Peru's merchant tail is priced for a market that already ended

Peru's weighted-average marginal cost ran at $29.5 per MWh in the twelve months to March 2025, a slight decline from the year before. Generators selling into the free market that same month were getting $67.7 per MWh for contracted volume, more than double, per COES data reported by Gestión. The same reporting carries COES's own base case: absent new generation or transmission, marginal cost reaches $150 to $200 per MWh by 2033. Most of the debt I've seen sized against uncontracted volume in Peru this year still prices the tail off the first number, not the third.

For more than a decade Peru ran what the market itself calls "sobreoferta eficiente," an efficient oversupply that kept spot prices low and made merchant exposure look like a rounding error. Gestión and Infobae both reported that era ending around mid-2025. The spot price did not spike the week it ended. What changed is the assumption underneath every merchant-tail model written since: that the next ten years look like the last ten.

Two prices, same market, same month

$29.5 and $67.7 are not two measures of the same thing. The first is the system's marginal cost, what the last unit dispatched costs to run, averaged across a quarter. The second is what free clients actually paid under bilateral contracts signed earlier, at whatever price cleared then. The gap between them is the premium the market has already put on certainty. Whoever sold uncontracted volume into the spot market that quarter settled near the first number. Their contracted neighbor collected the second.

That gap is also, awkwardly, the one lenders reach for when they discount a merchant tail hard. A debt sizing exercise that treats $29.5 as the realistic floor is treating the premium as something the market might withdraw at any time. Not as something it might widen.

Why the floor is the wrong floor

Project debt against a merchant tail typically gets sized to a conservative flat real price, close to a trailing average of spot, with modest escalation if any. That is sound practice in a market with no structural reason to move. It is a different exercise in a market whose own operator is on record saying the structural reason already exists, and that the fix for it is behind schedule.

I have made that case to a credit committee and lost. The counterargument deserves to win more often than it does: COES's $150 to $200 figure assumes nothing gets built, and a regulator forecasting the cost of inaction is not a neutral witness. It has every incentive to make the status quo look expensive enough to justify the next line. A bank sizing twenty years of debt should not take that number at face value, and it should not wave it off because the source has a motive either.

A merchant tail discounted to yesterday's spot price is not a conservative assumption. It is a bet that the operator's own forecast is wrong, made by people who will not be in the room when it either is or isn't.

What we think follows

Treat the merchant tail as two separate claims rather than one flat number. The near years, where today's $29.5-to-$67.7 spread is the best evidence available, earn the conservative discount lenders already apply. The out-years, where the gating constraint is the same transmission queue we have written about before, roughly 20,000 MW of concession-ready generation waiting behind a grid clearing barely 1,000 MW more, deserve a structure built to capture upside if that queue slips rather than one that flattens it to zero by assumption.

That points toward merchant tails priced in steps instead of a flat line, and toward deliberately retaining some uncontracted volume in later contract years rather than signing away every megawatt-hour at close for a marginally better leverage ratio today. Pushed too far the other way, this is merchant speculation wearing a project-finance structure, and that has ended badly for people running better models than mine.

Our own pipeline carries this exposure by date whether we choose to or not. Olmos Wind, roughly 135 MW with integrated battery storage in Lambayeque, and Frontera at a similar scale both target Ready-to-Build in 2028. Aurelion and Solar II, 300 MW each of bifacial solar in Moquegua, target 2028 and 2029. Each will spend real years of operating life inside whatever Peru's spot market has become once the current transmission queue clears, or doesn't, regardless of what any single offtake contract says on day one.

The open question

COES's forecast is explicitly conditional on nothing new getting built. Peru's own transmission queue, by COES's own numbers, says that condition is closer to the base case than the stress case through at least the early 2030s. If that holds, the $29.5 floor most debt is sized against today understates what those same megawatt-hours throw off in year twelve of a twenty-year asset. If it doesn't, if the promised lines clear on schedule, the forecast was a planning tool dressed up as a prediction, and the conservative discount was right all along. I don't know which one wins. I know which one the term sheets in front of me assume, by default, without anyone deciding it on purpose.

See the full active portfolio or read more Insights.