In 2024, Climeworks switched on Mammoth, a plant in Iceland it called the world's largest facility for pulling carbon dioxide straight out of the air, with a nameplate capacity of up to 36,000 tonnes a year [source: Climeworks, 2024]. A year later, Occidental's 1PointFive said its far bigger Stratos plant in Texas — designed for up to 500,000 tonnes a year — would start up by the end of 2025 [source: Hart Energy, 2025]. Corporate buyers, meanwhile, went on a spree: durable carbon-removal contracts reached 30.4 million tonnes in 2025, and Microsoft alone signed enough deals to more than double its cumulative commitments to some 45 million tonnes [source: CDR.fyi, 2025]. The industry that promises to reverse emissions, not just avoid them, is suddenly everywhere.
And yet the single most important number in the whole field is a small one. As of December 2025, the total amount of durable carbon dioxide removal (CDR) ever actually delivered and retired — captured, verified and counted — had only just crossed one million tonnes [source: CDR.fyi, 2025]. That is the tension this article is about. Carbon removal is real, it is scaling, and most climate scientists agree the world will need some of it. But the gap between what is pledged, what is purchased and what is verifiably captured is enormous, and reading that gap honestly is the difference between clear-eyed hope and expensive wishful thinking. None of this is investment advice; it is a map of a young and contested industry.
Table of Contents
- Why carbon removal is suddenly everywhere
- The three numbers that rarely match
- A field guide: from trees to machines
- Essential complement, or dangerous distraction?
- The carbon market's credibility reckoning
- What to watch
Why carbon removal is suddenly everywhere
The case for carbon removal starts with arithmetic that even its critics accept. The Intergovernmental Panel on Climate Change (IPCC), in its Sixth Assessment Report, concluded that carbon dioxide removal is "unavoidable" if the world is to reach net-zero, because some emissions — from aviation, heavy industry and agriculture — are extremely hard to eliminate, and something has to balance them out [source: IPCC, 2022]. Every modelled pathway that holds warming to 2°C or below includes at least some CDR. The panel was equally clear about the flip side: removal "cannot substitute" for deep, immediate cuts in emissions. It is the mop, not a licence to keep spilling.
The scale required is daunting. The Oxford-led "State of Carbon Dioxide Removal" report estimates that meeting the 1.5°C target implies removing roughly 7 to 9 billion tonnes of CO2 a year by mid-century [source: University of Oxford, 2024]. Today the world removes about 2 billion tonnes annually, but almost all of that comes from conventional methods — chiefly planting and managing trees. The "novel" methods that attract the headlines and the venture capital — engineered direct air capture, bioenergy with carbon capture, enhanced rock weathering, biochar — together account for only about 1.3 million tonnes a year, less than 0.1% of the total [source: University of Oxford, 2024]. Closing that gap means scaling novel removal 25- to 100-fold in under 25 years.
That combination — genuine necessity plus a vast unmet gap — is why money is flooding in. When a technology is both needed and barely deployed, it looks less like a cost and more like a market waiting to be built.
The three numbers that rarely match
If you take away one habit of mind from this article, make it this: whenever you see a carbon-removal figure, ask which of three very different things it measures.
The first is pledged capacity — what a plant is designed to remove. This is the number in press releases, and it is a claim, not a receipt. Mammoth's 36,000-tonne figure and Stratos's 500,000-tonne figure are both nameplate capacities [source: Climeworks, 2024; source: Hart Energy, 2025]. Nameplate is not net removal: Climeworks itself uses the language of a "carbon removal production waterfall" to describe how the headline capacity shrinks, step by step, into the smaller amount of carbon actually and permanently removed [source: Climeworks, 2024]. Real plants also ramp slowly. Mammoth was still installing its modular capture units through 2024, so in its first year it captured only a small fraction of that headline number. And capacity can slip: Stratos hit an unexpected snag during commissioning, and by mid-2026 its operator had not announced a revised start-up date [source: Carbon Herald, 2026].
The second number is purchased or contracted removal — tonnes a company has agreed to pay for, usually to be delivered years in the future. This is where the market looks explosive. Frontier, an "advance market commitment" launched in 2022 by Stripe, Alphabet, Shopify, Meta and McKinsey, doubled its guarantee to $1.8 billion in 2026, with more than 350 corporate buyers having purchased nearly four million tonnes over its lifetime [source: Frontier Climate, 2026]. Microsoft's buying dominates the market so completely that it accounted for roughly nine-tenths of global offtake in the first half of 2025 [source: CDR.fyi, 2025]. But a contract is a promise about the future, not carbon already out of the sky.
The third number — the one that matters most and gets quoted least — is delivered and verified removal: carbon actually captured, stored and independently counted. Here the totals collapse. Frontier's portfolio companies delivered only about 23,000 tonnes in 2025 [source: Frontier Climate, 2026]. And across the entire durable-CDR market, cumulative deliveries crossed just one million tonnes for the first time in December 2025 [source: CDR.fyi, 2025]. Set the numbers side by side and the shape of the industry appears: one year's contracts (30.4 million tonnes) were about thirty times all the durable removal ever delivered. None of this means the field is a fraud — early industries always sell futures before they ship product. But a purchase is not a removal, and a nameplate is neither.
A field guide: from trees to machines
"Carbon removal" is not one thing; it is a spectrum, and the key axis is durability — how long the carbon stays put. Research suggests that to truly offset fossil emissions, storage should last on the order of 1,000 years; storing carbon for only a century, one analysis found, would leave meaningful extra warming centuries later [source: Nature Communications Earth & Environment, 2024].
At the fast, cheap, familiar end are nature-based methods: reforestation and soil carbon. They are easy to scale and bring side benefits for ecosystems, but their durability is uncertain — a forest can burn, be logged or be ploughed back into the atmosphere. At the slow, costly, durable end are engineered and geological methods. Direct air capture (DAC) uses machines to filter CO2 from ambient air; bioenergy with carbon capture and storage (BECCS) grows biomass, burns it for energy and buries the emissions; both can inject CO2 into rock formations where it can stay for millennia. In between sit enhanced rock weathering, which spreads crushed basalt on fields so it reacts with CO2 and locks it into bicarbonates and minerals for over a thousand years, and biochar, which bakes plant waste into a stable charcoal that holds carbon for centuries [source: Nature Reviews Earth & Environment, 2025].
Durability comes at a price, in both money and measurement. The International Energy Agency puts the cost of engineered DAC at up to roughly $540 a tonne for solid-based systems today, with liquid-based systems somewhat lower — far above nature-based credits, though costs are expected to fall with scale [source: IEA, 2022]. Verification is its own frontier: enhanced weathering, for instance, is genuinely durable but notoriously hard to measure, because tracking the carbon as it dissolves and travels toward the ocean strains current sampling methods [source: Nature Reviews Earth & Environment, 2025]. This is why the market has been tilting toward the most durable, most measurable options even though they cost the most.
Essential complement, or dangerous distraction?
Here the debate turns genuinely contested, and honest coverage has to hold two views at once.
On one side, carbon removal is framed as an essential complement to emissions cuts — the only way to mop up the residual emissions that even an aggressively decarbonising world will still produce, and eventually to draw down the overshoot if temperatures breach 1.5°C. On this view, building the industry now, while it is small and expensive, is exactly how solar and batteries once climbed down their cost curves.
On the other side is the charge of "moral hazard," or "mitigation deterrence": the worry that the mere promise of future removal gives governments and companies an excuse to keep emitting today, slowing the cuts that matter most [source: WIREs Climate Change, 2023]. If a firm can buy its way to "net zero" with cheap offsets, why do the hard work of actually decarbonising? The evidence here is genuinely mixed. Some studies find signs that removal can crowd out mitigation; others find no such effect, or even the opposite — that learning about CDR makes people take the climate problem more seriously [source: WIREs Climate Change, 2023]. Researchers who study the risk tend not to conclude that removal is bad, but that it must be governed carefully: separate targets for cutting emissions and for removing carbon, so that one cannot quietly be substituted for the other [source: Carbon Gap, 2024]. The IPCC's own framing — necessary but not a substitute — is really a warning against exactly this substitution [source: IPCC, 2022].
The carbon market's credibility reckoning
The distinction between avoiding emissions and removing them is not academic; it is the fault line along which the carbon market has been shaking. In 2023, an investigation by The Guardian, Die Zeit and SourceMaterial reported that more than 90% of rainforest offset credits certified by Verra, the largest standards body, were likely "phantom credits" representing no real emissions reductions; the analysis suggested one major project type had overstated the threat to forests by around 400% [source: The Guardian, 2023]. Verra rejected the findings, with its markets chief calling it "absolutely incorrect" to say 90% of its credits were worthless [source: The Guardian, 2023]. Whichever way one reads the specifics, the episode shattered confidence in a whole class of cheap, avoidance-based offsets.
That crisis is a large part of why corporate money has migrated toward durable removal, where a tonne stored underground is at least conceptually easier to verify than a tonne of deforestation that supposedly did not happen. But the reckoning is not over — it has moved. Durable removal now faces its own quality questions: how to measure enhanced weathering, how to certify biochar permanence, how to be sure a delivered tonne is really durable. The lesson buyers took from 2023 was to demand measurement, and that same demand is now the pressure test for engineered removal. The credibility problem was not solved; it was inherited by a more expensive part of the market.
What to watch
Carbon removal sits on a real foundation. The IPCC says some of it is unavoidable, the required scale is vast, and durable methods genuinely can lock carbon away for the timescales that matter. But the same evidence counsels discipline. A nameplate capacity is a claim; a purchase is a promise; only a delivered, verified tonne is a removal — and by that strictest measure the industry has, so far, done very little. The 30-fold gap between one year's contracts and all-time deliveries is not necessarily a scandal, but it is the number to keep your eye on.
A few things are worth watching from here. Will delivered tonnes start to catch up with contracted ones, or will the gap keep widening? Will engineered plants like Stratos hit their nameplate capacity, and will DAC costs fall the way solar's did — or stall? Will the measurement standards for enhanced weathering and biochar mature enough to prevent a durable-removal version of the offsets scandal? And politically, will governments set separate, enforceable targets for cutting emissions and for removing carbon, so that removal complements the cuts instead of excusing their delay? The most useful habit a reader can carry into the next headline is simple: when someone quotes a carbon-removal number, ask whether it was captured, contracted or merely claimed.