Reading the 2026 Results: what the accounts say, where the demand went, and what has not been repaid

The argument in one paragraph

Eskom has recovered, and the recovery is real. The profit is real, the cash behind it is real, and the plant improvement that produced it can be traced to tonnes of coal not burnt and litres of diesel not bought. But the profit came from charging more rather than selling more, on a customer base that has been shrinking for a decade and is now shrinking faster. The audit remains qualified for another year. Roughly half a trillion rand of public money has gone into this company since 2008 and is not coming back, alongside a decade of economic growth the country did not have. Eskom has been stabilised. The account has not been settled, and those are different things.

1. The profit is real

Eskom reported a profit after tax of R30.3 billion for the year ended 31 March 2026, against R14 billion restated a year earlier.

The most obvious suspicion does not survive examination. Government provided R80 billion in March 2026, in the final month of the financial year, but it was recognised as a shareholder loan at year-end and approved for conversion to equity in August. It sits on the balance sheet, not in the profit.

The cash test is the one that matters and it passes. Operating cash flows fully funded the R45 billion capital programme, for the second consecutive year. A business reporting a profit while its cash goes backwards is telling you something different from one that pays for its own investment out of what it earns.

The cost savings are physical rather than accounting. Coal burnt fell from 106.2 million tonnes to 96.5 million. Spending on open-cycle gas turbines fell from R17.7 billion to R7.1 billion. Nuclear output rose from 8,409 gigawatt-hours to 11,618 as Koeberg returned. Energy availability improved to 65.16 percent from 60.60 percent, and the four days of load shedding recorded in the financial year all fell in its opening weeks. The last rolling blackouts were in May 2025, and by May 2026 Eskom had gone a full calendar year without them.

Anyone claiming the numbers cannot be believed has not read the statements.

2. It came from price, not volume

Revenue rose from R340.9 billion to R354.7 billion, a growth of 4.1 percent. In the same year Eskom implemented a standard tariff increase of 12.74 percent, and sales volumes fell 6.2 percent to 178,032 gigawatt hours.

Eskom charged roughly 13 percent more and collected roughly 4 percent more, because it sold considerably less electricity.

That is a materially different achievement from the one reported, and it raises the question the results do not answer: whether it can be repeated in a year when the tariff increase is smaller, and the volume decline is not.

3. The audit is still qualified

The auditors issued a qualified opinion for another year, now limited to the completeness of irregular expenditure reported under the Public Finance Management Act. Two previous qualifications, on losses arising from criminal conduct and on the accuracy of irregular expenditure reporting, have been removed.

The qualification means the auditors cannot satisfy themselves that Eskom has disclosed all the irregular expenditure it incurred. Not that the figure is wrong. That they cannot confirm it is complete.

The disclosed figure is around R4.9 billion, of which only about R28 million arose in the year under review. The overwhelming majority relates to historical contracts concentrated in two incidents from the state capture years. Almost no new irregular expenditure is being created, which cuts in Eskom’s favour.

It remains true that South Africa’s largest state-owned company has again failed to obtain a clean audit.

4. Eskom is selling less than it did during load shedding

In the year to March 2024, the worst year of load shedding on record, when the country lost close to six percent of the electricity it asked for, Eskom sold 183,311 gigawatt hours.

In the year to March 2026, when it lost 0.02 percent, Eskom sold 178,032.

Eskom sold less electricity in the year it kept the lights on than in the year it could not. Sales have declined at roughly two percent a year over the past decade, and for the first time in more than ten years the utility holds surplus capacity of two to three gigawatts. Electricity it can make and cannot sell.

Its chief financial officer calls the decline a serious signal.

5. Two different things are happening to demand

The first is that demand has moved behind the meter. Rooftop solar and embedded generation passed 8.3 gigawatts by June 2026, more than the total capacity of every independent power producer selling into the grid. Those installations were decided on in 2022 and 2023, are now paid for and depreciating on private balance sheets, and will not be removed because the grid has improved.

The second gets almost no attention. Some of the load did not move. It stopped.

Demand from energy-intensive industrial customers fell 9.7 terawatt-hours during the year, a decline of 22.5 percent. The Mozal aluminium smelter went into care and maintenance in March 2026 after its owners could not agree an electricity price. Two Glencore Merafe ferrochrome smelters were mothballed from January 2026. South Africa’s last manganese smelter shut its furnaces.

Those plants have not gone solar. They have stopped producing.

A country generating more of its own electricity is adapting. A country that has stopped smelting is shrinking. The 6.2 percent decline contains both, and only one is a success story.

6. The fixed cost problem

Eskom owns 30 power stations and more than 411,000 kilometres of power lines and cables. Almost none of that cost falls when a customer leaves. The line to a town costs the same whether the town draws from it or not.

Capital spending was R45 billion in the year under review and is planned to rise above R70 billion a year from 2029, with total investment of up to R343 billion over five years. A growing fixed cost base on a shrinking sales base.

There is one place the difference can go. It goes into the unit price, which raises costs for everyone still buying and strengthens the case for the next customer to leave. That is not a prediction. It is arithmetic, and Eskom’s leadership now states openly that double-digit increases are unaffordable.

7. Who left, and who remains

The customers who left are, almost by definition, those who could afford to leave. A household solar and battery system costs between R120,000 and R165,000 installed, against a median household income under R100,000 a year. Most of the 8.3 gigawatts sits on commercial rooftops.

This is the pattern of every failing public service in the country. When a system stops working, those with capital build a private version, and households end up paying twice: once through rates and tariffs, and again for the workaround.

Alongside national load shedding, Eskom has for years run load reduction: controlled cuts applied at peak hours to specific overloaded networks in areas affected by illegal connections and theft. Anyone living in an affected area knows it well. This year’s report adds the scale of it. Seven provinces have now been cleared, and more than 1.2 million customers have been relieved of scheduled cuts, with national elimination targeted for March 2027 and Gauteng the main outstanding area.

More than a year after national load shedding ended, over a million households were still having power cut on a schedule.

Those are the customers who remain. Many are on prepaid meters, which means they pay in full and in advance before a single unit flows, and they have no exit.

8. What it cost

In October 2024, the Director General of the National Treasury told Parliament’s Standing Committee on Appropriations that by the following year Eskom would have received R496 billion in bailouts since 2008. Roughly half a trillion rand into a single company over sixteen years, none of it available for anything else.

Eskom’s own report cites the CSIR estimate that load shedding cost the economy up to R2.8 trillion in 2023. The OECD puts the 2023 impact at 1.5 percentage points of growth, and the Reserve Bank reaches a similar conclusion. The economy grew 0.7 percent that year.

Every rand of that support was raised by the state at the state’s cost of funds and is now being serviced. Eskom’s debt also carries government guarantees running to hundreds of billions of rand, which remain until the underlying debt is settled. A guarantee does not appear as expenditure. It costs nothing until it costs everything.

For years Eskom was the largest single item in the state’s contingent liabilities, which is a substantial part of why South Africa was downgraded to below investment grade across all three agencies by 2020.

That mechanism has now reversed. In November 2025, S&P raised South Africa’s rating for the first time since 2009, citing reduced contingent liabilities tied to performance improvements at Eskom. Fitch followed in 2026. Eskom’s own rating was upgraded for the first time in over a decade. The country remains below investment grade.

9. What is still owed

Municipal arrear debt rose 17.9 percent during the year to R111.6 billion, and reached R119.9 billion by June 2026. It was around R35 billion in 2020. Eskom’s chief executive describes the level of municipal non-payment as a sovereign risk. On Eskom’s own projections, unchecked, it could reach R358 billion by 2031.

This is usually described as people not paying for electricity, and that is not what it is. Municipalities buy power and sell it on. In many cases, residents pay their municipality, and the municipality does not pay Eskom. Money is collected and not passed on.

Against that background, Eskom must raise around R88 billion of funding over five years and return to the capital markets on its own credit, without a government guarantee, having raised only R1.4 billion during the year under review. It must do so while selling less electricity each year than the year before, and through a change of leadership: the chief financial officer steps down in October 2026 after 24 years, with a successor approved but not contracted, and the chairman’s term ends in the same month.

The question this leaves

Eskom has done what it was asked to do. The plant runs better, the lights are on, the accounts are profitable and the country’s credit rating has improved partly because of it.

None of that returns the R496 billion, or the growth the economy did not have, or the businesses that closed.

If a utility can be repaired but the cost of its failure cannot be recovered, what should we now require of it in return?