Takeaways
- Commodity gas prices are violently volatile. The Henry Hub annual average moved from $2.03 per million British thermal units (MMBtu) in 2020 to $6.45 in 2022, then back to $2.19 in 2024, then up to $3.52 in 2025. That is a 218 percent rise, a 66 percent fall and a 61 percent rise inside six years.
- Household bills do not follow it down. When Henry Hub fell 61 percent between 2022 and 2023, the average United States residential delivered gas price rose 4.4 percent, from $14.75 to $15.40 per thousand cubic feet.
- Over six years the household price only went one way. Residential delivered gas rose 42 percent from 2020 to 2025 while the commodity underlying it rose 73 percent but spent most of that period below its starting point.
- Delivery charges, not molecules, are the reason. The commodity is a minority of a residential bill and the distribution component is set by regulated cost recovery, which is sticky upward and slow downward.
- This changes what a payback analysis should be measuring. A ground-source heat pump does not hedge the commodity. It removes exposure to the delivery charge, which is the part that has proven to be a one-way ratchet.
The number everyone quotes, and the number that matters
Henry Hub is the pricing point that gets reported. It is also the number least relevant to the person deciding whether to replace a furnace.
The United States Energy Information Administration (EIA) publishes both series. Set them side by side and they tell different stories.
| Year | Henry Hub spot ($/MMBtu) |
Change | US residential delivered ($/Mcf) |
Change |
|---|---|---|---|---|
| 2020 | 2.03 | — | 10.78 | — |
| 2021 | 3.89 | +92% | 12.18 | +13% |
| 2022 | 6.45 | +66% | 14.75 | +21% |
| 2023 | 2.53 | −61% | 15.40 | +4% |
| 2024 | 2.19 | −13% | 14.50 | −6% |
| 2025 | 3.52 | +61% | 15.34 | +6% |
Read the 2023 row again. The commodity lost nearly two thirds of its value and the delivered residential price went up. Read 2024: the commodity fell another 13 percent to what the EIA has described as the lowest annual average price in inflation-adjusted terms ever reported, and the residential price came down 6 percent, giving back less than a third of the previous year's increase. By 2025 it was back above its 2023 level.
The commodity round-tripped. The household bill did not.
Why the pass-through is asymmetric
None of this is a scandal. It is the structure of a regulated distribution business.
A residential gas bill has two broad parts. The commodity is passed through, usually close to cost and with a regulatory lag. The delivery charge recovers the cost of the pipes, the meters, the leak survey programme, the mains replacement programme and the return the utility is authorised to earn on that capital. That second part has no relationship to the spot price of gas at all.
When the commodity spikes, the whole bill spikes and it is visible. When the commodity collapses, only the commodity portion falls, and it can be more than offset by a rate case, a mains replacement surcharge or an approved increase in the authorised return. That is what the 2023 row in the table is showing.
The direction of travel in the delivery component matters more than most heating analyses acknowledge. Utilities across the country are replacing aging cast iron and bare steel distribution mains, and that capital enters rate base. A shrinking number of customers spread across the same or a growing pipe network means the fixed cost per remaining customer rises. This is the mechanism people refer to as the utility death spiral, and whatever one thinks of the term, its arithmetic is straightforward.
What this does to a payback calculation
The standard framing is that a ground-source heat pump is a bet that gas prices will rise. That framing is weak, because it is a bet on the most volatile and least predictable component of the bill.
The stronger framing is that a ground-source heat pump converts an operating cost with unbounded variance into a capital cost with known amortisation. The borefield does not get repriced. The heat exchange resource, the ground itself, does not have a tariff. What remains variable is the electricity to run the compressor and the circulation pumps, which is real exposure but a different and generally smaller one, and one that a customer can hedge further with on-site generation.
Three practical consequences follow for anyone building the model.
Do not escalate gas at a single rate
A single-rate escalation assumption, whether 2 percent or 6 percent, misrepresents the risk entirely. The six years in the table above contain no year that looks like a smooth escalation. Model a low, central and high scenario and report all three, because the spread between them is the actual finding.
Escalate the delivery charge separately from the commodity
They behave differently and lumping them together hides the asymmetry that makes the case. The delivery component is the one that has behaved as a ratchet.
State clearly what the customer is buying
Not cheaper heat necessarily, and not on every day of every year. Price certainty, and insulation from a bill component that has moved in one direction for six years. That is a defensible claim. "Gas will get more expensive" is a forecast, and forecasts of this particular commodity have a poor record.
A note from project work
In the feasibility work Gaiergy completed for the City of Vandalia, Illinois, the local gas price history was a material input to the case rather than background colour. That study documented a 137 percent increase in the applicable gas price over 2021 to 2025 and used a conservative 10 percent annual escalation in the central financial scenario against an observed 24 percent, alongside lower and higher escalation cases.
Those are local figures from a specific utility territory and they are considerably steeper than the national residential series above, which rose 26 percent over the same 2021 to 2025 window. They are cited here as an illustration of how far a single service territory can diverge from the national average, which is the argument for modelling the actual tariff rather than the headline.