In July, a pair of high-tech firms unveiled plans for a 1.0 gigawatt AI data center in Childress, Texas. The collaboration between Lancium, an energy technology and infrastructure company, and AI infrastructure provider Crusoe will birth a grid-connected campus spanning 270 acres.
Like similar projects worldwide, the data center will support the AI computational requirements of a leading technology company. Lancium is handling development and management, overseeing both grid power interconnection and ongoing energy orchestration for the campus.
This division of labor has already proven successful at scale in Abilene, home of the Lancium Clean Campus and Stargate 1, a digital infrastructure initiative slated to deploy a network of hyperscale AI data centers.
“Childress is a natural next step in our partnership with Crusoe,” said Michael McNamara, CEO of Lancium, in a July 15 press release. “Crusoe has been a great partner to build alongside in Abilene, and we're glad to bring that same model - and the jobs and investment that come with it - to the community of Childress. We are deeply committed to working with local stakeholders to advance a model of development that addresses the needs of the community.”
More broadly, the project highlights how hyperscalers are pairing on-site energy architecture directly into campus builds rather than relying on traditional utility hookups, noted Eric Gimon, a senior fellow with the Electricity Team at Energy Innovation, a non-partisan energy and climate policy think tank.
“I knew the Lancium folks before AI, because they were one of the few companies thinking about industrial parks with a flexible consumption profile,” said Gimon. (Neither Lancium nor CEO McNamara responded to interview requests for this story.)
In defining its position in AI, Lancium is now merging grid access with behind-the-meter resources such as battery storage, onsite generation and energy management software.
The flagship location in Abilene is part of a portfolio encompassing Childress as well as a forthcoming AI campus in Hall County. Last October, Lancium closed a $600 million debt financing deal for an Abilene site anchored by purpose-built substations and infrastructure.
Unlike a standard utility grid connection – which leaves AI developers subject to power outages and regulatory delays - behind-the-meter assets can turn the Abilene campus into an active energy ecosystem rather than a passive utility consumer, said Gimon.
“Having onsite battery allows you to do better ride-through, and will insulate you from the blips and blurps of the grid,” Gimon said. “Data centers can’t afford to have an outage, because that can ruin an entire batch of calculations. Developers need power on a consistent basis.”
New power generation
Construction in Abilene began in 2024. The project’s first phase - two buildings totaling about 200 megawatts - moved from groundbreaking to operations in a year. Crusoe then secured billions in financing to scale the site into a 1.2 GW campus spanning eight buildings.
This March, Crusoe announced plans to expand the Abilene development in support of large-scale Microsoft workloads, which would boost site capacity to roughly 2.1 GW. Reflecting a trend of multi-tenant “AI factories,” the Abilene design integrates solar resources and battery storage to enhance grid reliability and optimize carbon efficiency.
“What is needed here is a stack of different kinds of battery technologies, since lithium-ion batteries are not well suited to cycling several times an hour,” said Gimon. “Instead, high power/low-energy storage techs like LTO lithium - or Nickel-zinc batteries and kinetic solutions like flywheels - are used in concert with larger batteries. The batteries smooth out the jitter and turn the square steps into smoother ramps that energy supplies can deal with better.”
Meanwhile, Lancium said that natural gas turbines will serve as backup power for the next phase of its “clean campus.” According to the company, the site will leverage nearby wind resources while incentivizing greenfield renewable power development.
Clean, uninterrupted power is vital for a technology with relentless baseload energy demands - AI data centers have also introduced a new kind of electrical behavior into a grid system built for predictability, said Gimon.
“Other large loads are like a diesel truck that speeds and slows, whereas data centers can go from zero to 100 and vice versa very fast,” he said. “There are ways to manage these issues, it just takes time and money, which is hard in an environment where there’s pressure to keep building.”
Looking ahead, other regions of the country will be tasked to keep pace with the intensity and volatility of AI workloads. The question is no longer whether AI can scale, but how our electrical network can ascend alongside it.
Gimon points to Virginia’s “Data Center Alley” as a particularly apt example – a recent transmission line fault forced hyperscale facilities into backup power, removing over 3 GW of demand from the PJM Interconnection in seconds.
As a traditional power grid won’t tolerate these “crazy swing loads,” it’s incumbent upon data center leaders to seek a cleaner consumption profile. This could mean using natural gas – like the Tallgrass development in Wyoming – or Lancium’s multi-pronged energy effort in Texas, said Gimon.
“Gas can be part of the mix, but the cheapest resources are renewables and batteries, and they’re faster to build,” Gimon said. “That will get you off the ground faster than a gas-only solution.”
