US natural-gas deficit / AI-power (2028+) (new) Structural deficit ▲

Smith (Chronometer Partners, Invest Like the Best, Jul 21) — the central thesis after 18+ months modeling nearly every US gas well, pipeline and processing asset from the well level up: an historic US natural-gas deficit starting in 2028. The ~20 BCF/d of new production the system can physically deliver (rock + processing + gathering + interstate pipe) is already matched to the permitted 15→35 BCF/d LNG-export ramp, so probability-weighted AI-compute demand (P50 base case ~5 BCF/d — approvals + PPA + interconnection agreement; the P30/P0 tail 12-15 BCF/d by the early 2030s) can only be met out of working storage (~4 TCF) — which breaks materially below anything in history by mid-2028 and below all recorded evidence by 2029, making the gas price "convex and unbounded" ($8-10 starts shutting off spot LNG cargos; Russia-Ukraine and polar-vortex episodes are the only price precedents). Gas is >40% of US power and sets the marginal power price, so the pain lands as electricity prices in 2028-30 — the US consumer is the biggest loser, and the awful trade-off (exports vs AI vs consumer bills) feeds the NIMBYism already popping up. The tell is complacency: gas ~$3.50 with a flat curve to the mid-2030s, no rig response, EQT shutting in gas "for later," 2028 illiquid — the action starts within ~6 months as utilities roll forward and hedge 28, then "a knife fight to secure natural gas physical in 28 like we really haven't seen before." Longs: the best producers on trough multiples (EXE — ~70% of remaining core Haynesville at ~4× EBITDA, CEO-less and washed out; RRC — the highest-quality Appalachian upstream); solar yield-cos marking PPAs to much higher power prices for zero incremental capex (XIFR, CWEN) plus residential solar as the consumer's 10am-6pm hedge ("grows exponentially from here" even without tax incentives); the durable fix is large-scale AP1000 nuclear ~2033-34 (SMRs "science experiments") via CCJ (49% of Westinghouse — "deeply undervalued within Cameco"), BN (51%) and BWXT (US Navy nuclear, AP1000 dollar content). Losers: turbine/genset/fuel-cell makers adding capacity into the squeeze (CAT doubling Solar Turbines capacity by end-29 "at the exact wrong time," BE unable to source gas at 2 GW+ scale — fuel cells are backup gen, not baseload) and the hyperscalers, for whom energy goes from ~10% to 20-40% of compute cost by 2029 ("this sounds like DRAM two years ago" — slowly at first, then all at once). Fixes he'd deploy as czar: the US government building 2-4 AP1000s end-to-end to derisk the supply chain (China is building ~39 reactors, a third modeled on the AP1000), reinvigorated residential-solar incentives, and a 1-2 BCF/d Canada→MISO pipe (Canada has the deepest, richest economic gas resource, trapped behind pipe). Thesis-breaker: a step-function battery-technology change (sodium etc.) — "a watershed moment that I would welcome." Polomny (AIA free weekly, Jul 24): published Matt Smith's June letter in full — after 18 months of bottom-up work across producing basins, pipelines, processing, LNG exports and AI power projects, Smith argues US storage draws begin at unprecedented rates as early as 2028 and "by 2030, we are likely to exhaust working natural gas storage… the fuel everyone thinks is abundant is not… this does not portend a modest price increase." Polomny, as a generalist, can't refute it but keeps it on the radar — "my main thesis going forward is scarcity." 2026-SEP-10 (Toby McKenna, Rockpoint): the constraint is not just molecules but storage. ~20 Bcf/d of Gulf LNG exports, 5 Bcf into Canada and 0.5 Bcf to Mexico now ride on infrastructure built for heating load, and an LNG-linked customer reserving injection capability consumes roughly three times the space a conventional utility user does — so conventional share shrinks while volatility rises. The read-across template: the Gulf of Mexico took its first LNG cargo in 2014, showed nothing by 2015, and storage values there are now 300% of what they were. He calls AECO "on the precipice" of the same re-rating (explicitly not guidance). 2026-SEP-21 Hay: ~50 GW of gas-fired power under construction or planned ("some estimates are much higher"), plus Japan's ~18 GW of funded U.S. gas generation (9.2 GW in Portsmouth, Ohio), is the demand case for EQT as a Strong Buy.