Peru's power market runs on gas with twelve years left on the clock

Peru's Ministry of Energy and Mines put the country's proven natural gas reserves at 6.52 TCF this month, 95% of it concentrated in the Camisea lots, 88, 56 and 57. At current consumption that is a 12 to 13 year horizon, down 43% from the 11.5 TCF the country counted in 2004. In March, a rupture on the single pipeline that carries that gas to the coast already forced rationing and pushed thermal plants onto liquid fuel, months before anyone needed a reserves table to see where this was headed.

Every renewable PPA, every merchant tail model, every debt sizing exercise written against Peru's SEIN assumes a marginal cost set mostly by cheap gas. That assumption has a published expiration date now. Most financing models still treat it as a constant.

What a marginal-cost system is actually pricing

COES dispatches Peru's grid on marginal cost, and gas-fired plants set that price across most hours of most years. COES's own medium-demand projection for 2026 puts roughly 28% of national generation on gas. That is the denominator every uncontracted megawatt in the country gets valued against, whether it is a merchant tail after a PPA expires or a spot sale during a curtailment event.

The number has been stable for two decades because Camisea was treated as effectively permanent. It is not. Reserves fell from 11.5 TCF to 6.52 TCF between 2004 and now, and a Gas Energy Latin America projection cited in Peruvian trade press puts potential electricity cost increases at roughly five times current levels by 2040 if exploration does not replace what is being drawn down. Perupetro's own, more optimistic figures put the horizon closer to 20 to 35 years under different reserve and export assumptions. The spread between those estimates is itself the risk: nobody underwriting a 20-year PPA in Lima right now can tell you with confidence which number applies to year 15 of their debt term.

A debt sizing model that discounts the merchant tail off today's gas-anchored marginal cost is discounting against a number with a published expiration date.

The two risks collapsed into one price

Reserve depletion is a slow-moving number. The March 2026 pipeline rupture was not. A single rupture on the roughly 730 km line that carries Camisea gas to the coast forced a rationing protocol that prioritized households and essential services, and pushed thermal plants to burn liquid fuel at a materially higher cost. COES stated at the time that the system held enough capacity to meet demand without rolling blackouts, which is a fair point and worth conceding directly: reliability held. Cost did not. The marginal cost that every uncontracted megawatt in the country gets priced against moved, for the duration of the disruption, to whatever the liquid-fuel substitute cost that week.

That is the mechanism that matters for a project financier, more than the 12-year reserves number itself. A single pipeline is a single point of failure, and every failure on it reprices the system's marginal cost upward for its duration. A merchant-exposed asset generating during that window captures the spike. A model that sizes debt off an average marginal cost across normal years never saw the spike coming and will not see the next one either, because nobody has built a dispatch scenario for recurring pipeline failure into a 20-year cash flow.

What we think follows

Rising marginal cost is not obviously bad for renewable developers holding merchant exposure. A higher gas-driven price floor is upside for an uncontracted solar or wind asset selling into spot. The counterargument is just as real: Osinergmin and COES have shown a willingness to socialize reliability costs, reserve mechanisms and fuel subsidies across the generator base rather than let them show up purely as a price signal, and a renewable asset with no gas supply exposure could end up funding a stability mechanism for a fuel it does not burn. Which of those two effects dominates is not a settled question, and it should be modeled as a range rather than assumed away in either direction.

Our own pipeline sits inside this exact system. Olmos Wind, Frontera, Aurelion and Solar II are all Peru-based assets moving toward Ready-to-Build between 2028 and 2029, and their eventual spot exposure, whatever portion of it is uncontracted, will clear against whatever COES's marginal cost looks like in a system with materially less gas behind it than it has today. That is not a reason to avoid the market. It is a reason to model the dispatch curve honestly rather than extrapolate the last decade's cheap-gas average forward.

The open question

Nobody actually knows whether Peru replaces Camisea in time. Exploration incentives exist on paper, prospective resources in Ucayali and Madre de Dios run into the tens of TCF, and none of it is proven reserves yet. Perupetro's 2004-to-today decline curve argues for pessimism. The gap between prospective and proven gas is exactly where a country either finds its next Camisea or spends the 2030s importing LNG at a price nobody in Lima wants to put on a tariff filing.

We genuinely do not know which way that resolves, and we would treat anyone who tells you they do with some suspicion.

See the full active portfolio or read more Insights.