Humanity emits roughly 37 billion tonnes of carbon dioxide a year. By 2025 the Earth had warmed about 1.4 °C above pre-industrial levels, and current emissions pathways point toward roughly 2 °C by 2050. Every technology in this programme is a serious response to that. Every one of them, measured honestly against 37 gigatonnes, is at present a rounding error.
That is the organising fact, and it is why these six are grouped rather than treated as separate frontier topics. Read individually, each looks like a promising technology with an engineering roadmap. Read together and measured against the same denominator, a different picture emerges: the binding constraint across the whole field is not invention but rate of scaling, and the interventions get faster to deploy precisely as they get harder to justify. This programme is a peer to the AIHS feasibility study (S-001) and the engineered origins programme (P-004), and it opens Categories V and VI of the Frontier Research map.
Programme Index
- Generating power without carbon — commercial fusion
- Making things without carbon — zero-carbon industrial systems
- Turning sunlight back into fuel — artificial photosynthesis
- Taking carbon back out — carbon capture at scale
- Intervening in the climate directly — climate engineering
- Deciding who may intervene — geoengineering governance
A note on the flags
Every claim carries one of four honesty flags — the Institute's standard system. On this subject the flags do unusually heavy work, because the gap between demonstrated capability and deployed capability is wider here than almost anywhere else on the map:
Established Well-understood science. Frontier Active research, partly known. Speculative Theoretical, not yet demonstrated. Handwave Evocative, but without real basis.
The denominator
Established One number governs this entire programme: annual global carbon dioxide emissions of roughly 37 gigatonnes. Any technology proposed as a solution has to be measured against it, and the comparison is rarely made in public because it is unflattering to everyone.
Established Some worked examples from the parts that follow. The world's largest operating direct air capture plant has a design capacity around 36,000 tonnes a year — approximately one millionth of annual emissions. Europe's hydrogen-based steel capacity stood at roughly 2.5 million tonnes in 2025, against global steel production near 1.9 billion tonnes. The largest outdoor artificial photosynthesis demonstration covered about 103 square metres. No fusion plant has delivered electricity to any grid.
Frontier None of that means the technologies are worthless; several are progressing genuinely fast, and early-stage capacity figures are supposed to be small. It means that the interesting question about each is not does it work but what is its doubling time, and does that reach material scale before the carbon budget is spent. That question is asked in each part, and the answers differ sharply.
Part 1 · Generating power without carbon
Established Fusion is no longer a physics question — ignition at the National Ignition Facility in December 2022, since repeated, settled that net-energy fusion is physically achievable. Frontier It is now an engineering and economics question, with substantial private capital behind it: seventeen startups have each raised over $100 million, and the best-funded is building a compact high-field tokamak targeting burning plasma around late 2026 or early 2027. Established No commercial fusion plant has produced electricity for a grid, and no company is close to doing so; commercial output is projected for the early 2030s at the earliest. Frontier Which places fusion, on its own developers' timelines, after the decarbonisation deadline that is usually invoked to justify it.
Part 2 · Making things without carbon
Established Steel and cement are roughly a tenth of global emissions between them, and their emissions are chemical rather than merely energetic — you cannot electrify your way out of the carbon in the reaction. Frontier Hydrogen direct reduction replaces the carbon reductant in steelmaking and can save close to two tonnes of CO2 per tonne of steel; Scandinavian projects have shown it works. Established Capacity remains a fraction of a percent of global production, and commercial viability depends on policy support. Speculative Molten oxide electrolysis, which bypasses carbon entirely, is not expected to be commercially ready before the 2040s.
Part 3 · Turning sunlight back into fuel
Frontier Artificial photosynthesis uses sunlight to split water or reduce CO2 into fuels, and laboratory devices now exceed natural photosynthesis comfortably — a module-sized artificial leaf reported above 10% solar-to-hydrogen efficiency in 2025. Established Efficiency collapses with area: one standalone artificial leaf measured 8.4% at small scale and 2.7% when scaled to 441 square centimetres. Frontier Reducing CO2 is far harder than splitting water, with solar-to-fuel efficiencies for some monolithic devices below a tenth of a percent. Speculative Nothing in this field is close to commercial deployment.
Part 4 · Taking carbon back out
Established Direct air capture works and is expensive. Operating costs sit far above the $100 per tonne threshold usually cited as the viability target — one leading operator's co-chief executive put its own figure closer to $1,000 than $100 — and an independent analysis projects $230–540 per tonne even by 2050. Frontier Capacity is scaling: a 500,000-tonne-per-year facility in West Texas would, at design capacity, increase global DAC capacity by several hundred per cent from a single plant. Established That plant alone would still be roughly one seventy-thousandth of annual emissions. Frontier The sector's largest operator laid off staff in 2025 having raised over $800 million.
Part 5 · Intervening in the climate directly
Frontier Solar radiation modification — reflecting a fraction of incoming sunlight — is the only intervention here that could act on global temperature within years rather than decades, and it is cheap enough that a single state could attempt it. Established It addresses temperature, not carbon: ocean acidification continues, and stopping abruptly produces rapid warming, the well-characterised termination shock. Frontier Modelling suggests it might reduce climate risk globally while increasing it regionally, which is a distributional problem disguised as a technical one. Handwave It is not a solution to the carbon problem and nobody serious claims it is.
Part 6 · Deciding who may intervene
Established There is no binding international framework governing solar geoengineering research, experiments or deployment. What exists is a patchwork: a Convention on Biological Diversity moratorium from 2010, reaffirmed in 2024; statements under the London Convention; and national positions. Frontier Meanwhile the UK committed £56.8 million in 2025 to a programme including five outdoor experiments, a commercial venture raised $75 million toward deployment readiness, and at least one company has already released material into the atmosphere. Frontier Serious people disagree about whether a moratorium would help or entrench the problem, and the programme presents both cases.
How the six connect
The arc runs generate → substitute → recycle → remove → intervene → govern, and it has a property worth stating plainly: it is ordered by ascending desperation and descending consensus. Building clean generation is uncontroversial and slow. Decarbonising industry is uncontroversial and slower. Making fuel from sunlight is a laboratory science. Removing carbon already emitted is real, expensive, and tiny. Intervening in the climate directly is fast, cheap, and contested to the point that its governance is a separate discipline. Each step down the list is a response to the inadequacy of the ones above it — which is precisely why the last two exist, and precisely why they should be read in the context of the first four rather than on their own.