Actionable insights — The 2028 Natural Gas Crisis

11:59 1. Bottom-up supply modeling — build the ceiling well by well

1. Don't trust a company's stated "years of inventory." Digitize the actual acreage each producer controls as polygon shapes (lat-longs), and associate every producing well with the block it sits in.
2. Apply known well-performance parameters (initial rate + decline curve) to the remaining undeveloped locations to compute what can logically be produced — then stack all wells across all companies to a system-wide maximum deliverability (a flow ceiling, ~128-132 BCF/d here), not a resource-in-the-ground stock number.
3. Separate the two questions that get conflated: "is there enough molecules?" (yes) vs "can enough flow to market in this timeframe?" (the binding one). The existing captured inventory develops out in ~4-5 years.
4. Pressure-test any bull's "plenty of resource" claim by asking a producer for exact engineered locations on a map plus surface infrastructure + the capital to actually flow it — inventory that can't reach a pipeline isn't supply.

• Producers whose drilling-inventory claims exceed what mapped, engineered, infrastructure-connected locations justify — a red flag for supply that never shows up.

6:06 2. Probability-weight the demand — a P50 base case with a fat tail

1. For every proposed power source that would consume the commodity, assign a probability of actually being built (with an outside domain partner). Define a hard bar for the P50 base case: has approvals + a signed PPA (a buyer under contract) + an interconnection agreement in hand or in process.
2. Sum only the P50-qualifying projects into the base-case demand — the number you'll actually underwrite (~5 BCF/d of credible compute gas demand here).
3. Then compute the tail: relax the bar to P30/P0 (proposed but unpermitted) and watch the number balloon — "multiples" of the base case (12-15 BCF/d). The gap between base and tail is the risk that the system can't source.

• Projects graduating from P0/P30 toward P50 (a PPA signed, an interconnection granted) — each one moves demand from the tail into the base case.

22:21 3. Make storage the scoreboard — trajectory vs the all-history band

1. Treat working storage as the single nexus where supply and demand net out (here ~4 TCF, with a seasonal high/low band — draw in summer + winter, build in spring/fall).
2. Run your supply ceiling and probability-weighted demand forward year by year and plot the projected storage path against the entire historical range — the question isn't the price, it's "when does the line break below anything ever recorded?"
3. Price the consequence off the closest analogs to a genuine shortage (Russia-Ukraine $8-10/MCF, polar-vortex spikes) while flagging that those were transitory — a structural draw has no natural ceiling, so model the price response as convex/unbounded rather than a point target.

• Injection-season builds that fail to refill the band; the first year the storage trajectory dips under the historical low envelope.

21:04 4. Read the flat forward curve as the opportunity, not the truth

1. When your bottom-up work says a squeeze is coming, check whether the market agrees by looking at the forward curve and the rig count — a flat curve at a low spot price ($3.50, flat to the mid-2030s) plus a flat rig count means "no one is on it."
2. Diagnose why the curve is asleep: the back years (28+) are illiquid, so there's no flow to move them until a natural hedger (utilities) shows up — the mispricing is a plumbing artifact, not a considered view.
3. Corroborate with behavior that only makes sense if the crowd is complacent — producers shutting in gas "to save it for later," capital pouring into short-life assets, no one contracting forward supply.
4. Buy the asset the curve is mispricing while it's cheap (a producer at a trough multiple), sized to a re-rate you expect within ~6-12 months as the back curve wakes up — not a multi-year wait.

• The first utilities to start hedging/buying 28 gas — that's when the illiquid back years move; forward curves "up materially" within ~6 months.

26:21 5. Follow the marginal fuel — who gets a windfall when the input re-prices

1. Start from the market-clearing rule: in each power market the last, most expensive dispatched plant sets the price for everyone. Gas is that marginal fuel, so "as gas goes, power prices go."
2. Screen the dispatch curve for assets whose fuel is free and fixed (solar, then wind/hydro) — when the marginal price rises, their margin expands with no incremental capital.
3. Find the cleanest financial expression: owners that mark contracts to market at the higher price (utility-scale yield-cos re-pricing PPAs) capture the windfall without capex.
4. Extend the logic to the end consumer for a defensive analog — a household solar system is a hedge on the same peak-price risk (10am-6pm).

• Rising forward power prices in specific ISOs (PJM, ERCOT, MISO); yield-cos with near-term PPA re-pricing windows into that strength.

34:47 6. Read every power press release backwards — "think more gas"

1. Translate each "bring-your-own-generation" (BYOG) data-center or genset announcement into its fuel demand: a fuel cell or turbine order literally adds gas demand on top of your base case — so a press release from a genset/fuel-cell maker is a demand data point, "think more gas."
2. Use each unit's published energy efficiency (heat rate) to quantify exactly how much incremental gas it pulls, and add it to the model — don't treat these as "solutions," treat them as demand.
3. Then flip it to a short screen: the makers of that hardware are the tape's biggest recent winners and are adding capacity into the squeeze — a boom/bust echo of the early-2000s gas-plant overbuild. Fade the ones whose product stops making sense once gas is expensive (backup-only economics deployed as baseload).

• Turbine/genset order books still growing into 2028-29; the point where large-scale gas-gen orders "slow very meaningfully" as fuel-cost doubts set in.

54:33 7. The physical-supply / counterparty checklist — the DRAM "short-memory" analog

1. Adopt the analog: an under-invested capacity input goes "slowly at first, then all at once" (like DRAM) — the risk isn't gradual, it's a step-change once the illiquid part of the market wakes up.
2. For any business with the commodity as an input, run a three-question due-diligence: (a) do you know exactly where your physical gas comes from? (b) is physical supply locked up (not just financially hedged)? (c) who is your counterparty, and can they still deliver when the thing they're short moves violently ("imagine being short memory 18 months ago")?
3. Stress the model at the tail price, not the curve: budget for gas at $10+ (not $3.50) and re-run the cost stack — energy that's ~10% of hyperscaler cost today becomes 20-40% by 2029 if gas doubles/triples.
4. If the answer to any question is weak, the actionable output is defensive: demand a backup plan, favor names with secured supply, and treat unsecured backup-gen deployers as impaired.

• Utilities/hyperscalers beginning to contract forward gas supply; any counterparty visibly long capacity it can't physically source.

39:29 8. Pre-empt your own bull case — enumerate the disconfirming risks

1. After the model is built, go on a "listening tour" of the domain experts specifically to have them poke holes, then bake each common pushback into the base case so it can't surprise you.
2. Neutralize the obvious ones structurally: the "Permian has plenty of gas" pushback is answered by already including the 7+ BCF/d of Permian pipelines coming 26-30 and by noting more Permian gas needs sustained higher oil (which worsens the consumer thesis, not helps it).
3. Isolate the one risk you genuinely can't handicap — a step-function technology change (here, a battery breakthrough: sodium etc.) — and label it as the true kill-switch you'd "welcome," monitoring it explicitly rather than pretending it's priced.

• Commercialization of a non-lithium, long-duration battery chemistry at scale — the one development that could dissolve the deficit.