The Interventions Nobody Wants to Need
Emissions reduction is necessary and, on current trajectories, insufficient. That sentence contains the entire justification for this chapter.
The reason is physical rather than political. Carbon dioxide persists in the atmosphere for centuries; a substantial fraction of what has been emitted will still be there in a thousand years. Even if emissions stopped entirely tomorrow, temperatures would stay near current levels rather than returning to the previous baseline, and roughly another half degree is already committed by heat currently absorbed in the oceans.
Emissions cuts stop the problem getting worse. They do not reverse it.
Which leaves two further categories of action, both uncomfortable. Carbon dioxide removal addresses the cause by taking CO₂ back out of the air. Solar radiation management addresses one symptom by reflecting sunlight before it becomes heat. And adaptation accepts what cannot be prevented.
Climate engineering was taboo in policy circles for years, on the reasonable grounds that discussing it reduces pressure to cut emissions—a moral hazard argument that is genuinely valid and that has also produced two decades of not preparing for a contingency that is becoming more likely.
2026 Snapshot — Where the Technologies Stand
Carbon Dioxide Removal
Engineered removal currently runs below 0.01 gigatons of CO₂ per year. IPCC pathways consistent with 1.5°C require somewhere between 5 and 15 gigatons annually by mid-century.¹
That gap is roughly three orders of magnitude, and it is the most important number in this chapter. Every scenario in which warming is held near current levels depends on scaling an industry that today captures a rounding error, on a timeline of about twenty-five years.
Direct air capture costs $400–1,000 per ton today, with Climeworks at the upper end of that range and Carbon Engineering claiming a path toward $100–200 at scale; the IEA projects $125–335 per ton by 2030.² At present prices, removing a single gigaton would cost more than the GDP of most countries.
Solar Radiation Management
No deployment, limited research, and an active governance debate that has produced no framework.
The physical case is well established by accident. Mount Pinatubo's 1991 eruption injected roughly 20 megatons of sulfur dioxide into the stratosphere and cooled the planet by about half a degree for a year or two.⁵ The mechanism works; nature has demonstrated it repeatedly.
The cost is the disturbing part. Estimates for stratospheric aerosol injection sufficient to offset roughly a degree of warming run to a few billion dollars annually—Harvard's assessment suggests $2–8 billion per year.⁶ That is affordable for a mid-sized country, a wealthy individual, or a consortium. Unlike every other planetary-scale intervention, the barrier to solar radiation management is not cost or capability. It is exclusively governance, and there is none.
Adaptation
Developing countries need an estimated $140–300 billion annually for adaptation by 2030.³ Current flows are roughly $20 billion. This is an order-of-magnitude shortfall concentrated precisely on the populations least responsible for the problem and least able to finance the response.
Notable Players
Climeworks operates the largest direct air capture facilities, with the Orca plant in Iceland capturing about 4,000 tons annually and the larger Mammoth facility targeting roughly 36,000.⁴ These numbers are worth stating alongside the gigaton requirement: the world's leading facility captures what a few thousand Americans emit.
Carbon Engineering, in partnership with Occidental Petroleum, is building larger DAC hubs in Texas targeting 500,000 tons per year.¹⁰ That partnership is instructive and contested—the oil industry has both the subsurface engineering expertise and the capital, and also an interest in enhanced oil recovery applications that partially defeat the purpose.
Carbfix in Iceland mineralizes captured CO₂ in basalt formations, converting it to rock within years rather than storing it as pressurized gas. This solves the permanence problem that dogs geological storage.
Frontier Climate—an advance market commitment funded by Stripe, Alphabet, Shopify, Meta, and McKinsey—has committed over $1 billion to purchase carbon removal that does not yet exist.⁸ This is the most important market mechanism in the sector, because the constraint on early-stage removal is not technology but the absence of any customer.
Research and governance bodies including Harvard's solar geoengineering program, the Oxford Geoengineering Programme, the Carnegie Climate Governance Initiative, and the National Academies occupy the awkward position of studying something whose study is itself controversial.
Carbon markets divide between compliance systems (the EU Emissions Trading System, California's cap-and-trade) and the voluntary market, worth roughly $2 billion annually and beset by quality problems severe enough that the Integrity Council for Voluntary Carbon Markets exists specifically to address them.⁹
Carbon Dioxide Removal
Direct Air Capture
Air is drawn through a contactor containing a sorbent that binds CO₂; heat then releases it in concentrated form for storage or use.
The engineering is straightforward and the thermodynamics are unforgiving. Atmospheric CO₂ is about 0.04 percent, which means processing enormous volumes of air to collect a small quantity of gas, and the energy penalty of that dilution is irreducible. Current systems require roughly 2–3 megawatt-hours per ton, and that energy must be clean or the exercise is self-defeating.
The genuine advantages are that it is modular—capacity scales by building more units, which is a manufacturing learning curve rather than a research problem—and that its storage is verifiable and permanent in a way biological approaches are not.
Enhanced Weathering
Crushed silicate rock—typically olivine or basalt—spread across land or coastline reacts with CO₂ as it weathers, converting it to stable carbonates.
The theoretical potential is large, perhaps 2–4 gigatons annually, and the cost could be considerably lower than DAC at $50–200 per ton.⁷ Agricultural land is a natural deployment site, since the rock dust also improves soil.
The difficulty is measurement. Weathering happens slowly, in the ground, across large areas, and quantifying how much carbon has actually been sequestered is genuinely hard—which makes it difficult to sell and easy to overstate. Mining, crushing, and transport also carry emissions that must be netted out.
Nature-Based Approaches
Afforestation, reforestation, soil carbon practices, and wetland restoration could contribute perhaps 3–5 gigatons annually, with real co-benefits for biodiversity, water, and rural livelihoods.
They also carry a structural weakness that the carbon market has systematically underweighted: impermanence. A forest is carbon storage only for as long as it remains a forest, and forests burn, are logged, and die in droughts. Several large offset projects have literally burned down. Land also competes with food production, and total capacity saturates.
Nature-based removal is worth doing on its own merits. Treating it as equivalent to permanent geological storage in an accounting system is how offset markets lost their credibility.
Ocean-Based Approaches
Alkalinity enhancement, iron fertilization, and kelp cultivation all exploit the ocean's enormous carbon capacity. All remain at research stage, all have poorly understood ecosystem effects, and all face a governance problem: the open ocean belongs to no one, which means these interventions can be attempted without anyone's permission and without any framework for assessing consequences.
Solar Radiation Management
The Mechanism and the Temptation
Injecting reflective aerosols into the stratosphere would cool the planet within months. It is fast, cheap, and effective at the one thing it does.
It is also the most dangerous idea in this book, for reasons that are worth being precise about.
It does not address CO₂. Ocean acidification continues unabated, because acidification is a chemistry problem rather than a temperature problem. A world using SRM to hold temperatures down while emissions continue is a world with a stable climate and dying oceans.
Termination shock. Aerosols fall out within a couple of years, so the intervention must be sustained indefinitely. If it is masking two degrees of warming and stops—through war, economic collapse, or political change—that warming arrives over a few years rather than a century. Ecosystems that could have migrated in response to gradual change cannot respond to that. Beginning SRM creates an obligation that must be honored by every subsequent government, permanently.
Regional effects diverge. Global average cooling does not mean uniform cooling. Modeling consistently shows altered precipitation patterns, with monsoon systems that billions of people depend on among the most affected. An intervention that stabilizes temperature for one region while disrupting rainfall in another is not a shared benefit.
Attribution becomes impossible and blame does not. After deployment, every drought and flood anywhere will be attributed by someone to the intervention, correctly or not, and there will be no way to definitively establish otherwise.
The Governance Vacuum
No international framework governs deployment. No treaty prohibits it. No body has authority to authorize or forbid it.
Because it is cheap, a single state could deploy unilaterally, affecting everyone. That is not a hypothetical arrangement requiring new technology—it is the situation today.
Marine cloud brightening, which increases the reflectivity of low ocean clouds using sea salt spray, is more localized and more controllable, and correspondingly less studied and less effective. It may be the more governable option precisely because its effects are regional.
The uncomfortable conclusion is that the research and the governance framework are both needed, urgently, and that building the governance is the harder and more important task. A world that has not decided who may deploy is a world in which the decision will be made by whoever moves first.
Adaptation
Adaptation is not optional and receives a fraction of the attention that mitigation does.
Roughly half a degree of additional warming is already committed, and current warming is already producing impacts. The infrastructure, agriculture, and settlement patterns of the last century were designed for a climate that no longer exists.
The work is unglamorous and concrete: water storage and flood control; drought-tolerant crops and irrigation efficiency; building materials and designs that survive heat; early warning systems for heat waves; disease surveillance as vector ranges shift; sea walls, managed retreat, and restored natural barriers on coasts.
Three things are worth stating plainly.
The finance gap is the equity problem in its clearest form. Twenty billion dollars flowing against a $140–300 billion need, concentrated on countries that contributed least to the accumulated stock of emissions.
Adaptation has limits. Some coral reefs will not survive. Some low-lying land will not remain habitable. Some species cannot move fast enough. Adaptation reduces harm; it does not eliminate it, and pretending otherwise licenses inaction on mitigation.
Maladaptation is a real failure mode. Sea walls that protect one stretch of coast and accelerate erosion on the next. Irrigation that depletes aquifers. Air conditioning that raises emissions. Interventions that lock in vulnerability while appearing to reduce it.
The Path Forward
Near-Term Likely (2026–2032)
DAC capacity scales toward roughly a million tons annually—a thousandfold increase from today and still a rounding error against the requirement. Costs decline toward $200–300 per ton as manufacturing learning accumulates.
Advance market commitments continue to be the primary demand source, and compliance markets begin accepting durable removal, which would be the first genuine demand signal.
Nature-based approaches expand, with accounting standards tightening after a period of well-documented offset failures.
SRM research continues at small scale amid escalating controversy, and governance discussion intensifies without producing a framework.
Adaptation investment grows and remains far below need.
Plausible (2032–2040)
Engineered removal reaches perhaps 100 megatons annually at under $150 per ton, and becomes a recognizable industry rather than a set of demonstration projects.
An international SRM governance framework is established—most likely after a near-miss or a unilateral attempt rather than in anticipation of one.
If warming exceeds 1.5°C durably and impacts intensify, political pressure to deploy SRM becomes serious. This is the decision point the current absence of governance is setting up.
Adaptation becomes standard practice in infrastructure planning in wealthy countries, and remains underfunded everywhere else.
Wild Trajectory (2040+)
Gigaton-scale removal is achieved, net-negative emissions become possible, and atmospheric concentrations begin to decline for the first time in two centuries.
Or removal never scales, warming exceeds 2.5°C, SRM is deployed hastily under emergency conditions by an actor that did not wait for agreement, and the governance failure compounds the climate one.
Or neither: a 3°C world in which adaptation absorbs the difference unevenly, with the burden falling where it has always fallen.
Risks and Guardrails
Removal fails to scale. Guardrails: pursue a portfolio rather than betting on one method; fund research at levels proportionate to the role removal plays in every 1.5°C pathway; and, critically, do not build emissions plans that assume removal will arrive. Every scenario relying on large-scale future removal is borrowing against a technology that does not yet exist at scale.
Offsets substitute for reductions. This has already happened repeatedly. Guardrails: rigorous permanence and additionality standards; a clear accounting distinction between temporary biological storage and durable geological storage; and a hierarchy in which removal supplements reduction rather than substituting for it.
Unilateral SRM deployment. Guardrails: an international framework established before capability is needed rather than after it is used; transparency requirements for research; norms against deployment absent broad agreement; and monitoring capable of detecting deployment, since an intervention nobody can detect cannot be governed.
Termination shock. Guardrails: any deployment must be paired with a binding commitment to sustained emissions reduction and removal, so that the intervention has an exit rather than becoming permanent by default.
Adaptation inequality. Guardrails: climate finance that actually flows; loss and damage mechanisms with funding attached; and recognition that adaptation failure in vulnerable regions produces migration and instability that wealthy countries will experience regardless of their own adaptation spending.
The Deeper Questions
Is there a right to intervene? There is no pristine baseline to preserve; the atmosphere has been modified for two centuries. Declining to remove carbon is as much a choice as removing it. But intentional modification carries a responsibility that accidental modification did not, and the parties who would decide are not the parties who would bear the consequences.
Who decides? SRM would affect every person alive and could be initiated by one state. No existing institution has the authority or legitimacy to make that decision, and none is being built. This is the clearest governance gap described anywhere in this book.
Can technology solve what technology caused? Partially. Removal addresses the chemistry, and it will not by itself resolve the political economy that produced the emissions or the distributional questions about who pays. Technology expands the option set; it does not choose from it.
Conclusion
The uncontrolled experiment cannot be un-run. Cutting emissions stops it getting worse and does not undo what has been done.
Carbon removal addresses the actual cause, and the honest assessment is sobering: the leading facilities capture thousands of tons against a requirement measured in billions, at costs that would need to fall by roughly an order of magnitude. It is a real technology with a real learning curve and a genuinely uncertain destination, and every climate scenario that holds warming near current levels quietly assumes it succeeds.
Solar radiation management addresses a symptom, works quickly, costs remarkably little, and creates a permanent obligation while leaving ocean acidification untouched. Its barrier is not technical or financial. It is that no one has decided who may pull the lever, and the lever is within reach of many hands.
Adaptation is unavoidable, chronically underfunded, and concentrated on the people who did least to cause the problem.
None of this substitutes for emissions reduction. The portfolio is reduction first, removal to address the accumulated stock, adaptation for what arrives regardless, and radiation management as a contingency that should be governed before it is needed rather than after.
What has actually changed is the frame. For two centuries, humanity modified the climate without intending to. From here it will be modified deliberately or endured deliberately, and either way the choice is now an active one made by identifiable institutions. There is no longer a version of this in which nobody is responsible for the outcome.
Endnotes — Chapter 56
- IPCC pathways consistent with 1.5°C require approximately 5–15 Gt CO₂ per year of removal by 2050; current engineered removal is below 0.01 Gt per year.
- Direct air capture costs: Climeworks currently in the range of $600–1,000 per ton; Carbon Engineering claims potential for $100–200 per ton at scale; the IEA estimates $125–335 per ton by 2030.
- The UNEP Adaptation Gap Report estimates $140–300 billion per year is needed for adaptation in developing countries by 2030, against current flows of roughly $20 billion.
- Climeworks' Orca plant in Iceland has approximately 4,000 tons per year of capacity; the Mammoth facility targets approximately 36,000 tons per year.
- The Mount Pinatubo eruption (1991) injected roughly 20 Mt of SO₂ into the stratosphere, producing approximately 0.5°C of global cooling over one to two years—the clearest natural demonstration of the SRM mechanism.
- Harvard analysis suggests stratospheric aerosol injection sufficient for approximately 1°C of cooling would cost on the order of $2–8 billion per year, placing it within reach of a single mid-sized state or private consortium.
- Enhanced weathering has a theoretical potential of roughly 2–4 Gt CO₂ per year using crushed silicate rock such as olivine or basalt; real-world effectiveness is constrained by measurement difficulty and the emissions of mining and transport.
- Frontier Climate, funded by Stripe, Alphabet, Shopify, Meta, and McKinsey, has committed more than $1 billion in advance market commitments to purchase durable carbon removal.
- The voluntary carbon market was worth roughly $2 billion in 2023; persistent quality and permanence concerns led to the formation of the Integrity Council for Voluntary Carbon Markets to develop standards.
- Carbon Engineering's partnership with Occidental Petroleum targets DAC hubs in Texas at approximately 500,000 tons per year—the largest announced project, and one whose oil-industry involvement raises questions about end use.