Insight, Energy Access
Africa will install a record 17 GW of solar this year and three quarters of it is invisible to official statistics. We know about it because China publishes what it ships. That instrument works precisely as long as Africa fails to build its own panel industry.
Ember and the African Tech Futures Lab have published estimates showing that Africa will install a record 17 gigawatts of solar in 2026, up 45 per cent year on year and the third consecutive record, equivalent to around a hundred thousand panels a day. Thirty six of Africa's fifty four countries are expected to install record amounts, with nineteen countries growing more than 100 per cent year on year, including 544 per cent in the Democratic Republic of Congo, 282 per cent in Zimbabwe, 176 per cent in Egypt and 117 per cent in Zambia. South Africa's share of African installations will fall below a fifth for the first time since 2019.
This is not a marginal development and it should not be read as one. Senegal will add solar capacity between 2023 and 2026 equivalent to almost 80 per cent of its entire 2023 grid capacity, with the DRC and Kenya each adding more than half. The 2026 installations will generate around 23 terawatt hours a year, which is 2.3 per cent of Africa's electricity generation, slightly above the 2.2 per cent average annual demand growth of the previous decade.
The step change is clearest where Ember gives explicit year-on-year figures. In these four countries, 2025 annual installations approximately quadruple, triple or double in 2026.
| Country | 2025 solar additions (approx.) | 2026 solar additions (approx.) | 2025 to 2026 change |
|---|---|---|---|
| Congo (DRC) | ~200 MW | ~1,300 MW | +544% |
| Zimbabwe | ~105 MW | ~400 MW | +282% |
| Egypt | ~725 MW | ~2,000 MW | +176% |
| Zambia | ~275 MW | ~600 MW | +117% |
Year-on-year change is Ember's exact figure. Annual megawatt values are approximations read off Ember's chart above, given here to show the absolute scale that the percentage sits on top of.
The share-of-grid picture the record represents is worth naming directly. In ten countries, 2026 new solar adds more than a tenth of annual grid generation: Sierra Leone at 97 per cent, Togo at 24, Somalia and Djibouti at 21 each, DRC and Comoros at 14 each, Namibia and Liberia at 12 each, Chad at 11 and Lesotho at 10. Those ten countries are home to around 190 million people.
The analysis is careful, the methodology is transparent, and the finding is important. What follows is not a challenge to any of it. It is about three things the number cannot tell you, and the first concerns how it was produced.
Ember's estimates are not built from African statistics, because those largely do not exist. Official national solar capacity reporting could be found for only 36 of 54 countries, only 14 of those covered 2025, and only three, South Africa, Tunisia and Tanzania, publish solar data monthly or quarterly. Ember's 2025 estimate of 12.0 gigawatts is roughly double the International Energy Agency's 6.2 and more than double IRENA's 4.6.
The estimates are built instead from Chinese customs data, calibrated against the finding that around 73 per cent of Chinese exports are installed, with an average delay of about six months.
That is a resourceful and legitimate method, and it is the only one currently capable of seeing the phenomenon. It is also worth stating plainly what it means: the most consequential energy development on the African continent is visible primarily through the administrative records of the country selling the equipment.
The instrument works because 94 per cent of panels installed in Africa are still imported from China. It is an artefact of dependence. And it therefore has a property that deserves more attention than it has received: its accuracy degrades in direct proportion to Africa's success at building its own industry.
That is not hypothetical. African panel manufacturing is expected to quadruple to around 3.5 gigawatts in 2026, roughly a fifth of what the continent installs, led by new plants in Egypt and Tanzania. Ember also notes that large Chinese cell and wafer shipments to Africa appear to indicate re-shipment rather than panel manufacturing, which means the signal is already partly contaminated.
Follow that forward. A country that succeeds in manufacturing domestically stops importing finished panels, and disappears from the only dataset that could see its solar growth. A country that assembles panels from imported cells appears in the data as importing something that is not a panel. The better African manufacturing gets, the blinder the measurement becomes, and the gap will open first in exactly the countries with the most active industrial policy.
There is also a point worth making about who owns the evidence base. A continent whose energy transition is legible mainly through another government's trade statistics is in a weak position if those statistics become less detailed, less timely or less freely published. That is not a prediction. It is a dependency that currently has no backup.
The second thing the number cannot tell you is who is buying, and there is a strong clue in the economics.
Ember puts a year of Chinese panel imports at around 2.4 billion dollars, and notes that generating the same electricity from diesel would cost approximately that every three months. It also observes that Africa now imports more dollars of batteries than solar panels from China, led by Nigeria and the DRC.
Read those two facts together. Batteries are the tell. A household with no electricity connection and no generator does not begin its energy transition by buying a battery bank. A business that has been running a diesel generator for fifteen years does exactly that, because the generator is what the battery replaces.
The payback that is driving this boom is diesel payback. And diesel is owned by people who could already afford electricity, expensively: the factory, the cold store, the office block, the telecom tower, the clinic with a generator, the household in the suburb with a backup inverter.
That has a consequence which the phrase "record solar" tends to obscure. The solar boom and the electricity access gap are largely two different populations.
Ember's own decomposition supports this. Roughly 75 per cent of capacity added between 2023 and 2025 is distributed solar, meaning small scale, customer side installations, mostly rooftops, and only a quarter is explained by utility-scale and off-grid solar together, with off-grid growing fast but still small in this context. Customer side means there was a customer, with a roof, a business or a house, and usually an existing electricity bill or fuel bill to reduce.
None of this makes the boom less important. Reducing the cost of power for African industry and commerce is one of the more valuable things that could happen to the continent's economy, and it is happening faster than any policy achieved. It does mean that a number describing panels installed should not be used, without further work, to argue that the access problem is solving itself. Those are different claims requiring different evidence, and only one of them is supported here.
The third thing follows from the second, and it has a name in the energy access literature that deserves to be applied here.
Masera, Saatkamp and Kammen showed, against the then dominant energy ladder model, that households do not progress cleanly from one fuel to the next as income rises. They accumulate options and use several at once, choosing between them by cost, reliability and task. The pattern was named fuel stacking and it has held up across three decades of household energy research.
The extension we would make, and it is ours rather than theirs, is that firms stack supply the same way households stack fuels. A Nigerian factory with a grid connection, a diesel generator and a new rooftop array is not switching. It is holding a portfolio and dispatching from it: grid when it is available and cheap, solar in the middle of the day, battery through the evening peak, diesel in the rainy week and during the outage that lasts three days.
Two consequences follow, and both matter for anybody using these figures.
Capacity installed is not fossil fuel displaced. A megawatt of panels on a factory roof might displace 80 per cent of that site's diesel, or 30 per cent, depending on load shape, battery sizing and how much reliability the operator insists on holding in reserve. The 23 terawatt hour figure is a well founded estimate of what the panels can generate. It is not a measurement of what stopped being burned.
The generator does not leave. It is retained as insurance, which is rational, and its capital cost is already sunk. So the transition looks, from the site, like a large reduction in fuel purchases rather than a change of system. That is a good outcome. It is not the same outcome as displacement, and a plan that assumes the generator has gone will be wrong about emissions, about fuel imports and about resilience.
There is an unusually cheap test available here. Most of these countries publish refined product import volumes, and so do the exporters. Diesel imports set against estimated solar installations, country by country, over four years, would give a first empirical handle on how much of this capacity is displacing fuel and how much is being stacked on top of it. As far as we can establish, nobody has put those two series side by side. It would take one analyst a few weeks.
Joel Nana of the African Tech Futures Lab poses the question at the centre of this: whether these distributed assets remain a parallel power system compensating for grid failure, or get integrated in ways that produce wider benefits.
The reason that question is urgent is that utilities are forecasting demand against a base that is quietly eroding, and in most of these countries they cannot see it happening. We described the same dynamic in South Africa yesterday, where the departure of the largest and most creditworthy customers has been visible for years because the country has the reporting to see it. The rest of the continent is running the same process without the instruments.
The practical risks are specific. Thermal capacity procured under long term contracts against demand forecasts that do not include twenty gigawatts of invisible distributed solar becomes idle capacity that somebody still pays for, which is precisely the take-or-pay problem that has made Kenyan electricity expensive. Network investment gets sized for a load shape that no longer applies. And revenue shortfalls arrive as a surprise in the accounts rather than as a line in a forecast.
Distributed solar is not the cause of these problems. Invisibility is.
The fix is administrative rather than technical, and it is cheap. The panels already cross a border with a customs declaration attached, so national import registration by capacity is largely a matter of coding an existing form. Installer licensing and inverter registration at the point of connection cover most of the rest. Several African countries already have registration or permitting systems at an advanced stage.
The reason this has not happened is not incompetence, and it is worth naming honestly. Registration creates a legible object, and a legible object can be taxed, tariffed, levied or restricted. In several markets, part of why distributed solar has grown so fast is that it has grown outside any regime that could slow it down. A registration scheme designed as a revenue measure, or perceived as the first step towards one, will drive the market informal and destroy the data it was created to collect.
That is a real design problem and it is a social and political one rather than a statistical one. It is solved by separating the counting function from the revenue function institutionally and visibly, and by giving installers a reason to register that benefits them, such as warranty enforcement, access to finance or eligibility for grid connection. Which of those works is an empirical question about installer behaviour in a specific market, and it is answerable by asking installers.
It is the best available picture of something genuinely important, produced carefully, and it is far closer to the truth than the official statistics it corrects. Nobody should wait for better data before taking it seriously.
It is also a count of panels crossing a border, converted into capacity by a calibrated assumption, in a continent where the buyer, the prior situation and the displaced fuel are all unobserved. Three questions determine what it actually means, and none of them can be answered from customs records.
Who bought it. What they were doing before. What they still burn.
Those are field questions, and they are the ones we would ask. The reliability calculation a business or a household is actually making is the thing that decides whether this boom is an industrial cost reduction, an access transformation, or both in different places. It is not currently being measured anywhere, and it would not take much to start.
The Lab works on this in energy access, across the markets where the boom is fastest, and through field research designed to answer the questions a dataset structurally cannot.
If you are planning generation, financing a network, or writing an access strategy against these numbers, the three questions above are worth answering first.
This is an independent insight piece by Transitions Lab. For the Lab's applied work, see Energy Access & Off-Grid Systems. See also The Customers Who Can Leave on what happens when the largest customers exit the grid where the utility can see them, Paying for Power You Curtail on the take-or-pay bill that outlives the demand forecast, and Absorbing the Gap on the reliability calculation a firm or household is actually making. To discuss a study, see Contact.