What happens to demand when a cheaper substitute wins?

Demand is not a fixed physical requirement. It is a choice made at a price, and buyers who cannot cut consumption can often change what they consume. Substitution is the slow, reliable mechanism that caps how far any one commodity can run relative to its neighbours.

Substitution is everywhere, on an engineering clock

Real, well-documented examples:

• Power generation switches between natural gas and coal as their relative prices move — one of the fastest substitutions available, because the plants already exist and the operator simply dispatches whichever is cheaper.
• Autocatalysts have shifted between platinum and palladium in both directions as their relative prices inverted, though each redesign takes vehicle-model years.
• Aluminium substitutes for copper in some cables, heat exchangers and busbars when the price ratio stretches far enough — it conducts less well, but it is lighter and cheaper.
• Battery chemistries shift between nickel-rich and iron-phosphate formulations, changing the metal bill per vehicle substantially.
• Edible oils and grains substitute for one another in food and feed with meaningful cross-price elasticity.

Two consequences: a soft ceiling on the expensive commodity's relative price, and a floor under its substitute, whose demand rises just as the other's falls. Relative prices, not absolute ones, drive this.

Thrifting — the quieter cousin

Alongside switching material, users use less of the same material: thinner copper windings, less platinum loading per catalyst, higher-efficiency engines, better yields per tonne of fertiliser. Thrifting is permanent — nobody re-thickens the wire when the price falls — which is why demand forecasts built as "units × today's intensity" have a long record of overshooting. Intensity per unit falls after every sustained price spike.

The energy transition, described even-handedly

The transition is the largest current example of substitution, and it deserves precision rather than advocacy, because it cuts both ways:

• It is a demand risk for some commodities — the fuels being substituted away from in power generation and, over a longer horizon, transport.
• It is a demand source for others — copper and aluminium for grids and wiring, lithium and nickel for batteries, silver for solar cells, plus the steel, cement and diesel needed to build any of it.
• The pace is genuinely uncertain, and it depends on policy, technology cost curves, permitting, grid build-out and consumer adoption — none of which are forecastable with confidence. Projections from every direction have been wrong repeatedly, in both directions, and revised annually.

Two facts keep the analysis honest. First, the transition is itself commodity-intensive: building less commodity-intensive energy requires a great deal of metal up front. Second, the mechanisms of this course still apply to the new commodities: lithium was described as structurally short, then oversupplied within a few years, exactly as Unit 2's capex cycle predicts. Newness does not repeal elasticity.

What you can do is treat demand as a policy- and technology-conditional variable and watch the observables: installed generation capacity, vehicle fleet mix, grid capital spending, and metal intensity per unit. What you should not do is adopt one scenario as fact.

In the data

A ratio is only meaningful when both legs share a unit. Platinum and palladium, the two catalyst metals that substitute for each other, are both quoted in dollars per troy ounce:

On 29 September 2026 platinum stood at $1,705.30 and palladium at $1,219.00, a ratio of 1.40, a clean number that says how far apart the two substitutes have drifted. Gas against coal is harder: gas is priced per million BTU of energy, coal per tonne of rock, and no coal price is published alongside the gas one here, so that substitution can be reasoned about but not read off a page.

Try it now

1. Two substitutable energy sources are below: crude in dollars per barrel, natural gas in dollars per million BTU. Read a close off each on the same date and divide crude by gas. That ratio is dimensionless, which is the point — it is the exchange rate between two ways of buying the same heat.

2. Do it again at the widest and narrowest points you can find by eye. On energy content alone the two are worth about six to one; when the ratio stretches far past that, industrial users with dual-fuel burners have a reason to switch, and switching is what pulls it back. Mark where it stretched and where it snapped back.
3. Pick one commodity and write down its two most plausible substitutes and roughly how long a switch would take. That time is its demand elasticity. Then state the discipline in one line: demand forecasts that hold intensity constant ignore the market's own response — and the market always responds.