AI Power Demand Creates Structural Nat-Gas Squeeze: Long Midstream Before the Market Wakes Up

The thesis fails if (1) hyperscalers pivot aggressively to nuclear or battery storage, reducing gas-fired generation's share of incremental AI power demand, OR (2) new pipeline permitting is fast-tracked by federal policy (unlikely but possible under a deregulatory administration), adding capacity before the demand crunch materializes. The observable signal to watch: if utility integrated resource plans filed with state regulators in Virginia, Georgia, and Texas in H2 2026 show declining gas-fired generation additions in favor of storage-plus-renewables, the demand thesis weakens materially.

The strongest bear case is that US shale's supply elasticity is so high that any demand signal immediately calls forth enough new production to keep Henry Hub prices range-bound, eliminating the scarcity premium that would force pipeline operators to raise rates. The most fragile assumption in this thesis is that pipeline capacity is genuinely the binding constraint — if producers and shippers can reroute gas via existing underutilized routes or LNG import reversals, the bottleneck may not be as tight as the thesis implies. The data point that would flip conviction entirely: FERC pipeline capacity utilization data showing Southeast corridor pipes running below 85% utilization in Q3 2026 would indicate the bottleneck thesis is premature and the trade has no near-term catalyst.

The closest analogue is the 2005-2008 period when LNG import terminals were built across the US Gulf Coast in anticipation of a domestic gas shortage — only for shale to arrive and strand the assets. The mispricing lesson from that episode cuts both ways: (1) the market DID underprice the infrastructure bottleneck in the near term (pipeline stocks surged 2005-2008 as throughput demand exceeded capacity), but (2) the structural demand thesis was ultimately wrong because supply responded faster than expected. A second analogue: the 2021-2022 European gas crisis, where pipeline capacity constraints — not production shortfalls — drove TTF gas prices to 10x US Henry Hub. The rent accrued to whoever owned the pipe, not the field. The behavioral pattern at work here is the 'infrastructure lag' — humans systematically underestimate how long it takes to build physical networks and overestimate how quickly demand can be rerouted. This is a well-documented planning fallacy that repeats in every commodity supercycle.