Google splits its largest carbon buy across two delivery clocks

Google's largest carbon removal purchase to date, a Terradot agreement spanning more than 200,000 hectares of rice farms in southern Brazil, carries its structure in its delivery schedule: one million metric tons of CO2e of methane elimination, measured on a 20-year global warming potential, is due by 2030, and a second million tons of permanent carbon removal is due by 2040.

A single corporate treasury is underwriting two technologies on one landscape while booking a decade-long obligation for a removal process expected to scale only as the project develops, and the companies say the project is the first publicly announced at this scale to pair superpollutant elimination with durable carbon removal inside the same agricultural system. The claim is theirs, but the two horizons are what make the transaction hold together, with a million tons on each and delivery dates a decade apart.

Terradot will run Alternate Wetting and Drying across the paddies, an irrigation technique that periodically drains flooded fields rather than keeping them continuously submerged, and the dry spells remove the conditions that produce methane while cutting water consumption. Onto those same fields the company will spread naturally sourced crushed volcanic rock, which dissolves through enhanced rock weathering and draws carbon dioxide out of the atmosphere.

The sequencing is deliberate because methane clears the atmosphere far faster than carbon dioxide, so eliminating it buys temperature relief early, while CO2's centuries-to-millennia residence is what justifies a durable tranche on a 2040 clock; as the methane benefit tapers, the removal is expected to carry more of the load. Southern Brazil suits the model because warm, wet conditions accelerate weathering and raise the removal the rock can be credited with, and Google says the combination aims at "a net decrease in warming on every time horizon"—a portfolio argument applied to one project instead of a portfolio of them.

AWD can lower water use and input costs, and the crushed rock may improve soil health, so participants are paid in agronomic savings alongside carbon revenue, which should make enrollment stickier than a single-season rental—the difference between a program that can deliver against a 2030 clock and one that cannot. The capital here is a direct offtake priced against delivery milestones, not a green bond or a fund subscription; transition finance is being repriced by eligibility and control rather than by label, and a contract that writes its own measurement convention and its own delivery dates is a clean example of the trend.

The measurement convention deserves a closer look because the methane tranche is denominated in 20-year global warming potential, a window that loads more of the gas's near-term temperature effect into each credit. That choice is a pricing decision as much as a scientific one, and it is likely what puts the two tranches on the same headline footing—a million tons apiece—while they buy very different things on very different timelines. One tranche is a measurable reduction in a gas that clears quickly; the other is a long-dated option on a weathering curve.

As this publication has argued, the priced asset in corporate climate capital is increasingly certification rather than carbon, and this contract pushes the argument a step further. Google is buying two things a registry does not supply—a measurement window and a delivery date—and because the buyer set both, the corporate treasury is the early-stage underwriter of the method as much as of the field. The terms a buyer of Google's scale accepts tend to become the terms the next buyer is handed.

The two-product structure we flagged when the contract surfaced is the exportable part, and the hectare, not the tonnage, is the unit that matters. Tonnage is derived; hectares are enrolled, farmed, and measured. Terradot's footprint is both the production unit and the sales unit, which is why the durable tranche will track land long before it tracks credits. Google's rice-methane program now has two counterparties—the Mitti Labs offtake our September coverage described and this one—and each contract hardens the template for the next.

The second clock is the one to watch. The 2030 tranche rests on AWD, a technique with a known mechanism and a constraint measured in hectares, so the volume delivered is a function of how much acreage gets enrolled and verified. The 2040 tranche rests on how quickly crushed volcanic rock gives up its carbon in a warm, wet field, and if the weathering rate comes in under plan, the back half is where the contract is likely repriced. By then every developer selling into the same basin will be quoting Google's schedule back at the next treasury that calls.