Subtracting Carbon

Almost every confused argument about carbon removal originates in the same place: three quite different activities share a vocabulary, and the vocabulary conceals a difference of several orders of magnitude in both cost and difficulty.

Three things, not one

The first is point-source capture. Carbon dioxide is separated from the exhaust of an industrial process — a cement kiln, a steel plant — before it reaches the atmosphere. The gas stream at that point is between ten and twenty per cent carbon dioxide, hot and under some pressure. Separating a component present at fifteen per cent is a conventional chemical engineering problem, comparable to operations refineries have performed for a century.

The second is direct air capture. The same molecule is extracted from ordinary outdoor air, where its concentration is about 0.042 per cent — roughly four hundred times more dilute.

The third is biological and geochemical removal: growing trees, accelerating rock weathering, adding alkalinity to seawater, burying charred biomass.

Treating these as variants of one technology is the single most common error in public discussion, and the thermodynamics do not permit it.

Why dilution is decisive

Separating a mixture requires work, and the minimum work rises steeply as the target becomes rarer. Pulling carbon dioxide from a flue at fifteen per cent requires a theoretical minimum on the order of twenty kilojoules per mole. Pulling it from open air requires roughly five times that, and thermodynamic minima are not engineering targets — real processes consume several times the ideal.

The practical consequence is visible in the energy budgets. Direct air capture plants presently consume in the region of fifteen hundred to two and a half thousand kilowatt-hours of energy per tonne of carbon dioxide removed, much of it as heat for regenerating the sorbent. Point-source capture on a concentrated stream costs a small fraction of that.

This yields the first hard conclusion. If your electricity is not already effectively carbon-free and surplus to other demands, running direct air capture is close to self-defeating, because the generation required to run it emits a substantial fraction of what the plant removes. The technology is not an alternative to decarbonising the grid. It is something that only makes sense after you have done it.

The scale problem, stated honestly

Global emissions are around thirty-seven billion tonnes of carbon dioxide per year. Total operating direct air capture capacity worldwide is on the order of ten thousand tonnes per year.

That is a ratio of roughly one to four million. Announced projects would improve it to perhaps one to a few thousand by the early 2030s, assuming every announcement is built on schedule, which no industrial sector has ever achieved.

This is not an argument that the technology is useless. It is an argument about sequencing. A tonne not emitted and a tonne removed are equivalent in the atmosphere and are not remotely equivalent in cost: avoiding emission through efficiency, generation or fuel switching currently costs a small fraction of removing the same tonne afterwards. Any portfolio that spends on removal what it could have spent on avoidance is buying less climate benefit for the same money, and doing so deliberately.

Permanence, and the problem of verification

Where removal is genuinely needed, a second question governs whether it counts: how long does the carbon stay put?

Geological storage in deep saline formations is well understood, monitored by seismic survey, and reasonably regarded as permanent on any timescale that matters. Mineralisation — reacting the gas with basalt so that it becomes solid carbonate — is more permanent still.

Biological storage is different in kind. A forest holds carbon while it remains a forest. It is subject to fire, drought, disease, harvest and changes of policy, and the reversal risk is neither small nor independent of the warming the project is meant to address. A credit sold as permanent against a store that is conditional is not a conservative accounting choice; it is an error with a known sign.

Verification compounds the difficulty. Establishing that a removal occurred requires a counterfactual — what would have happened otherwise — and counterfactuals are unobservable. Repeated independent assessments of forest-based credit portfolios have found that a large share represented substantially less additional storage than claimed, typically because the baseline assumed a deforestation that was not in fact going to occur. The market's problem was not fraud in most cases. It was that the quantity being sold cannot be measured directly and must be inferred from an assumption that the seller chooses.

The legitimate case

Having stated the objections, the case in favour is real and should not be lost in them.

Some emissions are genuinely difficult to eliminate. Roughly half the carbon dioxide from cement is released by the chemical decomposition of limestone itself, not by the fuel, and no change of energy source removes it. Long-haul aviation, certain chemical processes and some agricultural emissions are similarly resistant. If a net-zero position is to mean anything arithmetically, those residual emissions must be balanced by genuine removals, and that requirement does not go away by being unpopular.

Further, reaching a stable temperature and then reducing it requires net-negative emissions at some point, which is impossible without removal capacity at a scale that does not currently exist. Capacity of that kind takes decades to build. Beginning now, at small volume and high cost, is how industries learn — provided the learning is not mistaken for the solution while it is happening.

The distinction that matters

There is a term for the central risk, and it is more precise than the usual accusation of hypocrisy: mitigation deterrence. The availability of a future removal option reduces present pressure to cut emissions, by making continued emission appear provisionally acceptable. The effect does not require anyone to act in bad faith. It follows from ordinary discounting: a cost deferred is a cost reduced, and removal converts an immediate obligation into a later one.

The defensible position is therefore narrow but clear. Removal is a complement to reduction and never a substitute. It should be purchased for residual emissions that have actually been reduced as far as available technology permits, from stores with verified permanence, at prices that reflect real cost rather than the cheapest available claim. And it should be reported separately from reductions, always, because a single net figure combining a tonne avoided with a tonne allegedly absorbed by a forest that may burn is not an accounting summary. It is a place to hide.