The soy pact that wasn't: how a nine-year moratorium fell apart and what 1.4 million hectares of forest look like

On 16 May 2026, the Amazon Soy Moratorium quietly died. The nine-year-old agreement between Brazilian grain traders, environmental NGOs and the federal government was supposed to expire on that date. What was not supposed to happen was the silence that followed. By the second week of July, no successor pact had been signed. The land market had already priced the gap in.

Modelling released on 17 July 2026 by a team at the University of São Paulo, working with researchers at the Potsdam Institute for Climate Impact Research, projects at least 1.4 million hectares (3.5 million acres) of additional deforestation in the Brazilian Amazon by 2036 as a direct consequence. That figure is conservative: the central scenario in the published paper is 2.4 million hectares by the end of the next decade, with an upper bound above 5 million. The carbon cost is roughly 400 million tonnes of CO₂ that the forest will not absorb over the same period, equivalent to the annual emissions of Canada.

This is not a story about a single document. It is a story about how a market-shaping agreement got outflanked by the politics it was never designed to survive, and what fills the vacuum when the only thing holding the line was a handshake between competitors.

The deal that held, until it didn't

The Amazon Soy Moratorium was signed in July 2006, after a Greenpeace investigation traced soy from newly cleared Amazon land into the supply chains of Cargill, Bunge, ADM and other major traders. The original signatories agreed not to purchase soy grown on land deforested in the Amazon biome after that date. The mechanism was unglamorous: satellite monitoring by Brazil's National Institute for Space Research (INPE), with traders committing to reject purchases from flagged properties. The arrangement was renewed in 2008, 2010, 2014, 2016, 2018, 2020 and 2024. Each renewal was a fight.

By the late 2010s, the soy frontier had largely shifted out of the Amazon and into the Cerrado, a savanna biome to the south and east that contains far more biodiversity than its open-grass appearance suggests. The agreement was no longer the brake on deforestation its authors hoped. But it remained a credible threat: a single round of European pressure could re-extend the rules, and Brazilian soy exporters had built compliance systems around the framework.

The 2024 renewal was the shortest on record. It covered a single crop year. When it lapsed in May 2025, the signatories issued a one-year bridge that expired, as scheduled, on 16 May 2026. This time, there was no bridge.

What the modellers actually said

The 17 July paper, led by ecologist-turned-modeller Raissa Guerra at USP, ran three deforestation scenarios against a counterfactual in which the moratorium remained in force. The 1.4 million hectare figure is the difference between the baseline path and the no-pact path under what the authors call the "moderate compliance" scenario, in which roughly half the post-2024 forest conversion that the moratorium would have blocked still occurs. Under "weak compliance," the gap rises to 2.4 million hectares. Under a worst-case "land rush" scenario in which the absence of a pact triggers speculative clearing ahead of any future regulation, the gap exceeds 5 million hectares by 2036.

The carbon figure, 400 million tonnes, is drawn from the moderate scenario. It uses IPCC Tier 1 biomass densities for the affected forest types and standard above-ground carbon-to-CO₂ conversion factors. The comparison, "equivalent to Canada's annual emissions," uses Canada's most recent inventory submission, roughly 540 million tonnes of CO₂ equivalent in 2024.

Guerra's team is careful about what they are not claiming. They do not attribute specific clearings to specific actors. They do not model indirect effects on cattle ranching, which is the dominant driver of Amazon deforestation. They note, almost in passing, that soy-driven clearing has historically been a smaller share of total Amazon loss than cattle, and that the moratorium's main effect was to keep soy's footprint manageable while the cattle frontier advanced.

That last point matters. The modellers are not saying 1.4 million hectares will be cleared by soy farmers. They are saying the absence of the pact removes one of the few mechanisms that constrained soy's expansion into standing forest, at a moment when global demand for Brazilian soy is at a record high.

The counter-narrative from Mato Grosso

The Brazilian agriculture lobby rejects the framing. Aبيان issued by the Brazilian Association of Vegetable Oil Industries (Abiove), which represented the traders' side of the moratorium, argues that the agreement had become redundant: by 2024, more than 99 percent of soy sourced from the Amazon biome was already compliant with the pact's deforestation cutoff. Mato Grosso, the state that produces roughly 17 percent of the world's soy, has its own state-level monitoring and a 2022 commitment to end illegal deforestation by 2030. The Brazilian сельскохозяйственный frontier, the argument runs, no longer needs a private-sector pact because state regulation and satellite enforcement now do the work.

There is a grain of truth there. INPE's PRODES and DETER systems did outlast the private monitoring regime that the moratorium institutionalised. Mato Grosso's CAR system does track rural properties in detail. The agronomic reality is also straightforward: most of the easy Amazon frontier was cleared in the 2000s, and the remaining forest sits on land that is harder to convert, less accessible, and often already claimed by smallholders or protected reserves.

The counter-argument, which Guerra's modelling takes seriously, is that the absence of a binding commitment lowers the political cost of clearing. Without the pact, there is no private-sector backstop when federal enforcement wobbles. And federal enforcement has wobbled. Deforestation alerts in the Amazon rose by roughly 12 percent year-on-year in the twelve months ending June 2026, according to INPE data summarised in the USP paper's appendix. The signal is weak and the data is contested, but the direction is consistent with the modelled scenarios.

What actually fills the gap

The more revealing question is what fills the regulatory vacuum. The European Union's Deforestation Regulation (EUDR), adopted in 2023 and phased in from late 2024, requires importers to prove that certain commodities, including soy, were not produced on land deforested after a 2020 cutoff. On paper, the EUDR is stricter than anything the moratorium ever required. In practice, its implementation has been pushed back twice, with the European Commission citing compliance-readiness concerns from member states and from trading partners including Brazil.

The current EUDR timeline places full enforcement at the end of 2026, with a risk-tiering system that delays checks on lower-risk origins. For Brazilian soy, the practical effect depends on whether Brazil is classified as low-, standard- or high-risk, and on whether the EU accepts the country's own certification systems as adequate. Brazilian officials, including the Minister of Agriculture, have argued that the country's state-level monitoring is sufficient.

The gap, in other words, is not being filled by nothing. It is being filled by a slower, more contested regulatory transition, in which the European market retains leverage but lacks the clean enforcement mechanism the moratorium provided. Brazilian traders still face reputational pressure from European buyers, and several have publicly said they will maintain moratorium-style screening on a voluntary basis. Whether that voluntary regime survives a weak soy price year, or a change of government in Brasília, is an open question.

The structural read

The moratorium was a textbook example of what economists call a private governance regime: an agreement between competitors, brokered by NGOs, that filled a regulatory gap by changing market access. It worked because the signers controlled the choke points, the grain silos in Santarém and the export terminals on the Amazon tributaries, and because European buyers cared enough to police compliance.

It is also a textbook example of the limits of that model. Private governance is brittle when the underlying political settlement shifts. Since 2023, the federal government in Brasília has been openly hostile to environmental restrictions it views as foreign-imposed, and the agribusiness caucus in Congress has grown more assertive. The 2024 renewal of the pact was a defensive action by traders trying to preserve market access, not an ambitious policy push.

What replaces it will not be a single document. It will be a patchwork: the EUDR, state-level CAR systems, voluntary trader commitments, satellite monitoring, and whatever the next Brazilian administration decides to enforce. The modellers' central estimate, 1.4 million hectares by 2036, is the cost of the patchwork being looser than the deal it replaced.

What to watch next

Three dates will determine whether the central scenario holds. First, the EU's formal risk classification of Brazil under the EUDR, expected by the end of 2026, will set the de facto enforcement bar for Brazilian soy in Europe. Second, Brazil's own national deforestation plan, due for revision in early 2027 under the terms of the country's updated NDC, will signal whether the federal government intends to backfill the moratorium with domestic regulation. Third, the 2026-27 Amazon dry season, which runs roughly July to October, will produce the first INPE alert numbers for a post-moratorium year. If the alerts run materially above the 2024-25 baseline, the moderate scenario starts to look optimistic.

The 1.4 million hectare figure is not a prediction. It is a model output, contingent on assumptions the authors make explicit. What it captures is the price of a nine-year-old handshake expiring without a successor, at a moment when the politics that produced it have moved on. The forest, as ever, absorbs the bill.

Desk note: this publication treats the soy-pact collapse as a market-governance story first and an environmental story second. The framing follows the modelling team's own: the variable in question is the absence of a private enforcement mechanism, not the underlying appetite for clearing. Sources are limited to the published paper, the Abiove statement, and EU regulatory filings; the analysis deliberately does not extrapolate beyond what those documents support.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://www.consilium.europa.eu/en/press/press-releases/2023/05/16/eu-deforestation-regulation-council-adopts-new-law-to-ban-products-linked-to-deforestation/
• https://en.wikipedia.org/wiki/Amazon_Soy_Moratorium