Insight, Energy Systems
South African miners are building around three gigawatts of their own renewable capacity. Every large customer that leaves the grid takes its share of the fixed costs with it, and the bill lands on whoever cannot follow. Whether this ends in a stronger system or a hollowed one is decided in tariff design.
South Africa's largest miners are building their own electricity system. Anglo American's Envusa platform targets around three gigawatts by 2030. Sibanye expects renewables to cover roughly 64 per cent of its South African electricity requirement by 2028. Exxaro's existing 68 megawatt Grootegeluk solar plant has reportedly reduced its grid dependence by around 30 per cent.
Read on its own, this is a good news story about corporate decarbonisation and about private capital building generation that the state has struggled to finance. It is also the clearest illustration available anywhere of a dynamic that will shape emerging market electricity systems for the next two decades, and the mining sector is simply where it becomes impossible to ignore.
The customers leaving the grid are the ones that were paying for it.
The utility death spiral entered energy economics in the early 2010s, most prominently through an Edison Electric Institute analysis of distributed generation, though the underlying logic is much older and applies to any network industry with high fixed costs and a declining volume base.
The mechanism is simple and self-reinforcing. A utility's costs are largely fixed: generation plant, transmission, distribution, debt service, staff. Its revenue is largely volumetric, recovered per kilowatt hour sold. When sales fall, fixed costs are spread across fewer units, so the tariff rises. The higher tariff improves the economics of self-supply, so more customers leave, so sales fall further.
The version usually discussed is about households with rooftop solar, and it is comparatively benign, because a household's departure is partial, slow and small. What is happening in South Africa is the aggressive version, and it differs in three ways that matter.
It is happening at the top of the load curve, not the bottom. Industrial and mining customers consume enormous volumes at high load factors. They are the customers over whom fixed costs were being spread.
The departures are large, fast and financed. A three gigawatt corporate renewable programme is not a household adding panels. It is a balance sheet decision made once, executed over a few years, and effectively irreversible.
The customers who cannot leave are the ones least able to absorb the consequence. Financial Mail's assessment of Eskom this year put it plainly: the most profitable customers, mines, manufacturers, commercial property groups and affluent households, have left for cheaper and more reliable private generation, leaving a customer base dominated by low usage households, indigent individuals and financially distressed municipalities.
The numbers behind that are not marginal. Eskom's demand fell 1.9 per cent, then 2.9 per cent, then 4.9 per cent across 2023, 2024 and 2025, and in the first 29 weeks of 2026 residual demand was down 9.1 per cent year on year and 15.6 per cent against the same period in 2022. The utility expected sales of 179,000 gigawatt hours for the financial year to March 2026, which would be the first time since 2000 that demand fell below 180,000. Behind the meter rooftop solar alone grew by more than five thousand megawatts between 2022 and 2026.
Tariffs have moved in the opposite direction throughout, rising more than 1,100 per cent since 2007, with further approved increases of 8.76 per cent in 2026 and 9.19 per cent in 2027 against much lower inflation.
That is not a forecast of a death spiral. It is a description of one in progress.
The mechanism is closed and it runs in one direction.
The loop is not the problem. The asymmetry in who can step out of it is: customers with capital exit, and customers without capital stay and pay the difference.
Pakistan's Jazz says it has connected around a thousand sites to 5G and is calling for grid improvements, because 5G equipment draws substantially more power than the 4G it replaces.
That is the defection decision arriving before it has been taken. Telecom operators across South Asia and Africa have run tower networks on diesel for two decades, precisely because the grid could not be relied upon. A generation of digital infrastructure with a higher and more continuous power draw will either be served by the grid or it will build around it, and the operator is currently asking, publicly, which it is going to be.
The general form is worth stating, because it is not about mining or telecoms.
Any customer whose electricity requirement is large, continuous and financeable will eventually build its own supply if the utility's price and reliability are worse than the alternative. Data centres, mines, smelters, cold chains, telecom networks and irrigation schemes all sit in that category. The only variable is whether they do it inside the system or outside it.
Whether mine specific power systems eventually strengthen the wider grid or create self contained industrial energy islands is the right question, and it has a technical answer, which is unusual and useful.
A corporate renewable project can be built in two fundamentally different ways.
Wheeled. Generation sits where the resource is good, the power travels over the national transmission network to the mine, and the user pays use of system charges for the privilege. The generation is private, the network is shared, and the user remains a paying customer of the transmission system it depends on.
Islanded. Generation sits at or near the load, connected behind the meter, with the grid used only as backup or not at all. Nothing is contributed to network costs, and the utility loses both the energy sale and the network charge.
These two look similar in a press release and are opposite in their effect on the system. Wheeling keeps the fixed cost base spread across the large users. Islanding removes them from it entirely.
Which one a company chooses is not a matter of corporate values. It is decided almost entirely by two tariff parameters: the use of system charge levied on wheeled power, and the standby or capacity charge levied on a self supplier that still wants the grid available when the sun is not shining.
Set those charges too high and wheeling is uneconomic, so every project islands, which is the worst outcome for the system. Set them too low and wheeling becomes a mechanism for avoiding fixed cost contribution while still relying on the network, which produces the same hollowing more slowly. The correct level is a genuinely difficult regulatory question and it is being decided, right now, in tariff schedules that almost nobody outside the sector reads.
Eskom's own published tariff structure shows the utility working on precisely this. The 2026/27 schedule increases the fixed portion of the Generation Capacity Charge from 20 to 30 per cent, with the remaining 70 per cent recovered through the energy charge, and that portion explicitly excluded from the energy credit provided under wheeling and net billing. In plain terms: more cost recovered through fixed charges that a departing customer cannot avoid, and a deliberate reduction in how much of the fixed cost a wheeling customer can escape.
Whether that calibration is right is arguable. That it is the actual lever, rather than any statement about industrial responsibility, is not.
There is a serious case that industrial self supply is the best thing that has happened to South African electricity.
Private capital is building generation that a fiscally constrained state could not fund. Large industrial users acting as anchor offtakers make renewable projects bankable in a way that a distressed utility's power purchase agreement no longer does. Miners securing cheaper, cleaner power lower the carbon intensity of minerals the world needs, and preserve industrial capacity that high tariffs were destroying. On the last point the evidence is grim: South African ferrochrome has contracted severely under electricity costs, and the industry has moved towards exporting raw ore rather than processed metal, which is the value capture problem we described earlier this week arriving through the electricity bill.
An industrial base that leaves the grid is better than an industrial base that closes. That is a real argument and it should not be dismissed.
But it settles the wrong question. Nobody sensible opposes corporate renewable procurement. The question is whether the fixed costs of a national network, built over a century and still required by everybody who cannot self supply, are recovered from a shrinking base of captive customers or shared across the users who actually depend on the system. That is a distributional choice, and at present it is being made by default through the technical detail of a tariff schedule rather than deliberately through policy.
South Africa also signed two loan agreements worth around 405 million dollars with the New Development Bank this week for domestic infrastructure.
Set that against the three gigawatts of private mining generation and the shape of the problem becomes visible. Public and concessional capital is being borrowed to maintain a system that the most creditworthy users are leaving, and the repayment will be recovered from a customer base with a declining share of the country's most valuable load.
None of the parties in this is behaving unreasonably. The miners are minimising cost and risk. The utility is defending its revenue. The development bank is funding infrastructure that needs funding. The outcome is nonetheless a system in which the people least able to pay carry a rising share of a fixed cost base, and no single decision in the chain produced it.
Three things, none of which are currently measured in a form anybody can use.
The split between wheeled and islanded capacity, tracked over time. This is the single most informative statistic about the future of the system and it is not routinely published. Aggregate private generation figures conceal the distinction entirely.
The incidence of the tariff, by customer type. Who is actually paying for the fixed cost base now, and how that has changed over five years. Distributional analysis of a tariff is straightforward arithmetic and it is rarely done in public.
What happens to the customers who cannot leave. Not average tariffs but disconnection rates, arrears, illegal connection, meter bypass and self disconnection among prepaid users. These are the observable symptoms of a cost being shifted onto people who cannot absorb it, and they are the measurements that would tell a regulator whether the calibration is working. They require asking households rather than reading a billing system.
The technical debate about wheeling charges and standby tariffs is where this is being decided. The evidence about who bears the consequence sits somewhere else entirely, in energy access and in the households for whom a reliability question becomes a livelihood question.
If you are financing network infrastructure in a system where the largest customers are leaving, the incidence question is worth answering before the next tariff determination.
This is an independent insight piece by Transitions Lab. For the Lab's applied work, see Energy Access & Off-Grid Systems and Local Manufacturing & Supply Chains. See also Stacking, Not Switching on the same defection dynamic across the continent that South Africa can see and most other utilities cannot, Paying for Power You Curtail on how price signals redistribute cost in a grid, Behind the Border on the infrastructure the public purse is being asked to fund, Absorbing the Gap on households as the shock absorber of last resort, Who Pays Decides What Gets Built on the same connection question when a hyperscaler builds its own supply, and The Load That Grows When It Is Hot on the defection dynamic in reverse when a large temperature-sensitive customer arrives on terms that shift cost onto the customers who cannot leave. To discuss a study, see Contact.