Direct air capture works as chemistry: machines pull CO2 out of the air,
and the carbon buried in Icelandic basalt stays there. But all the verified
plants in the world removed about 1,500 tonnes in 2025, the flagship plants run far
below their design, and it costs around $1,000 a tonne, against a promise
of 225 million tonnes a year by 2025.
Direct air capture works, at a scale too small to matter yet
As chemistry, yes: machines can pull carbon dioxide out of the air, and the CO2 buried in Iceland’s rock stays buried. As a climate solution, not yet. All the verified direct air capture in the world removed about 1,500 tonnes of CO2 in 20251, roughly one tonne for every 25 million tonnes emitted from fossil fuels2.
The flagship plants run far below what they were built for, the tonnes they do remove cost around $1,000 each, and the company behind them once promised 225 million tonnes a year by 202511114. That gap between promise and delivery is the story of direct air capture (DAC) so far.
That's all the verified direct air capture in the world removed in 2025: what fossil fuels emit in about 1.2 seconds
1,500 tonnes removed, 38 billion emitted
Fossil fuels released a record 38.1 billion tonnes of CO2 in 20252, about 1,200 tonnes every second, so all the verified direct air capture in the world removed roughly what fossil fuels emit in 1.2 seconds, by our own arithmetic. The plants being built now would bring that to about 0.55 million tonnes a year by 20301, still about 70,000 times less than a year’s fossil emissions.
Even among the newer ways of removing carbon, direct air capture is a sliver: 0.07% of the 2.04 million tonnes removed by novel methods in 2025, most of which came from biochar1. Almost all the carbon removal in the world, about 2.2 billion tonnes a year, still comes from forests and other conventional methods1.
The gap is widest against the promises. In 2017 Climeworks, the Swiss company whose two Icelandic plants account for all the verified removal in 20251, set a goal of capturing 225 million tonnes a year by 2025, nearly 1% of global emissions14. The 2025 result was about 150,000 times smaller, by our arithmetic.
The same gap runs through the market. Companies buy carbon removal credits, certificates for one tonne removed, often years in advance. By mid-2025 buyers had contracted 2.47 million tonnes of direct air capture credits, and 1,186 tonnes, about 0.05%, had been delivered8.
The emissions direct air capture is measured against: about 1,200 tonnes every second.
How a machine pulls CO2 out of the air, and why it is so expensive
The idea is simple. Fans push air over a material that binds CO2: Climeworks uses solid filters, while the Carbon Engineering design, now owned by the oil company Occidental, uses a liquid solvent1217. Heating the material releases the CO2 as a pure gas, which is then stored. In Iceland, Climeworks’ partner Carbfix dissolves it in water and pumps it into basalt deep underground6, where it turns to stone. In a pilot at the same site, using CO2 from a power plant and the same method, more than 95% of the injected CO2 had mineralised in under two years20. The storage half of the idea works.
The capture half is where it gets hard, and each step explains the next. CO2 makes up only a tiny fraction of the air, so a machine has to move and heat enormous volumes of air for every tonne it catches. That takes 1,500 to 3,000 kilowatt-hours of energy per tonne captured and stored18. Energy and equipment on that scale put the cost of Orca, the first Icelandic plant, at $1,000 to $1,300 a tonne11.
Against that, the main US subsidy pays $180 a tonne1, less than a fifth of today’s cost by our arithmetic, and buyers become scarce above $500 a tonne8. Even projected forward to a billion tonnes of capacity, the cheapest design is expected to cost $226 to $544 a tonne12. Our reading is that this is the difference from solar, whose panels got cheap by being made in their billions: a machine that must move and heat that much air gets a smaller discount from scale, and the projections above suggest as much.
The biggest plants run far below their design
The largest operating plant is Mammoth, in Iceland, designed to capture up to 36,000 tonnes a year5. In the first half of 2026 it removed 675 tonnes, on the company’s own figures4, about 4% of its design rate over a year by our arithmetic. Registry figures reported by the Icelandic outlet Heimildin put its first ten months at 105 tonnes9.
In June 2025 only 12 of Mammoth’s 72 planned collector units were running, and installation had stopped while engineers fixed problems10.
The smaller Orca plant, rated for up to 4,000 tonnes a year6, captured at most about 1,000 tonnes in any year, according to Heimildin’s reading of the registry and company accounts9. Its shortfall was large enough that the State of Carbon Dioxide Removal report, the main independent assessment of the field, had to revise its own earlier estimate down once registry data came in1.
Climeworks says the fixes are working. Mammoth’s output in early 2026 was more than five times the same period a year earlier, and its operating cost per tonne has fallen by more than half4. Those are company figures, not independently checked.
Direct air capture costs about $1,000 a tonne; the goal was $100
Climeworks’ founders once aimed far lower than today’s price.
The US government adopted the same goal in 2021, aiming for removal at less than $100 a tonne13. Here is where the price stands on each rung. Orca cost $1,000 to $1,300 a tonne11. Climeworks’ own target for 2030 is $400 to $600 a tonne of net removal, that is, after subtracting the plant’s own emissions7. A 2024 study that builds up costs from each part of the machinery projects $226 to $544 a tonne for the cheapest design, even after a billion tonnes of capacity has been built12. No independent projection reaches $100.
The IPCC’s earlier range of $100 to $300 came before that study, and even it puts direct air capture well above options like storing carbon in farm soils, which ranges from a net saving of $45 to a cost of $100 a tonne3.
The energy bill at climate scale
At 1,500 to 3,000 kilowatt-hours a tonne18, removing carbon at the scale climate scenarios discuss would need an enormous share of the world’s power. Removing 7 billion tonnes a year would take about 17,500 terawatt-hours, over half of all the electricity the world generated in 202418. That estimate comes from a paper arguing against large-scale direct air capture, and it assumes the middle of the energy range.
The competition for clean power is already real. Project Bison in Wyoming was paused because data centres and cryptocurrency miners were competing for the same clean electricity1. While clean power is scarce, every clean kilowatt-hour spent pulling CO2 from the air is one not spent replacing a fossil power station.
Oil recovery, subsidies and the greenwashing charge
It depends where the CO2 goes. Tonnes pumped into Icelandic basalt are real removals. But since July 2025, US law has paid the same tax credit, $180 a tonne for direct air capture, whether captured CO2 is stored for good or used to push more oil out of the ground151, and the State of CDR says using it for oil recovery would most likely mean net emissions1.
That is why Occidental draws the sharpest criticism. It bought Carbon Engineering for about $1.1 billion in 2023, is building the Stratos plant in Texas, and describes pumping CO2 into oil fields to recover more oil as critical to its long-term strategy17. Whether Stratos’s own CO2 goes to oil recovery or to storage is not settled in the sources this page uses.
A wider worry is that promises of carbon removal delay real cuts to emissions, which researchers call mitigation deterrence. Studies find the risk built into climate models; whether it happens in real policy is still argued over19.
Who is building it, and who is paying
Two companies dominate. Climeworks, the company whose co-founder Christoph Gebald set the $100 goal, runs the Icelandic plants, with Carbfix storing the CO2614. Occidental owns the Carbon Engineering technology and Stratos, designed to capture up to 500,000 tonnes a year17 but not yet running: its owner now expects operations in 202716.
The money comes from governments and a handful of buyers. The US set aside $3.5 billion for regional direct air capture hubs and had awarded $1.2 billion when, in October 2025, it cancelled 10 of the 21 projects; funding for the two headline hubs was restored by April 2026, but the State of CDR says uncertainty remains1. On the buying side, Microsoft alone has contracted 833,000 of the 2.47 million tonnes of credits sold so far8.
Why researchers still want it built
Even the need is unsettled. The IPCC’s pathways to 1.5 °C use anywhere from zero to 310 billion tonnes of direct air capture over the rest of the century, so some pathways need none at all3. The State of CDR’s most ambitious scenarios use from under 0.1 to 1.8 billion tonnes a year of it by 20501, somewhere between about 180 and 3,300 times the 0.55 million tonnes a year expected by 2030, by our arithmetic.
The case for building anyway is that some emissions will be hard or impossible to avoid, and removal will be needed to balance them.
Subsidies and one big buyer hold it up
Direct air capture runs on public money and a few buyers, and both have wobbled: the US hub cancellations in 2025, and a market where buyers thin out above $500 a tonne18. Stratos, the plant meant to show the technology at scale, has already slipped from its earlier timetable to operations in 202716. Our reading is that without costs falling several times over, demand beyond Microsoft and a few others will stay thin.
Millions of verified tonnes, or little changes
The first real test comes when Stratos starts running. Even at its full design, which neither Icelandic plant has come close to, it would remove about 0.001% of a year’s fossil emissions, what they release in about seven minutes, by our arithmetic172.
Meanwhile the problem is growing. Fossil emissions rose 1.1% to their record in 20252, and the gap between the carbon removal countries plan and what the Paris goals need has grown with them1. The evidence that would change this picture is delivered, verified tonnes in the millions, at a few hundred dollars each, with the CO2 stored for good. Plant announcements and credit sales don’t count until those tonnes exist.