31 The last thing we should talk about

Capturing carbon dioxide from thin air is the last thing we should talk about.

When I say this, I am deliberately expressing a double meaning. First, the energy requirements for carbon capture from thin air are so enormous, it seems almost absurd to talk about it (and there’s the worry that raising the possibility of fixing climate change by this sort of geoengineering might promote inaction today). But second, I do think we should talk about it, contemplate how best to do it, and fund research into how to do it better, because capturing carbon from thin air may turn out to be our last line of defense, if climate change is as bad as the climate scientists say, and if humanity fails to take the cheaper and more sensible options that may still be available today.

Before we discuss capturing carbon from thin air, we need to understand the global carbon picture better.

Understanding CO₂

When I first planned this book, my intention was to ignore climate change altogether. In some circles, “Is climate change happening?” was a controversial question. ¹ As were “Is it caused by humans?” and “Does it matter?” And, dangling at the end of a chain of controversies, “What should we do about it?” I felt that sustainable energy was a compelling issue by itself, and it was best to avoid controversy. My argument was to be: “Never mind when fossil fuels are going to run out; never mind whether climate change is happening; burning fossil fuels is not sustainable anyway; let’s imagine living sustainably, and figure out how much sustainable energy is available.”

However, climate change has risen into public consciousness, and it raises all sorts of interesting back-of-envelope questions. So I decided to discuss it a little in the preface and in this closing chapter. Not a complete discussion, just a few interesting numbers.

Figure 31.1. The weights of an atom of carbon and a molecule of CO₂ are in the ratio 12 to 44, because the carbon atom weighs 12 units and the two oxygen atoms weigh 16 each. 12 + 16 +16 = 44.

Units

Carbon pollution charges are usually measured in dollars or euros per ton of CO₂ so I’ll use the ton of CO₂ as the main unit when talking about percapita carbon pollution, and the ton of CO₂ per year to measure rates of pollution. (The average European’s greenhouse emissions are equivalent to 11 tons per year of CO₂; or 30 kg per day of CO₂.) But when talking about carbon in fossil fuels, vegetation, soil, and water, I’ll talk about tons of carbon. One ton of CO₂ contains 12/44 tons of carbon, a bit more than a quarter of a ton. On a planetary scale, I’ll talk about gigatons of carbon (Gt C). A gigaton of carbon is a billion tons. Gigatons are hard to imagine, but if you want to bring it down to a human scale, imagine burning one ton of coal (which is what you might use to heat a house over a year). Now imagine everyone on the planet burning one ton of coal per year: that’s 6 Gt C per year, because the planet has 6 billion people.

Where is the carbon?

Where is all the carbon? We need to know how much is in the oceans, in the ground, and in vegetation, compared to the atmosphere, if we want to understand the consequences of CO₂ emissions. ²

Figure 31.2. Estimated amounts of carbon, in gigatons, in accessible places on the earth. (There’s a load more carbon in rocks too; this carbon moves round on a timescale of millions of years, with a long-term balance between carbon in sediment being subducted at tectonic plate boundaries, and carbon popping out of volcanoes from time to time. For simplicity I ignore this geological carbon.)

Figure 31.2 shows where the carbon is. Most of it – 40 000 Gt – is in the ocean (in the form of dissolved CO₂ gas, carbonates, living plant and animal life, and decaying materials). Soils and vegetation together contain about 3700 Gt. Accessible fossil fuels – mainly coal – contain about 1600 Gt. Finally, the atmosphere contains about 600 Gt of carbon.

Until recently, all these pools of carbon were roughly in balance: all flows of carbon out of a pool (say, soils, vegetation, or atmosphere) were balanced by equal flows into that pool. The flows into and out of the fossil fuel pool were both negligible. Then humans started burning fossil fuels. This added two extra unbalanced flows, as shown in figure 31.3.

The rate of fossil fuel burning was roughly 1 Gt C/y in 1920, 2 Gt C/y in 1955, and 8.4 Gt C in 2006. (These figures include a small contribution from cement production, which releases CO₂ from limestone.) ³

Figure 31.3. The arrows show two extra carbon flows produced by burning fossil fuels. There is an imbalance between the 8.4 Gt C/y emissions into the atmosphere from burning fossil fuels and the 2 Gt C/y take-up of CO₂ by the oceans. This cartoon omits the less-well quantified flows between atmosphere, soil, vegetation, and so forth.

How has this significant extra flow of carbon modified the picture shown in figure 31.2? Well, it’s not exactly known. Figure 31.3 shows the key things that are known. Much of the extra 8.4 Gt C per year that we’re putting into the atmosphere stays in the atmosphere, raising the atmospheric concentration of carbon-dioxide. The atmosphere equilibrates fairly rapidly with the surface waters of the oceans (this equilibration takes only five or ten years), and there is a net flow of CO₂ from the atmosphere into the surface waters of the oceans, amounting to 2 Gt C per year. (Recent research indicates this rate of carbon-uptake by the oceans may be reducing, however.) ⁴ This unbalanced flow into the surface waters causes ocean acidification, which is bad news for coral. Some extra carbon is moving into vegetation and soil too, perhaps about 1.5 Gt C per year, but these flows are less well measured. Because roughly half of the carbon emissions are staying in the atmosphere, ⁵ continued carbon pollution at a rate of 8.4 Gt C per year will continue to increase CO₂ levels in the atmosphere, and in the surface waters.

What is the long-term destination of the extra CO₂? Well, since the amount in fossil fuels is so much smaller than the total in the oceans, “in the long term” the extra carbon will make its way into the ocean, and the amounts of carbon in the atmosphere, vegetation, and soil will return to normal. However, “the long term” means thousands of years. Equilibration between atmosphere and the surface waters is rapid, as I said, but figures 31.2 and 31.3 show a dashed line separating the surface waters of the ocean from the rest of the ocean. On a time-scale of 50 years, this boundary is virtually a solid wall. Radioactive carbon dispersed across the globe by the atomic bomb tests of the 1960s and 70s has penetrated the oceans to a depth of only about 400 m. ⁶ In contrast the average depth of the oceans is about 4000 m.

The oceans circulate slowly: a chunk of deep-ocean water takes about 1000 years to roll up to the surface and down again. The circulation of the deep waters is driven by a combination of temperature gradients and salinity gradients, so it’s called the thermohaline circulation (in contrast to the circulations of the surface waters, which are wind-driven).

This slow turn-over of the oceans has a crucial consequence: we have enough fossil fuels to seriously influence the climate over the next 1000 years.

Where is the carbon going

Figure 31.3 is a gross simplification. For example, humans are causing additional flows not shown on this diagram: the burning of peat and forests in Borneo in 1997 alone released about 0.7 Gt C. Accidentally-started fires in coal seams release about 0.25 Gt C per year.

Nevertheless, this cartoon helps us understand roughly what will happen in the short term and the medium term under various policies. First, if carbon pollution follows a “business as usual” trajectory, burning another 500 Gt of carbon over the next 50 years, we can expect the carbon to continue to trickle gradually into the surface waters of the ocean at a rate of 2 Gt C per year. By 2055, at least 100 Gt of the 500 would have gone into the surface waters, and CO₂ concentrations in the atmosphere would be roughly double their pre-industrial levels.

Figure 31.4. Decay of a small pulse of CO₂ added to today’s atmosphere, according to the Bern model of the carbon cycle. Source: Hansen et al. (2007).

If fossil-fuel burning were reduced to zero in the 2050s, the 2 Gt flow from atmosphere to ocean would also reduce significantly. (I used to imagine that this flow into the ocean would persist for decades, but that would be true only if the surface waters were out of equilibrium with the atmosphere; but, as I mentioned earlier, the surface waters and the atmosphere reach equilibrium within just a few years.) Much of the 500 Gt we put into the atmosphere would only gradually drift into the oceans over the next few thousand years, as the surface waters roll down and are replaced by new water from the deep.

Thus our perturbation of the carbon concentration might eventually be righted, but only after thousands of years. And that’s assuming that this large perturbation of the atmosphere doesn’t drastically alter the ecosystem. It’s conceivable, for example, that the acidification of the surface waters of the ocean might cause a sufficient extinction of ocean plant-life that a new vicious cycle kicks in: acidification means extinguished plant-life, means plant-life absorbs less CO₂ from the ocean, means oceans become even more acidic. Such vicious cycles (which scientists call “positive feedbacks” or “runaway feedbacks”) have happened on earth before: it’s believed, for example, that ice ages ended relatively rapidly because of positive feedback cycles in which rising temperatures caused surface snow and ice to melt, which reduced the ground’s reflection of sunlight, which meant the ground absorbed more heat, which led to increased temperatures. (Melted snow – water – is much darker than frozen snow.) Another positive feedback possibility to worry about involves methane hydrates, which are frozen in gigaton quantities in places like Arctic Siberia, and in 100-gigaton quantities on continental shelves. Global warming greater than 1 °C would possibly melt methane hydrates, ⁷ which release methane into the atmosphere, and methane increases global warming more strongly than CO₂ does.

This isn’t the place to discuss the uncertainties of climate change in any more detail. I highly recommend the books Avoiding Dangerous Climate Change (Schellnhuber et al., 2006) and Global Climate Change (Dessler and Parson, 2006). Also the papers by Hansen et al. (2007) and Charney et al. (1979).

The purpose of this chapter is to discuss the idea of fixing climate change by sucking carbon dioxide from thin air; we discuss the energy cost of this sucking next.

The cost of sucking

Today, pumping carbon out of the ground is big bucks. In the future, perhaps pumping carbon into the ground is going to be big bucks. Assuming that inadequate action is taken now to halt global carbon pollution, perhaps a coalition of the willing will in a few decades pay to create a giant vacuum cleaner, and clean up everyone’s mess.

Before we go into details of how to capture carbon from thin air, let’s discuss the unavoidable energy cost of carbon capture. Whatever technologies we use, they have to respect the laws of physics, and unfortunately grabbing CO₂ from thin air and concentrating it requires energy. The laws of physics say that the energy required must be at least 0.2 kWh per kg of CO₂ (table 31.5). Given that real processes are typically 35% efficient at best, I’d be amazed if the energy cost of carbon capture is ever reduced below 0.55 kWh per kg.

Now, let’s assume that we wish to neutralize a typical European’s CO₂ output of 11 tons per year, which is 30 kg per day per person. The energy required, assuming a cost of 0.55 kWh per kg of CO₂, is 16.5 kWh per day per person. This is exactly the same as British electricity consumption. So powering the giant vacuum cleaner may require us to double our electricity production – or at least, to somehow obtain extra power equal to our current electricity production.

If the cost of running giant vacuum cleaners can be brought down, brilliant, let’s make them. But no amount of research and development can get round the laws of physics, which say that grabbing CO₂ from thin air and concentrating it into liquid CO₂ requires at least 0.2 kWh per kg of CO₂.

Now, what’s the best way to suck CO₂ from thin air? I’ll discuss four technologies for building the giant vacuum cleaner:

1. chemical pumps;
2. trees;
3. accelerated weathering of rocks;
4. ocean nourishment.

A. Chemical technologies for carbon capture

The chemical technologies typically deal with carbon dioxide in two steps.

\[ \text{0.03\%~}\text{CO}_{\text{2}}\overset{\text{concentrate}}{\longrightarrow}\text{Pure~}\text{CO}_{\text{2}}\overset{\text{compress}}{\longrightarrow}\text{Liquid~}\text{CO}_{\text{2}} \]

First, they concentrate CO₂ from its low concentration in the atmosphere; then they compress it into a small volume ready for shoving somewhere (either down a hole in the ground or deep in the ocean). ⁸ Each of these steps has an energy cost. The costs required by the laws of physics are shown in table 31.5.

	cost (kWh/kg)
concentrate	0.13
compress	0.07
total	0.20

Table 31.5. The inescapable energy-cost of concentrating and compressing CO2 from thin air. ⁹

In 2005, the best published methods for CO₂ capture from thin air were quite inefficient: the energy cost was about 3.3 kWh per kg, with a financial cost of about $140 per ton of CO₂. ¹⁰ At this energy cost, capturing a European’s 30 kg per day would cost 100 kWh per day – almost the same as the European’s energy consumption of 125 kWh per day. Can better vacuum cleaners be designed?

Recently, Wallace Broecker, climate scientist, ¹¹ “perhaps the world’s foremost interpreter of the Earth’s operation as a biological, chemical, and physical system,” has been promoting an as yet unpublished technology developed by physicist Klaus Lackner for capturing CO₂ from thin air. Broecker imagines that the world could carry on burning fossil fuels at much the same rate as it does now, and 60 million CO₂-scrubbers (each the size of an up-ended shipping container) will vacuum up the CO₂. What energy does Lackner’s process require? In June 2007 Lackner told me that his lab was achieving 1.3 kWh per kg, but since then they have developed a new process based on a resin that absorbs CO₂ when dry and releases CO₂ when moist. Lackner told me in June 2008 that, in a dry climate, the concentration cost has been reduced to about 0.18–0.37 kWh of low-grade heat per kg CO₂. The compression cost is 0.11 kWh per kg. Thus Lackner’s total cost is 0.48 kWh or less per kg. For a European’s emissions of 30 kg CO₂ per day, we are still talking about a cost of 14 kWh per day, of which 3.3 kWh per day would be electricity, and the rest heat.

Hurray for technical progress! But please don’t think that this is a small cost. We would require roughly a 20% increase in world energy production, just to run the vacuum cleaners.

B. What about trees?

1 hectare = 10 000 m²

Trees are carbon-capturing systems; they suck CO₂ out of thin air, and they don’t violate any laws of physics. They are two-in-one machines: they are carbon-capture facilities powered by built-in solar power stations. They capture carbon using energy obtained from sunlight. The fossil fuels that we burn were originally created by this process. So, the suggestion is, how about trying to do the opposite of fossil fuel burning? How about creating wood and burying it in a hole in the ground, while, next door, humanity continues digging up fossil wood and setting fire to it? It’s daft to imagine creating buried wood at the same time as digging up buried wood. Even so, let’s work out the land area required to solve the climate problem with trees.

The best plants in Europe capture carbon at a rate of roughly 10 tons of dry wood per hectare per year ¹² – equivalent to about 15 tons of CO₂ per hectare per year – so to fix a European’s output of 11 tons of CO₂ per year we need 7500 square metres of forest per person. This required area of 7500 square metres per person is twice the area of Britain per person. And then you’d have to find somewhere to permanently store 7.5 tons of wood per person per year! At a density of 500 kg per m³, each person’s wood would occupy 15 m³ per year. A lifetime’s wood – which, remember, must be safely stored away and never burned – would occupy 1000 m³. That’s five times the entire volume of a typical house. If anyone proposes using trees to undo climate change, they need to realise that country-sized facilities are required. I don’t see how it could ever work.

C. Enhanced weathering of rocks

Is there a sneaky way to avoid the significant energy cost of the chemical approach to carbon-sucking? Here is an interesting idea: pulverize rocks that are capable of absorbing CO₂, and leave them in the open air. ¹³ This idea can be pitched as the acceleration of a natural geological process. Let me explain.

Two flows of carbon that I omitted from figure 31.3 are the flow of carbon from rocks into oceans, associated with the natural weathering of rocks, and the natural precipitation of carbon into marine sediments, which eventually turn back into rocks. These flows are relatively small, involving about 0.2 Gt C per year (0.7 Gt CO₂ per year). So they are dwarfed by current human carbon emissions, which are about 40 times bigger. But the suggestion of enhanced-weathering advocates is that we could fix climate change by speeding up the rate at which rocks are broken down and absorb CO₂. The appropriate rocks to break down include olivines or magnesium silicate minerals, which are widespread. The idea would be to find mines in places surrounded by many square kilometres of land on which crushed rocks could be spread, or perhaps to spread the crushed rocks directly on the oceans. Either way, the rocks would absorb CO₂ and turn into carbonates and the resulting carbonates would end up being washed into the oceans. To pulverize the rocks into appropriately small grains for the reaction with CO₂ to take place requires only 0.04 kWh per kg of sucked CO₂. Hang on, isn’t that smaller than the 0.20 kWh per kg required by the laws of physics? Yes, but nothing is wrong: the rocks themselves are the sources of the missing energy. Silicates have higher energy than carbonates, so the rocks pay the energy cost of sucking the CO₂ from thin air.

I like the small energy cost of this scheme but the difficult question is, who would like to volunteer to cover their country with pulverized rock?

D. Ocean nourishment

One problem with chemical methods, tree-growing methods, and rock-pulverizing methods for sucking CO₂ from thin air is that all would require a lot of work, and no-one has any incentive to do it – unless an international agreement pays for the cost of carbon capture. At the moment, carbon prices are too low.

A final idea for carbon sucking might sidestep this difficulty. The idea is to persuade the ocean to capture carbon a little faster than normal as a by-product of fish farming. ¹⁴

Some regions of the world have food shortages. There are fish shortages in many areas, because of over-fishing during the last 50 years. The idea of ocean nourishment is to fertilize the oceans, supporting the base of the food chain, enabling the oceans to support more plant life and more fish, and incidentally to fix more carbon. Led by Australian scientist Ian Jones, the ocean nourishment engineers would like to pump a nitrogencontaining fertilizer such as urea into appropriate fish-poor parts of the ocean. They claim that 900 km² of ocean can be nourished to take up about 5 Mt CO₂/y. Jones and his colleagues reckon that the ocean nourishment process is suitable for any areas of the ocean deficient in nitrogen. That includes most of the North Atlantic. Let’s put this idea on a map. UK carbon emissions are about 600 Mt CO₂/y. So complete neutralization of UK carbon emissions would require 120 such areas in the ocean. The map in figure 31.6 shows these areas to scale alongside the British Isles. As usual, a plan that actually adds up requires country-sized facilities! And we haven’t touched on how we would make all the required urea.

Figure 31.6. 120 areas in the Atlantic Ocean, each 900 km² in size. These make up the estimated area required in order to fix Britain’s carbon emissions by ocean nourishment.

While it’s an untested idea, and currently illegal, I do find ocean nourishment interesting because, in contrast to geological carbon storage, it’s a technology that might be implemented even if the international community doesn’t agree on a high value for cleaning up carbon pollution; fishermen might nourish the oceans purely in order to catch more fish.

Commentators can be predicted to oppose manipulations of the ocean, focusing on the uncertainties rather than on the potential benefits. They will be playing to the public’s fear of the unknown. People are ready to passively accept an escalation of an established practice (e.g., dumping CO₂ in the atmosphere) while being wary of innovations that might improve their future well being. They have an uneven aversion to risk.

Ian Jones

We, humanity, cannot release to the atmosphere all, or even most, fossil fuel CO₂. To do so would guarantee dramatic climate change, yielding a different planet…

J. Hansen et al (2007)

Avoiding dangerous climate change” is impossible – dangerous climate change is already here. The question is, can we avoid catastrophic climate change?

David King, UK Chief Scientist, 2007

The last thing, eighteen years on

A section added in the 2026 revision. MacKay’s title carries a double meaning — last in order of priority, and last as in a final resort — and he says so. Both readings have got stronger, and none of the eighteen years since has been kind to the technology.

The energy cost turned out worse than he assumed, not better

He works with a figure of 0.55 kWh per kilogram of carbon dioxide, roughly three times the thermodynamic floor of 0.2 that he derives, and observes that it “seems almost absurd” to talk about doing this at all. Real machines have now been built and measured:

kWh per kg of CO2
Thermodynamic minimum, as MacKay derives it	0.2
MacKay’s working assumption	0.55
Machines actually operating, 2025	1.5 to 2.5

So the real thing is three to five times worse than the assumption he already thought close to absurd, and eight to twelve times the physical limit. The fans alone, before any chemistry, account for 300 to 900 kWh a tonne.¹⁵

Redo his sentence with a real number. Neutralising a European’s 30 kg a day at 2 kWh per kilogram takes 60 kWh per day per person. MacKay’s own version reads: “16.5 kWh per day per person. This is exactly the same as British electricity consumption.” Britain now generates 11.2 kWh(e) per day per person. Sucking one European’s emissions out of the air would take more than five times Britain’s entire electricity generation per head — where in 2008 the two numbers were equal.

The gap widened from both ends: the machines are dearer in energy than he assumed, and the generation he compared them against has shrunk.

It has been built at a scale of about nine seconds

Twenty-seven direct air capture plants had been commissioned worldwide by the International Energy Agency’s last full count, with a combined operating capacity a little over 10 000 tonnes a year. Mammoth’s 36 000-tonne nameplate is not in that total and would dominate it if the plant performed; the paragraph below records what it actually captured. World energy-related emissions are about 35 800 million tonnes a year, which is 1135 tonnes every second.

The entire operating capacity in that count, running perfectly for a full year, offsets about nine seconds of world emissions.

The flagship is worse than that headline suggests. Climeworks’ Mammoth, in Iceland, is the largest such plant ever built, with a nameplate of 36 000 tonnes a year. In 2024 it captured 105 tonnes — about three-tenths of one per cent of design capacity. The company’s own co-chief executive put operating cost “closer to the $1000 per ton mark than the $100 per ton mark” — nearer £790 a tonne than £79, and the firm cut staff in 2025.¹⁶

The next machine is larger. STRATOS, in Texas, is designed for 500 000 tonnes a year, which would multiply that operating count roughly fiftyfold. At full output it would offset about seven minutes of world emissions annually.

The subsidy does not close the gap either

Chapter 23 records the American 45Q credit paying $180 (£142) a tonne for direct air capture — the most generous instrument for this anywhere in the world. Against an operating cost near $1000, it covers under a fifth. There is no jurisdiction in which building one of these machines is a commercial proposition, and none in prospect.

And emitting is free almost everywhere

The subsidy is only half the accounting. The other half is what it costs to put the carbon there in the first place, and the answer is usually nothing.

About 30% of the world’s carbon dioxide emissions carry a price of any kind. That is up from roughly 7% in 2010, most of the rise coming from China’s emissions trading system, which covers around 15% of global emissions on its own — though only about half of China’s own emissions, mostly the electricity sector. The corresponding statement is the one worth writing down: about 70% of the world’s emissions have no carbon price at all.

And where there is a price, it is mostly a small one. Most priced emissions are valued at $10 (£8) a tonne or below. Western European prices of $65 to $90 (£51 to £71) apply to about 5% of global emissions. Under 0.5% of all emissions face a price above $100 (£79) a tonne.¹⁷

Now set that against this chapter’s own figure of about $1000 (£790) a tonne to run a direct air capture plant, and against a number the chapter has not yet used. Estimates of the social cost of carbon — the damage a tonne actually does — cluster above $100 (£79) a tonne; a meta-analysis of 147 studies gives a median of $103 and a 2024 analysis $132.

So there are three numbers and they are each roughly an order of magnitude apart. Emitting a tonne typically costs nothing, or a few dollars. The damage it does is worth about a hundred. Undoing it costs about a thousand. No subsidy of the kind chapter 23 describes closes a gap of that shape, because the gap is not between the machine’s cost and its revenue. It is between the cost of un-emitting and the cost of emitting, and the second of those is zero for seven-tenths of the world’s carbon.

This is the sharpest form of the argument for a carbon price, and it is not the obvious one. It also bears on chapter 29, which finds that Europe’s price did useful work at the margin but called almost no carbon capture into existence, and concludes that MacKay’s ordering of price above regulation should be reversed. The arithmetic here explains only half of that. Direct air capture was never among the things any carbon price could have bought, at the $1000 a tonne this chapter records. But chapter 29’s puzzle is the cheaper case — capture on a flue, where chapter 23 records a subsidy that now exceeds the cost at the best sites. That one was within reach of prices that existed, and it still was not built, which is precisely why chapter 29 reaches for regulation rather than for a higher price. A price high enough to make direct air capture pay would have to be around $1000 a tonne — ten times the damage the tonne does, and a hundred times what most priced emissions currently pay. Long before it reached that level it would have made every cheaper abatement in this book pay, most of them many times over: the insulation of chapter 21, the heat pumps of chapter 7, the wind of chapter 4. The case for pricing carbon is not that it would finance this chapter’s machines. It is that this chapter’s machines are the very last thing it would finance — which is, once again, MacKay’s own ordering, arrived at from the other end.

And the cheap alternative he treats gently turned out to be largely fictitious

Section B of this chapter is trees, and MacKay is even-handed about them. What has happened since is that the voluntary offset market — the mechanism through which almost all tree-based carbon removal was actually financed — lost its credibility. A 2023 investigation concluded that more than 90% of the rainforest credits certified by the largest standards body were likely to be “phantom credits” representing no real reduction. The certifier disputed the methodology, and the dispute is not settled; the market repriced regardless, and several large buyers withdrew.¹⁸

So the position is worse than in 2008 in both directions: the machine is dearer than assumed, and the natural alternative it was supposed to compete with turned out to have been partly imaginary.

Which makes the title more right than he meant it

MacKay’s reasons for putting this last were the energy cost, and the worry that “raising the possibility of fixing climate change by this sort of geoengineering might promote inaction now.” Both have been vindicated in the only way that counts: the energy cost is larger than he assumed, and the possibility has been used exactly as he feared.

But there is a sharper version of the argument that he does not make, and this edition would add it. Carbon dioxide is about 0.04% of the atmosphere and roughly 10% of a power station’s flue gas — a concentration ratio of some two hundred and fifty to one, and that ratio is the whole difficulty. Chapter 23 records what happened to capture at the easy end: 0.14% of world emissions actually operating, after thirty years of expectation and with a subsidy that now exceeds the cost at the best sites.

If capture from a flue at 10% concentration is built at one part in seven hundred, capture from open air at a two-hundred-and-fiftieth of that concentration is not a policy option. It is a research programme. The useful question is not how to make direct air capture cheap. It is why the far easier version was not built when it could have been — and chapters 23 and 29 answer that, and the answer is not physics.

Notes

---

1. climate change… was a controversial question. Indeed there still is a “yawning gap between mainstream opinion on climate change among the educated elites of Europe and America” [voxbz].

2. Where is the carbon? Sources: Schellnhuber et al. (2006), Davidson and Janssens (2006).

3. The rate of fossil fuel burning… Source: Marland et al. (2007).

4. Recent research indicates carbon-uptake by the oceans may be reducing. www.timesonline.co.uk/tol/news/uk/science/ article1805870.ece, www.sciencemag.org/cgi/content/abstract/1136188, [yofchc], Le Quéré et al. (2007).

5. roughly half of the carbon emissions are staying in the atmosphere. It takes 2.1 billion tons of carbon in the atmosphere (7.5 Gt CO₂) to raise the atmospheric CO₂ concentration by one part per million (1 ppm). If all the CO₂ we pumped into the atmosphere stayed there, the concentration would be rising by more than 3 ppm per year – but it is actually rising at only 1.5 ppm per year.

6. Radioactive carbon… has penetrated to a depth of only about 400 m. The mean value of the penetration depth of bomb ¹⁴C for all observational sites during the late 1970s is 390±39 m (Broecker et al., 1995). From [3e28ed].

7. Global warming greater than 1 °C would possibly melt methane hydrates. Source: Hansen et al. (2007, p1942).

8. Shoving the CO₂ down a hole in the ground or deep in the ocean. See Williams (2000) for discussion. “For a large fraction of injected CO₂ to remain in the ocean, injection must be at great depths. A consensus is developing that the best near-term strategy would be to discharge CO₂ at depths of 1000–1500 metres, which can be done with existing technology.” See also the Special Report by the IPCC: www.ipcc.ch/ipccreports/srccs.htm.

9. Table 31.5. Inescapable cost of concentrating and compressing CO₂ from thin air. The unavoidable energy requirement to concentrate CO₂ from 0.03% to 100% at atmospheric pressure is kT ln 100/0.03 per molecule, which is 0.13 kWh per kg. The ideal energy cost of compression of CO₂ to 110 bar (a pressure mentioned for geological storage) is 0.067 kWh/kg. So the total ideal cost of CO₂ capture and compression is 0.2 kWh/kg. According to the IPCC special report on carbon capture and storage, the practical cost of the second step, compression of CO₂ to 110 bar, is 0.11 kWh per kg. (0.4 GJ per t CO₂; 18 kJ per mole CO₂; 7 kT per molecule.)

10. In 2005, the best methods for carbon capture were quite inefficient: the energy cost was about 3.3 kWh per kg, with a financial cost of about $140 per ton of CO₂. Sources: Keith et al. (2005), Lackner et al. (2001), Herzog (2003), Herzog (2001), David and Herzog (2000).

11. Wallace Broecker, climate scientist. . . www.af-info.or.jp/eng/honor/hot/enrbro.html. His book promoting artificial trees: Broecker and Kunzig (2008).

12. The best plants in Europe capture carbon at a rate of roughly 10 tons of dry wood per hectare per year. Source: Select Committee on Science and Technology.

13. Enhanced weathering of rocks. See Schuiling and Krijgsman (2006).

14. Ocean nourishment. See Judd et al. (2008). See also Chisholm et al. (2001). The risks of ocean nourishment are discussed in Jones (2008).

15. Operating direct air capture plants consume roughly 1500–2500 kWh per tonne of carbon dioxide captured, combining electrical energy for fans, pumps and compression with thermal energy for sorbent regeneration; the split and the total vary substantially between the solid-sorbent and liquid-solvent approaches, and figures are often reported for only one of the two energy streams, which is a frequent source of confusion in the literature. Fan energy alone has been estimated at 300–900 kWh per tonne, and around 370 kWh per tonne for the Climeworks Orca plant specifically. Two cautions. Comparing these totals against MacKay’s 0.55 kWh/kg is not quite like for like, since low-grade heat is not the same commodity as electricity and Iceland’s geothermal supply makes it unusually cheap there; on a pure exergy basis the gap would be narrower than three to five times. And these are first-of-a-kind machines, so improvement should be expected — the point is the starting position, not a claim that it cannot fall.

16. Climeworks’ Mammoth plant near Hellisheiði, Iceland, began operating in May 2024 with a nameplate capacity of up to 36 000 tonnes of CO₂ a year; reporting in 2025 put its 2024 capture at about 105 tonnes, and its predecessor Orca at a comparably small fraction of a 4000-tonne nameplate. Commissioning years are not representative of steady-state performance and the company has disputed characterisations of the plant as failing; what is not disputed is the operating cost, which its co-chief executive described as nearer $1000 than $100 a tonne, or the staff reductions announced in 2025. World direct air capture capacity of a little over 10 000 tonnes a year across 27 commissioned plants is the International Energy Agency’s, most of those plants being demonstration units rather than commercial operations. The seconds-and-minutes comparisons use energy-related CO₂ of 35 800 Mt a year from the Energy Institute’s Statistical Review of World Energy 2026, which is 1135 tonnes a second.

17. Carbon-pricing coverage and price levels are from Our World in Data’s How much are people across the world paying for their carbon emissions?, drawing on Dolphin and Merkle (2024) and the World Bank’s carbon pricing dashboard (2025): about 30% of global CO₂ emissions priced in 2024 against roughly 7% in 2010, China’s system covering some 15% of global emissions but only about half of China’s own, most priced emissions at or below $10 a tonne, Western European prices of $65–90 over about 5% of emissions, and under 0.5% of all emissions facing a price above $100. Social-cost-of-carbon figures are the median of $103 per tonne across 147 studies in a published meta-analysis and $132 in a 2024 analysis; the literature is wide and the number is sensitive to the discount rate above almost everything else, so “above $100” is the defensible statement rather than any single value. Three cautions on the coverage figure. Coverage is not stringency — an emission inside a trading system whose permits clear near zero is counted as priced, which is why the share and the level have to be read together and why the 30% on its own would flatter the position. Free allocation of permits, which is extensive in the European system for trade-exposed industry, means the headline price is not what every covered emitter actually pays at the margin. And the comparison drawn in the text is between a price per tonne emitted and a cost per tonne removed, which are the same unit but not the same transaction: a price changes behaviour at the margin, whereas the removal cost has to be paid in full on every tonne. The $1000 per tonne for direct air capture is this chapter’s own operating figure, and the $180 45Q credit is chapter 23’s. Currency conversions use the mid-2026 rate of 1.27 dollars to the pound given in chapter I, to two significant figures.

18. A joint investigation by The Guardian, Die Zeit and SourceMaterial published in January 2023, drawing on peer-reviewed analyses of Verra-certified rainforest projects, concluded that more than 90% of the credits examined were likely to be “phantom credits” not representing genuine emissions reductions. Verra rejected the analysis and criticised the counterfactual baselines used; subsequent academic work has supported the direction of the finding while disagreeing about magnitude, and Verra has since revised its methodologies. This edition records that the credibility of the market collapsed and that buyers withdrew, which is not in dispute, rather than adjudicating the percentage. Note that the finding concerns avoided-deforestation credits specifically, not tree planting or forest restoration, which raise different questions about permanence and additionality.