The cushion is back — and industry is no longer under it

In 2025 Brazil sold the world US$348.7 billion — the largest export figure in its history, US$9 billion above the previous record. Three products hit record volumes in the same year: 416 million tonnes of iron ore, 108 million tonnes of soybeans and 98 million tonnes of oil.

In the same country, in the same year, manufacturing accounted for less than 11% of GDP — against 13.6% a decade earlier.

The two facts coexist without ever touching, and that is precisely what needs explaining. Because there was half a century in which they were the same thing.

What an LSE thesis had measured

The article that gave rise to this investigation laid out the argument of Nicolas Grinberg, of the London School of Economics: Brazilian industry was profitable for half a century without ever being competitive, because part of its profit did not come from the factory. It came from the soil.

The mechanism had concrete parts: an overvalued exchange rate, import tariffs and subsidized credit. Together, they transferred to industry a slice of the wealth generated by farmland and mines and paid for by the rest of the world. By Grinberg's calculation, between 1955 and 1980 roughly a third of the profits of Brazilian industry came from that surplus — from people who owned no land at all. In the following period, that slice fell to 6.5%.

The thesis explains the decline by the shrinking of the cushion itself. And this is where its data stops: the mid-2000s.

The cushion didn't shrink. It got fatter

If the explanation were only the size of the surplus, the next twenty years should have handed the slice back to industry. That is not what the trade figures show.

The wealth Grinberg describes as the source of the subsidy is not smaller. It is at the largest scale ever recorded. And manufacturing, which was the destination of that subsidy, is at its lowest share of GDP in decades: 10.7% in 2023 and 10.8% in 2024, according to IBGE's National Accounts, against 13.6% in 2013.

Nor is the sector collapsing — it grew 3.8% in 2024, after contracting 0.5% in 2022 and 1.3% in 2023, and closed 2025 essentially flat. What it is not doing is keeping up with the size of the economy. Much less with the size of what the country sells abroad.

The hypothesis: the plumbing, not the source

The reading this investigation proposes — and it is Sip Dölyn's, not the thesis's — is that the right question is not about the size of the surplus. It is about the plumbing.

The mechanism Grinberg describes was not automatic. It depended on three specific institutions that made the wealth of the soil pass through the factory before turning into profit: a managed exchange rate that made imported machinery cheaper and exported commodities dearer; a tariff wall that guaranteed the domestic market; and directed credit at below-market interest rates.

• The managed exchange rate gave way to a floating rate in 1999.

• The tariff wall came down with the trade opening of the early 1990s.

• Directed credit lost relative scale and came to coexist with a high policy rate.

None of these changes was a miscalculation: all of them were deliberate decisions, with reasons of their own, and several of them solved real problems. But it is the combined effect, on this specific mechanism, that matters here. The surplus from the soil is still being generated and is still being sold. It simply no longer passes through industry on the way.

If this reading is right, it turns part of the debate around. The Brazilian discussion about deindustrialization tends to revolve around Dutch disease — the idea that a commodity boom pushes up the currency and kills industry through the exchange rate. What Grinberg's mechanism suggests is something different and more uncomfortable: the commodity boom once sustained Brazilian industry, and did so for a long time. What changed was not the sign of the boom. It was the dismantling of the piping that connected one to the other.

What this article cannot claim

A caveat worth more than a tidy conclusion: this investigation has not redone Grinberg's calculation for 2011-2025. Redoing it would require his methodology, including the choice of when the Brazilian exchange rate would be "undistorted" — an assumption the author himself discusses, and about which he warns that he probably underestimated the magnitude of what he measured.

What is here is something else, and more modest: two official series placed side by side, and the question they raise when no one keeps them apart. The share of the primary surplus in Brazilian industrial profit today is a number nobody has published. It ought to exist.

Until it does, the portrait stands: a country that has never sold so much of what it takes out of the ground, and an industry that has never weighed so little. Half a century ago, those two sentences described the same piece of machinery. Today they describe two countries sharing the same map.