# High-Energy Repair

_A hundred-year program to clean up the air, water, and soil through engineering._

We can return atmospheric CO₂ to pre-industrial levels within a hundred years. We can clean forever chemicals out of the water supply. We can rebuild the topsoil we've degraded. This is no longer research. It is an engineering program with known methods, costs, and dates.

The research phase proved the cleanup is physically possible. It also proved it requires an exorbitant amount of energy.

Even if we reach true net zero tomorrow, roughly 2,000 gigatons of legacy CO₂ will keep warming the planet for millennia. Net zero stops the bathtub from filling. It does not drain it. Draining it requires engineered carbon removal at 30 gigatons per year, sustained for more than half a century.

Of the removal methods, direct air capture is the one that scales without limit and falls along a manufacturing cost curve. DAC runs $400 to $1,000 per ton today. At two-cent electricity it drops to $80 to $200 a ton, within striking distance of the $100 mark where the removal math makes sense.

At that price, the whole program pencils out at about $3T per year, globally. That's a lot of money, but it's less than half the fuel rent the world pays the fossil fuel industry every year now.

The cleanup is bottlenecked on one variable: the price of clean electricity.

## The Accidental Inflection

The energy is arriving for reasons that have nothing to do with climate.

AI is the demand shock. Microsoft, Google, Amazon, and Meta are pouring hundreds of billions into advanced nuclear, enhanced geothermal, and grid buildout. Compute supremacy demands firm, around-the-clock power. The grid cannot deliver. They are not decarbonizing. They are racing each other. Their offtake contracts absorb the upfront capital costs of every credible clean energy technology.

Enhanced geothermal is the supply shock no one saw coming. The oil and gas industry spent two decades perfecting horizontal drilling and fiber-optic sensing for shale gas. That toolkit is exactly what enhanced geothermal needs to scale anywhere on Earth. Drill deep enough, hit hot rock, run a closed loop. Fervo Energy is building a 400-megawatt plant in Utah with techniques the shale industry built for the Permian.

Solar is the supply shock that already happened. China decided cheap energy was a strategic weapon and built the world's first electrostate. In 2024 alone, China installed more solar than the United States has built in its entire history. Utility-scale solar is now the cheapest electricity ever produced.

All three forces ride Wright's law: every cumulative doubling of production drives manufactured energy costs down a predictable percentage. The curve does not plateau at parity with fossil fuels. It keeps falling. Extractive energy moves the other way. Easy reserves are gone. Every new barrel costs more than the last. The curves crossed in 2024. They will not cross back. The rest of the manufactured stack — advanced fission, fusion, enhanced geothermal — is on the same path. Hydro, legacy gigawatt-class fission, and bioenergy ride alongside as the incumbent clean-firm fleet, maintaining rather than expanding.

Generation is mostly solved: 85.6% of new generating capacity worldwide in 2025 was renewables (IRENA, _Renewable Capacity Statistics 2026_). Delivery is the bottleneck. About 2,290 gigawatts of generation and storage — nearly twice the entire existing US grid — sit waiting in interconnection queues that run five to seven years (LBNL, _Queued Up: 2025 Edition_).

> Every form of manufactured energy follows the green curve. Every form of extractive energy follows the gray. Only the slope changes. ([Cost curves chart](/embed/cost-curves).)

Below two cents per kilowatt-hour, the economic logic of most industrial processes inverts. Synthesizing hydrocarbons becomes cheaper than drilling for them. Pulling CO₂ from the air becomes cheaper than letting it accumulate. Desalinating seawater becomes cheaper than rationing fresh water.

## The Lever

Policy can compound the inflection. The data center buildout already requires transmission upgrades, expedited permitting, and grid investment that ratepayers alone cannot fund. Couple those public investments to a carbon removal obligation on large new loads. One buildout, two outcomes: cheaper electricity for ordinary Americans, and demand pull that drives DAC down its cost curve. This is the place where the program needs consensus, because the atmosphere is the one thing no single buyer owns.

## Four Pillars

Four engineering pillars, sequenced across a century.

### P1 · Energy Production

Solar is the cheapest electricity ever produced and still falling. Advanced fission, enhanced geothermal, and long-duration storage are scaling on hyperscaler balance sheets. Below two cents, every downstream pillar inverts.

[Read more about Energy Production →](/sectors/energy-production)

### P2 · Net Zero Emissions

Roughly two-thirds of emissions come from combustion. That share electrifies as electricity gets cheap. The remaining one-third needs new chemistry and biology — cement, fertilizer, livestock, land use. Each has a closing cost gap as the electrification cascade compounds.

Princeton's Net-Zero America study (Larson, Jenkins, Greig _et al._, 2021) maps five pathways to a US net-zero by 2050. The pathways disagree on the end state, but every one front-loads the same no-regrets moves: build clean generation fast, expand the grid, and electrify what people drive and what heats their buildings. Variable wind and solar can't run a modern economy alone — they need firm, dispatchable complements, what Jenkins and colleagues call clean firm power (Sepulveda et al., _Joule_ 2018).

For the hard half (steel, cement, shipping, aviation), green hydrogen unlocks near $1/kg — the DOE Hydrogen Shot's 2031 target — when clean electrons fall below about $20/MWh (≈2¢/kWh).

[Read more about Net Zero →](/sectors/net-zero)

### P3 · Carbon Removal

Net zero stops the bathtub from filling. Draining it requires roughly 30 gigatons per year of engineered removal by the 2060s, sustained for half a century. The three methods whose costs collapse as electricity gets cheap — direct air capture, ocean alkalinity enhancement, and mineralization — alone carry ~38 gigatons per year of theoretical scale potential. Biochar is the leader by delivered tonnage today (~100 kilotons per quarter in 2025); enhanced rock weathering and biomass burial are additive on top.

On the demand side, the market is forming. Frontier (Stripe, Alphabet, Shopify, Meta, McKinsey) has now committed $1.8 billion in advance market commitments, nearly double its earlier pledge: a $915 million tranche announced June 17, 2026 brought in Anthropic as the first AI company to join the buyers' group. That is committed demand, the revealed willingness to pay that pulls supply into being, not removal delivered. Microsoft has signed multi-megaton offtakes across DAC, mineralization, biochar, and forestry, anchoring early commercial-scale durable-CDR contracts. The 45Q tax credit pays up to $180 per ton for engineered removal.

[Read more about Carbon Removal →](/sectors/carbon-removal)

### P4 · Infrastructure Resilience

The warming through 2050 is already locked in. Water systems for PFAS and drought. Cold chain for food security as ambient temperatures climb. Grid resilience for a higher-variance climate. Insurance markets are repricing climate risk faster than policy is — Swiss Re reports a global insurance protection gap of $181 billion in 2024, already past where the same publisher had projected we'd be by 2030. Where insurance withdraws, capital follows toward hardening.

[Read more about Infrastructure →](/sectors/infrastructure)

---

The Hoover Dam, the Interstate, the Apollo Program, rural electrification. We did those on smaller economies and worse tools. This is doable. It has always been doable. What changed is the price of electricity.