Main

Carbon capture and storage (CCS) plays a key role in climate mitigation pathways, yet its feasibility is vigorously debated1,2,3. The recent interest in CCS4,5,6, including negative emissions technologies—direct air capture (DACCS) and bioenergy with CCS (BECCS)—is reflected in plans to increase CCS capacity eight-fold from 2023 to 20307. However, 10 years ago, a similar wave of CCS plans5 largely failed8,9. Can the new push bring CCS on track10,11,12,13 for the Paris climate targets?

Answering this question requires overcoming three challenges. The first is anticipating how many CCS plans are likely to succeed. The second is projecting medium-term growth of CCS, given the uncertainty about the drivers of, and barriers to, its uptake14,15. The third is estimating feasible long-term growth rates that depend on the size of the future CCS market16,17.

We address these challenges by building on the tradition of using empirical evidence18,19,20,21,22,23,24,25,26 from historical technology analogues or reference cases27,28. Using advanced policy-driven technologies as reference cases, we contribute with three methodological innovations. First, we analyse historical failure rates of planned projects to estimate feasible near-term (5–10 years) CCS deployment. Second, we use this estimate to project a range of medium-term (10–20 years) CCS expansion, assuming quasi-exponential growth typical of early stages of technology deployment. Finally, we estimate the feasible range of long-term (20–80 years) CCS growth rates based on the peak growth rates of historical analogues. Thus, we develop an approach for projecting the deployment of emerging policy-driven technologies across the first three phases of the technology life-cycle—the formative phase29,30,31, the acceleration phase19 and the stable growth phase19. Finally, we compare our findings to CCS growth in the three recent IPCC scenario ensembles32,33,34 and estimate the feasible range of CO2 captured and stored with CCS over the 21st century.

We find that only a handful of climate mitigation pathways (10%, IPCC categories C1–C4) depict CCS capacity growth compatible with even the most optimistic assumptions when (1) CCS plans double by 2025 and their failure rate decreases by half; (2) CCS expansion in 2030–2040 is as fast as solar power expansion was in the 2010s or nuclear power expansion was in the 1960s and 1970s; and (3) CCS grows over the following decades as fast as the growth of nuclear in the 1970s and 1980s. Only 33% of pathways meet the first two constraints and only 26% meet the last one. Virtually all pathways that meet all three constraints depict <200 GtCO2 captured and stored by 2070 and <600 GtCO2 by 2100 (at the 95th percentile). Under the less realistic assumptions of a doubling of CCS plans by 2025, a zero failure rate and growth similar to that of flue-gas desulfurization (FGD), this amount could increase to 400 GtCO2 by 2070 and 1,100 GtCO2 by 2100, which still stands in contrast to a large number of 1.5 °C- and 2 °C-compatible pathways, which envision up to 700 GtCO2 captured and stored by 2070 and 1,400 GtCO2 by 2100.

Growth phases of policy-driven emerging technologies

The growth of new technologies starts with a formative phase30,31,35, when “the technology is tested, refined and adapted to market conditions”29. At the end of the formative phase—typically between 0.1% and 2.5% of the final market17,19,22,30,35 (A. Jakhmola, J.J., V. Vinichenko and A.C., manuscript in preparation)—a technology ‘takes off’, driven by increasing returns36. This leads to an acceleration phase with quasi-exponential growth19,21,36. Eventually, countervailing forces such as resource availability and socio-political resistance dampen the acceleration in the stable growth phase, where the growth peaks and is no longer accelerating19,22,37, which is eventually followed by slow-down and saturation due to market limits38. Feasibility constraints in the different phases reflect different growth mechanisms, and therefore require separate assessments (Fig. 1). We use the feasibility space approach27,39, which compares a target case—in this case, future CCS growth—with reference cases40,41 appropriate for each growth phase (Table 1 and Methods).

Fig. 1: Method for projecting the feasible deployment of policy-driven technologies along the phases of technology growth using feasibility spaces.
figure 1

To construct each feasibility space, we use a tailored set of metrics and reference cases most appropriate to each of the first three phases of the technology life-cycle—formative, acceleration and stable growth (Methods and Table 1). For the formative phase, we project feasible CCS deployment (Gt yr−1) based on project plans and their failure rates; for the acceleration phase, the acceleration rate of reference technologies; and for the stable growth rate, we use the maximum growth rate at the inflection point of the S-curve normalized to the market size. This approach can be applied not only to global but also to national and regional targets, as well as to other climate mitigation and energy technologies. Error bars are used to illustrate the uncertainty in feasible deployment over time.

Full size image
Table 1 Reference cases and metrics used to construct feasibility spaces of CCS deployment in the formative, acceleration and stable growth phases
Full size table

For the formative phase, which we find can be complete by 2030, we construct a feasibility space based on near-term CCS plans and historical failure rates of past CCS (Supplementary Note 1) and early nuclear power plans. For the acceleration phase, which we assume would occur in 2030–2040, we construct a feasibility space using the compound annual growth rate (CAGR) metric derived from reference cases of historical nuclear, solar and wind power growth at similar levels of market penetration. For the stable growth phase in the post-2040 period, we construct a feasibility space using maximum annual capacity additions normalized to the market size18,19 derived from reference cases of nuclear, wind and solar power growth at the global, regional and national levels. Subsequently, we map the 1.5 °C- and 2 °C-compatible IPCC Sixth Assessment Report (AR6) pathways10,32 (Methods) onto the three feasibility spaces and calculate the amount of CO2 captured and stored in the pathways that meet the feasibility constraints. Finally, we use FGD, which is similar to the CO2 capture component of CCS, as a reference case for a sensitivity analysis representing the most optimistic (albeit relatively unrealistic) assumptions about CCS growth (Supplementary Note 2).

More CCS plans and fewer failures needed for climate targets

Recent reports indicate that if all current project plans are realized, operational CCS capacity would reach 0.34 Gt yr−1 by 20305,7. However, we have seen ambitious plans before; the initially promising first wave of CCS plans failed to meet expectations despite a number of supportive policies8,9,42 (Fig. 2 and Extended Data Fig. 1). If all plans from the first wave had been realized, today’s operational capacity would be around 0.27 Gt yr−1 (Supplementary Table 1), instead of the paltry 0.04 Gt yr−1 operational today.

Fig. 2: Historical development and prospects for near-term CCS deployment.
figure 2

a, Operational (dark) and planned (light) CCS capacity in 2002–2022, by sector43 (Methods). b, Operational capacity in 2030 based on current operational capacity (black bars) and current (2022) plans under different failure rates (grey bars) compared to the IPCC AR6 1.5 °C- (n = 218) and 2 °C-compatible (n = 423) pathways (coloured bars illustrate the medians; error bars show the IQR32) and recent IEA NZE pathways (coloured dots)12,13. c, Feasibility space of CCS deployment in the formative phase depicting operational CCS capacity (Gt yr−1) (Supplementary Table 6) in 2030 as a function of CCS plans (y axis) and their failure rates (x axis). The hatched zone represents all observations within the feasibility frontier and thus those consistent with empirically grounded assumptions about near-term CCS plans and failure rates (crosses, Table 1). The shading shows that this frontier is fuzzy or, in other words, not binary27,39. The isolines show different combinations of planned capacity and failure rates that lead to the same operational CCS capacity in 2030. The IEA NZE pathways12,13 are indicated by red and orange isolines, and the median CCS capacity in the IPCC AR6 pathways32 are shown with blue (1.5 °C-compatible) and green (2 °C-compatible) isolines.

Source data

Full size image

The CCS capacity installed by 2030 depends on the plans announced by 2025 because CCS projects take, on average, 5 years from announcement to completion. Under an optimistic assumption that the planned capacity will double between 2022 and 2025, the 2030 capacity could reach 0.6 Gt yr−1 (Methods). However, how much CCS will be actually installed also depends on the failure rate—the share of planned capacity that is not realized. To determine a realistic range of failure rates, we first calculate the historical failure rate of CCS projects from 1972 to 2018 (88%)43 (Supplementary Tables 1–4 and Methods). This industry-wide failure rate varies across sectors and subsectors44,45,46,47, with CCS plans in the first wave dominated by the electricity sector, which failed at a rate of over 90%. In the current wave, the planned projects are more diversified (Fig. 2). Based on historical subsector-specific failure rates, we estimate that the aggregate failure rate of today’s plans would drop to 76% (Methods). Finally, we measure the failure rate of early nuclear power plans in the United States (45%), which was a policy-driven technology developed in the wake of the oil crises of the 1970s48,49,50. At the start of this expansion, nuclear power contributed 2% of global and 3% of US electricity markets, which was more advanced than CCS is today, yet still close to the formative phase. Given the capital intensity51, lumpiness52 and complexity53,54 of nuclear power, we consider it as a close technological analogue of CCS and use it as an optimistic reference case.

With the capacity announced by 2022 and the historical failure rate (88%), CCS capacity in 2030 would be 0.07 Gt yr−1. If the planned capacity doubles between 2023 and 2025 and the failure rate drops to that of nuclear power in the United States (45%), CCS capacity would reach 0.37 Gt yr−1 by 2030. We consider this the upper feasible bound. By contrast, the IPCC AR6 1.5 °C pathways10,32 envisage a median CCS capacity of 0.9 Gt yr−1 (interquartile range (IQR) 0.4–1.5) by 2030. The International Energy Agency (IEA) net-zero emissions (NZE) scenario13 envisages an even higher 1 Gt yr−1 capacity (1.7 Gt yr−1 in the 2021 edition12), which is almost definitely out of reach given the current project timeline and likely failure rates. However, the 2 °C pathways envision a median CCS capacity of 0.3 Gt yr−1 (IQR 0.04–0.6), which could be achieved by expanding the planned capacity to 0.4–0.6 Gt yr−1 by 2025 and reducing the failure rate to 45–60% (Fig. 2).

CCS growth should accelerate at least as fast as wind power

As we show, it is realistic for CCS capacity in 2030 to reach 0.07–0.37 Gt yr−1 capacity. This would amount to 0.3–1.8% of the market potential, or all CO2 emissions in sectors where CCS plans have been announced (Methods), thus resulting in the take-off and subsequent quasi-exponential growth characteristic of the acceleration phase. Can such growth bring CCS on track for climate targets by 2040? To answer this question, we project a range of feasible 2040 CCS capacity values based on: (1) the feasible range of the 2030 capacity estimated in the previous section; and (2) feasible year-on-year growth rates in the acceleration phase in 2030–2040, derived from the reference cases of three policy-driven technologies—nuclear, wind and solar power—at similar levels of market penetration (Methods).

We estimate the range of CCS capacity achievable by 2040 to be around 0.95–4.3 Gt yr−1 (Table 2). The upper end of this range can only be reached under optimistic assumptions about CCS deployment in the formative phase (0.37 Gt yr−1) and acceleration in 2030–2040 comparable to that of nuclear power in the 1970s (Fig. 3 and Supplementary Fig. 1), or if CCS accelerates as fast as FGD did in the 1980s. Under these optimistic assumptions, CCS capacity in 2040 would be above the median of the 1.5 °C pathways (3.8 Gt yr−1). The median of the 2 °C pathways (2.4 Gt yr−1) could be reached under more modest, albeit still optimistic, assumptions about CCS deployment in the formative phase (0.26–0.37 Gt yr−1) and with acceleration comparable to that of wind power in the 2000s (Table 2 and Fig. 3).

Table 2 Feasible upper bounds of CCS capacity in 2030 and 2040 estimated from reference cases and compared to the IPCC AR6 pathways
Full size table
Fig. 3: Feasibility space of CCS deployment for the acceleration phase in 2030–2040 compared to IPCC AR6 pathways.
figure 3

CCS capacity and market penetration achieved by 2030 (x axis) versus a 10-year moving CAGR in 2030–2040 (y axis). The maximum feasible capacity in 2030 makes up the feasibility frontier along the x axis (0.37 Gt yr−1 or 1.8% of the market potential) (Fig. 2c). Acceleration rates for the reference cases make up the three feasibility frontiers for CCS acceleration in 2030–2040, with the black lines showing the historical acceleration rates of nuclear (1961–1978), wind (1995–2017) and solar (2008–2022) power as reference cases for the CCS (Table 1). Dashed lines illustrate the continuation of these reference cases under higher than realistic CCS capacity by 2030. The dark red line shows the historical acceleration of FGD (1972–1986) as a reference case for the capture component of CO2. The hatched zone represents all observations within the feasibility frontier and the shading shows that this frontier is fuzzy, or in other words not binary27,39. The 1.5 °C- and 2 °C-compatible pathways10,32 are shown as dots and their distributions form the two-dimensional (2D) density plot (from white to yellow). Blue and green isolines show different combinations of the two metrics that lead to the median CCS capacity in the 1.5 °C- and 2 °C-compatible pathways, respectively (Table 2), regardless of feasibility considerations. The x axis of this figure is cut off at 2.1 Gt yr−1 (10% market penetration), thus excluding 47 1.5 °C pathways (20%) and 33 2 °C pathways (8%) with a CCS capacity of up to 21 Gt yr−1 by 2030. Density plots are constructed from the entire sample of pathways (n = 218 for 1.5 °C, n = 423 for 2 °C).

Source data

Full size image

Despite being in line with the realistic range of 2040 CCS capacity, most pathways fall outside the feasibility frontier when we account for the timing of acceleration (Fig. 3). A total of 76% of 1.5 °C- and 42% of 2 °C-compatible pathways depict unrealistically fast growth by 2030, with an additional 14% of both 1.5 °C- and 2 °C-compatible pathways requiring unrealistically fast acceleration from 2030–2040. Only 10% of 1.5 °C- and 44% of 2 °C-compatible pathways are located within the feasibility frontier for the formative and acceleration phases. Even under a less realistic assumption that the growth of CCS will accelerate as fast as for FGD, only 18% of the 1.5 °C-compatible pathways would be within the feasibility frontier.

In most pathways CCS grows faster than nuclear at its peak

After acceleration, technologies enter a stable growth phase when annual additions peak at the maximum growth rate (G) at the inflection point of the S-curve19 (Fig. 1). We find that the median values of the maximum annual additions of CCS capacity were similar (0.4–0.5 Gt yr−1 added annually) across the 1.5 °C-, 2 °C- and 2.5 °C-compatible pathways (Supplementary Figs. 2 and 3). What varies is when the fastest growth is achieved: in the 1.5 °C-compatible pathways, it occurs around 2045, in the 2 °C-compatible pathways around 2055 and in the 2.5 °C-compatible pathways around 2065.

We use policy-driven low-carbon technologies—nuclear, wind and solar power—as reference cases for growth rates at the stable growth phase. We complement global observations with regional and national observations where the growth of these technologies had already peaked (Extended Data Table 1)19,55. To compare maximum annual capacity additions across reference cases and mitigation pathways, we normalize G to the size of the market—the total electricity supply for the reference cases, and the sum of the gross CO2 emissions in sectors with capturable emissions plus negative emissions from BECCS and DACCS for CCS (Methods).

In most IPCC pathways, the use of fossil fuels declines earlier than the expansion of negative emissions, which means the potential CCS market may decline over time (Supplementary Fig. 4 and Extended Data Fig. 2). To account for market uncertainties, we normalize the CCS growth rates to both the maximum market size in 2022 (G2022) and to the market size at the time when the maximum growth rate was achieved (GTMax). Under both normalizations, the CCS growth in most 1.5 °C and 2 °C pathways is faster than the historical growth of nuclear, wind and solar globally. Under normalization to the 2022 market (G2022), only 26% of the 1.5 °C and 2 °C pathways depict global CCS growth consistent with the peak growth of global nuclear power (and ≤1% with the global growth of solar or wind). Under normalization to the future market (GTMax), only 5% of the 1.5 °C and 2 °C pathways depict global CCS growth consistent with the peak growth of global nuclear power (Extended Data Table 1 and Extended Data Fig. 3).

In feasible pathways <600 GtCO2 is captured by 2100

Only 10% of climate mitigation pathways (14% of the 2 °C- and 4% of the 1.5 °C-compatible pathways) meet all the feasibility constraints for CCS growth, and those that do capture and store considerably less CO2 over the 21st century. In fact, vetting the IPCC AR6 mitigation pathways for the feasibility of CCS growth reduces the upper bound of CO2 captured and stored by 2070 from 685 Gt to 201 Gt (both at the 95th percentile) (Supplementary Fig. 5), and by 2100 from 1,428 Gt to 589 Gt (both at the 95th percentile), even if the more relaxed long-term growth metric (G2022) is used (Fig. 4 and Extended Data Table 2). The effect is especially pronounced for the most stringent 1.5 °C-compatible pathways with no or limited overshoot (IPCC category C1), which typically rely on the early and rapid growth of CCS, and within which only the low-energy demand pathway (ref. 56) without any CCS satisfies all the CCS feasibility constraints.

Fig. 4: The effect of feasibility constraints on the long-term cumulative capture and storage of CO2 in 2030–2070 and 2030–2100 in IPCC AR6 pathways.
figure 4

The x axis contains different groups of mitigation pathways before and after imposing feasibility constraints, including feasible by 2030 under realistic project plans and failure rates, feasible by 2040 under acceleration rates similar to those of the reference cases, and feasible in the long term under maximum annual growth similar to the growth of the reference cases when normalized to capturable emissions today (G2022) and capturable emissions when the maximum growth is achieved (GTMax). ‘Most optimistic’ illustrates a group of pathways with relaxed constraints—0% failure rate and a doubling of plans, acceleration and stable growth (G2022) of FGD (Extended Data Fig. 4 and Supplementary Table 7). ‘All pathways’ is n = 218 for 1.5 °C in the left-hand panels and n = 423 for the 2 °C pathways in the right-hand panels. Violins and boxplots illustrate the cumulative CO2 capture and storage by 2070 and 2100 (in Gt) in these groups of pathways (left-hand y axis)—boxplots show the IQR, with the median marked by a black line and whiskers extending from the IQR range to the non-outlier minimum and maximum. Grey bars illustrate the share of pathways in each group (right-hand y axis). In groups with fewer than 10 pathways, we use dots rather than violins and boxplots.

Source data

Full size image

We test the sensitivity of these findings under 36 combinations of CCS plans, failure rates, near-term growth acceleration and long-term maximum growth rates, constrained at different levels (Supplementary Tables 7 and 8). The cumulative CO2 captured in IPCC pathways under the vast majority (95th percentile) of these sensitivity runs does not exceed 290 Gt by 2070 or 1,000 Gt by 2100. Out of all 36 sensitivity runs, the most optimistic one is that CCS plans double, their failure rate drops to zero and, subsequently, CCS deployment accelerates and grows as fast as FGD. Even in this case, only about half of the Paris-compatible pathways meet the constraints, with a median capture of 200 Gt (95th percentile at 398 Gt) in 2030–2070 and 592 Gt (95th percentile at 1,078 Gt), which is still considerably lower than in the full suite of AR6 pathways (95th percentile at 685 and 1,428 Gt, respectively) (Fig. 4 and Extended Data Fig. 4). We consider these relaxed constraints to be highly unrealistic, given the historical failure rates and because CCS is a more complex technology than FGD, with higher capital costs and larger infrastructure requirements, such as pipelines and storage, that potentially face more public opposition (Supplementary Note 2 and Supplementary Table 9).

Discussion and conclusions

To delineate feasible trajectories of future CCS deployment, we develop an analytical approach based on technology diffusion theories and empirical evidence from CCS history and other policy-driven technologies. This study advances recent efforts to use empirical evidence for assessing the feasibility of low-carbon technologies15,18,19,21,57,58 through developing specific feasibility assessments for three separate phases of technology diffusion—formative, acceleration and stable growth—each with distinct policy challenges.

The first challenge is to increase the number of planned CCS projects while reducing their failure rates for the technology to take off. We show how realistic assumptions about failure rates, based on the history of CCS and other historical benchmarks, can identify a feasible upper bound of CCS capacity in 2030 (0.37 Gt yr−1). This analytical approach could be used for projecting the formative phase deployment of other policy-driven technologies, but more research is needed on failure rates and their dynamics.

The second challenge is to rapidly accelerate CCS after it has taken off, which we assume will start by 2030. We use the year-on-year growth rates of solar, wind and nuclear power as reference cases to define the upper bound for CCS capacity by 2040 (4.3 Gt yr−1). The literature is not consistent on how to use year-on-year growth rates in projecting technology diffusion, sometimes deriving them from formative phase data with erratic growth. We argue that year-on-year growth rates are most meaningful when derived from post-take-off reference cases with similar deployment levels and in conjunction with realistic estimates of initial deployment (in our case, CCS capacity in 2030). Policy support at this stage will require a strong emphasis on creating a market for CCS, either through taxing or regulating CO2 emissions (Supplementary Notes 3 and 4).

The final challenge is maintaining long-term CCS growth after its likely stabilization beyond 2040. Most 1.5 °C- and 2 °C-compatible pathways depict CCS capacity additions that are faster than the global growth of nuclear power and recent growth rates of wind and solar power. Nuclear power grew in the context of a rapidly expanding electricity demand and the oil crises of the 1970s50,59,60, while Europe’s renewable energy growth is also partially in response to an energy crisis55. Thus, nuclear and renewables may be too optimistic as reference cases for CCS, which is not being driven by an energy crisis.

Our findings contribute to the debate about the potential role of CCS in climate mitigation1,2,61,62,63. The integrated assessment modelling community has addressed concerns about the outsized role of CCS in mitigation pathways2,62 by developing scenarios, including illustrative mitigation pathways, with low or no (BE)CCS10,34,56,64,65,66,67. At least three of the illustrative mitigation pathways (low demand, shifting pathways and gradual strengthening) from the recent IPCC AR6 report meet, or nearly meet, our feasibility constraints by relying on alternative mitigation strategies, such as suppressing demand, changing diets or shifting strong climate action towards the end of the century. In general, more recent pathways feature a somewhat less ambitious use of CCS (Extended Data Fig. 5, Supplementary Note 5 and Supplementary Table 10). The efforts to depict more-realistic CCS deployment in mitigation pathways has been supported by a strategy of vetting scenarios68,69 using thresholds for the maximum CO2 storage68 or decadal growth in electricity generation with CCS69.

Our more granular analysis identifies three additional constraints on CCS deployment—reaching take-off by 2030, achieving rapid acceleration in the 2030s and ensuring sustained long-term growth. Our findings are more conservative than those of previous studies18,70,71 because we consider potential failures of planned CCS projects and use reference cases that are more similar to CCS in terms of their complexity, capital-intensity (Supplementary Note 3), land use and public acceptance (Supplementary Note 2) than FGD18. We show that 10% of the 1.5 °C- and 44% of the 2 °C-compatible pathways depict realistic CCS capacity in 2040, which strongly correlates with the amount of CO2 cumulatively captured and stored in the long term (Supplementary Fig. 6). The additional constraints on long-term growth further reduce the number of feasible pathways to 1–4% of the 1.5 °C and 3–14% of the 2 °C pathways (depending on the assumptions about the size of the future market) (Fig. 4, Supplementary Note 6, Supplementary Figs. 7–11 and Supplementary Table 11) and reduce the (non-outlier) maximum amount of CO2 that can feasibly be captured and stored by 2100 to <600 GtCO2 (Fig. 4). This has significant implications for global carbon budgets.

Among the many applications of CCS, BECCS and DACCS have attracted a lot of attention because they are both essential for achieving negative emissions and thus climate targets70,72,73,74. Ref. 73 shows that BECCS and DACCS should complete the formative phase by 2040 to keep net-zero goals within reach. However, BECCS and DACCS together only account for 10% of current CCS plans72 (Fig. 2 and Supplementary Table 1), and while their development is important, it is the success of the remaining 90% of CCS projects that will be pivotal in enabling negative emissions technologies that would rely on capture, transportation and long-term storage developed for other CCS applications. Our results illustrate the challenges of ensuring the take-off and expansion of this wider and more diverse technological ecosystem that will be key to the success of BECCS and DACCS.

More broadly, our research contributes to the analysis of the growth of policy-driven emerging technologies. We show how appropriate growth metrics and reference cases make it possible to develop empirically grounded feasibility spaces and feasibility frontiers for distinct phases of technological growth, and thus project its deployment based on realistic and transparent assumptions. This approach can be applied not only to global but also to national and regional targets (Fig. 1, Supplementary Note 7, Extended Data Figs. 6 and 7 and Extended Data Table 3), as well as to other climate and energy technologies that are currently in the early stages of development.

Methods

CCS projects database

To study current and historical CCS project plans and their failure rates, we build a dataset of completed, failed and currently planned commercial (at least 0.1 Mt yr−1 capacity) CCS projects starting from 1972—the completion year of the first integrated CCS project (Terrell natural gas processing plant). For each project, we code the capture rate, project announcement and completion year, facility status (for example, active, failed, planned), facility operation start and end years, CO2 storage type (for example, enhanced oil recovery or dedicated geological storage), sectoral and subsectoral application, country and region (Supplementary Tables 3 and 4). Data on completed and failed projects was collected from annual Global CCS Institute reports as well as dormant and existing databases listing planned CCS projects at different points in time78,79,80,81 (full list in Supplementary Table 2), whereas data on currently planned projects was gathered primarily from the recently published (March 2023) IEA Carbon Capture, Utilization and Storage Projects Database7. These sources were complemented with a systematic Google search, described in Supplementary Note 1.

To facilitate the future use of our database in combination with the IEA Carbon Capture, Utilization and Storage Projects Database, we define our variables in line with the IEA’s database and filtered our project entries to meet the minimum project capacity of 0.1 Mt yr−1. To facilitate the use of the database in combination with the IPCC AR6 Scenarios Database32, we align relevant variables with the AR6 scenario variables related to CCS. The resulting database is available in ref. 43.

IPCC pathways used in this study

For analysing future trajectories of CCS deployment in the AR6 ensemble, we use mitigation pathways classified under IPCC AR6 scenario categories 1 (‘Below 1.5 °C with no or limited overshoot’), 2 (‘Below 1.5 °C with high overshoot’), 3 (‘Likely below 2 °C’), 4 (‘Below 2 °C’) and 5 (‘Below 2.5 °C’)10. We further group these categories into 1.5 °C- (categories 1 and 2), 2 °C- (categories 3 and 4) and 2.5 °C- (category 5) compatible mitigation pathways. Because our analysis compared future deployment trajectories for different temperature targets, we exclude model families that did not produce pathways for each of the three groups in the IPCC AR6 scenario ensemble, including TIAM (40 pathways), C-ROADS (5), EPPA (5) and MERGE (1).

Before proceeding with our analysis, we also checked scenario data for consistency with the actual CCS capacity in 2020. In the resulting sample of pathways, CCS deployment in 2020 rarely reflected its actual capacity (~35 MtCO2), with the majority of pathways reporting zero for 2020. We exclude 30 pathways where the CCS capacity in 2020 is reported to be more than 50 MtCO2 (some pathways reported up to 500 MtCO2 in 2020). Our final sample was 840 pathways—218 1.5 °C-, 423 2 °C- and 199 2.5 °C-compatible pathways.

Our analysis of CCS deployment covers a range of CCS technologies (Extended Data Fig. 1 and Supplementary Table 1), including those leading to negative emissions (BECCS and DACCS). Some inconsistencies related to these technologies were found among scenario outputs between model families. For example, some model families report DACCS deployment as a separate variable, others as a separate value that is also included in the total CCS variable. Some model families report DACCS and BECCS as negative values, others as positive. We harmonize these outputs and use the resulting sum of BECCS, DACCS and CCS capacity in fossil fuel and industry further in our analysis as total CCS capacity.

Lastly, we append each scenario data time series (2020–2100) with historical CCS capacity data from 2000–2020, thus replacing the zeros in 2020 (and further, if less than the actual 2020 capacity) with the latest value from our CCS Projects Database (35 MtCO2). The resulting dataset thus contains the 1.5 °C–2.5 °C pathways (840 pathways) with decadal CCS deployment values from 2000 to 2100.

Defining the current and future CCS market potential

In this study, we define the market for CCS technologies as the annual gross global CO2 emissions in sectors where CCS is technologically applicable. To estimate today’s CCS market, we calculate CO2 emissions from the Emissions Database for Global Atmospheric Research (EDGAR) v.6 database82 in sectors and subsectors with recorded CCS plans (that is, 'Energy systems' and 'Industry' sector variables, excluding 'Other' emissions). The resulting sum of ‘capturable’ CO2 emissions, which we defined as M2022, results in 21 GtCO2.

To estimate future changes in the CCS market based on scenario data, we first calculate the sum of the CO2 emissions in these CCS-applicable sectors for each decade up to 2100 as the sum of the following scenario variables: Emissions∣CO2∣Energy∣Supply, Emissions∣CO2∣Energy∣Demand∣Industry, Emissions∣CO2∣Industrial Processes, Emissions∣CO2∣Waste and Emissions∣CO2∣Other.

These values are reported as final or net (that is, after carbon sequestration and removal). To calculate the size of the market (equation (1)), we convert the net CO2 emissions to gross as the sum of the former with the overall CCS capacity in the fossil-fuel-based and industrial sectors (carbon sequestration), emissions reductions achieved through non-CCS negative emissions technologies and negative emissions CCS technologies (carbon removal). Thus, the first part of equation (1) corresponds to the amount of CO2 entering the atmosphere in year t (in sectors with capturable emissions) before being captured or offset. To that, we add the BECCS and DACCS capacity as negative emissions delivered by CCS technologies in year t (in mitigation pathways). For an illustration of the approach, see Supplementary Fig. 4.

$$M_{t} = \overbrace{E_{t} + {\rm{CCS}}_{n,t} + {\rm{NET}}_{t} + {\rm{CCS}}_{ind,t}}^{{{\rm{gross}}}\,{{\rm{CO}}}_{2}\,{{\rm{emissions}}}\,{{\rm{in}}}\,{{\rm{year}}}\,{{\rm{t}}}} + \underbrace{{\rm{CCS}}_{n,t}}_{{{\rm{BECCS}}}\,{\rm{and}}\,{{\rm{DACCS}}}}$$
(1)

where:

  • Mt is the CCS market size in year t;

  • Et is the net CO2 emissions in sectors where CCS technologies can be applied in year t;

  • CCSn,t is the sum of the biogenic and atmospheric (that is, negative) CO2 emissions captured and stored in year t;

  • NETt is the sum of the other negative CO2 emissions in year t; and

  • CCSind,t is the sum of the fossil-based and industrial CO2 emissions captured and stored in year t.

The dataset with these variables calculated for each pathway is available in ref. 43. Due to missing variables, 68 pathways (~10% of the 1.5 °C- and 2 °C-compatible pathways) failed to produce this metric.

Reference case selection

Selecting reference cases is central to the feasibility space method27. In this study, we use three main sets of reference cases (Table 1), one for each of the first three phases of the technology life cycle (Fig. 1). Given the centrality of the policy support shaping the expansion of CCS42, we use reference cases of technologies whose deployment was strongly shaped by policy support—nuclear power50,59, as well as solar and wind power83,84,85. These are capital-intensive technologies that have also all faced public opposition86 similar to the public opposition that has already started plaguing CCS projects87 (Supplementary Note 2). These reference technologies provided evidence of feasible CCS growth in each phase.

Given the market immaturity, capital intensity and lumpiness9,52,53 of CCS, we identify the project failure rates of large-scale policy-driven technologies at relatively low market penetration levels as reference cases for the failure rate of CCS projects in the formative phase. For the two less optimistic failure rates, we use both the historical failure rate of all CCS projects (88%) and a CCS subsector-adjusted rate based on current project plans (76%, see below). These failure rates are comparable to the recent failure of other emerging large-scale technologies, such as floating offshore wind power (>90%) and new applications of solar (mainly concentrated solar power, 66%)88,89 (Supplementary Table 5). For the optimistic case, we use the historical failure rate of nuclear power in the United States (45%) in 1972–198248, when it accounted for 3% of the national electricity supply. We consider this as an optimistic benchmark for CCS projects, given that nuclear power was more commercially mature at that point.

In the acceleration and stable growth phases, we use the reference cases of nuclear, solar and wind power. In addition to being policy-driven and capital-intensive, the historical development of these technologies spans a wide range of socio-political contexts that can constrain or facilitate future CCS growth. This set of technologies varies widely in terms of modularity, complexity and the degree of customization required52,53,54, all of which might be relevant for CCS, given the variety of potential applications (for example, from the most complex BECCS to the more modular DACCS) and infrastructure needs. We validated the resulting feasibility spaces in the acceleration and stable growth phases with studies using the same19,55 or similar70 growth metrics in a wider set of technologies or contexts.

In addition, we use the historical growth of FGD as a reference case for the CO2 capture component of CCS18,90, FGD being an end-of-pipe technology without the transportation, storage, public opposition, high project costs or competitiveness challenges that CCS faces (Supplementary Note 2 and Supplementary Table 9). However, given the technological similarity to gas capture, it could be used to verify whether the capture component of CCS imposes additional constraints on the feasible speed of CCS deployment. In addition, we use evidence from FGD acceleration and growth in the sensitivity analysis on the number of IPCC pathways that clear all the feasibility constraints and the amount of CO2 they envision capturing (Fig. 4, Extended Data Fig. 4 and Supplementary Tables 7 and 8).

Feasibility space for the formative phase

The formative phase is shaped by the current state of the technology and how volatile its growth is. We thus construct a feasibility space for the operational CCS capacity in 2030 (Fig. 2) using today’s planned capacity as a starting point. In projecting the operational capacity in 2030, we use two variables that shape growth volatility for emerging policy-driven technologies in the formative phase—planned capacity and its failure rate.

For the first variable, we assume the CCS plans to no more than double (to 600 Mt yr−1) from 2023 to 2025 (annual growth ~42%). This is in line with the annual growth rates in 2021 and 2022, which have been steadily declining since the start of the second wave, from 108% in 2019 to 43% in 2022. Due to the small number of projects planned to start operations by 20255, we assume possible early project success to be insufficient to ignite industry interest beyond this ceiling by 2025. We also assume 5 years for project implementation, which is the average project delivery time of active and completed CCS projects. This is in line with recent estimates of 4 to 7 years91. Nevertheless, we account for a potentially longer project delivery time, which is characteristic of large-scale projects51,92, by using the currently planned capacity as the feasible minimum.

For the second variable, we calculate the share of planned capacity that fails to become operational by 2030 using three reference cases (Table 1). For the first failure reference case, we estimate the historical failure rate of all CCS projects (88%, equation (2)).

$${F}_{{\mathrm{hist}}}=\sum \frac{{\mathrm{CC{S}}}_{{\mathrm{f}},1972-2017}}{{\mathrm{CC{S}}}_{{\mathrm{p}},1972-2017}}\times 100 \%$$
(2)

where:

  • Fhist is the historical CCS failure rate;

  • CCSf,1972−2017 is the capacity of failed projects between 1972 and 2017; and

  • CCSp,1972−2017 is the capacity of planned projects between 1972 and 2017.

For our second failure reference case, we calculate a subsector-adjusted failure rate (76%), which is an average of the historical subsectoral failure rates (Supplementary Table 1) weighted by their share in 2022 project plans. First, we calculate a failure rate for each subsector (equation (3)). Then, we calculate the average failure rate weighted by the share of each subsector in current project plans (equation (4)). If the application of CCS is new (for example, BECCS electricity), with only a few projects planned historically (n < 5), we use the historical failure rate of all CCS projects (88%).

$${F}_{s}=\sum \frac{{\mathrm{CC{S}}}_{{\mathrm{f}},s,1972-2017}}{{\mathrm{CC{S}}}_{{\mathrm{p}},s,1972-2017}}\times 100 \%$$
(3)
$${F}_{{\mathrm{adj}}}=\sum {F}_{s}\times \frac{{\mathrm{CC{S}}}_{{\mathrm{p}},s,2022}}{{\mathrm{CC{S}}}_{{\mathrm{p}},2022}}\times 100 \%$$
(4)

where:

  • Fs is the historical failure rate in subsector s;

  • CCSf is the capacity of failed projects between 1972 and 2017 in subsector s;

  • CCSp is the capacity of planned projects between 1972 and 2017 in subsector s;

  • Fadj is the subesector-adjusted failure rate of current CCS plans;

  • CCSp,s,2022 is the planned capacity of projects in subsector s in 2022; and

  • CCSp,2022 is the total capacity of planned projects in 2022.

Finally, for the third CCS failure rate reference case, we use a decadal failure episode of nuclear power deployment in the United States (1972–1982), as documented in ref. 48. The year when the first failures occurred was 1972, so some projects that failed from 1972 to 1982 had been planned before 1972. During this time period, we calculate a failure rate of 45% for all planned nuclear power plant projects. We find this episode to be a fitting reference case for the current plans for CCS globally for several reasons. First, in 1972, nuclear power production was still concentrated in a few pioneering countries led by the United States, and the global share of nuclear power in electricity production was less then 2.5%, which means it was close to the formative phase. Second, CCS and nuclear power share several similarities in terms of technological characteristics, such as lumpiness52,53, the need for customization53,54 and capital intensity9.

Feasibility space in the acceleration phase

The level of technological development during the acceleration phase depends on the rate of growth and the initial level from which growth starts. We construct a feasibility space for the acceleration phase, with the potential market penetration in 2030 for the latter and the CAGR between 2030 and 2040 for the former. We measure the 10-year CAGR for three reference cases of policy-driven emerging technologies—wind, solar and nuclear power—from the time they reached similar levels of market penetration compared to what can realistically be achieved by CCS in 2030 (0.3–1.8%). For nuclear power, this period corresponds to 1961–1968, when nuclear power production had grown from 0.15% (current market share of CCS) to 1.8% (maximum achievable market share of CCS by 2030) of its maximum production in 2006. For comparison, in 1961, there were 15 operational nuclear reactors worldwide75. For wind power, which has not yet reached its maximum capacity, we took the period between 1995 and 2007, when wind power production grew from 0.06% to 0.7% of global electricity production, which roughly corresponds to 0.15% and 1.8%, respectively, assuming the market potential of wind power is about 40% of the electricity supply19. For market penetration rates from 0.3–1.8% in these two reference cases, achieved every year in selected periods (above), we estimated the acceleration rate for the following 10 years (equation (5)). For solar power, it has been less than 10 years since it gained 1.8% of its global market potential, therefore it forms a narrow segment of the feasibility frontier (from 2008, when it gained 0.15%, up to 2012, when it gained 1.1% (ref. 77), assuming the market potential of solar power is about 40% of the electricity supply19). Taken together, three time series form three y-axis feasibility frontiers of CCS acceleration in Fig. 3. From different combinations of x- and y-axis metrics, we estimate the feasible range of CCS deployment by 2040. In addition, we measure the acceleration of FGD normalized to the total coal power capacity, as a reference case for deployment of the CO2 capture component of CCS.

To map the IPCC climate pathways onto the feasibility space for acceleration, we calculate the market penetration level of CCS in 2030 by dividing the CCS capacity in 2030 in each pathway (CCS2030) by the current amount of capturable emissions in sectors where CCS projects are currently planned (M2022). Then, we calculate the CCS growth rates in the acceleration phase in 2030–2040 using equation (5). Taken together, these two variables formed the 2D density plot (R package ‘ggdensity’93) for CCS acceleration in mitigation pathways (Fig. 3).

$${A}_{{\mathrm{CCS}}}=\left({\left(\frac{{\mathrm{CC{S}}}_{t+10}}{{\mathrm{CC{S}}}_{t}}\right)}^{\frac{1}{10}}-1\right)\times 100 \%$$
(5)

where:

  • ACCS is the growth rate in the acceleration phase; and

  • CCSt is the CCS capacity (including BECCS and DACCS) in year t.

Feasibility space in the stable growth phase

In the stable growth phase, the mechanisms supporting growth are balanced out by those slowing it down, which can be measured with the metric G, introduced in ref. 19. We construct a feasibility space for the stable growth phase based on the maximum growth rate and when this growth rate was achieved. To measure the G implied in the IPCC pathways, we fit the Gompertz (equation (6)) and logistic growth (equation (7)) models to a CCS deployment time series in 2030–2100 combined with the historical data on operational CCS capacity in 2000–2020. Because we are only interested in the stable growth rate, we truncate the resulting time series at the maximum annual CCS capacity to increase the goodness of fit in pathways where CCS deployment starts to decrease after stable growth and saturation. In such pathways, CCS is never phased out by the end of the century.

$${f}_{{\mathrm{gmp}}}(t)=L{e}^{-{e}^{-k(t-{t}_{0})}}$$
(6)
$${f}_{{\mathrm{log}}}(t)=\frac{L}{1+{e}^{-k(t-{t}_{0})}}$$
(7)

where:

  • e is a constant approximately equal to 2.718;

  • k is the growth constant; and

  • t0 is the inflection point.

For the logistic curve, the inflection point is located at 50% of the asymptote, L. For the Gompertz curve, the inflection point is located at 37% of the asymptote, L.

The G for the maximum annual addition of CCS capacity (GtCO2 yr−2) in each pathway is calculated as follows:

$${G}_{{\mathrm{gmp}}}=\frac{Lk}{e}$$
(8)
$${G}_{{\mathrm{log}}}=\frac{Lk}{4}$$
(9)

To account for the uncertainties concerning the size and dynamics of the potential CCS market, which can either expand due to the increased use of fossil fuels or adoption of negative emissions technologies, or shrink due to climate change mitigation action, we normalize G to the market size. We use two normalization options—today’s market size (M202282) and the market size at the inflection point (MTMax). For the latter, scenario outputs only contain decadal data, which does not allow precise measurement of the market size in the year of maximum growth. To obtain approximate estimates of this value, we assume the linear development of emissions trajectories within each decade (equation (10)).

$${M}_{{\mathrm{TMax}}}={M}_{t}+\frac{{M}_{t+10}-{M}_{t}}{10}\times ({\mathrm{TMax}}-t)$$
(10)

where:

  • TMax is the year of the inflection point (and maximum growth);

  • t and t + 10 are the years between which TMax is located (for example, 2030 and 2040); and

  • Mt is the market size at year t.

These two normalization options result in two maximum growth rate variables—G2022 and GTMax. Normalization to market size enables us to compare these growth rates with global, regional and national reference cases for nuclear, solar and wind power, where the maximum growth was normalized to the size of the market. In the cases where solar and wind power are still accelerating, GTMax could not be reliably estimated, and therefore we instead use a similar metric of maximum growth, R3—the average added production over the last three years normalized to the average market size over those years (equation (11),19,55). These instances are reported in Extended Data Table 1, where TMax is not reported. In addition, we measure the R3 of FGD (due to data availability18), normalized to the total coal power capacity as a reference case for the maximum growth of the CO2 capture component of CCS. Extended Data Fig. 3 displays the IQR ranges for the GTMax estimated from cases where the stable growth phase had already been reached19,55.

$$R_3=\frac{\left(\frac{{P}_{t}-{P}_{t-3}}{3}\right)}{\left(\frac{{M}_{t}+{M}_{t-3}}{2}\right)}\times 100 \%$$
(11)

where:

  • R3 is a metric used instead of GTMax for reference cases where the maximum growth has not yet been achieved;

  • Pt is the production in year t; and

  • Mt is the market size in year t.

For regional reference cases, we report the maximum growth rate of nuclear power deployment in the Organization for Economic Cooperation and Development region (with membership as of 1990). For solar and wind power deployment, we estimate the maximum growth for the Organization for Economic Cooperation and Development, the 27 countries of the European Union and Asian regions, with the latter two reported in Extended Data Table 1 as the fastest for wind and solar, respectively. For national reference cases in this table, we estimate the maximum growth rate in countries with electricity markets of around 100 TWh (or more), which we consider to be the most relevant for climate change mitigation and this analysis. In Extended Data Table 1, we report regional and five national reference cases with the fastest maximum growth rate together with global values. In Extended Data Fig. 3, the IQRs of the maximum growth rates in the national reference cases are estimated based on all cases, regardless of the size of the electricity market.

We also calculate the duration of CCS diffusion from 10% to 90% of its maximum capacity, envisaged in mitigation pathways, or Δt (equations (12) and (13)).

$$\Delta {t}_{{\mathrm{gmp}}}=\frac{\ln \left(\frac{\ln (0.1)}{\ln (0.9)}\right)}{k}$$
(12)
$$\Delta {t}_{{\mathrm{log}}}=\frac{\ln (81)}{k}$$
(13)

Vetting of mitigation pathways using feasibility constraints

In vetting the IPCC AR6 climate mitigation pathways with the proposed metrics, we combine the three feasibility spaces for analysing the feasible range of CCS deployment in the formative, acceleration and stable growth phases. Our upper constraints for each step are:

  • Formative phase: up to a 45% failure rate and a doubling of planned CCS capacity by 2025. This constraint results in the maximum feasible CCS capacity by 2030 (0.37 Gt yr−1).

  • Acceleration phase: the acceleration rate of CCS in 2030–2040 is never faster than the highest acceleration observed for any of our reference cases in the acceleration phase (wind, solar and nuclear power).

  • Stable growth phase: up to a 1.45% maximum annual growth rate. This global maximum annual growth rate was achieved by nuclear power in 1984 (average for Gompertz and logistic model fits). This constraint represents the maximum feasible CCS growth rate that can occur any time after take-off.

We measure the cumulative amount of CO2 captured and stored by 2070 and 2100 in the IPCC AR6 climate mitigation scenarios by summing the linearly extrapolated capacity of CCS technologies between each decade, starting from 2030.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.