Direct Air Capture (DAC) is the use of engineered equipment — fans, sorbents, solvents, or electrochemical cells — to separate carbon dioxide from ambient air, concentrate it, and hand it off to either permanent storage or industrial use. Unlike point-source capture at a power plant or cement kiln, DAC has no flue gas to draw from: it works against an atmospheric concentration of roughly 425 ppm, which is the central reason it is both energy-intensive and, when paired with geological storage, one of the most durable carbon dioxide removal (CDR) options available. For buyers and policymakers building portfolios with century-plus permanence, DAC sits alongside mineralization and bio-oil sequestration as a small-but-growing share of the durable removals market.
How it works
Two thermodynamic families dominate today. Solid-sorbent DAC uses amine-functionalized materials, metal-organic frameworks, or calcined limestone packed into contactors. Air is pushed across the sorbent, CO₂ binds chemically or physically, and the sorbent is then regenerated by some combination of heat (typically 80–120 °C for amines, ~900 °C for limestone), vacuum, and steam — the “temperature-vacuum swing adsorption” (TVSA) cycle whose life-cycle assessment was characterised by Deutz and Bardow (2021). Liquid-solvent DAC absorbs CO₂ into an alkaline solution (hydroxide or carbonate), then drives it off thermally or, increasingly, via an electrochemical pH swing.
Energy demand is the binding constraint. Comparative techno-economic work by Sabatino et al. (2021) puts heat-and-electricity requirements between roughly 5 and 10 GJ per tonne of CO₂ for current designs, with costs ranging across an order of magnitude depending on assumptions about heat source, capacity factor, and learning rates. A broader review by Erans et al. (2022) lays out the process-engineering trade-offs — contactor pressure drop, sorbent degradation, water balance — that separate a workable pilot from an economic plant. Newer sorbent classes (MOFs, electrochemical cells, moisture-swing materials) are surveyed in Sodiq et al. (2022) and in the porous-materials review by Ding et al. (2021).
Who’s doing it
The supplier list is now in the high dozens. A non-exhaustive cut illustrating the range of approaches:
- Climeworks operates the only two commercial-scale solid-sorbent plants running today: Orca (4,000 t/yr) and Mammoth (nameplate 36,000 t/yr) in Iceland, both paired with Carbfix basalt mineralization.
- Heirloom uses calcined limestone as the sorbent — a slow carbonation–calcination loop — at a facility in Tracy, California, with stored CO₂ routed to mineralization partners.
- Origen also runs a lime cycle, but built around a proprietary zero-emission kiln rather than amine sorbents.
- Avnos builds a “hybrid” moisture-swing system that co-produces roughly five tonnes of water per tonne of CO₂ and avoids external thermal input.
- Mission Zero and Phlair are both pursuing electrochemical routes — Mission Zero with an aqueous solvent, Phlair with a hydrogen-driven pH swing — pitched at intermittent renewable power.
- RepAir takes the electrochemistry further, using a solid-state cell with no heat or solvent at all.
- Octavia Carbon is the most-developed non-OECD pure-play, running solid-sorbent units on Kenyan geothermal power and injecting into Rift Valley basalt.
- Deep Sky is not a technology developer but a project aggregator, hosting multiple third-party DAC units at Canadian sites with geological storage.
- Atoco, spun out of Omar Yaghi’s lab, is commercialising MOF and COF sorbents — earlier in the development curve than the operators above.
Manufacturers like Carbyon, DACMA, Skytree and Sirona sell equipment rather than removal tonnes, reflecting an emerging split between hardware vendors and project developers.
Durability and verification
Durability for DAC is, in principle, the cleanest story in CDR: once CO₂ is dissolved in formation brine or mineralized in basalt, the geophysical retention horizon is measured in centuries to millennia. The qualifier is “in principle”. Leakage risk is essentially the same as for conventional CCS — well integrity, caprock characterisation, induced seismicity — and the body of operating evidence remains modest. The measurement, reporting and verification (MRV) chain has to track three things: how much CO₂ was actually pulled from air (metered at the contactor outlet), how much was injected (metered at the wellhead), and the embodied emissions of the energy and materials used to do it. The last item is where life-cycle assessments such as Deutz and Bardow (2021) and the PRospective EnvironMental Impact asSEment framework (2022) matter: a DAC plant on a marginal gas grid can have an effective removal efficiency well below 1, and registries differ on how strictly they require additionality of clean power.
Mineralization pathways (Climeworks/Carbfix, Octavia/Cella) shorten the monitoring tail because the CO₂ is fixed as carbonate within roughly two years; saline-aquifer storage requires longer-duration monitoring plans. Puro.earth, Isometric and Verra each publish DAC methodologies, but tonne accounting still varies between them, and a single project may issue credits under different rules in different vintages.
Open questions
A reader returning in a year should look for: whether any plant beyond Mammoth is reliably operating above 50% of nameplate; whether electrochemical and moisture-swing systems clear pilot scale with credible energy numbers; what the first post-IRA cost curve looks like in the U.S. after the DOE DAC Hubs reshuffle; how many tonnes have actually been delivered against the roughly 10 million tonnes of forward DAC offtake already contracted; and whether MRV protocols converge on a single defensible standard for grid-electricity accounting. The pathway’s eventual share of the durable CDR market depends less on chemistry than on the answers to those.