Brazil's curtailment number hides where the real risk sits

Brazil curtailed roughly a fifth of its solar and wind output in 2025, an estimated BRL 6.5 billion, about $1.2 billion, in lost revenue, per Volt Robotics data reported by pv magazine. That is a national average. In the Northeast, where most of the country's wind and solar is actually sited, curtailment ran closer to 27% for solar and 16% for wind in 2025, concentrated in Bahia and Ceará, per Rystad Energy. The number showing up in investor decks is the first one. The number that should be showing up in debt sizing is the second.

Averages do that. They are supposed to. The problem is what gets averaged away: the Northeast is both Brazil's cheapest wind and solar resource and its tightest transmission corridor to the load centers in the Southeast. Blending that region into a national figure makes the risk look smaller than it is, in exactly the place where the capacity factors made the project pencil in the first place.

A number that blends away its own risk

Wood Mackenzie's own forecast does not expect that gap to close. It puts Northeast submarket curtailment at roughly an 11% decade-average rate through 2035, against about 2% for the Southeast and Centre-West over the same stretch. That is not a temporary dislocation while new lines get built. It is a structural split between the region with the resource and the region with the grid capacity to use it, and the model says it widens before it narrows.

A debt model that takes a national curtailment haircut and applies it to a Northeast asset is importing the performance of a region the asset does not operate in. It is the same category of error as reading a change in output as a level, just committed one layer up, at the geography rather than the number.

The mechanism: a volume obligation meets a congested feeder

Most bilateral contracts signed in Brazil's free market commit the seller to deliver a fixed annual volume of energy, not a "take whatever the plant produces" arrangement. Brazil settles across four submarkets at different spot prices once the interconnection limit between them binds, which is precisely the condition that triggers curtailment orders in the Northeast in the first place.

A curtailment order and a resource shortfall cost the generator the same megawatt-hour. Only one of them was ever in anybody's P50 case.

When ONS instructs a Northeast plant to back down for grid reasons, the contract's volume obligation does not shrink with it. The seller still owes the buyer the committed energy and has to make up the difference by buying in the spot market it is actually exposed to, at whatever price that market clears when the curtailment hits. That cost sits outside the generation forecast entirely. It lives in the contract, not the resource assessment, which is exactly where most financing models stop looking.

What we think follows

Northeast wind and solar being the cheapest power in Brazil does not make it the cheapest risk. Lenders and offtakers underwriting a Northeast asset should price curtailment as a submarket-specific input with its own trajectory, not a single blended figure carried over from a national report. Contracts still being negotiated are the place to push for curtailment-sharing language tied specifically to grid-cause dispatch instructions, separated out from resource-cause shortfalls that the seller should reasonably keep bearing.

There is a fair counter to that. An offtaker with no visibility into Brazilian submarket mechanics has limited appetite to share a risk it cannot monitor, and a seller who pushes too hard on curtailment-sharing can simply lose the deal to a competitor willing to eat it. That tension is real, and it is why this sits in the term sheet rather than in a disclosure footnote.

We do not have a project in Brazil's Northeast, and the comparison stops at the mechanism. But the same pattern, the best resource and the worst grid risk landing in the same place, is part of why we treat storage as part of the asset rather than something added later. Olmos Wind, the most advanced project in our pipeline at approximately 135 MW in Lambayeque, carries battery storage in the design from the outset, with Ready-to-Build targeted for the first quarter of 2028.

The open question

EPE and ONS both have Northeast-Southeast transmission reinforcement in their planning pipeline. Whether that build-out arrives on a timeline that arrests the curtailment path Wood Mackenzie is projecting, or whether the region with Brazil's cheapest power simply keeps getting the least reliable delivery path regardless of how fast lines get permitted and built, is not something I have a confident answer to. It is the question that decides whether this is a financing problem for the next three years or the next decade.

See the full active portfolio or read more Insights.