Coca Codo Sinclair: Ecuador’s largest power plant and the erosion moving toward it

CELEC, the state generator, publishes a daily monitoring report on the regressive erosion of the Coca river (CELEC). Since about August 2025 those reports had placed the erosion front at Km 3+600, 3.6 km downstream of Coca Codo Sinclair’s intake; the last to do so is the report of November 2 (CELEC). The report of November 3 says the front “advanced” to Km 3+200, 3.2 km from the intake, and the report of November 4 repeats it (CELEC, November 3; CELEC, November 4).

The reports describe Km 3+200 as a zone of incipient erosion, where sediment is continually being reworked and the channel is changing direction and slope, and they record about 0.40 m of bed deepening at Km 3+600. Both say the intake shows no structural damage and that the plant ran without interruption.

The step came without a large flood. The highest flow in CELEC’s reports of late October was 490 m³/s, on October 28 (CELEC); the April 2025 flood that moved the front 1.4 km peaked at 1,440 m³/s (Primicias). Four hundred metres is small against the roughly 17 km the front has climbed since 2020, but it leaves the front 200 m from Km 3.0, where CELEC’s nearest planned work, a quarry and a diversion of the river toward a hard point on its right bank, is still at the design stage (CELEC).

The reports’ own bookkeeping is loose. Each one counts the days the front has gone without advancing; from September 30 to November 2 every report printed the same figure, “417 días”, while the count from early August 2025 passed 450 (CELEC, October 28; CELEC, November 2). The report of November 3 drops the counter. Its table of works still lists the Km 7.8 dike, partly destroyed in May, as “Concluida” (completed), and says in the past tense that it “formó parte” (formed part) of the erosion-control system (CELEC).

Coca Codo Sinclair takes its water from the Coca river where the Quijos and the Salado meet, on the Amazon slope of the Andes. A dam about 31 m high diverts the river through desanders into a 24.8 km tunnel, which ends at a small compensating reservoir. From there two penstocks drop about 600 m to an underground powerhouse with eight Pelton units (Contraloría; Andritz).

Sinohydro built it under a USD 1,979.7 million engineering, procurement and construction contract signed in October 2009, 85% of it financed by a China Eximbank loan of USD 1,682.7 million at 6.9%. Construction began in July 2010; a tunnel collapse in December 2014 killed 13 workers; the plant was inaugurated on November 18, 2016, about ten months late (AidData). The final cost is quoted at USD 2,500–2,763 million, plus about USD 600 million for the transmission lines (Primicias, 2021; Primicias, 2026; Expreso).

The loan is being paid for still. It carried 6.9% fixed over 15 years, with 5.5 years of grace, and required Sinosure credit insurance for which Sinohydro paid a USD 172.5 million premium; in 2017 the tax authority rejected the deduction of that premium and found USD 198 million of taxes pending from the contractor. Ecuador paid seven instalments of USD 93.49 million from 2017 to 2020, deferred principal for 16 months in 2020 and rescheduled in September 2022, at 6.35% with maturity in February 2029 and twelve semi-annual payments of USD 74.98 million; USD 899.8 million of principal was outstanding in November 2022 (AidData). If that schedule has been kept, about USD 300 million remains, due between March 2027 and September 2028.

The size was a choice. The 1992 feasibility study sized the plant at about 864 MW, and the 1,500 MW rating was adopted without new studies, according to consultant Ricardo Buitrón (Primicias). The eight turbines take 278.5 m³/s at full load (IWP&DC); the tunnel that feeds them was designed for 222 m³/s (Contraloría). The difference is what the compensating reservoir supplies, for about four hours. The plant was built with a conveyance that cannot feed its own turbines for more than a few hours a day, and its yearly averages since 2023 sit within 15% of the 1992 figure.

The reservoir holds hours of water, not days, so the plant generates what the river brings. Full output needs about 222 m³/s at the intake; below that the plant loses about 50 MW for every 10 m³/s less (Primicias). Averaged over each year, it has delivered 56–65% of its nameplate.

A wet year does not lift that share much. 2025 brought 44% more water than 2024, but the plant produced about 6% more: floods bring sediment and spill, and the tunnel caps what the plant can take. Over 1,350 days from January 2023 to November 2026, on the 521 days when the river carried more than 300 m³/s, nine days in ten stayed at or below 1,192 MW. The best day on record averaged 1,227 MW, on January 27, 2025.

The 1,500 MW are real only for a few hours at a time, when the compensating reservoir tops up the tunnel’s flow for the evening peak: 1,342 MW at 21:00 on September 27, 2026, for instance (Primicias). In the dry season the plant falls well below its yearly average: 567 MW in October 2024, when the river dropped to 96 m³/s (Primicias), and 653 MW in March 2026.

On November 5 the river carried 184.4 m³/s and the plant generated 19,976 MWh, about 832 MW on average and a fifth of the 100,684 MWh delivered on the national grid that day (CENACE).

Floods carry more sediment than the desanders can remove, and CELEC then closes the intake gates to protect the turbines. CELEC’s unit manager counted 28 such shutdowns in 2024 and 31 in 2025 (Primicias); another count gives 33 for 2025 (Primicias). The longest lasted about 96 hours, around July 1–4, 2025 (Primicias).

On May 21, 2026 a flood of about 2,800 m³/s at the intake left the plant at 5% of its normal output and brought power cuts from Quito to the coast (El Mercurio; Primicias). On September 14 sediment left three of the eight units running for four hours, and CENACE ordered load-shedding in Quito (Primicias). On October 28 the river rose to 490 m³/s, most of it went past the intake, and the plant averaged about 363 MW for the day (CELEC; CENACE).

The plant is also hard to take out of service. A two-day shutdown to clean sediment from the intake, planned for June 2026, was moved to the last four months of the year after CENACE judged that losing all 1,500 MW at once was not feasible with the reserves available (Primicias).

The distributors are the eight steel manifolds that carry water from the penstocks to the turbines. The first fissures in them were found in 2014, two years before the plant started: 7,648 of them (Primicias). The Contraloría, the state comptroller, confirmed that count in 2018, some of the cracks up to 38 cm long, and blamed unapproved materials, poor quality control and poor welding procedures (Ecuador Chequea). The same report warned that a distributor failure could flood the powerhouse and endanger about 70 workers, and said the state would refrain from accepting the plant (Primicias).

Repair campaigns ran in 2015, 2018, 2019 and 2021 (Primicias). A dispute board ruled on August 9, 2020 that Sinohydro must pay for an independent root-cause study and remediate the distributors with the TÜV SÜD method tried on distributor 8, where 154 defects were repaired and 15 persisted, and it found no evidence that CELEC’s preferred carbon-fibre lining would fix them. By July 2021 more than 7,000 fissures had been repaired and more than 800 new ones found (Primicias).

A confidential CELEC report of January 27, 2022 described the fissures in all eight distributors as irreparable: Sinohydro had not built them to the ASME standard the contract required, water was seeping through exterior cracks on distributors 2 and 8, and the alternatives were a lining or a new powerhouse cavern, which the Contraloría had costed at about USD 1,010 million in 2018 (Primicias). That year CELEC reported 17,661 fissures to the National Assembly (El Universo). The sources reviewed for this article contain no later count and no published assessment of what happens if the cracks grow.

What remains in hand is a USD 36 million guarantee on the distributors that expires in May 2029 (Primicias). Repairs cost about USD 5 million every two years, and who pays for them is undecided (Primicias); in April 2026 the vice-minister of energy said the cost of repairing them “no lo hemos evaluado” (we have not evaluated it) (El Universo). The Contraloría’s civil-liability finding over the construction, USD 170.9 million, with 27 parties sanctioned including Sinohydro and the supervising consortium, was unpaid as of March 2023; the same body found that USD 80 million in fines were never charged to the contractor, and that of USD 305.7 million of findings issued to 2018, USD 295.1 million were dismissed after an October 2018 resolution (Expreso). A finding is an administrative act, not a court ruling.

The erosion began when the San Rafael falls, about 20 km below the intake, collapsed on February 2, 2020 (Primicias). With the falls gone, the river began cutting into its own bed, and the head of that cut has been moving upstream ever since. Its cause is disputed: a sediment model by the Escuela Politécnica Nacional attributed a 42% rise in the erosion rate to the plant, which traps the river’s sediment and was built without the bottom gates of its original design, while the national risk management agency calls it a natural phenomenon (Plan V). About 500 million tonnes of sediment have moved since 2020 (Expreso).

The front stalls for months and then jumps in floods; five of its six advances from 2020 to May 2026 came in June or July, the Amazon slope’s wet season (Primicias). In 2025 it moved about 2.5 km in two flood episodes, from 6.1 km before the April flood to 3.6 km by August (Primicias). The November 2026 step came outside that season.

Little stands in its way, and what has been built has a poor record. The first defence, in 2020, was an emergency dike of eleven concrete containers for USD 278,320; the Contraloría found in 2021 that it worked only up to about 290 m³/s and “no tuvo efecto” (had no effect) (Primicias). A screen of 297 piles, 24 m deep, at Km 1.2, completed in 2023, is the first of 14 planned (Primicias; Primicias).

The permeable dike at Km 7.8 was built where the erosion had already passed. Its tender opened in November 2023 at a reference price of USD 17.3 million; a technical report of March 28, 2024 found the river bed about 11 m below the top of the dike’s design and called for a full review of the studies, and commission reports that year recommended voiding the tender. It was awarded in September 2024 anyway, to Consorcio SEMACC (Expreso). Two experts say the front had moved upstream of Km 7.8 before construction began (Primicias). The dike is about 200 m long on a river terrace about 900 m wide (Expreso) and was designed for 3,900 m³/s, a 25-year flood; CELEC’s model puts the flood of May 21, 2026 at 4,363 m³/s at its section (Primicias). It cost about USD 19–20 million and was partly destroyed 38 days after it opened (Primicias). The minister said the contractor’s USD 20 million in policies still applied because the state had not formally received it (Primicias); on November 3 CELEC’s list of works showed no repair or replacement (CELEC).

Assessments by the US Army Corps of Engineers (USACE) indicate that at least four such structures are needed (The Cuenca Dispatch). On November 4 CELEC listed the Km 3.0 work at design and a stepped spillway at Km 1.2 at feasibility; its crews were placing riprap, building spur dikes and training the channel with their own machinery (CELEC).

The erosion has cost more than electricity. It has broken the SOTE and OCP oil pipelines and cut the E45 road: Petroecuador’s internal report counts at least 13 interruptions of oil transport from 2020 to 2025, about 23.5 million barrels not moved and USD 1,124 million of lost revenue (Primicias). A permanent pipeline route on the river’s right bank, at about USD 200 million, is to be tendered in 2027 (Primicias). Two geologists, Alfredo Carrasco and Bernardo Beate, put the erosion’s total losses to May 2026 at USD 4,700–5,500 million (Expreso); that is an estimate, not an audited figure, and it is larger than the plant cost.

The estimates differ. USACE said in February 2024 that the erosion could reach the intake works in two to five years (Primicias). The energy minister, Juan Carlos Blum, said in June 2026 that studies put it there in 2027 (Primicias). CELEC said in April 2025 that the intake ran no risk (Primicias). None of these sources gives a probability.

A simple comparison shows what is at stake. A flood season like 2025’s, which moved the front about 2.5 km, would leave it about 0.7 km from the intake. That is arithmetic on a past season, not a forecast: in 2026 the high-flow season passed without moving the front at all, and its next step came in the dry season.

No contingency plan for the intake has been published. In June the minister called moving it very costly and said it would be studied from scratch (Primicias). The documented alternative is a new intake upstream on the Quijos: at least 18 months of studies and three years of building, USD 400–500 million or more, or about USD 700 million on a 2022 CELEC estimate (Primicias). It would lose the Salado’s water, 30–40% of the flow on one academic estimate (El Oriente); on the plant’s recent output that would mean roughly 550–650 MW on average, years from now.

Nobody else carries the risk. The state insurer excluded erosion from the plant’s policies in January 2021, at the demand of foreign reinsurers, when the intake works were valued at USD 400 million (Primicias). The PowerChina operating contract excludes erosion damage too, and leaves mitigation with CELEC (El Diario); the minister has said that if the plant cannot run, that contract “no tendría sentido” (would make no sense) (Primicias).

There is a second risk at the far end of the plant. USACE also warned that sediment building up at the powerhouse could put it out of operation permanently within seven years (Primicias). Whatever happens to the plant, the rest of the system would have to cover it: of the 1,650 MW of firm thermal capacity the energy ministry ordered on July 28, 2026, 126.5 MW was in service by September 16 (Primicias).

CELEC filed an arbitration claim of USD 580 million against Sinohydro in May 2021 (Primicias); Sinohydro counterclaimed in 2024. On July 8, 2025, during the president’s state visit to China, the foreign minister announced that Ecuador would drop the case and that PowerChina would pay USD 400 million by the end of 2026 (El Universo). What was signed is smaller. The consent award of March 30, 2026 has Sinohydro pay USD 200 million in four instalments of USD 50 million and provide USD 200 million of supplier credit, tied to purchases from the contractor’s group; of the nine acceptance conditions in the construction contract, Sinohydro had met four, and the rest “se reputarán irrevocablemente cumplidas” (shall be deemed irrevocably fulfilled); and defects from before acceptance pass to an annex, not public, of a new operation and maintenance contract with PowerChina (Primicias; KCH). CELEC signed final acceptance of the plant on April 17 (Ecuador Chequea).

On the day of acceptance CELEC said the distributor guarantees “se mantendrán en custodia” (will remain in custody) (Ecuador Chequea). On May 8 it returned USD 98.9 million of them, bringing the guarantees released to USD 154.1 million of USD 190.1 million; only the USD 36 million policy on the distributors is held, until the PowerChina contract is signed (Primicias). That 25-year contract was first put at about USD 60 million a year (El Universo), then at USD 46 million, about USD 1,150 million over its term, against a 2009 estimate by the plant’s consultant Electroconsult of about USD 18 million a year (Primicias). It was still unsigned in August, with December as the target (Primicias). The sources reviewed for this article do not confirm that any instalment has been paid.

Taken together: a USD 580 million claim and five unmet acceptance conditions were exchanged for USD 200 million in cash that no public source yet shows received, USD 200 million of credit to be spent with the same group, and a 25-year contract worth about USD 1,150 million to that group. The liability for the fissures was moved into a document the public cannot read, not settled. That is this article’s reading of the reported terms; the award itself is not public.

On August 28, 2026 a first-instance tribunal of the Corte Nacional de Justicia convicted former president Lenín Moreno and 19 others of bribery linked to the plant’s contract. Prosecutors had alleged about USD 76.1 million in bribes, about 4% of the construction contract, paid from 2009 to 2018 through consultancy and representation contracts and moved through Panama, Switzerland, Spain, Belize, China and the United States. Moreno and four others, among them two Sinohydro executives and a former Chinese ambassador who later acted for the company, received five years; reparations total more than USD 163 million, of which Moreno’s share is USD 58,026 (El Comercio; Primicias). The ruling is not final. As of late September the written sentence had not been notified and no appeal had been filed (El Universo).

On its record, Coca Codo Sinclair is an 850–950 MW plant in an average year, 550–850 MW in a dry month, and close to zero for hours whenever the Coca floods. Planning on its 1,500 MW, or even on 1,200 MW, overstates the country’s firm supply by several hundred megawatts.

The intake is the larger uncertainty. While the front holds, the plant keeps those figures; if the erosion reaches the intake works, the country’s largest plant would be out for years, and the fallback would give about two thirds of today’s output. The November advance does not settle when that could happen, but it shows the front can move outside the flood season.

Three things about the plant are not known publicly: the state of its eight distributors since 2022, what CELEC would do if the intake were lost, and whether any of the settlement’s cash has arrived. Each is a question of record-keeping, not engineering, and each could be answered by the agencies that hold the documents.

• CELEC’s daily erosion reports (CELEC): whether the front holds at Km 3+200, and the bed at Km 3+600.
• The Coca’s flow at the intake and the plant’s output, on CELEC Sur’s production page.
• Any tender or contract for the Km 3.0 works, a decision on the Km 7.8 dike, and any published contingency plan or decision on a new intake.
• A fissure count or condition report on the distributors after 2022, and the annex of the PowerChina contract that covers them.
• The PowerChina contract, due by December, and the payment of the award’s instalments.
• Notification of the written sentence in the bribery case, and any appeal.
• The April–July 2027 high-flow season on the Amazon slope.

The erosion front, the river’s flow at the intake and the protective works come from CELEC’s daily monitoring reports, read from October 20 to November 4, 2026, and its project page (CELEC). Daily output and river flow come from CELEC Sur’s hourly readings, averaged by day; days without a reading are left out, and these averages run 2–5% above the contractor’s figures for 2024 and 2025. Single-day figures for output and national generation are CENACE’s. The ceiling figures come from the 1,350 days with both readings from January 1, 2023 to November 6, 2026. The output after a move of the intake is an estimate that scales recent output to 60–70% of the flow. The remaining loan principal is arithmetic on the 2022 rescheduling as AidData reports it, not a Ministry of Finance figure. The comparison of tunnel and turbine flows uses the Contraloría’s design figure and the turbine maker’s trade press. Press reports are cited where they are the only public source, and official statements and estimates are attributed to whoever made them.

Output data run to November 7, 2026, and the erosion front to CELEC’s report of November 4. The agencies that publish these figures can revise them.