Matthew Smith — Natural Gas: The Next Bottleneck - [Invest Like the Best, EP.483]
What this covers
Patrick talks with Matthew Smith, founder and CIO of Chronometer Partners, whose team spent 18 months modeling nearly every US gas well, pipeline, and processing asset. The conversation starts with history. Shale abundance after 2010 made the US a surplus producer and opened the way for LNG exports. Smith's claim is that the squeeze was set in motion before AI arrived, because export projects are project-financed and permitted years ahead, and capacity is already heading from about 15 to 35 Bcf/d. He also separates flow from stock: the US is not short of gas in the ground, but of the rate at which it can be developed, processed, and moved.
The middle of the discussion works through the physical bottlenecks. Processing plants take two to three years to build. Almost no major interstate pipelines have been added recently. Canada has limited storage, and Permian gas depends on oil prices. Smith explains why electricity, not gas, is where the pain shows first: gas is the marginal fuel that sets power prices. He also discusses why cutting contracted LNG cargoes is hard, since foreign utilities depend on them.
The last stretch turns to responses, several of them contested. Smith favors large AP1000 reactors over SMRs, which he calls largely science experiments. He sets Vogtle's cost overruns against China's reactor buildout and calls for a government-led demonstration. He also argues that residential solar with batteries protects households from peak pricing, and that hyperscalers are underpricing gas supply risk, as the DRAM market did before its shortage.
Smith argues that US natural gas faces a structural deficit starting in 2028, as AI power demand and LNG exports draw down working storage, making upside price risk convex and unbounded. He concludes that the market is complacent and that large-scale nuclear and residential solar are the necessary responses.
- Pre-AI LNG export commitments already take US exports from about 15 to 35 Bcf/d by 2030, while modeled supply growth is about 20 Bcf/d
- AI gas-fired power adds about 5 Bcf/d in a base case and 12 to 15 Bcf/d in an extreme case, pushing storage to historic lows by 2029 to 2030
- Forward curves are flat in the 2028 to 2030 period because the market assumes gas abundance, so the risk is not priced
Tightening energy constraints raise the stakes for efficiency, prompting listeners to imagine solutions.
- Performance per watt is likely to become a compute metric that matters more as energy constraints tighten.
“Performance per watt is probably a compute metric that we're going to care more and more about.”
- The host hopes listeners will imagine solutions and create urgency so the US emerges more resilient, capable, and efficient.
“I hope, as the US has been very good at doing historically, that lots of people listen and start to imagine solutions”