Industrial developer BaRupOn acquired roughly 700 acres in Liberty County, Texas, to expand into chemical manufacturing after securing federal support tied to domestic industrial capacity. But when executives learned the site would face about $35 million in grid interconnection costs and could not secure substantial utility power until 2029, the project reoriented around behind-the-meter energy and AI infrastructure.
“We thought power is easy,” BaRupOn COO Balaji Tammabattula told Data Center Knowledge. “The thing that we did not think about was the biggest problem.”
Amid years-long interconnection delays and rising upgrade costs, more AI developers are pursuing behind-the-meter power to keep build schedules on track. BaRupOn’s Liberty County pivot shows how these constraints are reshaping site selection, financing, and hyperscaler appetite.
From COVID Test Kits to AI Power Infrastructure
The pivot caps a rapid evolution for the company. BaRupOn started as a medical diagnostics manufacturer during the COVID-19 pandemic and says it supplied roughly 250 million test kits before turning to chemicals used in medical and industrial markets. As the company scoped a methacrylate chemical manufacturing facility in Liberty County, power constraints became the defining risk.
BaRupOn initially sought about 25 MW to run manufacturing operations tied to a Defense Department industrial base expansion initiative. After approaching utilities, executives concluded that interconnection delays and upgrade costs would stall the project for years. The company shifted to an on-site generation strategy, starting with natural gas.
BaRupOn says it subsequently signed gas supply agreements with Kinder Morgan and currently holds permits for roughly 60 MW of on-site generation, with long-term plans that could scale substantially higher through additional gas, solar, and potentially nuclear capacity.
In parallel, the company is building a 200,000-square-foot data center on the Liberty County site while continuing construction of the chemical manufacturing facility nearby.
Hyperscalers Have Change of Heart on Power
What changed most over the past year, according to Tammabattula, is hyperscaler willingness to reconsider off-grid or behind-the-meter deployments. “Last year, when we were having these conversations with hyperscalers, they were not interested in anything behind the meter,” he said. “Those conversations we thought were a dead end – we’re getting back into those conversations.”
The shift reflects mounting pressure across the AI infrastructure market as developers race to secure power capacity fast enough to support next-generation clusters.
“Three hundred megawatts was a lot of power five years ago,” Tammabattula said. “Today it is nothing.”
AI Demand Collides with Grid Timelines
The Liberty County project illustrates how AI infrastructure growth is colliding with the utility and transmission timelines.
Rob Gramlich, analyst and founder of Grid Strategies, said lengthy interconnection queues and delays for large-load projects are increasingly common as demand surges.
BaRupOn executives said they selected Liberty County because it offered land, water access, rail connectivity, and proximity to the Port of Houston. Power availability, however, became the defining constraint. “We realized that the power is much more valuable on the data center side than the manufacturing side,” Tammabattula said. The company now plans to use the site for both manufacturing and AI infrastructure development.
Nuclear Emerges as Long-Term Pathway
BaRupOn recently completed a feasibility study with NANO Nuclear Energy exploring whether advanced microreactors could eventually support the Liberty County campus.
Executives stressed the nuclear portion remains conceptual and likely years away if it moves forward. Still, BaRupOn Chief Strategy Officer Derek Matthews argued that conventional power pathways are struggling to meet the scale, emissions requirements, and reliability demands emerging around large AI campuses.
Matthews said Liberty County falls within a Texas nonattainment region with strict emissions permitting requirements that complicate large-scale gas generation deployments.
“You get to this point of like, ‘Well, what do you do?’” Matthews said. “You can’t burn gas.”
While the company evaluated solar and other alternatives, Matthews said land requirements and grid limitations at scale posed challenges.
He argued that smaller modular microreactors may ultimately prove more practical for data center environments than traditional gigawatt-scale nuclear facilities because operators can take individual units offline for maintenance and refueling.
“All these other groups are buying up Three Mile Island or all these other large reactors,” Matthews said. “Power is so desperate right now.”
BaRupOn says its long-term power planning could eventually exceed 1 GW if nuclear, solar, and additional gas infrastructure move forward. For now, the immediate challenge mirrors a broader AI infrastructure market problem: securing enough power fast enough to attract anchor tenants, secure financing, and support future compute deployments.
“The real infrastructure is not developing fast enough compared to the need,” Tammabattula said.