About Contact
Thursday, May 14, 2026

South Africa – ESKOM


South Africa Energy  ·  Eskom P116410  ·  MDB Reform Platform

Does the World Bank Learn from Project Failures? (Part 3) — One Project. One Utility. $18.1 Billion. Rated Moderately Unsatisfactory.

The Largest Single-Project Failure in the World Bank’s Africa Portfolio. Nine Consecutive Unsatisfactory ISRs. $1.7 Billion Disbursed While the Project Was Failing. No PPAR.

📄
Full Paper — 26 Pages (PDF) Draws on the ICR (106 pages), IEG ICRR, all 21 ISRs, Bank-approved procurement plans, the Inspection Panel Investigation Report, the Zondo Commission, Betrayal of the Promise, Shadow State, de Ruyter’s Truth to Power, the IEG Master Database of 10,542 rated projects, and the complete ISR ratings timeline.
↓  Download Full Paper (PDF)
$18.1bnTotal project cost at completion. Medupi alone: $17.5bn. Original estimate: $13.9bn. Cost overrun: +$4.3bn (+31%).
$5.4bnMedupi cost overrun. $12.0bn → $17.5bn. Larger than the Angola DPF, Ethiopia education, and Niger MTI portfolios combined.
MUIEG Outcome: MU. Bank Performance: MU. Quality at Entry: MU. Quality of Supervision: MU. ERR halved.
0PPAR recommended. No field evaluation. $1.8M Armenia project: Yes. $3.75bn Eskom: 19-page desk review.

The Climate Champion That Financed Coal

In October 2008, the Bank endorsed its Strategic Framework for Climate Change. $6.4 billion for Climate Investment Funds. An 87 per cent increase in renewable energy financing. Zoellick pledged “low carbon growth.” The US Treasury proposed guidelines requiring “full consideration of alternatives” before approving a coal plant. Eighteen months later — April 8, 2010 — the same institution approved its largest-ever loan: $3.75 billion for Medupi, a coal-fired power plant emitting 25–30 million metric tons of CO2 annually — more than 60+ of the world’s lowest-emitting countries combined. Five shareholder countries abstained. Over 200 civil society organisations opposed. When the disbursement opportunity presented itself, the climate credentials were set aside.

The Corruption Was Already in the Public Record

The Bank was not the first to arrive. Construction at Medupi had started in 2007 — three years before the Bank’s involvement. The design was fixed. The procurement was already compromised. In 2007, Hitachi’s boiler contract involved the ANC’s investment arm, Chancellor House, which held a 25 per cent stake in Hitachi Power Africa. The Eskom board chairman simultaneously sat on the ANC Finance Committee. In 2015, Hitachi paid a $19 million fine to the US SEC under the Foreign Corrupt Practices Act. The boilers that resulted from this corrupted procurement turned out to be defective. The Bank appraised in 2009. It approved in 2010. The corruption was already in the public record. Two days before the Board voted, the Inspection Panel received a complaint from Lephalale communities. The Panel’s Investigation Report found non-compliance. The project proceeded.

In the Bank’s Own Words — from the ICROn the virtual design: ‘Eskom used a virtual design based on the Majuba power plant, constructed about 10 years earlier… outdated codes and standards.’ On the boilers: ‘The boiler furnace is too small for the combustion of South African coal — a situation that cannot be rectified.’ On coal testing: ‘No documentation was ever provided to the World Bank. Had this been done, the too small furnaces could have been identified.’ On the Board’s exception: ‘The Board agreed, on an exceptional basis, to finance contracts that were awarded without following the World Bank’s guidelines.’

The Record: Appraisal vs. Completion

MeasureAt AppraisalAt CompletionVariance
Total project cost$13.86bn$18.13bn+$4.27bn (+31%)
Medupi component$12.05bn$17.47bn+$5.43bn (+45%)
IBRD loan disbursed$3.75bn$3.16bn$590M cancelled
ERR22.7%12.8%Halved
Closing dateOct 2015Jun 20216 years late
Restructurings06Dec 2014 – Jun 2020
Unplanned outages28% of timeBelow all benchmarks

A note on the financing figures

The Bank’s own IBRD loan was $3.75 billion ($3.16 billion disbursed). The IEG database records $9.13 billion — the Bank’s loan plus co-financing from the AfDB, EIB, JICA, the Clean Technology Fund, and bilateral lenders. The total project cost at completion was $18.1 billion. All three figures appear in the paper.

State Capture: The Zondo Commission Record

The state capture literature — the Zondo Commission, Betrayal of the Promise, Shadow State, the GuptaLeaks investigations — documents what happened to Eskom in precise detail. R74 billion ($10.6bn) in annual procurement became the most valuable prize in South Africa’s political competition for rents. The board was captured: almost every new appointee had documented Gupta links. The CEO visited the Gupta compound 19 times in four months. Coal contracts escalated from R161/ton to R550/ton. The R659 million overnight prepayment funded the Gupta purchase from Eskom’s own coffers. R14.7 billion ($920 million) in Eskom contracts were found by the Zondo Commission to be “afflicted by State Capture.”

Not a Bystander: The Bank’s Procurement Role

The Bank was not a passive observer. It was actively participating in the procurement architecture of a captured institution. The Bank-approved procurement plan for P116410 was revised six times — each revision submitted to and approved by Bank staff. The Bank conducted Prior Review on all works contracts above $15 million: reviewing bid evaluations, assessing contract awards, clearing procurement decisions. Contracts worth tens of millions of dollars were signed during 2015–2018 — the exact period the Zondo Commission found Eskom’s procurement was being manipulated by politically connected networks.

The Complicity QuestionIf you clear and approve every procurement plan, if you conduct Prior Review on every major contract, if you sign off on the bid evaluations and the contract awards — you are not a bystander. You are a participant. And if the procurement environment turns out to have been captured, the institution that approved the procurement has questions to answer. The Board had already agreed ‘on an exceptional basis’ to finance contracts awarded without following the Bank’s own guidelines. The Bank’s procurement review function either failed to detect the capture — in which case its oversight was structurally inadequate — or it detected it and chose to continue. Either conclusion is damning.

Four Country Directors oversaw the project: Ruth Kagia, Guangzhe Chen, Paul Noumba Um, and Marie Françoise Marie-Nelly. Three Regional Vice Presidents for Africa: Obiageli Ezekwesili, Makhtar Diop, and Hafez Ghanem. Makhtar Diop presided over every one of the nine consecutive Unsatisfactory ISRs — and subsequently became VP Infrastructure and then Managing Director of IFC. The word ‘state capture’ does not appear in the ICR. The word ‘Gupta’ does not appear. The Bank’s 106-page completion report does not name the crisis.

The Game Theory: Why No One Stopped It

The TTL who pushed the largest IBRD loan in history through Board approval subsequently rose to one of the most senior infrastructure positions in the institution. The RVP who presided over nine consecutive Unsatisfactory ISRs subsequently became Managing Director of IFC. The project failed. The careers advanced. The sovereign guarantee ensured the Bank would be repaid regardless of what happened to Eskom, to the communities of Lephalale, or to the South African taxpayers. Nobody bears the cost of failure except the borrowing country. In the Nash equilibrium of the approval culture, pushing for approval is individually rational regardless of the development outcome.

No PPAR: The Evaluation That Will Not Be Conducted

ProjectCountryCommitmentRatingPPAR?
Energy Efficiency (P116680)Armenia$1.8MMSYes
Food Crisis Response (P113219)Sierra Leone$3.0MMSYes
Fish Resources (P105881)Senegal$3.5MHUYes
Electricity Enhancement (P073477)Senegal$44.1MUYes
Eskom Investment Support (P116410)South Africa$3,750MMUNo

Annex: The ISR Split-Screen Timeline (Key Rows)

#DateDOIPDisbursedExternal events
01Jun 2010SS$0MBoard approved Apr 8. IP complaint filed Apr 6. US/UK/NL abstained.
03Dec 2011SMS$569MInspection Panel finds non-compliance: water, air quality.
06Oct 2013UU$1,333MCRASH. Both ratings drop to Unsatisfactory.
09Jun 2015UU$2,292MHitachi pays $19M FCPA fine. Boiler defects detected.
14Mar 2018UMU$3,028MNine consecutive U-rated ISRs end. Zondo Commission launched.
16Apr 2019MSMU$3,048MDO jumps U → MS in one cycle. Underlying problems not resolved.
21Aug 2021MUMU$3,132MProject closing. $18.1bn total cost. IEG: no PPAR.

The full 21-row timeline is in Annex A of the paper.


Series: Does the World Bank Learn from Project Failures?

PartCountry–SectorCommitmentStatus
1Nigeria Water$1.8bnPublished
2Angola DPF$2.2bnPublished
3South Africa Energy (Eskom)$9.13bnThis paper

Companion Papers

Institutional Power Architecture and Portfolio Distortion at the World Bank — How MTI controls the DPF instrument, manages the Finance Ministry relationship, and records the worst outcome rate of any GP at scale.

Why the System Does Not Learn: A Game Theory Analysis — The Nash equilibrium of the approval culture. Why no single actor inside the institution has the incentive to stop a project that is too large to cancel.

Browse by Topic