Henry Hub natural gas annual-average spot crosses each of $4 / $5 / $7 / $10 per MMBtu

---
title: Henry Hub natural gas annual-average spot crosses each of $4 / $5 / $7 / $10 per MMBtu
status: draft
dimensions: ["utilities","commodities","energy","housing"]
horizon: medium
trigger: Henry Hub natural gas spot price annual average (calendar year) crosses each of {$4, $5, $7, $10} per MMBtu, per EIA STEO retrospective data.
timeline: {"p10":2028,"p50":2031,"p90":2038}
confidence: medium
sub_gates: [{"slug":"natgas-annual-avg-4","p50":2028,"why":"Likely 2027-2029. EIA May 2026 STEO has 2026 at $3.50 and 2027 at $3.18 (revised down). Crossing $4 annual average requires the LNG demand pull (Plaquemines+CC Stage 3+Golden Pass all fully ramped, ~20 Bcf/d export rate) plus an AI data-center demand pulse of 2-3 Bcf/d to outrun the Permian associated-gas + Haynesville response. WoodMac sees $5 by 2030, implying $4 likely crosses 2-3 years earlier on the path. Cold winter (Jan 2026 hit $7.72 monthly) could pull a single year average to $4 even without structural pull."},{"slug":"natgas-annual-avg-5","p50":2031,"why":"Median analyst target (WoodMac 2030 SPO base case at $5/MMBtu). Requires LNG exports above 22 Bcf/d (Rio Grande, Port Arthur, Plaquemines Phase 2 online), AI gas demand at 4-6 Bcf/d, and Haynesville growth tapering. WoodMac thesis: marginal supply shifts from cheap Permian associated gas to more expensive non-associated Haynesville/Marcellus drilling, driving higher clearing price. JPM near-term consensus ($3.50 sufficient) lags; Goldman/RBC mid-decade picks up the bull case."},{"slug":"natgas-annual-avg-7","p50":2034,"why":"Bull case requiring sustained demand surge (AI hyperscaler buildout running full WoodMac scenario, 30 Bcf/d total US demand growth by early 2040s pulled forward) AND a supply shock (Hormuz-style LNG disruption permanent, Russian gas full phase-out by end-2026 forcing more Atlantic LNG, Qatar Ras Laffan capacity scarred 5+ years). 2022 averaged $6.45 on Ukraine war alone — $7 sustained for a year needs that magnitude of shock plus AI demand. Last $7+ annual print was 2008 ($8.85)."},{"slug":"natgas-annual-avg-10","p50":2042,"why":"Tail scenario — combined supply shock + sustained demand surge; reminiscent of 2008 ($8.85) or 2022 ($6.45) but sustained for an entire calendar year. Would require a Persian Gulf LNG corridor permanently disrupted, Russian gas zero, US production capped (water/permit/pipeline), AND data center demand at IEA Lift-Off scenario. Approximately the price level where coal-to-gas switching unwinds (coal becomes economic again) and demand destruction in industrial sector kicks in hard — economically self-limiting near this level."}]
history: [{"date":"2026-05-25T00:00:00.000Z","p10":2028,"p50":2031,"p90":2038,"why":"Initial estimate from initial research. Canonical tier $5 is anchored to Wood Mackenzie 2030 SPO base case + analyst consensus reset post-AI-data-center-demand revelation. P10 (2028) requires a cold winter + AI/LNG pull-forward; P90 (2038) requires WoodMac thesis fully wrong, gas demand growth absorbed by record Permian/Haynesville production at sub-$4 clearing. $4 tier P50 = 2028 (likely 2027-2029 path); $7 tier P50 = 2034 (bull case requires sustained shock); $10 tier P50 = 2042 (tail scenario, demand-destruction self-limiting near this level)."}]
cross_gate: [{"other":"residential-solar-storage-0.04","relation":"competes","strength":"medium","note":"Cheap residential solar+storage erodes peak-shaving and behind-the-meter gas demand. If rooftop solar+battery delivers <$0.04/kWh, distributed gas peakers face structural displacement. But effect on gross gas demand is small at residential level — utility-scale solar+storage matters more, and even then the dispatchable-baseload value of CCGT for AI data centers (need 99.99% uptime) is largely solar-irreducible. Weakens this gate's $7-$10 bull tiers materially, $4-$5 tiers marginally."},{"other":"ai-agent-30pct-knowledge-work","relation":"enabled_by","strength":"strong","note":"AI compute scaling drives data center electricity demand; gas peakers and CCGTs are the cheapest fast-permit option for new AI capex. Meta Hyperion Louisiana (10 gas plants, 7 GW, $11B in plants alone — March 2026 deal); xAI Memphis/Southaven (46 portable + 41 permanent turbines, 1.2 GW); McKinsey: AI = 70% of new data center capacity by 2030; Berkeley Labs: 74-132 GW demand growth by 2029. If AI demand fires (P50 2029 on agent gate), gas demand pull is real. If AI capex stalls (P90 2034 on agent gate), $5+ tiers slip out 2-4 years."},{"other":"smr-first-oecd-deployment","relation":"competes","strength":"weak","note":"SMRs would eventually displace some gas-fired generation for AI data center baseload (Meta/Oklo, Google/Kairos, Amazon/X-Energy hyperscaler PPAs). Very slow displacement curve — first OECD SMR P50 = 2032; meaningful capacity 2035+. Gas peakers stay in the mix for load-following + reliability backup. Substitution effect is real but mostly post-2035."},{"other":"commodity-wti-downside","relation":"correlates","strength":"weak","note":"WTI downside (back toward $50-60) means less Permian drilling, less associated gas, modestly bullish Henry Hub. But effect is smaller than the demand-side LNG/AI pull. EIA Feb 2026 STEO modeled WTI at $53 (2026) / $49 (2027) and saw HH at $4.31/$4.38 — i.e., low oil correlated with modestly higher gas — but the May 2026 update with higher Permian GOR has gas at $3.50/$3.18 even with elevated oil. The pure WTI signal is weak."},{"other":"global-economy-explosive-growth","relation":"enabled_by","strength":"medium","note":"Higher GDP growth = higher industrial + residential energy demand + faster electrification. The 30 Bcf/d total US gas demand growth WoodMac assumes for early 2040s requires sustained 2%+ real GDP growth. If explosive-growth fires (AI-driven productivity), AI compute demand pulls gas; if growth disappoints, both LNG export utilization and US power demand decelerate."},{"other":"humanoid-10m-households","relation":"enabled_by","strength":"weak","note":"Household humanoid robots add marginal residential electric demand; AI cloud inference for humanoid brains adds data-center demand. Long-cycle, small near-term effect on gas."},{"other":"autonomous-resource-frontier-positive-roi","relation":"competes","strength":"weak","note":"Frontier mining/space resource extraction could expand US/global metals supply, lowering renewable-build cost and indirectly displacing gas peakers. Multi-decade horizon; not material."}]
key_dependencies: [{"factor":"AI hyperscaler data center buildout","kind":"gate","direction":"accelerates","linked_gate":"ai-agent-30pct-knowledge-work","impact":"AI compute scaling drives 3-12 Bcf/d of incremental gas demand by 2030 (xAI, Meta Hyperion); $5-$7 tiers are conditional on this demand firing on schedule."},{"factor":"LNG export capacity ramp 2026-2028","kind":"market","direction":"accelerates","linked_gate":null,"impact":"Plaquemines, Corpus Christi Stage 3, Golden Pass, and Rio Grande LNG together add ~6+ Bcf/d by end-2027, pulling US production toward structural tightness needed for $4-$5 tiers."},{"factor":"Permian associated gas supply response","kind":"market","direction":"delays","linked_gate":"commodity-wti-downside","impact":"Record Permian production growth (+6-10% YoY) and new pipeline capacity absorbs demand pull, keeping Henry Hub suppressed at $3.18-$3.50 through 2027 and pushing $4 tier into 2028."},{"factor":"Residential solar and storage cost crossover","kind":"gate","direction":"delays","linked_gate":"residential-solar-storage-0.04","impact":"Sub-$0.04/kWh solar+storage erodes peak-shaving and behind-the-meter gas demand; most damaging to $7-$10 bull tiers, marginal effect on $4-$5 modal scenario."},{"factor":"SMR deployment for data center baseload","kind":"gate","direction":"delays","linked_gate":"smr-first-oecd-deployment","impact":"First OECD SMR P50 2032 with meaningful hyperscaler displacement by 2035+; limits structural gas demand in the $7-$10 tail tiers but minimal effect on near-term $4-$5 path."},{"factor":"European gas market tightness and TTF-JKM premium","kind":"market","direction":"accelerates","linked_gate":null,"impact":"EU storage at 4-year lows, TTF at $18+/MMBtu, and Russian phase-out pull US LNG cargoes Atlantic, sustaining high export utilization and tightening Henry Hub supply-demand balance."},{"factor":"Gas turbine supply chain bottleneck","kind":"market","direction":"delays","linked_gate":null,"impact":"CCGT costs up 66%, turbine prices up 195% from 2019 with waitlists into early 2030s; physically constrains how fast data-center gas demand can materialize, slipping demand-side bull case 2-4 years."}]
external_calibration: {"metaculus":null,"manifold":null,"expert_consensus":"EIA STEO May 2026: 2026 $3.50 / 2027 $3.18 (down 4.4% / 11.5% from April; Permian GOR-driven). EIA Feb 2026 STEO had projected 2026 $4.31 / 2027 $4.38 (pre-Permian-pipeline assumption). Wood Mackenzie May 2025 SPO: $5.00/MMBtu by 2030, $6.00 by 2035 (real terms). BNEF Jan 2026 forecast: range-bound $3.30-$3.45 through 2026-2033 (lower for longer thesis). Rystad Dec 2025: 2026 avg $4.30, 2027 flat. J.P. Morgan: $3.50/MMBtu sufficient to clear AI+LNG demand. Haynes Boone Fall 2025 Bank Price Deck: $3.19-$3.23 long-term consensus. Forward curve (NYMEX) flat at $3.26 through end-2026, $4.00+ Dec 2026 contract — backwardation reversal pricing winter risk.","market_size":"Total US natural gas market ~$300-400B/yr at $3.50/MMBtu × 90-95 Bcf/d × 365. Global gas+LNG trade ~$500B/yr. Henry Hub is the benchmark for ~$1T+ of forward contract volume; a $1/MMBtu sustained move translates to ~$100B/yr global welfare transfer between producers and consumers."}
last_updated: "2026-05-25T00:00:00.000Z"
sources_count: 37
---

## TL;DR

I put the canonical **$5 annual-average P50 at 2031** — about 5 years out from today (May 25 2026) — that Henry Hub spot price averages $5/MMBtu or above for a full calendar year. The thesis: the EIA May 2026 STEO has 2026 at **$3.50** and 2027 at **$3.18** (revised down 4.4% / 11.5% from April on Permian gas-to-oil-ratio (GOR) surprise + cheap associated-gas response), but Wood Mackenzie's spring-2025 Strategic Planning Outlook flags a structural inflection from data-center + LNG demand that lifts Henry Hub to **$5/MMBtu by 2030 and $6 by 2035** (real terms) [1][2]. The Big Three drivers — Plaquemines/Corpus Christi 3/Golden Pass/Rio Grande LNG ramp, AI data-center direct gas peakers (xAI Memphis/Southaven 1.2 GW, Meta Hyperion Louisiana 7 GW), and OECD coal-to-gas retirement — collectively pull 4-8 Bcf/d of incremental demand into a market currently at ~120 Bcf/d production. The supply response (Haynesville +1.2-1.6 Bcf/d/yr, Permian associated gas +1.4-2.0 Bcf/d/yr) clears the early-decade pull, but the late-decade marginal supply requires deeper Haynesville drilling at $5+ break-evens [3][4]. **Confidence: medium** — directional thesis is well-supported, but timing is highly sensitive to (a) AI buildout pace, (b) LNG export utilization, (c) Permian pipeline build-out, and (d) winter weather (Jan 2026 monthly hit $7.72 on Winter Storm Fern alone).

| Tier | Drivers | P50 year | Status as of 2026-05 |
|------|---------|----------|----------------------|
| $4   | LNG ramp + AI demand pulse + cold winter | 2028 | Open; 2026 tracking $3.50 |
| $5   | WoodMac base case; sustained AI+LNG pull | 2031 | Open; canonical tier |
| $7   | Bull case; demand surge + supply shock | 2034 | Open; last seen 2008 ($8.85 ann avg) |
| $10  | Tail; combined shock + demand-destruction limited | 2042 | Open; no historical print |

## Current state (as of 2026-05-25)

**Spot**: Henry Hub averaged **$2.82/MMBtu on May 11 2026** (most recent EIA weekly print) [5]. Daily prints range $2.70-$2.92 through May. Mid-May 2026 cash spot reclaimed $3 on East Coast heat wave; June 2026 NYMEX futures pushed to a 7-week high above $3 on heat + LNG demand + Iran-war headlines [6][7].

**Monthly progression 2026**:
- Jan 2026: $7.72/MMBtu (cold + Winter Storm Fern) [8]
- Feb 2026: $3.62
- Mar 2026: $3.04
- Apr 2026: $2.77
- May 2026 MTD: ~$2.80-2.92

**Storage**: End-of-winter (March 2026) inventories at **1,908 Bcf** — 4% above the five-year average, despite the 2,020 Bcf withdrawal (4% more than the 5-year average) [3]. EIA projects end-October 2026 inventory will be **7% above the 5-year average** on rising production and weaker-than-expected demand. Storage glut is the proximate reason for the EIA's May STEO downgrade.

**Production**: Lower-48 marketed production averaged **117.2 Bcf/d in 1Q26** (+4% YoY); EIA forecasts **118.9 Bcf/d in 2026, 124.0 Bcf/d in 2027** [3][9]. Permian gas production projected to grow 10% in 2027 as new pipeline capacity (Aspen Katy Hub 3 Bcf/d, Matterhorn Express expansion) alleviates the Waha Hub negative-price phenomenon (Waha averaged below zero 8 of last 9 months) [3].

**LNG exports**: ~17 Bcf/d in 2026 (EIA), rising to 18 Bcf/d in 2027 [10]. Plaquemines (Venture Global, 3.8 Bcf/d full Phase 1+2) ramping; Corpus Christi Stage 3 (Cheniere, ~1.4 Bcf/d incremental) Train 6 in commissioning May 2026, Train 7 expected fall 2026 [11][12]; Golden Pass LNG (Exxon/QatarEnergy, 2.1 Bcf/d) shipped first cargo April 20 2026, ramping through 2027 [13]; Rio Grande LNG Phase 1 (NextDecade, 2.1 Bcf/d) targeting first LNG H1 2027 [10][14].

**Forward curve**: NYMEX strip averages **$3.26/MMBtu through end-2026** with December 2026 contract above $4 — backwardation pricing winter risk premium [15]. 2027-2030 strip flat at $3.30-$3.50, NOT pricing in the WoodMac bull case.

## Demand drivers

**1. AI data center direct natural gas demand (the breakaway driver).** Multiple credible 2026 datapoints:

- **xAI Memphis/Southaven**: Operating 46 portable gas turbines (~1.2 GW combined) at Colossus 1 (Memphis) + Colossus 2 (Southaven, MS); permitted for an additional 41 permanent turbines [16][17][18]. Powers Anthropic's full Colossus-1 lease announced May 2026 + xAI's Grok training. Currently subject to NAACP lawsuit over unpermitted "temporary" turbines (Clean Air Act violations, motion for preliminary injunction May 6 2026) [19][20].
- **Meta Hyperion Louisiana**: 10 gas-fired power plants funded by Meta via Entergy, totaling **7+ GW** for the $27B AI campus in Richland Parish [21][22]. Power plants alone cost ~$11-16B; gas plants would increase Louisiana's grid by **30%+**; once built, 90% of Louisiana electricity will be gas-generated. Approved on accelerated 8-month review (LPSC April 15 2026, 4-1 vote). Fortune calls it "the largest single power request in state history."
- **East Daley / S&P Global / Wells Fargo aggregate**: midstream companies forecast **3-12 Bcf/d incremental gas demand from data centers by 2030**; S&P Global base case is **3-6 Bcf/d**; bull case 10-12 Bcf/d [23].
- **Berkeley Labs**: US data center electricity demand could rise **74-132 GW by 2029** (up to 12% of US consumption) [23].
- **IEA Energy & AI April 2025**: global data center electricity consumption doubles to 945 TWh by 2030; natural gas adds **175 TWh of generation** to meet data center demand globally [24]. US share dominant.
- **JPMorgan modeling**: AI power demand alone adds **1.4 Bcf/d by 2027 and 6.2 Bcf/d by 2030** [25].
- **Behind-the-meter "hidden" gas demand** (Oxford Smith School Apr 2026): on-site gas turbines for data centers do NOT show up in EIA electric-sector statistics (classified as industrial); could be **0.5-2.0 Bcf/day by 2028** of off-grid demand invisible to standard reporting [26].

**2. LNG export capacity ramp.** The "wave" through 2026-2028 [10][11][12][13][14]:

- Plaquemines LNG (Venture Global, ~3.8 Bcf/d Phases 1+2): Phase 1 fully ramped 2025; Phase 2 advancing through 2026.
- Corpus Christi Stage 3 (Cheniere, +10 Mt/y / ~1.4 Bcf/d): Train 6 producing LNG May 2026, Train 7 expected fall 2026 — full 25 Mt/y capacity by end-2026.
- Golden Pass LNG (Exxon/QatarEnergy, ~2.1 Bcf/d): Train 1 first cargo April 20 2026; ramping through 2027.
- Rio Grande LNG (NextDecade, ~2.2 Bcf/d Phase 1): First LNG H1 2027.
- Port Arthur LNG (Sempra, ~1.8 Bcf/d Phase 1): late 2027-2028; Sempra requesting cooldown cargoes May 2026.
- Combined nominal incremental: **~6+ Bcf/d by end-2027** vs 2025 baseline of 15 Bcf/d.
- EIA AEO 2026: US LNG capacity rises from 14.9 Bcf/d (2025) to **27.7 Bcf/d by 2030** [27].

**3. European replacement of Russian pipeline gas + Iran-war LNG disruption.** EU emerged from winter 2025-26 with storage at **27.7% full on March 31 2026** — 4-year low [28]. Goldman Sachs **lifted Q2 2026 TTF target to €72/MWh** (~$22/MMBtu) on cold weather + Qatar Ras Laffan disruption from Hormuz [29]. JKM-TTF spread blew out to $3/MMBtu in March, settled into $1.5-2/MMBtu range in May, pulling US LNG cargoes Atlantic→Pacific. European traders **bidding €100/MWh ($30/MMBtu) on winter 2026-27 TTF options** [30]. Russian gas full phase-out target end-2026 maintains structural demand pull. Effect on Henry Hub: indirect through LNG arbitrage (US producers gain higher netbacks → more drilling), but US domestic prices remain anchored by Permian + Haynesville supply response.

**4. Coal-to-gas retirement (OECD).** Coal share of US generation: 17% (2025) → 16% (2026) → 15% (2027) per EIA STEO [31]. Each 1% coal retirement = ~0.5-1.0 Bcf/d incremental gas demand. Globally, China + India coal expansion offsets OECD retirements at gross level, but the Atlantic basin gas pull is real.

**5. Cold-winter optionality.** January 2026 monthly average of $7.72 (Winter Storm Fern) demonstrates that a single cold winter can pull annual average toward $4 even absent structural demand growth. The 2008 record annual average of $8.85 was a hot-summer + cold-winter + Hurricane-Gustav combo; 2022's $6.45 was Russia-Ukraine driven. Single-year tail risk is meaningful for the $4 tier.

## Supply drivers

**Permian associated gas growth (the bearish counterweight).** Permian gas production projected at **29.2 Bcf/d in 2026 (+6% YoY) and ~32 Bcf/d in 2027 (+10%)** per EIA May 2026 STEO [3][9]. Key factors:

- Permian rig count up to ~415 (highest since November 2025) on war-elevated oil [32].
- Gas-to-oil ratios continuing to rise — wells producing more gas as fields mature.
- Pipeline relief: Aspen Katy Hub (3 Bcf/d), Matterhorn Express expansion, and Whistler ADCC (1.7 Bcf/d to Corpus Christi) come online 2026-2027.
- Waha Hub negative pricing (averaged sub-zero 8 of last 9 months) is the symptom; new pipeline capacity is the cure. Once Permian gas can move, it lifts Henry Hub from below (more supply) but also relieves the Permian bottleneck (better netbacks).

**Haynesville (the LNG-export response).** Production growth: +1.2 Bcf/d in 2026 (to 15.6 Bcf/d, +8.3%) and +1.6 Bcf/d in 2027 (to 17.1 Bcf/d, +10.3%) [33][34]. Haynesville is the price-elastic basin — operators (Comstock, Aethon, Expand Energy) cut activity below ~$3.50 and ramp above $4. The 2026-2027 ramp assumes the EIA's Feb-2026 $4.31/$4.38 price (which the May STEO has since revised down to $3.50/$3.18); at $3.18 sustained, Haynesville growth slows materially [33].

**Marcellus / Appalachia (the structurally cheap but pipe-constrained basin).** Production growth: +0.3 Bcf/d 2026, +0.5 Bcf/d 2027 to ~37 Bcf/d (still ~32% of total US dry gas). Mountain Valley Pipeline online since June 2024 added takeaway. Future growth pipeline-bound — every new project (MVP Boost, Williams REA Project, Enbridge New England expansion) faces 2-4 years of permitting friction. ~970 trillion cubic feet of technically recoverable resource sits in Marcellus alone — supply is not the constraint, infrastructure is.

**Total US production trajectory**: marketed gas 116 (2025) → 119-121 (2026) → 124-126 (2027) Bcf/d. **Record highs both years.** The supply-demand math: net demand growth +0.6 Bcf/d (2026) and +2.5 Bcf/d (2027) per EIA Jan 2026 base case [10] — but the May 2026 revision lifts supply faster (+1.1 Bcf/d 2026, +2.6 Bcf/d 2027) so the storage glut persists.

**LNG bottlenecks / pipeline capacity constraints.** Permian gas needs to reach the Gulf Coast LNG terminals; pipeline capacity is binding. ADCC Pipeline (Whistler JV, 1.7 Bcf/d Agua Dulce → Corpus Christi) operational. Aspen Katy Hub 3 Bcf/d FID May 2026 [12]. Permian-Gulf Coast capacity expected adequate by end-2027, then bottleneck moves to Haynesville Gulf Coast lateral capacity.

## Per-tier reasoning

- **$4 (P50 2028)**: From a $3.50 (2026) / $3.18 (2027) base, $4 requires ~25% upward revision. Drivers: (a) LNG exports above 20 Bcf/d sustained, (b) AI data center direct gas demand at 2-3 Bcf/d (xAI + Meta + early hyperscaler peakers fully online), (c) coal retirement at ~1 Bcf/d/yr pace, (d) one cold winter pull. Realistic 2027-2029 path; 2028 is the modal year. Single-year tail risk (e.g., another Winter Storm Fern + Hormuz LNG redirect) could pull 2026 itself to $4 if Q2-Q4 prints firm to $3.50-$4.00 range. **The EIA's own pre-Permian-revision Feb 2026 STEO had 2026 at $4.31** — so the gate is genuinely on the knife's edge for the next 12-24 months.

- **$5 (P50 2031, canonical tier)**: Wood Mackenzie's May 2025 Strategic Planning Outlook base case — "Henry Hub prices will rise from the current $3.50 to average $5.00/MMBtu by 2030, and $6.00 by 2035" [1]. WoodMac thesis: gas producers require $5+ to incentivize sustained non-associated drilling once Permian associated-gas growth flattens (~2030). This aligns with: total US gas demand +30 Bcf/d by early 2040s (vs WoodMac old base case +13 Bcf/d); LNG exports at 27-30 Bcf/d; data center demand at 4-8 Bcf/d (S&P/East Daley range); coal at ~12% of generation (down from 17%). The $5 print stays sticky once it crosses — long-cycle structural rebalancing, not a single-event spike.

- **$7 (P50 2034)**: Bull case. Last $7+ annual average was 2008 ($8.85, post-Hurricane Gustav + oil-driven energy demand). 2022's $6.45 was Russia-Ukraine driven. To average $7 for a full year requires (a) WoodMac late-decade scenario fully realized, (b) Asia LNG pull persistent (Qatar Ras Laffan permanently scarred OR Iran war persistent), (c) AI demand at IEA Lift-Off case ($1,008 TWh global data center demand 2030 vs 945 base), (d) Permian supply growth tapped by 2032 (Tier-1 Permian inventory exhausts), (e) Haynesville growth requiring $6+ to incentivize ultra-deep drilling. Plausible early-to-mid 2030s.

- **$10 (P50 2042)**: Tail. No historical year has averaged $10/MMBtu. Annual peak was 2008's $8.85. To reach $10 annual requires (a) sustained Persian Gulf LNG disruption (Qatar offline 5+ years), (b) Russian gas zero permanent, (c) US production capped (water-supply / Permian Tier-2 / permit) at <130 Bcf/d, (d) AI demand at IEA Lift-Off + Energy Bottleneck Resolved case. **Self-limiting near this level**: coal-to-gas switching reverses (coal becomes economic), industrial demand destruction kicks in hard (chemicals, fertilizers, glass), residential switching to heat pumps accelerates, LNG export economics break down (US Henry Hub at $10 + liquefaction $4 + shipping $1 = $15+ delivered, vs Qatar marginal cost ~$2). Likely later-2030s+ event, if ever.

## Counter-arguments

1. **Record US production absorbs the demand pull.** EIA forecasts 124 Bcf/d in 2027 (record high); Rystad sees 130+ Bcf/d by Dec 2027. WoodMac itself notes "ample gas resource" — the bull case requires producers to NOT respond. If Haynesville/Permian operators ramp aggressively at $4-5 price signals (as Comstock/Aethon publicly signaled), the bull supply-shortage thesis breaks. JPM consensus is explicit: **$3.50/MMBtu sufficient to clear AI+LNG demand**.

2. **AI data center demand may be overstated.** McKinsey says AI = 70% of new data center capacity by 2030, but Industrial Info Resources tracks **4,500 announced data center projects worldwide ($2.63T)** with widespread expectation that "not all will be built as planned." Hyperscaler capex commitments are real, but realized GW-on-the-ground frequently underdelivers (utility queues, gas turbine shortages — turbines now have waitlists into early 2030s [35]). IEA Headwinds case: data center demand 833 TWh (2030) vs base 945 TWh — 12% lower.

3. **Gas turbine bottleneck slows the gas demand pull.** BloombergNEF reports gas turbine prices **up 195% from 2019**; CCGT plant cost up 66% in 2 years to $2,157/kW; build time +23% [35]. Turbine shortage actually constrains how fast data centers can take gas demand. If the supply chain doesn't keep up, the demand-side bull case slips out by 2-4 years.

4. **Residential solar+storage and SMR substitute over the long run.** Cheap solar+storage at <$0.04/kWh delivered (2033 P50) erodes peak-shaving gas demand. SMRs at $80-150/MWh FOAK (2030-2032 first OECD deployment) attack data center baseload. Combined post-2032 effect: 2-5 Bcf/d of gas demand displacement vs counterfactual. Material but slow.

5. **Renewables uptake.** EIA STEO May 2026: solar share of US generation rises from 7% (2025) → 8% (2026) → 9% (2027); wind 11% → 11% → 12%. Combined VRE share goes from 18% → 21% by 2027. The structural decarbonization of the grid (driven by IRA-era tax credits, lower-cost solar, BNEF utility-scale solar LCOE forecast falling 30% by 2035) is a chronic headwind to gas-fired generation growth.

6. **AI buildout slowdown.** A credible AI capex pullback (training plateau, regulatory clamp, or recession) shaves 3-8 Bcf/d off late-decade gas demand. The agent gate's P90 = 2034 implies meaningful probability mass on AI capability stalling — which directly feeds through to gas demand.

7. **Demand destruction at $7+**. The 2022 European industrial demand destruction (-90 bcm since 2021) was concentrated in chemicals, fertilizers, ceramics, glass. At Henry Hub $7+, similar US industrial demand destruction kicks in (fertilizer plants idle, chemical chains close), self-limiting upside. The $10 tier is mathematically reachable only on a permanent supply-shock + sustained AI/LNG pull combination.

## Cross-gate dependencies

**Strong enabler — `ai-agent-30pct-knowledge-work`**: The single largest demand-side driver. Hyperscaler AI capex (Meta Hyperion 5 GW IT load + 7 GW gas; xAI Colossus 2 GW; Anthropic Colossus-1 lease) is justified by AI training + inference scaling. If AI agent capabilities cross 30% knowledge-work threshold (P50 2029), Meta/Google/Amazon/Microsoft/xAI capex commits stay sticky → gas demand pull is real. If AI plateaus (P90 2034), gas demand decelerates 2-3 Bcf/d below WoodMac base case. **The $5-$7 tiers of this gate are conditional on AI demand firing.**

**Medium substitutes — `residential-solar-storage-0.04`**: Cheap residential solar+storage erodes behind-the-meter gas demand and shaves peak-load gas dispatching. But the dispatchable-baseload role of CCGT for AI data centers requires 99.99% uptime, which solar+storage doesn't yet provide at scale. Substitution is real on the $7-$10 tail but weaker on the $4-$5 modal scenario.

**Weak substitutes — `smr-first-oecd-deployment`**: SMRs eventually displace gas baseload for data centers (Meta/Oklo, Google/Kairos, Amazon/X-Energy, MSFT TBD all explicitly target hyperscaler PPAs). First OECD SMR P50 2032; meaningful displacement 2035+. Mostly post-2035 effect — slips the $7-$10 tiers out, modest effect on $4-$5.

**Medium correlate — `commodity-wti-upside`**: High WTI = more Permian drilling = more associated gas = bearish Henry Hub at the margin. The May 2026 EIA STEO downgrade ($4.31 → $3.50 for 2026) was directly attributable to Permian GOR upside from war-elevated oil drilling. Counter-direction: oil-driven inflation spillover into all-energy prices, JKM/TTF pull on US LNG arbitrage. Net 2026 effect on Henry Hub: slightly negative (supply-side dominated). 2027+ effect could flip if war-elevated oil persists AND Permian Tier-1 exhausts.

**Medium enabler — `global-economy-explosive-growth`**: GDP growth elasticity of energy demand ~0.6-0.7. WoodMac's 30 Bcf/d demand growth-by-2040s requires sustained 2%+ real GDP. Pre-IRA EIA modeling was at 13 Bcf/d (lower GDP path). Explosive-growth feedback into gas demand is real but indirect.

## Evidence and sources

1. [Wood Mackenzie — US Gas Sector Set to Benefit as Henry Hub Prices Poised to Climb](https://www.pgjonline.com/news/2025/june/wood-mackenzie-us-gas-sector-set-to-benefit-as-henry-hub-prices-poised-to-climb) — Alex Beeker/Dulles Wang, May 2025: $5.00 by 2030, $6.00 by 2035. Accessed 2026-05-25.
2. [Wood Mackenzie — North America gas strategic planning outlook 2050](https://www.woodmac.com/news/opinion/north-america-gas-strategic-planning-outlook-a-glimpse-into-2050/) — 30 Bcf/d demand growth by early 2040s (vs old 13 Bcf/d); HH to $6/Mcf in 2040s. Accessed 2026-05-25.
3. [EIA Short-Term Energy Outlook May 12 2026](https://www.eia.gov/outlooks/steo/) — 2026 $3.50 / 2027 $3.18; 2026 production 118.9 Bcf/d, 2027 124.0 Bcf/d; storage 1,908 Bcf end-March; Permian +6% in 2026 / +10% in 2027. Accessed 2026-05-25.
4. [Rigzone — USA EIA Lowers Henry Hub Price Forecast for 2026, 2027](https://www.rigzone.com/news/usa_eia_lowers_henry_hub_price_forecast_for_2026_2027-20-may-2026-183738-article/) — May 20 2026: 4.4% / 11.5% cuts; Q2 2026 $2.83/MMBtu; Q4 2026 $3.31; Q4 2027 $3.32. Accessed 2026-05-25.
5. [FRED — Henry Hub Natural Gas Spot Price (DHHNGSP)](https://fred.stlouisfed.org/series/DHHNGSP) — May 11 2026 print $2.82/MMBtu; May 8 $2.75. Accessed 2026-05-25.
6. [NGI — Spot Natural Gas Prices Extend Gains as Henry Hub Climbs](https://naturalgasintel.com/news/spot-natural-gas-prices-extend-gains-as-henry-hub-climbs-northeast-slides/) — May 19 2026: HH reaches highest level since March; heat lifts Southeast/Midwest. Accessed 2026-05-25.
7. [NGI — Heat Wave Snaps Natural Gas Spot Prices out of Spring Doldrums](https://naturalgasintel.com/news/heat-wave-snaps-natural-gas-spot-prices-out-of-spring-doldrums/) — May 18 2026: HH cash reclaims $3; East Coast heat. Accessed 2026-05-25.
8. [EIA — Natural Gas Spot Price History](https://www.eia.gov/dnav/ng/hist/rngwhhdm.htm) — Monthly: 2026 Jan $7.72, Feb $3.62, Mar $3.04, Apr $2.77; 2025 avg $3.53; 2022 $6.45; 2008 $8.85. Accessed 2026-05-25.
9. [EIA Press Release May 12 2026](https://www.eia.gov/pressroom/releases/press588.php) — Energy market indicators; 17 Bcf/d LNG exports 2026, 18 Bcf/d 2027; 40% gas share of generation 2026. Accessed 2026-05-25.
10. [EIA Today in Energy — Henry Hub natural gas spot prices to fall slightly in 2026 before rising in 2027](https://www.eia.gov/todayinenergy/detail.php?id=67004) — Jan 2026 STEO original view: 2027 sharp rise to $4.60; LNG +9% (2026) +11% (2027); ramp of Plaquemines / CC3 / Golden Pass. Accessed 2026-05-25.
11. [NGI — Cheniere's Corpus Christi LNG Stage 3 Expansion Nearly Complete](https://www.naturalgasintel.com/news/chenieres-corpus-christi-lng-stage-3-expansion-nearly-complete/) — May 22 2026: Train 6 producing LNG; Train 7 fall 2026; full 25 Mt/y. Accessed 2026-05-25.
12. [NGI — Late-Stage Commissioning Begins at Corpus Christi Expansion, Golden Pass](https://naturalgasintel.com/news/late-stage-commissioning-begins-at-corpus-christi-expansion-golden-pass-after-key-ferc-approvals/) — Feb 25 2026: US LNG feed gas nearing 20 Bcf/d. Accessed 2026-05-25.
13. [Gas Processing & LNG — Golden Pass LNG in Texas set to load first export cargo](https://gasprocessingnews.com/news/2026/04/golden-pass-lng-in-texas-set-to-load-first-export-cargo-amid-record-feedgas-intake/) — April 1 2026: HL Sea Eagle to load April 20; 434 MMcf/d feedgas; Train 1 ~6 Mt/y full capacity. Accessed 2026-05-25.
14. [Energy News Beat — New Wave of US LNG Projects Could Add Over 6 Bcf/d of Export Capacity](https://energynewsbeat.com/energy-news-beat-publishers-note/new-wave-of-us-lng-projects-could-add-over-6-bcf-d-of-export-capacity/) — April 16 2026: US capacity 14 → 28 Bcf/d by 2029; LNG export trajectory. Accessed 2026-05-25.
15. [AGA — Natural Gas Market Indicators May 15 2026](https://www.aga.org/research-policy/resource-library/natural-gas-market-indicators-may-15-2026/) — Forward curve $3.26 through end-2026; Dec 2026 above $4; June 2026 contract up 8.2% since rollover. Accessed 2026-05-25.
16. [DCD — Musk's xAI gets go-ahead for 41 natural gas turbines in Mississippi](https://www.datacenterdynamics.com/en/news/musks-xai-gets-go-ahead-for-41-natural-gas-turbines-in-mississippi-to-power-colossus-data-centers/) — March 2026: 1.2 GW permitted at Southaven, MS for Colossus 2/3. Accessed 2026-05-25.
17. [DCD — xAI deploys 19 natural gas turbines at Colossus 2 in Southaven](https://www.datacenterdynamics.com/en/news/xai-deploys-19-natural-gas-turbines-at-colossus-2-data-center-in-southaven-mississippi-report/) — May 14 2026: 19 new turbines, 500+ MW; 46 total at site. Accessed 2026-05-25.
18. [Time / Canary Media — Elon Musk's xAI adds more unpermitted gas generators](https://time.news/elon-musks-xai-adds-more-unpermitted-gas-generators-for-data-centers-canary-media/) — May 15 2026: 46 temporary + 41 permanent gas units at Memphis; 100,000 H100 GPU cluster. Accessed 2026-05-25.
19. [Mississippi Free Press — NAACP Asks Judge to Shut Down xAI Gas Turbines in Southaven](https://www.mississippifreepress.org/xai-gas-turbines-causing-harm-in-southaven-naacp-argues-in-motion-to-shut-them-down/) — May 8 2026: NAACP preliminary injunction motion; 27+6 unpermitted turbines. Accessed 2026-05-25.
20. [DCD — Lawsuit launched against Musk's xAI over 'illegal' gas turbines](https://www.datacenterdynamics.com/en/news/lawsuit-launched-against-musks-xai-over-illegal-gas-turbines-at-memphis-data-center/) — May 21 2026: NAACP filing; 1,700 tons NOx/year; Clean Air Act violations. Accessed 2026-05-25.
21. [Fortune — Meta orders 10 gas-fired power plants for Hyperion AI campus](https://fortune.com/2026/03/27/meta-hyperion-10-gas-power-plants-louisiana-entergy/) — March 27 2026: 10 plants, enough for 5M homes; Jordan Blum Energy Editor. Accessed 2026-05-25.
22. [Winbuzzer — Meta Funds 7 Gas Plants for Its Largest AI Data Center](https://winbuzzer.com/2026/03/31/meta-funds-seven-gas-plants-largest-ai-data-center-xcxwbn/) — March 31 2026: $27B Hyperion; 7+ GW combined; Louisiana grid +30%. Accessed 2026-05-25.
23. [DCD — Welcome to gas land: How natural gas is powering the US AI data center boom](https://www.datacenterdynamics.com/en/analysis/welcome-to-gas-land-how-natural-gas-is-powering-the-us-ai-boom/) — May 6 2026: 74-132 GW demand growth (Berkeley); 3-12 Bcf/d incremental (S&P/East Daley); 20 GW NC/GA/VA new gas; 3.3 Bcf/d new pipelines. Accessed 2026-05-25.
24. [IEA — Energy and AI](https://iea.blob.core.windows.net/assets/40a4db21-2225-42f0-8a07-addcc2ea86b3/EnergyandAI.pdf) — Global DC demand 945 TWh 2030; gas +175 TWh; 15-27 GW US onsite gas by 2030; 65-household power density per fridge-sized rack. Accessed 2026-05-25.
25. [Interactive Brokers / JPMorgan — Revising The Gas Outlook Higher](https://www.interactivebrokers.com/campus/traders-insight/securities/commodities/revising-the-gas-outlook-higher/) — JPM: AI adds 1.4 Bcf/d (2027), 6.2 Bcf/d (2030); LNG +9.4 Bcf/d through 2030; $3.50 sufficient; 22.4 Bcf/d exports at 86% utilization. Accessed 2026-05-25.
26. [Oxford Smith School — Impact on Power and Commodity Markets](https://www.smithschool.ox.ac.uk/sites/default/files/2026-04/Impact-on-Power-and-Commodity-Markets.pdf) — Behind-the-meter gas demand 0.5-2.0 Bcf/d invisible to EIA electric-sector data; 3,000 MW DC onsite gas = 0.5 Bcf/d. Accessed 2026-05-25.
27. [Energy News Beat — What Does the Demand for Natural Gas and LNG Look Like for the Next 20 Years?](https://energynewsbeat.co/ai/what-does-the-demand-for-natural-gas-and-lng-look-like-for-the-next-20-years/) — May 22 2026: AEO2026 LNG 30 Bcf/d by 2050; 27.7 Bcf/d capacity by 2030; AI 3-8 Bcf/d. Accessed 2026-05-25.
28. [OPIS — Storm Brewing Over Europe Natural Gas Market as Buyers Compete](https://www.opis.com/resources/energy-market-news-from-opis/storm-brewing-over-europe-natural-gas-market-as-buyers-compete/) — May 21 2026: EU storage 27.7% on Mar 31 (4-year low); TTF $18.21/MMBtu; JKM-TTF spread; Asian pull. Accessed 2026-05-25.
29. [Ainvest — Goldman Sachs Lifts TTF Gas Forecasts on Cold Weather, Storage Draw, Qatar Disruptions](https://www.ainvest.com/news/goldman-sachs-lifts-ttf-gas-forecasts-cold-weather-storage-draw-qatar-disruptions-2605/) — May 8 2026: Q2 2026 TTF €72/MWh (+14%); prompt JKM $20/MMBtu; end-Mar storage 16%. Accessed 2026-05-25.
30. [Energy News Beat — Europe's Gas Market Braces for Winter Shock](https://energynewsbeat.com/imports/europes-gas-market-braces-for-winter-shock-traders-betting-on-prices-doubling/) — May 6 2026: TTF €100/MWh options winter 2026-27; Russian gas full phase-out end-2026; 185 bcm EU LNG imports forecast 2026. Accessed 2026-05-25.
31. [EIA — Today in Energy: Fossil generation could rise with faster-than-expected growth in data center power demand](https://www.eia.gov/Todayinenergy/detail.php?id=67344) — High-growth scenario: 7.3% gas-fired generation growth 2025-27 vs 1.7% base; +$0.50/MMBtu delivered price; ERCOT +105 BkWh. Accessed 2026-05-25.
32. [EIA — STEO archive Mar 2026](https://www.eia.gov/outlooks/steo/archives/mar26.pdf) — Permian rig count; Winter Storm Fern -3.6 Bcf/d Jan 2026 production drop; recovery in Feb. Accessed 2026-05-25.
33. [The Center Square — Haynesville forecast to lead U.S. shale growth in next two years](https://www.thecentersquare.com/national/article_a06157db-82b2-4eda-9b4c-f98904ee8024.html) — Feb 19 2026: Haynesville +1.2 (2026) +1.6 (2027) Bcf/d; 17.1 Bcf/d 2027; Comstock 19 western Haynesville wells. Accessed 2026-05-25.
34. [Oil & Gas Journal — EIA: US natural gas production to hit record highs in 2026-27](https://www.ogj.com/general-interest/economics-markets/news/55357813/eia-us-natural-gas-production-to-hit-record-highs-in-2026-27) — Feb 2026 STEO: 120.8/122.3 Bcf/d 2026/27; HH $4.31/$4.38; 69% from Appalachia + Haynesville + Permian. Accessed 2026-05-25.
35. [TechCrunch — Data center demand drives 66% surge in natural gas power plant costs](https://techcrunch.com/2026/04/27/data-center-demand-drives-66-surge-in-natural-gas-power-plant-costs/) — April 27 2026: CCGT cost +66% to $2,157/kW; turbine prices +195% from 2019; waitlists into early 2030s; 40 → 106 GW data center demand 2035. Accessed 2026-05-25.