Research findings — Energy Sankey

F1 · 2026-07-03 · status: implemented (D121 scorecard)

Fuel-share of delivered cost — the exposure metric nobody re-quantified

Of every dollar a customer pays for delivered energy, how many cents are indexed to a volatile commodity price? A metric this engine can compute natively — and, per the July 2026 research pass, one with no rigorous quantification anywhere since 2008.

1 · What the metric is

Every generation source splits its cost into parts that are fixed once built (capital recovery, fixed O&M) and parts that track a commodity market forever (fuel). Wind, solar, operating nuclear, geothermal, hydro, and storage sit almost entirely in the first bucket. Gas and coal sit heavily in the second: at $4/MMBtu, fuel is $30.40 of the $64 new-build combined-cycle LCOE in this project's own research values — 48% — and the share only grows as a plant ages past capital recovery.

Fuel-share of delivered cost aggregates that exposure across the whole mix, per year, per scenario: the fraction of the wholesale dollar that re-prices when Henry Hub moves. It turns "clean power is a hedge" from a slogan into a measurable, scenario-comparable quantity.

≈21¢ fuel-indexed

≈79¢ fixed once built

2050 default: gas ≈ 40% of generation × $30.4/MWh fuel ≈ $12 of the $59.4 wholesale At 2022's average ($6.45/MMBtu) the fuel slice adds ≈ +$7/MWh; at the Aug-2022 peak ($9.85), gas fuel alone is $75/MWh — more than the entire low-end LCOE of new wind or solar.

2 · Why it's volatile enough to matter

A 3× swing inside five years; 2022's spike was LNG/Europe-driven and the coupling is structural now. Coal, by contrast, barely moves ($1.96–2.54/MMBtu over the same window; SEDS CLEID) — which is why the metric's variance is almost entirely a gas story.

3 · The gap

The canonical treatment is Bolinger & Wiser (LBNL), Renewable Energy as a Hedge Against Fuel Price Risk: locking in gas price certainty via futures cost ~$0.50/MMBtu over 10-year horizons — an implicit hedge premium for zero-fuel generation of roughly $3–5/MWh. That work is 2008-era. A deliberate search (2026-07-03; see docs/altitude_integration_plan.md §A.3) found no rigorous 2024–2026 re-quantification.

Why this is an opportunity Every serious energy model outputs LCOE. Almost none output price-exposure composition per scenario. This engine carries every term natively — computing it is bookkeeping, not new modeling.

4 · The metric

exposure(y) = Σ_{f ∈ {coal, gas}} primary_quads(f, y) × delivered_price(f, y) / ( grid_output_TWh(y) × wholesale(y) )

primary_quads(f, y)	The allocator's mix — already in every projection artifact's links. Quads × $/MMBtu × 10⁹ = fuel-indexed $B directly; no heat-rate needed in the numerator.
delivered_price(f, y)	EIA SEDS CLEID (coal) / NGEID (gas) histories through 2024, then compounded at the scenario's price-path rate. L1, snapshot committed.
wholesale(y)	The engine's per-year `wholesale_avg` diagnostic.

Status: implemented. The metric shipped in the pre-registered thesis scorecard (D121, docs/thesis_scorecard.md + webapp/data/scorecard.json). First numbers, US default projection: ~17% of wholesale dollars fuel-indexed in 2030; 13.9% at 2050 on a flat real gas price, 25.5% under a +5%/yr path. The clean-favored scenario shaves ~0.9pp off either — structural de-indexing, measured.

5 · What it is not

Not a hedging model (no futures curves or basis risk — it measures exposure, not the cost of insuring it). Not hourly. Petroleum excluded (~0.1 Q). Biomass feedstock deliberately excluded (too heterogeneous to index). One line on a five-line scorecard, never a verdict by itself — high exposure is a risk posture some buyers rationally accept for speed.

Sources: research/generation/natural_gas.md + coal.md (heat rates, price anchors, SEDS histories); committed projection artifacts (mix, wholesale); Bolinger & Wiser, LBNL (hedge premium); docs/altitude_integration_plan.md §A.3 (the 2026-07-03 search). Related: docs/mythbusters.md M2.