Island Roadhouse Data Centers has unveiled the design approach for an integrated, low-water, atomic-powered data center campus planned in Missouri, outlining a vision that places on-site power generation, advanced liquid cooling and industrial heat reuse at the center of its architecture.
The design directly addresses what has become the defining constraint for AI infrastructure: access to power, and the time required to secure it. According to Lawrence Berkeley National Laboratory's "Queued Up: 2026 Edition" published in June 2026, roughly 8,200 projects representing 1,312 GW of generation capacity were waiting in United States interconnection queues at the end of 2025, with the median time from interconnection request to commercial operation exceeding five years for projects completed that year. The International Energy Agency's April 2026 report "Key Questions on Energy and AI" projects global data center electricity consumption will roughly double from 485 TWh in 2025 to 950 TWh by 2030, noting that developers in the United States are increasingly turning to on-site generation as grid connections lag. Water availability adds another layer of pressure, with Lawrence Berkeley National Laboratory's December 2024 "2024 United States Data Center Energy Usage Report" estimating that U.S. data centers directly consumed approximately 66 billion liters of water in 2023.
The Island Roadhouse campus is designed to operate entirely behind the meter, with on-site generation carrying the compute load and the atomic-served campus core. The planning basis relies on advanced atomic reactors configured in an N+1 arrangement for reactor-level redundancy, with heat from the reactors converted to electricity through an indirect closed supercritical carbon dioxide Brayton cycle. The campus is not planned as a power exporter, and no grid export is contemplated.
"Power and cooling are not two problems. They are one system," said David Kinsman, Chief Technology Officer of Island Roadhouse Data Centers. "We are engineering the reactors, the power conversion, the liquid cooling and the heat network as a single campus. The constraint on AI infrastructure is no longer the chip. It is power, and the years it can take to get power delivered. We designed this campus so that the power is generated where the compute sits, behind the meter."
The high-density data halls are built around closed-loop warm-water direct-to-chip liquid cooling, with dry heat rejection serving as the primary and ultimate campus heat sink. The campus targets a low-water design with an annual water-use-effectiveness planning goal. Thermal energy storage is planned as an optimization asset, and the campus has no grid-scale battery — its only planned energy store is thermal. Waste heat will be made available to third-party industrial process and other non-industrial users through a campus thermal network, with thermal capacity reservation, metered thermal energy and thermal interconnection forming the planned service model. Those users would finance, build, own and operate their own facilities. The campus is designed to reject all of its required heat independently, making heat reuse an optimization and sustainability benefit rather than an operational dependency.
"Every data center produces heat, and most of them pay to throw it away," Kinsman said. "Our thermal design treats heat as a product. The campus is designed to sell data center waste heat through a metered thermal network so users can take the heat we would otherwise reject. This is what the next generation of sustainable data centers will look like."
Island Roadhouse sells long-term, power-backed, liquid-cooling-ready data center capacity and facility services. Customers bring and operate their own compute, and the company does not sell electricity, cloud compute or GPU-as-a-service as its base business. Phase 1 is planned to offer up to 154 MW of sellable IT load, targeting commercial operation in 2034, with a Phase 2 expansion planned under the campus master plan.
