The problem was known, documented, and monitored. Then it became an emergency.
Brazil's power grid faced a documented and foreseeable capacity shortfall, yet the auction designed to address it arrived late enough to transform a planned procurement into something resembling an emergency. What followed — fragmented products, elevated price ceilings, and a heavy concentration of thermal contracts — raises enduring questions about whether the solution was proportionate to the problem, or whether regulatory choices quietly shifted the burden of those decisions onto electricity consumers for decades to come.
- A capacity gap flagged as early as 2022 went unaddressed long enough that by 2026 the auction resembled a scramble rather than a strategy, shrinking the field of bidders and narrowing the range of technologies that could compete.
- One thermal product cleared with a discount of just 0.01 percent — a near-absence of competitive pressure in a market already dominated by a handful of incumbent players.
- Regulators raised the price ceiling hours before the auction, then offered contradictory justifications: that competition was strong enough to keep prices down, yet weak enough that the auction would fail without a higher ceiling.
- The auction contracted roughly 16.5 gigawatts of fossil fuel thermal capacity, far exceeding the 9.6 gigawatts the official energy plan had envisioned, while faster and more flexible alternatives like battery storage and demand response remained largely untested.
- Official demand forecasts projected 2026 peak load at 113 gigawatts; the actual figure came in near 101 gigawatts — a gap of more than 11 gigawatts that, compounded by planning margins, risks locking structural overcapacity into the grid for a generation.
Brazil's power system had a real problem. The National Electric System Operator had been documenting a structural capacity gap since 2022, and the country's official energy planning documents acknowledged it year after year. What the system lacked was not awareness — it was timely action.
By the time the auction finally took place in 2026, the delay had already reshaped the market. International experience with capacity procurement is consistent on this point: the longer you wait, the fewer bidders show up, the less technological diversity you get, and the weaker the competitive pressure on prices. What fills the void instead are incumbent players and thermal plants that can be built quickly. The predictable problem of 2022 had become the emergency purchase of 2026.
The auction was structured around eight separate products — a design that, in a market already concentrated among a small number of major players, thinned competition rather than deepened it. The results were telling. One thermal product cleared with a discount of just 0.01 percent. Hours before bidding opened, regulators raised the price ceiling substantially, then offered a justification that contradicted itself: competition was strong enough that the ceiling wouldn't matter, yet weak enough that the auction would fail without raising it. In oligopolistic markets, high ceilings don't just set limits — they coordinate expectations.
The technology mix that emerged raised further questions. Of the roughly 19.5 gigawatts contracted, about 16.5 gigawatts came from fossil fuel thermal plants, predominantly natural gas. The official 2035 Energy Development Plan had envisioned something different — a complementary blend of hydroelectric, renewables, storage, demand response, and flexible thermal generation, with thermal needs estimated at around 9.6 gigawatts through 2031. The auction more than doubled that figure.
Underlying everything was the demand forecast. The plan had projected 2026 peak demand at 113 gigawatts. The actual figure reported by the grid operator came in near 101 gigawatts — a gap exceeding 11 gigawatts. Capacity mechanisms are acutely sensitive to peak demand estimates: overestimate the peak, and you contract for too much power, and that excess sits in the system for decades, raising costs for everyone. The grid operator had also noted that the system's critical hours clustered between 7 p.m. and 11 p.m., yet the auction permitted thermal contracts with startup times of up to 12 or 18 hours — plants that cannot respond to the bottlenecks the system actually faces.
The structural need for more capacity was genuine. The question that lingers is whether the regulatory choices made along the way — the delay, the fragmentation, the ceiling, the thermal concentration, the demand projections — solved that problem efficiently, or quietly transferred its costs onto electricity consumers for a generation.
Brazil's power system needed more capacity. Everyone knew this. The National Electric System Operator had been flagging the problem since 2022. The Energy Research Company's planning documents acknowledged it year after year. The structural gap between what the grid could deliver and what it would need to deliver was documented, monitored, institutional fact.
Yet the auction to address this gap didn't happen until 2026. By then, what should have been a measured, competitive procurement had become something closer to an emergency purchase. The delay itself changed the market. International research on capacity adequacy is clear on this point: when you wait too long to contract for power, you lose bidders, you lose technological diversity, you lose the downward pressure that competition creates. What you gain instead are incumbent players and quick-to-build thermal plants. The problem that was predictable in 2022 became a scramble in 2026.
The auction itself was fragmented into eight separate products. In a market already dominated by a handful of major players, this kind of segmentation works against competition rather than for it. It thins out the field of rivals in each individual product. The results bore this out. One thermal product—UTE-2027—cleared with a discount of just 0.01 percent. That is not the signature of robust competition. It is the signature of a market where few players are bidding and little is being contested.
Then there was the question of the price ceiling. Hours before the auction, regulators raised it substantially. The logic offered in defense of this move contained an internal contradiction. Officials argued that competition was strong enough that the ceiling wouldn't actually influence final prices. Yet they also argued the ceiling had to be raised significantly or the auction would fail to attract bids. If competition were truly robust, those two statements cannot both be true. In oligopolistic markets—which this one is—high price ceilings stop being administrative parameters and become implicit pricing signals. They coordinate expectations. The precedent that mattered most was Belo Monte, an earlier auction where competitive concerns led to public scrutiny, oversight from the audit court, and caution in setting the ceiling. That was the opposite approach.
The technology mix that resulted from the auction raised its own questions. The auction contracted roughly 19.5 gigawatts of capacity. About 16.5 gigawatts came from fossil fuel thermal plants, mostly natural gas. The official planning document—the 2035 Energy Development Plan—had sketched a different picture. It showed power coming from a mix of hydroelectric, renewables, storage, demand response, and thermal generation. The idea was complementarity, not dependence on a single technology. The plan itself had identified a need for about 9.6 gigawatts of flexible thermal capacity through 2031. The auction contracted more than 16 gigawatts of thermal. A recent law shifted some of those obligations to small hydroelectric plants, but the thermal concentration remained stark.
Underlying all of this was a question about the demand forecast itself. The 2035 plan projected peak demand would reach 109.2 gigawatts in 2025 and 113 gigawatts in 2026. The actual numbers reported by the grid operator came in significantly lower: about 106 gigawatts in 2025 and only 101 gigawatts in 2026. The gap in 2026 exceeded 11 gigawatts. More important than a single year's miss was what the data suggested about the trajectory. The plan assumed demand would keep climbing after a surge in 2024. Instead, the peak load flattened and then fell. Capacity mechanisms are extremely sensitive to peak demand. Small overestimates compound into structural overcapacity because planners build in safety margins and probabilistic supply criteria. If you overestimate the peak by 11 gigawatts, you contract for too much power, and that excess sits in the system for decades, raising costs for everyone.
There was also a mismatch between what the grid operator said it needed and what the auction actually bought. Official diagnoses emphasized the system's growing need for flexibility—the ability to ramp up or down quickly, to respond dynamically to changing conditions. The critical hours, the grid operator noted, tended to cluster between 7 p.m. and 11 p.m. Yet the auction allowed contracts for thermal plants with startup times of up to 12 hours for natural gas and 18 hours for coal. Those plants cannot respond to the actual bottlenecks the system faces. The grid operator and the energy company both acknowledged that other tools existed to build flexibility: battery storage, demand response programs, operational improvements, even the return of daylight saving time. The energy company had explicitly warned that capacity auctions conducted before fully deploying existing flexible resources could create oversupply and unnecessary cost to consumers.
None of this is an argument that Brazil's power system didn't need more capacity. The structural need was real and documented. The argument is whether the regulatory design that was chosen actually solved the problem efficiently, competitively, and in proportion to what the system actually faces. A massive, long-term commitment to thermal generation based on demand forecasts that appear to have been overstated by more than 10 gigawatts creates risks that will ripple through electricity bills for decades. It locks in a technology choice when the grid operator itself was calling for flexibility and diversity. It does so in a market where competition appears thin and where the tools to build that flexibility through other means were never fully deployed.
Notable Quotes
In oligopolistic markets, high price ceilings stop being administrative parameters and become implicit pricing signals that coordinate expectations.— Analysis of auction design and market structure
The grid operator had explicitly warned that capacity auctions conducted before fully deploying existing flexible resources could create oversupply and unnecessary cost to consumers.— Energy Research Company (EPE) assessment
The Hearth Conversation Another angle on the story
The grid operator knew about this problem in 2022. Why did it take four years to hold an auction?
That's the core of the controversy. The need was documented and monitored. But the auction became an emergency measure instead of a planned procurement. When you delay like that, you lose competitive pressure. Bidders know you're desperate. Incumbent players have an advantage.
So the delay itself changed who could win and how much they could charge?
Exactly. International research is clear on this. Delayed capacity auctions reduce competition and favor technologies that are already built or can be built quickly. Thermal plants fit that description. Renewable alternatives need more development time.
The auction was split into eight different products. Why does that matter?
In a concentrated market with few major players, fragmentation works against you. It thins out the bidders in each product. You end up with situations where one product clears with almost no discount at all. That signals weak competition.
What about the demand forecast? How far off was it?
The plan projected 113 gigawatts of peak demand in 2026. The actual peak was 101 gigawatts. That's an 11-gigawatt miss. Capacity mechanisms are extremely sensitive to peak demand. Small overestimates become structural overcapacity that sits in the system for decades, raising costs.
Did the auction at least buy the kind of flexibility the grid operator said it needed?
No. The grid operator identified critical hours between 7 p.m. and 11 p.m. But the auction allowed thermal plants with startup times of 12 to 18 hours. Those plants can't respond to the actual bottlenecks. And the grid operator had already identified other tools—storage, demand response—that weren't fully deployed.
So the auction locked in a technology choice before exploring alternatives?
Yes. And it did so based on demand forecasts that appear overstated, in a market where competition looks thin, using a regulatory design that favors incumbents. The structural need for capacity was real. The question is whether this was the right way to solve it.