Fermi America, developer of what it characterises as the world’s largest behind-the-meter artificial intelligence private grid campus in partnership with the Texas Tech University System, has secured an agreement with Energy Transfer to deliver firm natural gas supply to Fermi’s HyperGrid campus located outside Amarillo, Texas. Energy Transfer operates approximately 140,000 miles of pipelines and associated energy infrastructure across 44 states, positioning it as one of the largest and most diversified midstream energy companies in North America.
Infrastructure Connection and Timeline
Under the agreement, Fermi America will connect to Energy Transfer’s pipeline infrastructure immediately south of the campus site, accessing natural gas supplies required for power generation serving artificial intelligence applications. The pipeline interconnection project is expected to enter service in the first quarter of 2026, requiring minimal capital expenditure by Fermi America.
Rendering of the Fermi America™ Private Grid Campus.
The proximity of existing Energy Transfer pipeline infrastructure to the Amarillo campus site provides advantageous conditions for establishing gas supply connectivity with limited infrastructure investment and a relatively short development timeline compared to scenarios requiring extensive new pipeline construction.
Strategic Positioning and Power Generation
Toby Neugebauer, co-founder and CEO of Fermi America, characterised the Amarillo campus as one of the premier sites nationally for clean energy sources, emphasising that this positioning derives initially from its qualities as an elite natural gas location. He stated that the Energy Transfer agreement secures a firm natural gas supply necessary to generate the first two gigawatts of clean, reliable power for AI data centre customers.
The emphasis on firm gas supply reflects requirements for reliable, dispatchable power generation serving data centres with continuous operating requirements and limited tolerance for power interruptions. Firm gas contracts provide supply assurance not subject to curtailment during periods of high demand or supply constraints that can affect interruptible gas supplies.
Campus Scale and Development
The HyperGrid campus encompasses 5,236 acres outside Amarillo in the Texas Panhandle region. The scale of the development reflects the substantial land, power, and infrastructure requirements of large artificial intelligence computing facilities, which consume significantly more power per unit area than traditional data centres due to the intensive computational requirements of AI model training and inference operations.
The behind-the-meter private grid structure indicates that Fermi America is developing dedicated power generation serving the campus rather than relying on electricity purchases from the public utility grid. This approach provides control over power supply reliability, pricing certainty, and potentially advantageous economics for extremely large power consumers.
Supply Chain and Execution Strategy
The Energy Transfer agreement advances Fermi’s objective of bringing online the balance of long-lead gas-fired generation supply chain assets secured earlier in 2025. This statement indicates that Fermi has previously secured commitments for power generation equipment, which typically involves substantial lead times for manufacturing and delivery of gas turbines, heat recovery steam generators, and associated equipment.
Neugebauer emphasised Fermi’s commitment to actionable timelines and velocity of execution, suggesting a focus on rapid project development compared to typical timelines for large-scale power generation and data centre projects. Speed of execution has become increasingly important as AI companies race to secure computing capacity supporting rapidly growing demand for AI model development and deployment.
AI Data Centre Power Requirements
Artificial intelligence workloads, particularly training of large language models and other advanced AI systems, require substantially more computing power and electrical energy than traditional data centre applications. Individual AI training clusters can consume tens or hundreds of megawatts, with the largest facilities potentially requiring gigawatts of power capacity.
This extraordinary power demand has created challenges for AI companies seeking to expand computing capacity, as the available power supply in traditional data centre markets has become constrained. Purpose-built facilities with dedicated power generation address this constraint by developing new power supply rather than competing for limited grid capacity in established markets.
Natural Gas and Clean Energy Positioning
Fermi America’s characterisation of natural gas-fired generation as providing “clean, reliable power” reflects the positioning of gas generation as lower carbon compared to coal whilst providing dispatchable generation complementing renewable energy sources. Modern combined-cycle gas turbine plants achieve higher efficiency and lower emissions than older generation technologies whilst providing flexibility to adjust output, responding to variable renewable generation or load requirements.
The Texas Panhandle region offers both natural gas supply access through pipeline infrastructure serving regional gas production and exceptional renewable energy resources, particularly wind generation. Neugebauer’s statement regarding clean energy sources suggests potential for the campus to incorporate renewable generation alongside gas-fired capacity, though the Energy Transfer agreement specifically addresses gas supply for thermal generation.
For more information visit www.FermiAmerica.com
