer. The resulting platform can be used to tailor each microgrid’s power mix to specific data center needs without a complete redesign.
That software backbone, according to Hansen, is “the most important thing” for CleanAI’s customers — it reduces technical and operational risk and operates project timelines. “I haven’t found a data center developer who can ramp the compute faster than we can ramp the power,” he said.
Choosing a location
For existing data centers that are unable to expand capacity due to grid interconnection challenges — or for those that already have the land but are stuck in a long interconnection queue — CleanAI has a secondary offering it calls “bridge to grid.” For that model, the company will design and build a microgrid at a site chosen by the data center developer, and fully power the existing capacity behind the meter in the interim.
There are restrictions of course, largely based on location. “If the facility is downtown in a major metropolitan area there’s nothing we can really do to help them,” Hansen said. But for the majority of facilities that are on the outskirts of cities or in more rural areas, with room to build solar arrays within around ten miles, CleanAI can help them expand capacity.
At present though — about eight months in — CleanAI is largely focused on greenfield sites. The company currently has a portfolio of five, 1 GW “phase one” projects in Texas, near Dallas, Houston, San Antonio, and Temple. (There are also a handful of “phase two” projects in the pipeline, Hansen said, located in the southwest of the U.S., but they’re in the very early stages of development.)
Locations are a key selling point for the data center customers CleanAI is already contracted with, as well as for those in the pipeline. “A lot of the customers we talk to say ‘oh, you’ve got a solar, battery, gas hybrid project, I hope it’s not in West Texas!’” Hansen said. When they find out the site is near Dallas, for example, their tune changes.
Identifying those first five sites, he added, was a key part of the market research that went into building CleanAI in the first place. “At my prior employer we had a lot of solar projects, dozens of them. But none of them were in locations where a data center wanted to be,” Hansen said. “We realized we had to take a fresh approach, and site fresh projects where data centers wanted to build.”
