4000 words from Claude. Only firm conclusion: success at a appropriate cost is unproven. This is not a backgrounder, not a factchecked article. I’m still doing research.
My gut is that carbon capture is not working out very often. But some experts and a lot of companies remain hopeful. Not my expertise, at least yet. Dave
Carbon capture has been sold for two decades as the technology that lets the world keep burning fossil fuels and making cement while still hitting net zero. The pitch is simple: grab the CO2 before it reaches the atmosphere, or pull it straight out of the air, and bury it for good. The empirical record — more than 100 operating projects, a growing pile of independent audits, and almost 30 years of monitoring data from the first storage site — tells a messier story than the pitch does. Some of it is genuinely encouraging. Much of it is not.
The scale problem, first
Start with the numbers everyone agrees on. The Global CCS Institute’s Global Status of CCS 2026 report, using data as of July 31, 2026, counts 113 commercial carbon capture and storage (CCS) projects in operation worldwide, with combined capture capacity of 68 million tonnes of CO2 per year (Mtpa) — up 6% from the year before. A further 56 Mtpa is under construction, and the global pipeline of announced projects totals 522 Mtpa. The IEA’s own February 2026 project baseline puts operating capacity slightly lower, at 62.5 Mtpa, with 52.4 Mtpa under construction and 471.9 Mtpa planned (operating and planned figures aren’t additive, since they track overlapping phases of the same projects).
Whichever count you use, it’s worth holding next to global emissions. Energy-related CO2 emissions hit roughly 38.4 billion tonnes in 2025, per IEA data. That means today’s entire global CCS fleet captures somewhere around 0.15–0.2% of annual energy emissions — a figure that has barely moved off that mark since 2014, when 22 projects captured about 40 Mt. The IEA’s net-zero modeling has historically called for roughly 1 gigatonne of CCS capacity by 2030 and far more by mid-century; even the most optimistic “everything gets built” pipeline scenario for 2030 (about 430 Mtpa) lands at under half that. Project lead times don’t help: the IEA puts the average span from design to commissioning at about six years.
The geographic picture is lopsided. The Americas lead with 63 operating projects and roughly 48 Mtpa of capacity (the US alone accounts for close to half of global operating capacity). Europe, despite heavy policy support, operates only a few Mtpa today, though it has the largest build-out pipeline if announced projects proceed. China is the fastest-growing newcomer — more than 120 planned and operational CCUS demonstration projects as of early 2026, with roughly 9.4 Mtpa of capture capacity and 5.9 Mtpa of injection capacity, according to the government-affiliated China Carbon Capture, Utilization and Storage Annual Report. Sinopec’s Qilu-Shengli project, online since 2022, remains China’s flagship megaton-scale effort, feeding captured CO2 into enhanced oil recovery at the Shengli oilfield.
What “capture” usually means in practice: enhanced oil recovery
A detail that gets lost in capacity tallies: most of the CO2 captured worldwide isn’t going into dedicated storage at all. It’s going into oilfields. Estimates vary by year and source, but a consistent picture emerges from the Global CCS Institute, EIA-adjacent analysts, and multiple industry surveys: somewhere between 70% and 90% of CO2 captured historically has been used for enhanced oil recovery (EOR), where it’s injected into depleted reservoirs to push out more crude. Some of that CO2 stays underground, which counts toward storage totals — but the entire point of the exercise is producing more oil, which is then burned, releasing new emissions. A 2025 peer-reviewed analysis in Energy & Environmental Science found that under even optimistic assumptions, CO2 injected via EOR displaces far less reservoir pore space than the oil it helps extract eventually re-emits when burned — meaning “carbon-neutral oil” from EOR is not achievable in most conventional reservoirs. Dedicated geological storage with no oil-production component — the kind climate modeling actually assumes — remains a minority of global CCS activity, though it’s the fastest-growing segment of the pipeline.
The flagship-project record: more misses than hits
If you want to know whether CCS works as advertised, the most useful evidence isn’t capacity statistics — it’s what happens at individual, audited projects once they’re actually running.
Gorgon, Australia. Chevron’s Gorgon LNG project, paired with what was billed as the world’s largest CCS facility, was approved on condition that it capture and inject roughly 80% of the CO2 extracted from its gas fields — about 4 million tonnes a year. It has never hit that target. Water-handling and equipment problems plagued the system from its 2019 start; in FY2020–21 it stored 2.26 Mt against a 4 Mt target, and the shortfall over its first five years ran to roughly 5.23 million tonnes. Chevron and partners ExxonMobil and Shell agreed to buy carbon offsets estimated at up to A$200 million (roughly US$184 million at the time) to cover the gap, and underperformance has continued into subsequent reporting periods, according to IEEFA’s ongoing tracking. One industry commentator summarized the pattern bluntly in a 2021 Reuters/Bloomberg account: the frequent inefficiencies at Gorgon are “typical of the technical risks involved in CCS projects.”
Petra Nova, Texas. The only commercial coal-plant carbon capture retrofit in the US was mothballed in 2020 after three years of operation, officially because of a crashing oil price (the captured CO2 was sold for EOR). An IEEFA post-mortem argued the economics were shaky well before that: NRG and the Department of Energy never disclosed precisely what it cost per tonne to capture CO2 there; IEEFA’s own estimate of roughly $60/tonne is an extrapolation from general DOE cost statements rather than a Petra-Nova-specific disclosure. JX Nippon restarted the plant in 2023 as what the company itself called a research exercise. Houston Chronicle reporting on the restart noted the plant’s performance had improved over its original run, but its economics had not.
Boundary Dam, Canada. SaskPower’s coal-plant retrofit — the first of its kind, online since 2014 — was designed for a 90% capture rate. A 2024 IEEFA analysis of nine years of operating data found an average capture rate of 57%, driven by both mechanical breakdowns (including a severe 2021 stretch when the unit was down for weeks at a time) and the fact that SaskPower’s actual contractual targets with EOR buyers are lower than the original 90% headline figure — utility spokespeople have put the real operating target closer to 75–80%.
Quest, Canada. Shell’s hydrogen-production CCS project near Edmonton has stored roughly 9 million tonnes of CO2 since 2015 at lower-than-expected cost, and Shell cites it as a success story. But its capture rate — around 77% — remains below the original 90% target, and a 2020 Global Witness analysis found that between 2015 and 2019 the project emitted more CO2 (7.5 Mt) than it captured and claimed credit for (5 Mt) once the full emissions footprint of the hydrogen plant was counted. A 2024 Greenpeace investigation separately alleged Quest had claimed carbon credits for CO2 that was never actually captured; Shell has not directly addressed that specific allegation in the public sourcing reviewed for this piece, though the company maintains Quest has been a success story overall.
Sleipner and Snøhvit, Norway — the clear counter-examples. Not every long-running project tells a story of shortfalls. Sleipner, operating since 1996, is the world’s longest-running dedicated storage project and has become the reference case the whole industry points to: roughly 1 Mt injected per year into the Utsira Sand saline aquifer, with over 20 years of repeated seismic monitoring showing the CO2 plume behaving close to model predictions and no evidence of leakage. Snøhvit (operating since 2008) is similarly cited as “performing close to capture capacity” in IEEFA’s own comparative scorecard — one of only three projects the group found meeting that bar, alongside Shell’s Quest-adjacent Alberta operations in specific years. Between them, Sleipner and Snøhvit had stored more than 22 million tonnes by the end of 2017.
In Salah, Algeria — a caution story with a reassuring ending. Between 2004 and 2011, roughly 3.8–4 million tonnes of CO2 were injected into a depleted gas reservoir. Monitoring (including satellite InSAR measuring millimeter-scale ground deformation) detected signs that injection pressure had fractured part of the lower caprock, and operators halted injection in 2011 out of caution while they reassessed the geomechanics. The reassuring part: years of subsequent analysis, including Lawrence Livermore National Laboratory review work, found no detectable leakage of CO2 out of the storage complex despite the fracturing — the extra buffer layers in the 950-meter-thick caprock held. It’s a genuine data point for “monitoring can catch problems before they become leaks,” which is a different claim from “nothing ever goes wrong.”
What the broader statistical record shows
Individual project stories are illustrative, but several systematic studies put numbers on the pattern:
- A widely cited 2021 study in Environmental Research Letters (UC San Diego researchers, examining the US NETL project database) found that more than 80% of proposed US commercial CCS projects have ended in failure, with the power sector worse still — around 90% failure. Gas-processing CCS, the most mature application, fared much better: more than 70% of proposed projects there succeeded.
- A 2021 Energy Policy study (Wang et al.) examined 263 global CCUS projects announced between 1995 and 2018 and found 43% were cancelled or put on hold; among large-scale pilot and demonstration plants specifically (>0.3 Mt/year), the cancellation/hold rate was 78%.
- A 2024 Nature Communications Earth & Environment feasibility analysis calculated a historical CCS project failure rate of 88% across 1972–2018, dropping to an estimated 76% when adjusted for the current project mix by subsector.
- IEEFA’s own tracking of 13 “flagship” global projects found 10 of 13 failed or underperformed against their designed capacity, several by large margins. A separate IEEFA fact sheet counted three outright project failures, five projects underperforming materially, two refusing to publish performance data, and only three performing close to nameplate capacity.
None of these studies claims CCS can’t work in specific, narrow applications — gas processing, in particular, has a strong track record, largely because separating CO2 is already part of that process regardless of climate goals. The pattern across the literature is that CCS retrofitted onto power generation specifically has the worst record, and that announced, pre-construction projects fail or stall at very high rates industry-wide.
Direct air capture: smaller, newer, and running well behind its own numbers
Direct air capture (DAC) — pulling CO2 straight from ambient air rather than a smokestack — is the newest and most closely watched front, and the empirical gap between nameplate and delivered performance has been especially stark.
Climeworks’ Orca plant in Iceland (online since September 2021, 4,000 tonnes/year design capacity) has never come close to that figure: Icelandic outlet Heimildin’s reporting, corroborated by Climeworks’ own certified-credit data, put cumulative captured volume at roughly 1,058 tonnes between 2021 and late 2024 — under 12% cumulative utilization, with a peak single year around 1,000 tonnes. Mammoth, Climeworks’ much larger follow-up plant (36,000 tonnes/year nameplate, operating since May 2024), captured just about 105–119 tonnes of verified net removal in its first ten to twelve months — roughly 0.3–1% of nameplate. Climeworks has since reported real improvement: net removal reportedly rose to 675 tonnes in the first half of 2026, more than five times the 119 tonnes captured in the same period of 2025, which the company attributes to sorbent and mechanical upgrades. That’s a genuine trajectory improvement, even if the absolute numbers remain tiny relative to design capacity. Climeworks laid off roughly 22% of its staff in May 2025 amid the slower-than-promised ramp.
1PointFive’s Stratos facility in Texas — built by an Occidental Petroleum subsidiary using Carbon Engineering’s liquid-solvent technology, and marketed as the world’s largest DAC plant at a designed 500,000 tonnes/year — is a genuine reporting discrepancy rather than a settled data point: some 2026 coverage describes it as having reached commercial operation and design capacity, while other trade coverage describes a slipped startup timeline and delivered volumes well short of nameplate. We were not able to independently confirm actual delivered tonnage from a primary source, and that gap is flagged here rather than resolved on secondary sourcing.
On cost, essentially every public source agrees DAC remains far from the oft-cited $100/tonne policy target. Climeworks has stated its current operating cost is close to $1,000/tonne, with a stated 2030 target of $250–350/tonne on next-generation sorbent technology. 1PointFive’s liquid-solvent approach is generally estimated cheaper at scale — roughly $300–900/tonne depending on the source and whether it reflects contracted offtake price or underlying cost. The US Department of Energy’s long-standing $100/tonne threshold for “climate-scale viability” is, in Climeworks’ own public statements, not realistically achievable on any near-term roadmap.
Point-source capture costs and the cement/steel frontier
Point-source capture costs vary enormously by sector, mostly because CO2 concentration in the flue gas differs so much. Gas processing and ethanol plants, where CO2 is already highly concentrated, capture for well under $50/tonne in some cases. Coal and gas power plant retrofits have historically run $60–100+/tonne (Petra Nova’s estimated $60/tonne is at the low end of credible power-sector estimates). Cement and steel — so-called hard-to-abate sectors responsible for a combined 15% or so of global CO2 emissions — are harder and more expensive, generally cited in the $100–150+/tonne range, because a large share of cement’s CO2 comes from the chemical calcination reaction itself, not just fuel combustion.
The most interesting recent data point here may be a genuine good-news story: Heidelberg Materials’ Brevik plant in Norway, inaugurated June 18, 2025, became the world’s first commercial-scale cement-plant carbon capture facility, designed to capture about 400,000 tonnes/year (roughly 50% of the plant’s emissions) using amine-based solvent technology from SLB Capturi. Multiple industry trackers describe the project as having stayed on schedule and on budget through construction and into commissioning — a claim that traces mainly to Heidelberg Materials’ and SLB Capturi’s own project updates rather than an independent audit, but one that stands in notable contrast to the power-sector track record above regardless. It’s part of Norway’s state-backed “Longship” CCS value chain, which ships liquefied CO2 by sea to the Northern Lights storage terminal for injection beneath the North Sea. Heidelberg began shipping “evoZero” near-zero-carbon cement to European customers in October 2025. It’s one project, barely a year into operation, so it’s far too early to call it proof that cement CCS has solved the problems that have dogged power-sector retrofits — but it is the cleanest real-world counter-example to the failure-rate literature currently available.
The policy and money picture
In the US, the 45Q tax credit — $85/tonne for dedicated geological storage, recently raised to parity ($85/tonne) for utilization/EOR projects under 2025’s One Big Beautiful Bill Act — remains the central federal subsidy, and survived the Trump administration’s broader rollback of clean-energy tax credits, reportedly due in part to direct lobbying from Occidental Petroleum. At the same time, the same administration cancelled an estimated $3.7 billion in Department of Energy carbon-capture and industrial-decarbonization grants in May 2025, terminated the federal awards underpinning two flagship DAC regional hubs in October 2025, and cancelled more than $7.5 billion in Infrastructure Investment and Jobs Act awards tied to roughly 55 carbon-management-relevant projects over the course of 2025 — a genuinely disjointed policy picture industry trade press has struggled to characterize as either pro- or anti-CCS. Separately, Industrial Info Resources counted 26 US carbon-capture-related projects worth roughly $27 billion cancelled or put on hold in 2025 for a mix of policy and economic reasons.
In Europe, the EU’s Innovation Fund and Norway’s direct state financing of Longship remain the primary vehicles; the UK’s Track-1 CCS clusters (HyNet, East Coast) continue to move through contracting stages more slowly than originally scheduled. China’s buildout is overwhelmingly state-enterprise-driven (Sinopec, CNOOC, China Energy Investment) and tied explicitly to EOR revenue and the national 2030 peak/2060 neutrality targets, with a stated ambition of 30 Mt/year capacity by 2030 — roughly triple current levels.
Where the climate models — and their critics — stand
The IPCC’s Sixth Assessment Report synthesis treats some level of carbon capture, utilization and storage as effectively unavoidable in net-zero pathways, alongside direct air capture with storage (DACCS) to counter residual hard-to-abate emissions; the IEA has said reaching net zero would be “virtually impossible” without CCUS. Median IPCC-pathway estimates run to several hundred gigatonnes of cumulative CO2 captured and stored by 2100.
That framing has sharp critics. Climate Analytics, in a widely circulated December 2023 briefing timed to COP28, argued that pathways consistent with holding warming to 1.5°C “in the most sustainable manner” rely on CCS to a far more limited degree than industry-aligned scenarios suggest, and warned that heavy reliance on fossil CCS risks becoming what the group called an “86 billion tonne carbon bomb” — a rhetorical framing, not a measured quantity, but one signaling real disagreement within the climate-science and policy community about how much weight CCS should be asked to carry versus faster fossil-fuel phase-out.
The bottom line, empirically
Carbon capture demonstrably works as a geological phenomenon: Sleipner’s nearly 30-year monitoring record and In Salah’s no-leakage outcome (despite a real caprock scare) are solid evidence that CO2, once safely down, mostly stays down, and that monitoring technology can catch problems. What the empirical record does not support is the industry’s historical marketing claim that capture systems reliably hit their advertised capture rates on the first or even the fifth try. Flagship power-sector and LNG-sector projects have missed targets more often than they’ve hit them, often by 30–50 percentage points, and multiple independent academic studies put the historical failure rate for proposed CCS projects at 76–88%. DAC remains an order of magnitude above cost targets and, for its highest-profile plants, has so far delivered a small fraction of nameplate capacity — though Climeworks’ most recent trajectory data suggests real, if slow, improvement. The newest point-source success story, Brevik, is genuinely encouraging but one data point. Most of the CO2 captured worldwide still ends up pushing more oil out of the ground rather than in dedicated storage. All of those things can be true at once, and they are.