Pipe the Fuel

Natural gas is the bridge fuel of the AI buildout. The fastest path to new dispatchable power for data centers is a gas-fired turbine (L21), and every turbine needs a pipeline connection. This layer is about the physical infrastructure that delivers fuel to the power plants — the interstate and intrastate pipeline networks that move natural gas from production basins to generation sites.

Williams Companies operates Transco, the largest-volume natural gas pipeline in the United States, running from the Gulf Coast to the Northeast. Their Southeast Supply Enhancement project directly serves the growing data center power demand in the Virginia-Carolinas corridor — the densest data center market in the world. Kinder Morgan transports approximately 40% of all US natural gas and handles 40% of US LNG export supply. Their pipeline network is the arterial system that makes gas-fired power generation possible at the scale data centers require.

The Power Grid Reckoning trend creates this layer's structural importance. The US electrical grid was not designed for the concentrated, baseload demand pattern that AI data centers impose. When a hyperscaler announces a 1GW data center campus, the question is not just "where does the electricity come from?" but "where does the fuel come from to generate that electricity?" If the nearest gas pipeline lacks capacity, no amount of capital can build a gas turbine fast enough — because the pipeline permitting and construction timeline is 3-7 years, often longer than the power plant itself.

This creates a geographic constraint on the AI buildout. Data centers can only be built where three things converge: electrical grid interconnection capacity, fiber connectivity, and fuel supply. Pipeline capacity is increasingly the binding constraint in regions where grid connections exist but gas throughput is maxed out. The Permian Basin, Marcellus Shale, and Gulf Coast corridors have abundant gas supply — making them natural hubs for the next wave of AI data center campuses.

The investment framing is unusual for AI infrastructure. KMI and WMB are traditional energy midstream companies trading at utility-like multiples with 5-6% dividend yields. The market has not fully priced in the structural demand increase from AI power generation. These are toll-road businesses — they collect fees on every cubic foot of gas that flows through their pipes, regardless of the commodity price.