A plant in Iceland pulls carbon dioxide out of ordinary outdoor air, then locks it underground as stone. That idea used to sound like a science-fair demo. In a September 2026 update, Climeworks said upgrades at its Mammoth site more than doubled capture performance on refurbished units, cut operating cost per ton by more than half, and pushed first-half 2026 net removals to 675 tons — more than five times the same period in 2025.
Direct air capture, or DAC, means machines fan ambient air through filters that chemically bind CO₂. Heat then releases a concentrated stream that can be stored. Mammoth sits next to the Hellisheiði geothermal plant in southwest Iceland. Geothermal partner ON Power supplies low-temperature heat and electricity. Storage partner Carbfix dissolves the captured CO₂ and injects it into basaltic rock, where it mineralizes into solid carbonate — permanently, on human timescales.
Think of it as an industrial lung. Fans pull in dilute outdoor air. A solid sorbent — the filter material — grabs CO₂ molecules that make up only a few hundred parts per million of that air. When the filter is full, operators heat it enough to release a richer CO₂ stream, then send that stream to storage. The hard parts are energy, filter lifetime, and reliability in weather that does not care about your spreadsheet.
What Mammoth was designed to do
Climeworks broke ground in June 2022 and started operations in May 2024. The company billed Mammoth as about ten times larger than its earlier Iceland plant, Orca. Design nameplate capacity is up to 36,000 tons of CO₂ per year once the full modular plant is running. At startup, twelve of seventy-two planned collector containers were on site. Nameplate is a peak engineering figure, not a promise of certified tons delivered every year. Downtime, weather, maintenance, and “grey” emissions from running the plant all shrink net carbon dioxide removal (CDR) relative to that headline number.
Climeworks’ own May 2024 startup release was careful about sequencing: first CO₂ captured with a fraction of the planned boxes installed, then continued build-out. Modular containers are the point. You can add capacity in chunks, swap upgraded guts into existing frames, and learn from one module before you bet the next megaton hub on untested hardware. That is also why early underperformance was so visible. The boxes were there; the filters and mechanics inside them were still learning.
Independent project registries and on-the-ground reporting have been blunt about the gap. Early certified credit volumes sat far below nameplate while Climeworks iterated on sorbent chemistry and hardware. A January 2026 geoCDR summary noted that full-scale ramp-up for Mammoth was still projected years out, even as the company pushed Generation 3 filter materials meant to capture more CO₂ with less energy and longer lifetimes. Latitude Media’s site visit reporting likewise stressed that only a dozen of the planned seventy-two containers were fully operational years after groundbreaking — a reminder that “world’s largest” can describe ambition as much as instantaneous output.
The September 2026 performance claim
In its performance update, Climeworks co-founder and co-CEO Jan Wurzbacher described eighteen months of operating data. Two collector containers with new sorbent and mechanical upgrades had run for more than six months. Those units, the company said, were reaching the intended peak daily run rate of about 1.37 tons of captured CO₂ per container — roughly double the originally deployed containers — with day-to-day swings from weather and operations.
Net CDR is the customer-facing metric: tons permanently removed after losses and plant emissions are subtracted. Mammoth produced 675 tons of net CDR in the first half of 2026, versus 119 tons in the first half of 2025. Capacity factor on the upgraded containers was described as about 40–50%, which Climeworks compared with typical solar (around 20–25%) and wind (around 30–40%) capacity factors. Operating cost per ton fell by more than 50% over the prior year, according to the same update.
Read those figures as engineering progress, not climate victory laps. Six hundred seventy-five net tons in six months is still a tiny slice of the 36,000-ton nameplate story. The useful signal is relative: same site, same company, same Iceland weather — much higher net removal and much lower operating cost after sorbent and mechanical changes. Bloomberg’s September 2026 coverage framed the same shift as roughly a sixfold jump in capture alongside the cost cut, matching the company’s first-half comparison.
Basalt like this is why Iceland can turn captured CO₂ into stone: CarbFix dissolves the gas and injects it into reactive volcanic rock, where it mineralizes.
Those numbers still leave Mammoth far below its ultimate nameplate. The near-term plan is narrower and more practical: roll the upgraded sorbent and mechanical package across all twelve collector containers in one plant module by the end of 2026, then begin testing next-generation DAC hardware at Mammoth in early 2027. Lab claims in the same update include a tenfold improvement in sorbent lifetime and a fourfold process densification inside the same module size — pathways to fewer boxes and less frequent filter replacement if they survive real Icelandic weather.
Why basalt storage matters
Capture is only half the story. CarbFix’s mineralization approach mixes CO₂ with water and injects it into reactive basalt. Within years, much of the carbon becomes carbonate minerals inside the rock. That is different from injecting supercritical CO₂ into deep saline aquifers and monitoring it for decades. Iceland’s geology and cheap geothermal heat make the pairing unusually favorable. Other DAC projects, including large U.S. hubs still in development, will need their own power, heat, and storage solutions — and those choices dominate cost and public acceptance.
Puro.earth’s facility notes for Mammoth put a finer point on nameplate versus reality. Even when the plant is complete, effective net removal after losses and grey emissions has been estimated around 22,000 tons per year — still substantial for a commercial DAC site, and still far below the raw capture headline. Buyers of removal credits care about that net number, plus third-party verification. Marketing slides care about nameplate. Serious scale-up has to live in the gap between them.
Cost remains the industry’s binding constraint. Analysts and registries still treat DAC as expensive relative to many nature-based removals, and early Mammoth credit volumes showed how hard first-of-a-kind plants are. A more-than-halved operating cost and doubled unit throughput do not by themselves prove gigaton readiness. They do show that field data, not just slide decks, can move the cost curve when a company is willing to keep the plant running while it redesigns the guts.
What to watch next
Three checkpoints matter more than any single press figure. First, whether the twelve-container module upgrade through late 2026 reproduces the two-container results at plant scale. Second, whether certified third-party credit issuances track the company’s net CDR claims as volumes rise. Third, whether next-generation hardware tested in 2027 keeps cutting energy use and sorbent replacement without new reliability surprises.
There is also a policy and finance layer. Climeworks has pointed to U.S. Department of Energy megaton hub work, including Project Cypress in Louisiana, as the next scale jump beyond Iceland. Those hubs will not inherit Iceland’s geothermal free lunch. They will have to prove heat, power, and storage stacks that work in different geologies and electricity markets. Mammoth’s job, in that frame, is to be the harsh outdoor laboratory that kills bad designs early.
Mammoth will not, by itself, bend the global emissions curve. Even a fully ramped 36,000-ton plant is tiny next to annual fossil CO₂. The useful question is whether repeated, measured upgrades at operating plants can drive costs into a range where policy, buyers, and project finance can stack megaton hubs. Climeworks’ September 2026 numbers are a step in that direction — still early, still expensive relative to the climate problem, but no longer stuck at pilot-plant anecdote.
As an Amazon Associate I earn from qualifying purchases. If you buy through links on this page, I may earn a commission at no extra cost to you.
Further reading
Drawdown: The Most Comprehensive Plan Ever Proposed to Reverse Global Warming — Paul Hawken’s edited survey of climate solutions, including carbon removal pathways, written for non-specialists.
How to Avoid a Climate Disaster — Bill Gates on the breakthroughs still needed for zero emissions — including hard-to-abate sectors and carbon removal.