When 10 gigatons isn’t enough

E​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​ngineers sometimes joke about answers that have absurdly large first steps. “How do we solve the energy crisis? Step 1: Build a Dyson sphere.” Are we taking this same approach to global warming? Is our step one unrealistic, and is our dogged adherence to this first step holding us back from seriously considering other possible climate solutions?

The net-zero balancing act

In the climate policy world, step one is net zero—a term that basically means humans end or at least counterbalance all of their carbon dioxide (CO2) emissions. We dumped about 42.2 billion tons (gigatons) of CO2 into the Earth’s atmosphere last year alone. In a net zero world, this number drops to zero. At this point—so goes the thinking, anyway—temperatures will level off at around 1.5°C above preindustrial levels (but no more than 2°C) and our global warming worries will be mostly over.

So far, however, our effort to reach net zero has almost entirely centered around reducing emissions, with very little attention paid to counterbalancing emissions through removals as well. In our reduction portfolio, we have been trying to switch to renewable sources of energy, improve the efficiency of homes and industries, drive electric cars, and so on. For the smidgeon of CO2 left over—from so-called hard to abate emissions—we have contemplated a wide variety of promising carbon dioxide removal (CDR) solutions, but none of these has been deployed yet at climate relevant scale.

In the meantime, our emissions-first approach hasn’t worked so well. Both temperatures and emissions keep going up, despite generations of ambition and over a decade now of corporate and government net zero plans and agreements. Emissions are actually higher today than they were when the landmark Paris Agreement was signed in 2015, which was supposed to give us a realistic roadmap for reaching net zero.

Climate policy enthusiasm is flagging, too. A pall of eco-fatalism is settling as the magnitude of the climate challenge ahead is leading companies, governments, and even longtime climate advocates to reconsider whether the sacrifices being called for are realistic, economically advisable, or likely to succeed. Notable figures like Bill Gates are arguing we should focus more attention on alleviating present-day suffering than solving increasingly apocalyptic climate scenarios; leading companies like Microsoft are sharply curtailing their purchases of carbon removal even as they build natural-gas-powered AI infrastructure; and in the United States—once the world’s center of CDR investment and innovation—the Trump administration has gutted clean-energy and the federal climate-research enterprise. Even Paris Agreement commitments are lagging. At present, 45 countries are 18 months delinquent on updating their NDCs, which were due in February 2025; some are reportedly not even returning phone calls.

Still, policymakers remain unrelenting in their focus. It’s emissions reductions or bust, which is fine, maybe, except that we don’t also have a viable Plan B for keeping global temperatures in check in case our emissions goals don’t pan out. We have adaptation plans galore, plus radical (at least in terms of public acceptance) ideas for cooling our planet through solar radiation management and other geoengineering solutions, plus an increasing focus on capturing industrial methane and carbon emissions. But fundamentally, we’ve put all our global warming eggs into the emission reduction basket.

What if emissions don’t come down any time soon? And even if the Paris Agreement’s goal of net zero is achievable, what then? Given what we’ve learned since the Agreement was signed about the consequences of warming—and from the increasingly climate-disrupted world we’re already experiencing—what evidence is there that reaching 1.5°C and simply staying there is a good idea, rather than trying to push temperatures back down toward their historic preindustrial level?

In this context, it isn’t controversial to note that we need to begin removing carbon dioxide from our planet’s atmosphere. CO2 is the primary driver of global warming, and its concentrations have been increasing rapidly in recent years, driving temperatures higher along with it. CDR, therefore, has been gradually entering the mainstream of climate policy conversations, albeit mostly through nature-based approaches like planting more trees, enhancing soils, or capturing the emissions from decaying plant material (a process known as biochar). These conversations represent progress, but only to a point. What is very controversial today, still, is to suggest we should focus on CDR at least as intently as we have been focusing on emissions reductions.

Why we’ve aimed low

What we have been doing instead is quite the opposite. Since it was first proposed as a necessary climate repair action in the earliest IPCC reports, CDR has been relegated to a minor subset of a few scattered national and subnational mitigation policies—a distant second to emissions reductions, supported largely through incentives and project-level support rather than through comprehensive government programs designed to deliver CDR at climate-relevant scale. Inasmuch, our historical goals for CDR have been imbued with assumptions about timing, funding, and political support that might not necessarily reflect what we actually need today to save our planet.

Consider the popular Climate Action Tracker (CAT) model, for example. In its latest update (from November of 2025), this model suggests that if we stay on our current “business as usual” emissions and policy trajectory, our Earth might warm to 3.3°C above preindustrial levels by the year 2100. But CAT is mostly looking at the reduction side of the balance beam, not the removal side; this is an emissions-pathway model, asking how far warming falls as governments reduce emissions. Even in this model’s best case temperature outcome, the amount of CDR called for is capped at 8.6 Gt/y (of which 3.6 Gt/y is through planting trees and restoring forests), with emissions cuts achieving the vast majority of the temperature control. The clear message from policymakers is that emissions reduction is the only way we’ll be able to keep global temperatures in check.

The reasons for our recalcitrance about CDR are likely four-fold. First and foremost is our net zero mindset. Almost all our climate models—not just CAT—assume the world’s total anthropogenic emissions (i.e., emissions caused or made by humans) need to begin dropping rapidly in the near future, even though the reality has been quite different. Burning fossil fuels currently adds about 38.1 Gt of CO2 pollution to our atmosphere every year (about 90% of the overall total), increasing at a rate of about 1.1% annually. Another 4.1 Gt of anthropogenic CO2 comes from net land-use change, such as deforestation, forest degradation, logging, and peat loss; this figure has been declining by around 10% annually. Together, these two components added 42.2 Gt of anthropogenic CO2 emissions to the Earth’s atmosphere in 2025, growing at a rate of about 0.3% per year over the last decade. Reality notwithstanding, we are invested in the Paris Agreement’s net zero goals, and will keep pushing almost exclusively on the emissions reduction lever, whether or not we should.

Second, and probably feeding into our emissions-first mindset, is the foundational influence of prominent voices like former US Vice President Al Gore who have long portrayed CDR as a moral hazard—as a course of action that will weaken pressure for near-term emissions reductions and prolong our fossil-fuel use. This concern is well represented in the literature (see, for example, analyses by Asayama, Brad and Schneider, Carton et al., and Markusson) and helps explain why CDR is often framed as a complement to deep emissions cuts rather than as an alternative pathway. This caution, however, has also narrowed the policy question at hand from “What combination of actions will most reliably and effectively control temperatures?” to “How little CDR can we use while achieving an emissions pathway already assumed to succeed?”

The third reason global policymakers have been reluctant to embrace CDR so far is a lack of confidence in how quickly markets and technology will be able to rise to the challenge. Even CDR evangelists have largely concluded that CDR as we know it can’t possibly deal with global warming on its own without significant emissions reductions first, so we have deemed it somewhere between impractical and impossible to scale CDR beyond certain levels. Instead, we are only looking for solutions that can close a small gap—hard to abate emissions, or the residual CO2 that remains after emissions reductions have brought us closer to net zero. In this sense, CDR isn’t an equal partner in our fight against global warming, or even in our attempt to reach net zero, but more of an aside. The actual heavy lifting has been, is now, and will likely remain (until the global warming crisis convinces us otherwise) almost entirely centered around emissions reductions.

The fourth and final main reason why our CDR goals have been marginalized is that we’ve already settled on a removal goal that might in fact be far too low. Specifically, climate researchers and policymakers frequently note that the world should try to reach around 10 gigatons per year (Gt/y) of CDR capacity by the year 2050. We are a very long way from achieving this, but 10 Gt/y has nevertheless come to function as something of a finish line. If we can piece together enough projects across direct air capture, biochar, enhanced rock weathering, soil carbon, and other approaches, and engage enough companies and even countries in this effort, then we’ll be able to find a way. This 10 Gt/y number doesn’t represent a scientific consensus about how much CDR we’ll need for all time periods and scenarios, however. In a recent article published in the CDR preprint server CDRXiv, University of Washington climate researcher Rebecca Neumann traced the history of this number through a wide variety of scientific, government, NGO, industry and media sources and concluded that the primary driver of this variation is not what is required to keep our planet’s temperature stable, but how quickly we assume fossil fuel emissions will be reduced. Scenarios that permit greater continued reliance on fossil fuels also require more CDR later; scenarios with deeper emissions reductions now require much less CDR later. The 10 gigaton number, therefore, is not a standalone geophysical requirement, but a pathway-dependent conclusion derived from assumptions about future emissions reductions, technology, cost, deployment speed, environmental constraints, and policy choices. And yet we see 10 Gt/y mentioned everywhere in research, media, and official policy work, as though this number represents some sort of scientific consensus about what the future of CDR requires.

Taken together, as much as we may need CDR to counterbalance the effects of our continued fossil fuel use, we have a long way to go before CDR can come to the rescue—not only technically in terms of whether CDR can in fact evolve into the tool we need, but also politically in terms of CDR overcoming decades of bias and marginalization. At minimum, CDR needs to be part of our climate plan; what we don’t know is how much CDR we’ll actually need if we look honestly at our CO2 emission trendlines and our progress to-date on reaching net zero.

The ZERO model

What happens in a world where we sidestep all our baked-in biases and assumptions about CDR and just do the math? CO2 emissions aren’t coming down. How much CDR will it take to keep global temperatures under control if CDR is the only tool we have to work with (outside of solar radiation management, which is a different paper), assuming we’ll be able to do this (casting aside all our judgements about cost, technology, political roadblocks and so on)? We’ll call this wildly pessimistic and yet wildly optimistic model “ZERO,” which stands for the “zero emissions reductions outlook” model. For simplicity, we’ll hold total anthropogenic CO2 emissions constant at 42.2 Gt/y instead of applying the 0.3% annual growth rate. In effect, this assumes that continued growth in fossil CO2 emissions will be eventually offset by reductions elsewhere, including declining net land-use emissions, so that the global total stops rising but doesn’t fall. The point of our ZERO model is to develop an order-of-magnitude sense of what our CDR needs might be in a persistently high-emissions world, not to calculate this requirement to the decimal point.

The following table shows the outcomes, with various levels of CDR deployed starting in the year 2050 (we’ll call this particular variant the ZERO-S model, with the S signifying “step change,” going from essentially zero CDR now to 10 Gt/y or more in 2050; this isn’t realistic, but again, the point is to see the big picture). Temperature estimates were generated using the multi-decadal approximation model published in Myles Allen et al.’s 2024 geological net zero paper, which combines IPCC’s well-established Transient Climate Response to Cumulative Carbon Emissions (TCRE) model with the temperature response to non-CO2 radiative forcing. The results align reasonably well (plus or minus a few tenths of a degree) with Oxford’s FaIR simple climate model; through 2100, the mean absolute difference between the Allen calculation and the FaIR ensemble median is about 0.10°C across the eight CDR scenarios, with a maximum difference of about 0.25°C (the gap grows at longer horizons, as we would expect). The annex section of this article describes this methodology in more detail, and a spreadsheet is linked where researchers can repeat these calculations.

As you can see from these estimates, if we don’t bend our emissions curve before 2050 and also wait until 2050 to start deploying CDR at scale, then it’s essentially futile to try removing only 10 Gt/y. Instead, CDR levels of around 50 Gt/y or more are required to flatten our planet’s long-term warming curve (and that’s assuming our total anthropogenic CO2 emissions stay at current levels and that other feedback mechanisms don’t kick in—both unwarranted assumptions). Indeed, even 50 Gt/y is irresponsible insofar as it leaves our planet hovering above 2°C for the next 100-plus years. In the face of persistently high CO2 emissions, only CDR levels nearing 100 Gt/y begin returning temperatures toward the preindustrial baseline on a roughly century-scale horizon, which is to say, within the next five generations or so, rather than expecting our battle against global warming to rage on for centuries to come.

ZERO-S TEMPERATURE TRAJECTORIES

(2) CUMULATIVE NET ANTHROPOGENIC CO2 BY MEASUREMENT YEAR (GT)

(3) AVERAGE TEMPERATURE CHANGE PER DECADE BY INTERVAL (°C/DECADE)

(1) Net anthropogenic CO2 = anthropogenic CO2 emissions minus active anthropogenic CDR. Natural land and ocean sinks are not subtracted. Allen et al. distinguish these passive sinks from active removals because passive uptake must remain available to draw down legacy CO2.

(2) Cumulative net anthropogenic CO2 changes throughout the trajectory and is therefore shown in a separate companion table rather than as a single value in the main table. The 2050 starting value is about 3,785 Gt: approximately 2,730 Gt was emitted through 2024, 42.2 Gt in 2025, and another 1,013 Gt from 2026 through 2049 if emissions remain at 42.2 Gt/y. Thereafter, each 25-year measurement adds 25 years of net anthropogenic CO2. These values are an anthropogenic emissions/removals ledger, not the amount of CO2 physically present in the atmosphere. In the most aggressive removal scenario the ledger becomes negative after the historical anthropogenic total has been erased. Values beyond that point should not be interpreted literally without a dynamic carbon-cycle model.

(3) Temperature change per decade also changes over time and is therefore shown in a separate companion table rather than as a single rate in the main table. Each value is the average rate over the corresponding 25-year interval, calculated as temperature at interval end minus temperature at interval start, quantity divided by 2.5 decades. The underlying temperature calculation for each interval uses Allen et al.’s multi-decadal approximation.

(4) Assumes 1.37°C of human-induced warming in 2025, increasing to 2.02°C by 2050 by extending the current 0.27°C/decade warming rate as a deliberately pessimistic pre-2050 stress-test assumption. CDR is assumed to reach the stated scale in 2050.

(5) Temperatures after 2050 are calculated in 25-year intervals using Allen et al.’s approximation. See annex for details.

(6) The Allen calculation was pressure-tested against FaIR using its 841-member calibrated and constrained parameter ensemble and the same post-2050 emissions, removals, and non-CO2 forcing assumptions. Agreement is close through 2100 and remains directionally similar through 2150, although uncertainty and model divergence increase with time and with large sustained negative emissions.

What if we don’t suddenly deploy 100 Gt/y in 2050 but instead work up to this level, starting with, say, 5 Gt/y in 2030-40, increasing to 10 Gt/y in 2050, and then adding 20 Gt/y of removal capacity every decade until we reach 100 Gt/y? How long would our removal efforts take in this case? We’ll call this variant the ZERO-G model, where the G stands for “gradual” CDR deployment.

ZERO-G MODEL TEMPERATURE TRAJECTORIES

In this scenario, our global temperatures will keep climbing until around the year 2080, peaking at 2.25°C before turning downward to around 1°C by around the year 2145 (and continuing lower thereafter, although our model can’t predict exactly how much or how fast). These outcomes are similar to the ZERO-S model for CDR levels below 60 Gt/y. By deploying CDR gradually over the coming century and reaching a peak deployment capacity of 100 Gt/y by 2100, we will experience several decades of temperatures at 2.25°C or higher. Still, barring any irreparable damage from these conditions (which again is certainly not a given) or interventions like solar radiation management to help keep our temperatures under control in the interim, this CDR level will put us on a glidepath to relatively rapid climate restoration.

How does the ZERO model compare with State of CDR estimates?

The latest State of Carbon Dioxide Removal (SoCDR) report helps explain where the familiar 7–10 Gt/y CDR figure is actually used and why it should not be mistaken for a long-term estimate of how much carbon removal our climate may ultimately require. SoCDR groups its findings and recommendations into three main scenario families: Targets & Pledges, Highest Ambition, and Delayed Ambition. The so-called Targets & Pledges scenario is based on current national climate goals; Highest Ambition explores an aggressive, rapid path to net-zero; and the Delayed Ambition scenario shows the consequences of postponing emission cuts, requiring higher reliance on future carbon removal. Each scenario is evaluated across several integrated assessment models, producing a range of individual pathways.

In SoCDR’s Highest Ambition scenario, the median total CDR across pathways rises from about 2.9 Gt/y in 2030 to 3.9 Gt/y in 2035 and 8.8 Gt/y in 2050. Net-zero CO2 is reached around mid-century, although the exact year varies among models. At this point, the median amount of CDR is about 10.9 Gt/y, roughly enough to cancel the positive CO2 emissions that still remain. Emissions reductions are assumed to have done most of the work needed to reach net zero. Cumulatively, about 84% of this achievement comes from cutting emissions and about 16% from CDR.

But SoCDR does not conclude that 10.9 Gt/y will be a sufficient level of CDR forever into the future. Instead, under its Highest Ambition scenario, CDR rises to a median level of 15.3 Gt/y by 2100, with some model pathways reaching 24.7 Gt/y. If aggressive emissions reductions are delayed by ten years, the median rises to 23.6 Gt/y under SoCDR’s Delayed Ambition scenario, with some model pathways reaching as high as 28.2 Gt/y.

This comparison becomes more revealing when we look at net-negative CO2 flow rather than gross CDR. For example, again, in SoCDR’s Delayed Ambition scenario, 23.6 Gt/y of CDR is being deployed by 2100 (the median level), but some of this removal is still needed to offset continuing residual emissions; the net-negative CO2 flow at this time is roughly 15 Gt/y. Achieving the same net-negative rate in the ZERO-S model requires about 57 Gt/y of CDR (because 42.2 Gt/y must first be used to cancel continuing emissions). Yet these two worlds produce very different cooling outcomes. In the ZERO-S world, a net-negative rate of about 15 Gt/y leaves the global average temperature at roughly 2.14°C in 2100, versus around 1.5°C in the SoCDR world.

This difference can be attributed to starting temperatures and forcing assumptions. For starting temperatures, SoCDR models assume that CO2 emissions will come down and CDR will begin scaling to climate-relevant levels before 2050 (the ZERO-S model assumes neither), so temperatures at 2050 are a few tenths of a degree cooler. With regard to forcing, SoCDR’s Highest and Delayed Ambition scenarios model deeply mitigated worlds in which both CO2 and non-CO2 forcing decline, while ZERO models stress-test a world in which gross CO2 emissions remain at 42.2 Gt/y and non-CO2 forcing remains high. These two approaches therefore arrive at similar net-negative CO2 flows, but under very different physical conditions.

COMPARISON OF RECOMMENDATIONS AND TEMPERATURE PROJECTIONS IN SOCDR SCENARIOS VS. ZERO MODELS

Comparing these approaches is important because the State of CDR report makes it abundantly clear that the amount of CDR ultimately deemed necessary is a direct reflection of our assumptions about emissions reduction, which reinforces Neumann’s assessment about the origins of the 10 Gt/y recommendation. As well, this comparison makes it clearer that SoCDR’s headline mid-century numbers are in fact not estimates of the maximum amount of CDR society might need, but are instead medians in a deployment approach that aims to reduce global average temperatures to 1.5°C within a reasonable period of time, and that this approach also requires doubling or even tripling CDR levels by 2100.

Finally, it’s important to note that SoCDR’s estimates are consistent with Paris-aligned mitigation pathways—that is, they do not model a “zero progress” world as a worst case outcome—and they are not designed to stabilize temperatures if global emissions remain high, or restore our climate to its preindustrial temperature. SoCDR’sHighest Ambition scenario spends a median 55 years above 1.5°C; the Delayed Ambition scenario spends 68 years above it. SoCDR does not ask how much CDR would be required to drive temperatures any lower.

Perhaps future agreements will supersede Paris in order to properly address climate restoration, but for now at least, our “1.5°C should be fine” approach—while it seems exceptionally aggressive by our current standards—recommends less CDR than our climate needs for restoration by a factor of three, five or even ten. CDR levels in a slowly decarbonizing or non-decarbonizing world, or in a world that we’d like to have closer to 0°C than 1.5°C within the next few hundred years, will require a level of CDR ambition that far exceeds even our most ambitious current plans.

Has this question already been modeled?

Not in this precise form, as far as can be determined. There is a substantial literature on high-emissions pathways, delayed mitigation, overshoot, negative emissions, and the risk that anticipated CDR can substitute for emissions reductions (see, for example, Keller et al., MacDougall, Tokarska and Zickfeld, Carton et al., and Smith et al.). Some recent studies examine this substitution problem more directly. Ampah et al., for example, found that high reliance on CDR can sustain substantially higher residual fossil-fuel and industrial emissions and delay net zero, while Bindl, Edwards, and Cui found that planning for high CDR under a 2°C target can produce minimal fossil-fuel phaseout before mid-century. MacIsaac et al. further show that balancing additional fossil-fuel CO₂ emissions with reforestation-based removals does not necessarily produce the same climate outcome as avoiding those emissions in the first place. What appears less developed is the narrower stress test posed by the ZERO model: holding gross CO2 emissions near today’s level and asking how the required scale of CDR changes when the objective is temperature stabilization or restoration rather than compliance with net-zero goals.

Can we do this?

The answer is probably not yet. We can only get so far with the CDR technologies we’re currently considering since the construction and operating costs blow past the multi-trillion dollar mark annually even at the 10 Gt/y level. If we start thinking in terms of 100 Gt/y instead, we may need to come up with new solutions, not just better versions of the solutions we already have.

These new solutions might include largely unexplored ideas like passive solid-state conversion, so novel that it hasn’t even reached pilot stage yet, or direct ocean removal. Our “improved” solutions might include accelerating investigation into ocean iron fertilization and ocean alkalinity enhancement, which both have enormous removal potential and remarkably low costs (relative to other methods) but which are also among the least tested solutions to-date (at least at scale).

We also need to start investing heavily in solutions that are most likely to succeed at this kind of scale. Like the 10 Gt/y figure itself, conventional wisdom holds that our CDR future will likely employ a basket of solutions, with different CDR approaches deployed in different parts of the world. But if any one of these new or improved approaches can demonstrate the capacity to quickly and affordably remove ten-plus gigatons per year without the need for massive construction, pipelines, and trillion dollar budgets, then our conversation about the highest best role for CDR in our fight against global warming might change dramatically.

The good news is that surely our CDR capabilities will improve over the coming decades and centuries, as will our CDR ambitions. And the global policy mindset will also change, albeit slowly. For example, the IPCC’s Sixth Assessment Report, issued in 2022, concluded that CDR is unavoidable for achieving net-zero CO2, and this conclusion is increasingly reflected in subsequent UN assessments like UNEP’s 2025 Emissions Gap Report. As well, several governments—although still not many—are beginning to institutionalize some of the mechanisms needed to make CDR work at scale, including durable finance, clearer regulation, and legal integration.

But in the end, how much CDR we actually need will depend on the temperature goals we set. And we haven’t done this yet. Instead, we’ve set emissions goals, and as these goals have failed, we have let our temperature goals fail as well. The tail has been wagging the dog.

The bottom line question is this: If fossil fuel use doesn’t come down despite our best efforts, should we continue to persist in our generations-long fight to the exclusion of all other approaches, or should we also start developing, in earnest, a viable Plan B to keep our planet cool? (This is rhetorical, by the way.) From a policy perspective, we need to be realistic about whether our current approach has worked. If we need to change course, then the longer we delay, the more perilous our coming journey will be.

Annex: ZERO calculations

The figures in this paper are derived from the multi-decadal temperature approximation developed by Oxford physicist Myles Allen and others for their 2024 Nature article, “Geological Net Zero and the need for disaggregated accounting for carbon sinks.” This formulation keeps the familiar IPCC TCRE relationship between cumulative CO2 emissions and warming, but also allows the calculation to include the slow adjustment of temperature to radiative forcing and a separate pathway for non-CO2 forcing. The spreadsheet linked below shows how equations were applied and temperature estimates were calculated. The same emissions and forcing assumptions were then run through FaIR as a pressure test of the analytical results. ChatGPT was used to build the formulas, pull the FaIR model from GitHub, run the analysis, and create the Excel spreadsheet.

For each 25-year interval after 2050, temperature change is calculated using Allen et al.’s approximation:

ΔT = k_E[ΔG + (ρ_F − ρ_E)G_avg Δt] + k_F[ΔF + ρ_F F_avg Δt]

where

• ΔT = temperature change over the interval
• k_E = temperature response per unit of cumulative CO2
• ΔG = change in cumulative net anthropogenic CO2
• ρ_F = slow temperature-adjustment rate
• ρ_E = slow carbon-cycle-adjustment rate
• G_avg = average cumulative net anthropogenic CO2 over the interval
• Δt = length of the interval
• k_F = temperature response per unit of radiative forcing
• ΔF = change in non-CO2 radiative forcing
• F_avg = average non-CO2 radiative forcing over the interval

Net anthropogenic CO2 in this calculation means gross anthropogenic CO2 emissions minus active anthropogenic CDR. Natural land and ocean sinks are not subtracted from this figure or from cumulative anthropogenic emissions. This follows the model’s distinction between active removals and passive uptake: natural sinks already respond to the excess CO2 in the atmosphere and must remain available to draw down the legacy burden. The cumulative figures in the companion table are therefore an anthropogenic accounting ledger, not an estimate of how much CO2 is physically present in the atmosphere. This also avoids the mistake of assuming that because land and ocean currently absorb roughly half of annual emissions, only half of gross emissions need to be balanced by active CDR.

Finally, non-CO2 warming is represented as radiative forcing. We use the non-CO2 component of the SSP3-7.0 high-forcing trajectory from the Reduced Complexity Model Intercomparison Project (RCMIP), shifted by a constant amount so that its 2025 value matches the latest assessed anthropogenic forcing balance: 3.10 W/m2 of total anthropogenic effective radiative forcing minus 2.37 W/m2 from CO2, or about 0.73 W/m2 from non-CO2 influences. The resulting non-CO2 forcing values used here are approximately 1.02 W/m2 in 2050, 1.38 in 2075, 1.74 in 2100, 1.87 in 2125, and 1.85 in 2150. Only this non-CO2 forcing trajectory is borrowed from SSP3-7.0; CO2 emissions are held at 42.2 Gt/y throughout the experiment and CDR is varied independently.

As a pressure test, we ran the same post-2050 CO2 emissions, CDR levels, and non-CO2 forcing trajectory through FaIR, a physically based reduced-complexity climate model that converts emissions to concentrations, radiative forcing, and temperature while allowing the carbon-cycle response to change with cumulative uptake and warming. We used FaIR’s published 841-member calibrated and constrained parameter ensemble and rebased each run to the same 2.018°C starting temperature in 2050 so the comparison tests the post-2050 response rather than differences in historical initialization. The results are encouraging. In 2100, the central Allen calculation gives 3.42°C with no CDR, 2.47°C with 42.2 Gt/y of CDR, 2.07°C with 60 Gt/y, and 1.17°C with 100 Gt/y; the corresponding FaIR medians are 3.18°C, 2.42°C, 2.08°C, and 1.26°C. Across all eight scenarios, the mean absolute difference in 2100 is about 0.10°C and the maximum is about 0.25°C. By 2150 the mean difference increases to about 0.18°C and the maximum to about 0.51°C, which is consistent with Allen et al.’s warning that the analytical approximation becomes less reliable over longer periods and under large sustained negative emissions. FaIR is itself an emulator rather than a full Earth-system model, so this agreement should be viewed as a pressure test, not a validation of precise future temperatures. No separate temperature increments are added for albedo, cloud, water-vapor, permafrost, or other feedbacks; much of the fast climate response is represented in the calibrated response parameters, while slower Earth-system feedbacks remain an important source of uncertainty. For this reason, the table stops at 2150 and does not assign specific restoration dates to the net-negative scenarios.

Selected references

Allen, Myles R., David J. Frame, Pierre Friedlingstein, et al. 2025. “Geological Net Zero and the Need for Disaggregated Accounting for Carbon Sinks.” Nature 638: 343–350. doi:10.1038/s41586-024-08326-8.

Climate Action Tracker. 2025. Little Change in Warming Outlook for Four Years; New 2035 Climate Targets Make No Difference. November 13, 2025.

Edwards, Morgan R., Oliver Geden, Matthew J. Gidden, William F. Lamb, Jan C. Minx, Gregory F. Nemet, Stephen M. Smith, et al. 2026. The State of Carbon Dioxide Removal—3rd Edition. OSF. doi:10.17605/OSF.IO/ZRD65.

Fankhauser, Sam, Stephen M. Smith, Myles Allen, Kaya Axelsson, Thomas Hale, Cameron Hepburn, et al. 2022. “The Meaning of Net Zero and How to Get It Right.” Nature Climate Change 12: 15–21. doi:10.1038/s41558-021-01245-w.

Friedlingstein, Pierre, Michael O’Sullivan, Matthew W. Jones, Robbie M. Andrew, Dorothee C. E. Bakker, Judith Hauck, Peter Landschützer, et al. 2026. “Global Carbon Budget 2025.” Earth System Science Data 18: 3211–3288. doi:10.5194/essd-18-3211-2026.

Intergovernmental Panel on Climate Change (IPCC). 2018. Global Warming of 1.5°C: An IPCC Special Report on the Impacts of Global Warming of 1.5°C above Pre-industrial Levels and Related Global Greenhouse Gas Emission Pathways. Geneva: IPCC.

Intergovernmental Panel on Climate Change (IPCC). 2022. Climate Change 2022: Mitigation of Climate Change. Contribution of Working Group III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. doi:10.1017/9781009157926.

Leach, Nicholas J., Stuart Jenkins, Zebedee Nicholls, Christopher J. Smith, John Lynch, Michelle Cain, Tristram Walsh, Bill Wu, Junichi Tsutsui, and Myles R. Allen. 2021. “FaIRv2.0.0: A Generalized Impulse Response Model for Climate Uncertainty and Future Scenario Exploration.” Geoscientific Model Development 14: 3007–3036. doi:10.5194/gmd-14-3007-2021.

MacDougall, Andrew H. 2013. “Reversing Climate Warming by Artificial Atmospheric Carbon-Dioxide Removal: Can a Holocene-Like Climate Be Restored?” Geophysical Research Letters 40 (20): 5480–5485. doi:10.1002/2013GL057467.

Neumann, Rebecca B. 2026. “How Carbon Dioxide Removal Lost Its Way: Tracing the Origin and Transformation of the 10-Gt Durable CDR Target.” CDRxiv preprint, May 19, 2026.

Pett-Ridge, Jennifer, Sara Kuebbing, Allegra C. Mayer, Susan Hovorka, Hélène Pilorgé, Sarah E. Baker, Simon H. Pang, Corinne D. Scown, et al. Roads to Removal: Options for Carbon Dioxide Removal in the United States. Livermore, CA: Lawrence Livermore National Laboratory, 2023. LLNL-TR-85290. https://doi.org/10.2172/2301853.

Smith, Chris, Donald P. Cummins, Hege-Beate Fredriksen, Zebedee Nicholls, Malte Meinshausen, Myles Allen, Stuart Jenkins, Nicholas Leach, Camilla Mathison, and Antti-Ilari Partanen. 2024. “fair-calibrate v1.4.1: Calibration, Constraining, and Validation of the FaIR Simple Climate Model for Reliable Future Climate Projections.” Geoscientific Model Development 17: 8569–8592. doi:10.5194/gmd-17-8569-2024.

Smith, Chris, Lennart Ramme, Christopher D. Wells, Ada Gjermundsen, Hongmei Li, Tatiana Ilyina, Adakudlu Muralidhar, et al. 2026. “Overshoot and (Ir)reversibility to 2300 in Two CO2-Emissions Driven Earth System Models.” Earth System Dynamics 17: 893–911. doi:10.5194/esd-17-893-2026.

Tokarska, Katarzyna B., and Kirsten Zickfeld. 2015. “The Effectiveness of Net Negative Carbon Dioxide Emissions in Reversing Anthropogenic Climate Change.” Environmental Research Letters 10 (9): 094013. doi:10.1088/1748-9326/10/9/094013.

United Nations Environment Programme (UNEP). 2025. Emissions Gap Report 2025: Off Target. Nairobi: United Nations Environment Programme.

United Nations Framework Convention on Climate Change (UNFCCC). 2015. Paris Agreement. Bonn: UNFCCC.