Unilever's new array covers about thirty per cent of its Nairobi factory's electricity. That figure is not caution, it is what Kenya's net-metering rules make worth building, and it shows where the value in commercial solar actually accrues.
Unilever spent about KSh70 million on an 800-kilowatt solar system at its Nairobi manufacturing plant, which is expected to supply roughly thirty per cent of the factory's electricity and to cut its energy bill by around KSh30 million a year. The plant's monthly electricity spend falls from about KSh12 million to between KSh7 million and KSh8 million. The boilers moved from heavy fuel oil to biomass at the same time, and the company puts the combined reduction in the factory's emissions at about forty per cent against a 2023 baseline.
On the savings as reported, the simple capital payback is a little over two years, before financing costs. That number will be quoted widely this week as evidence that industrial decarbonisation in East Africa has become commercially obvious.
It is the wrong number to study. The interesting one is thirty per cent.
A factory roof in Nairobi can usually carry several times 800 kilowatts. Solar resource is not the constraint, roof area is rarely the constraint, and at a payback of two years capital is not obviously the constraint either. If the economics were as good as the headline implies, the rational system would be much larger.
It is not larger because of what happens at the meter. Kenya's Energy (Net-Metering) Regulations, 2024 admit renewable systems below 1 MW, with commercial and industrial installations permitted up to that ceiling, but they price exported electricity at a credit worth fifty per cent of each unit exported. Credits may be carried forward within the licensee's financial year and are then forfeited. In September 2026 the regulator went further and amended Kenya Power's tariff schedule so that unauthorised injection into the distribution network, described as "dumping", is measured and charged at the applicable base tariff.
Read those three provisions together and the design problem changes shape. An exported kilowatt-hour is worth half an imported one, and only if the same site later draws enough grid power inside the same financial year to absorb the credit. Export beyond that is worth nothing, and export without an approved agreement is worth less than nothing. The asset is therefore not sized to the roof, nor to annual consumption, nor even to the daytime load. It is sized to the floor of the daytime load, the quantity the factory is confident of consuming at the moment the panels are producing it, on the worst plausible production day.
Thirty per cent is not timidity. It is the regulation, expressed in kilowatts.
This is where the usual framing of distributed solar, as a technology becoming cheap, stops being useful. Teece's 1986 account of profiting from technological innovation90027-2) asks two questions about any innovation: how well can the innovator appropriate the returns, and who controls the complementary assets needed to deliver it. Where the appropriability regime is weak and the complementary assets are specialised, the returns accrue to whoever holds the complementary assets rather than to whoever holds the technology.
Commercial and industrial solar in Kenya is close to the limiting case. The appropriability regime for a module is effectively nil: panels are a traded commodity, and the inverter and mounting market is competitive. Orb Energy manufactures its own panels in Bengaluru, which is a cost position rather than a moat. What is scarce is everything around the panel: the credit that lets a mid-sized processor take the asset without the capital, the approval and interconnection agreement, the metering arrangement, the maintenance presence that makes a fifteen-year claim credible, and the load data needed to size the system to that daytime floor rather than to the roof.
Seen that way, the $2 million KawiSafi Ventures has put into Orb Energy's African operations is not money for panels. Orb has installed close to 15 MW of rooftop capacity and serves more than sixty commercial and industrial customers in East Africa, and it sells systems either outright or through financing structures that reduce the customer's upfront requirement. The capital is buying the capacity to carry customer paper. That is the complementary asset, and it is the part of the stack that does not arrive by container.
Kenyan industrial tariffs are estimated at $0.18 to $0.23 per kWh. The comparators in the same analysis are Uganda at about $0.125, Tanzania at $0.08 to $0.09, South Africa at $0.03 to $0.085, Egypt at $0.02 to $0.03 and Ethiopia at $0.01 to $0.05. The Kenya Association of Manufacturers, whose Regulatory Audit Report produced those figures, also records manufacturing at 7.6 per cent of GDP in 2024 against an average near ten per cent between 2011 and 2016, and concludes that "instead of industrialising, Kenya is deindustrialising."
Against a tariff of twenty cents, a factory that self-generates a third of its load is not buying an environmental attribute. It is buying back about a third of a cost line on which it is losing to Tanzania and Egypt. That is why the Unilever installation belongs in an industrial-policy conversation rather than a sustainability report, and why the list of Kenyan firms doing the same, Bamburi Cement, Mabati Rolling, Bidco Africa, British American Tobacco and Carbacid among them, reads like a list of the country's largest electricity bills rather than its most ambitious climate strategies.
The wider evidence on industrial self-generation supports reading it as a defensive purchase. About 53.3 per cent of firms in sub-Saharan Africa self-generate, the highest share in the world, and roughly 29.4 per cent of the electricity firms use in the region comes from their own plant. Firms in the region report losing about 8.4 per cent of annual sales to outages, against a global average of 5.2 per cent. Diesel self-generation has always been a hedge that a firm pays for twice, in capital and in fuel. Behind-the-meter solar is the first version of that hedge whose running cost falls rather than rises, which is a change in kind and not only in degree.
The arithmetic that makes sense inside the factory shows up on the other side of the meter as a revenue problem. Commercial and industrial self-generation in Kenya reached 676.6 MW after businesses added 72.8 MW in the first half of 2026, against captive capacity of about 575 MW in December 2024. Over the year to June 2026, Kenya Power's sales to new customers fell by KSh1.07 billion, with consumption by new users down twenty per cent to 161.7 GWh.
The utility's stated concern is technical rather than commercial. Kenya Power has reported that variable renewables already supply about 34 per cent of the energy mix at daytime peak, and argues that approval and visibility over connection points, metering and equipment are what let it manage frequency, voltage and the safety assumptions its technicians work under. Both things are true at once, and that is precisely the difficulty: the rule that protects the grid is also the rule that caps the asset, and the regulator has no instrument that separates the two.
The published numbers describe an intention. Four things decide whether it holds, and none of them are visible from Nairobi's business pages.
Whether the saving survives the second year. Self-consumption ratios drift as production schedules change. A system sized to a 2026 load floor can be oversized by 2028 if a shift pattern moves, and the surplus is then worth half, or nothing.
Who carried the credit risk, and at what price. A two-year payback on capital is not a two-year payback on a financed asset. The spread between those two numbers is where the complementary-asset argument is settled empirically.
How long the approval took. If interconnection approval takes longer than procurement, the regulatory interface is the real lead time in the project, and it should be reported as such.
What the plant manager would do differently. Load floors are known in the maintenance office long before they are known in a feasibility model. Asking is cheaper than modelling, and usually more accurate.
For manufacturers, the practical conclusion is narrow and useful: before sizing an array, read the export and forfeiture terms and the current tariff schedule, then size to the load you can prove you consume. For investors, the conclusion is that the scarce thing in this market is not generation, it is the bundle of credit, approval and service around it. For regulators, the conclusion is that the rule written to protect the grid is currently also the rule setting the ceiling on industrial self-supply, and that those two jobs deserve separate instruments.
If this reads close to a decision you are weighing, we would like to hear about it. A short note is enough to start.
The Lab is an independent research team, not an advocacy organisation. We do not campaign, endorse products, or accept commissions where a finding is expected in advance. Where a partner's own theory of change is what needs testing, that is a legitimate engagement, and the Lab tests it honestly.