Matthew Smith argues the US faces a historic natural gas shortfall from 2028, as AI data centers and LNG exports draw more gas than production and pipelines can deliver.
Smith says working gas storage could be exhausted by 2030, making upside price risk unbounded and convex, with electricity prices hit hardest in 2028 through 2030.
Smith names Expand Energy, which he says controls about 70% of remaining core Haynesville wells and trades at four times EBITDA despite having no CEO, as a winner, while US consumers will likely pay higher bills.
Natural gas shortage could hit the US by 2028
The US is heading toward a natural gas squeeze that the market has not priced. LNG exports are locked in by contracts and financing, and AI data centers add demand just as storage, processing and pipelines cannot keep pace. Smith concludes that prices face convex upside, and that producers and solar owners gain while consumers pay.
Takeaways
US natural gas faces a historic supply shortfall. Gas supplies over 40% of US power. Smith argues that from 2028 AI data centers and LNG exports will need more gas than the country can produce and deliver. Working storage of about 4 TCF could run out by 2030, and the market is tightening within six months.
Price upside is unbounded and convex. Storage falling below all historical levels by 2029 leaves no buffer. Smith says gas could structurally double or triple without any weather event, and electricity prices would be hit hardest in 2028 through 2030. Forward prices sit at only $3.50 to $3.60.
Delivery, not the resource, is the binding constraint. The resource in the ground is well mapped. The limits are depletion, processing that takes two or three years to build, gathering, and interstate pipelines. Only Mountain Valley has been completed in roughly a decade, so new supply cannot arrive quickly.
LNG exports cannot simply be stopped. Export capacity goes from about 15 Bcf a day to a scheduled 35 Bcf a day by 2030, while US production could add only about 20. Contract law, tens of billions in project financing, and allies relying on US gas make curbing exports unrealistic.
AI and distributed generation add large new demand. Smith's base case counts about 5 Bcf a day of AI-related demand, using only projects at 50% probability or better. An extreme scenario reaches 12 to 15 Bcf a day. Investors have put tens of billions into distributed power that will burn gas.
Gas producers and solar owners win; consumers lose. Smith favors Expand Energy, with about 70% of core Haynesville wells, four times EBITDA and no CEO, plus Range. Solar owners like XPLR and Clearway gain because fuel is free. Consumers and distributed generation makers like Bloom and Caterpillar lose.
Large-scale nuclear is the only long-term fix. Smith favors Westinghouse AP1000s, owned by Cameco and Brookfield. Vogtle 3 and 4 cost three times as much and took 15 years. He suggests the government build two to four using the $260bn Loan Programs Office. Small modular reactors remain unproven.
Remedies Smith proposes
Build a pipeline of 1 to 2 Bcf a day from Canada into the Midwest and wheel the gas around the country.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Have the government build two to four AP1000 reactors to de-risk the nuclear supply chain.
Encourage residential solar with batteries, which becomes economic as power prices rise.
Companies should lock in physical gas supply now, since hyperscaler energy costs could rise from about 10% to 20-40% of the cost of doing business.
Engineering and construction firms trading at 25 times cash flow should diversify beyond gas plant construction.