Uganda's electricity is cheap. Its interruptions are not — and nothing on your bill says so.
Most conversations about Ugandan industrial power are about price. They are aimed at the wrong number. Uganda has surplus generation and some of the lowest industrial tariffs in the region. What it has instead is instability — and the distributor has now said so itself, in Parliament. The cost of that instability lands in scrapped batches, restart time, damaged drives and missed delivery dates, and no invoice you receive contains any of it. This page is about how to put a number on it.
Not a shortage. A reliability problem.
There is enough power
Uganda's reliable capacity sits near 1,550 MW against a February 2026 system peak of about 1,337 MW. In the same month the country exported some 270 MW to Kenya, Rwanda and Tanzania, and Kenya has asked for more. Karuma's 600 MW came fully online in 2024. This is not a country rationing scarce electricity.
And it is inexpensive
ERA's approved tariff for July–September 2026 is UGX 308.1/kWh for large industrial and UGX 207.7/kWh for extra-large, against 363.8 for medium industrial and 562.1 commercial. By regional standards those are low industrial rates. ERA resets them quarterly, so check the current schedule before building a case on them.
What the distributor told Parliament. In August 2026, UEDCL's acting managing director told the Public Accounts Committee (Central) that since UEDCL took over distribution from UMEME, the number of outages had "almost doubled" and the time to restore supply had risen from about 12 hours to nearly 20. Government has put roughly US$129 million into the network across the 2025 and 2026 financial years. This is not a complaint from a manufacturers' lobby — it is the utility's own evidence, on the record, three weeks before this page was written.
Independently, the Economic Policy Research Centre's Business Climate Index for April–June 2025 surveyed 1,152 firms and found about four in ten reporting that the distribution handover had hurt their operations, with more frequent outages, unstable voltage and degraded power quality the leading complaints. UEDCL's own reported outage rate of "about 2%" still works out at roughly 15 hours a month.
An interruption is the one event that erases its own evidence.
A tariff increase is easy to act on because it arrives as a number, on paper, monthly. An interruption arrives as an absence. Consider what actually happens on a Ugandan plant when the feeder drops at 14:40 on a Thursday:
- The line stops. Nobody records the time, because everyone is dealing with the stop.
- The genset starts, or it doesn't. If it does, the changeover gap is unrecorded. If it doesn't, the reason is discovered later and attributed to something.
- Work in progress is lost — a batch that was mid-cook, a mould that set, a run that has to be re-sequenced. It is written off as production variance.
- Restart takes time nobody counts, because restart is normal work.
- The drives and motors take a voltage event. It doesn't fail today. It fails in eleven months, and the failure is recorded as a bearing.
- An order ships late. That is recorded — as a logistics problem.
By the end of the month every one of those costs has been absorbed into some other category. The electricity bill, meanwhile, is lower than it would have been, because you consumed less grid power. The single event that cost you the most all month is the one that made your energy costs look better.
Six components, and you already know four of them.
You do not need a study. You need one line instrumented and a month of data. The cost of an interruption on a manufacturing site decomposes into six parts, and most plants can estimate four of them today:
- Lost production time = outage duration + changeover gap + restart ramp. The last two are usually larger than the first and almost never counted. Multiply by your contribution margin per hour on that line — not revenue, margin.
- Scrap and rework — the work in progress that cannot be recovered. Highly process-dependent: near zero on a discrete assembly line, very high on anything with a thermal or fermentation profile.
- Standby fuel and wear — litres burned per outage hour, plus service intervals that are consumed in hours rather than in months. Diesel at generator-hour rates is the visible part; the maintenance schedule you are pulling forward is not.
- Equipment life — voltage events accumulate against motors, drives and control electronics. This is real, it is hard to attribute, and it is the reason plants with unstable supply replace VSDs sooner than the datasheet says.
- Labour paid for output not produced — a shift on site during an outage still costs a shift.
- Delivery reliability — the hardest to quantify and often the largest in the long run, because it shows up as a customer who stops calling rather than as a line in a ledger.
Instrumenting the line gives you components 1, 3 and 5 automatically and to the minute. Component 2 follows from the production record. Components 4 and 6 stay estimates — but they are estimates anchored to a measured event count instead of to an impression.
Why we don't publish a shilling figure here. Any per-outage cost we quoted would be a number from someone else's factory. The whole argument of this page is that the figure is specific to your line, your margin and your process — and that it is measurable rather than assumable. A supplier who tells you what an outage costs your plant before seeing your plant is describing their own marketing, not your operation.
The event, the response, and the consequence.
- Grid presence and quality at the intake. Every interruption timestamped with a start and an end, plus voltage and frequency behaviour around it — because "unstable voltage" was among the leading complaints in the EPRC survey and it damages plant without ever fully interrupting it.
- Changeover behaviour. How long the gap was, whether the genset started, whether it started first time, and how long it took to accept load. Start failures are found in the record rather than in the next outage.
- Generator run hours and fuel. Litres per outage hour and per kWh, which is what turns standby cost from an invoice total into a rate you can act on — and it makes fuel loss visible, which on many sites pays for the monitoring on its own.
- Production stop and restart, per line. Linked to the grid event by timestamp, so an interruption's production consequence is attached to the interruption rather than sitting in a separate report.
- Energy sub-metering. Consumption per line and department, giving you cost per tonne and separating grid kWh from generator kWh — which are very different costs wearing the same unit.
- A record that survives the outage. Nodes are battery-buffered and log locally. Monitoring that dies with the supply produces a blank at precisely the moment worth examining.
The six components of an interruption, and the measurement for each
| Component | How it is usually estimated | How it is measured |
|---|---|---|
| Lost production during the outage | "About two hours" | Mains presence + counts: exact minutes, exact units |
| Restart and ramp-up | Forgotten | Time from power return to rated speed, per event |
| Scrap and rework | Guessed at month end | Reject counts in the restart window |
| Generator fuel and wear | A monthly fuel bill | Litres and run-hours per outage |
| Damaged equipment and voltage events | Discovered later | Voltage log at the incomer |
| Overtime and missed deliveries | In the payroll, unattributed | Events matched to shift records |
Context: reliable capacity ~1,550 MW against a 1,337 MW peak — Uganda is not short of power; the constraint is distribution, where UEDCL itself told Parliament outages "almost doubled" and restoration lengthened to ~20 hours after the handover.
Three conversations you can't have without it.
With UEDCL
An independent, timestamped log of interruptions on your feeder — count, duration and restoration time — is a materially different basis for a service conversation than a recollection that it happens a lot. It is also the evidence base if you are pursuing a dedicated feeder or an upgrade.
With your own board
Standby generation, a dedicated supply, solar, storage — each is a capital decision that currently gets made on impression. A measured annual interruption cost turns the question from "should we spend on power resilience" into "which option pays back, and in how long".
With your customers
Delivery reliability is easier to defend, and easier to price, when you can show what caused a late shipment and what you have done about it. Uncertainty is what customers charge for.
There is also a regulatory door worth knowing about. The Electricity (Supply of Electricity in Bulk to Specified Consumers) Regulations, 2022 allow qualifying consumers to buy power in bulk. Legal commentary puts the eligibility threshold at an average demand of 1,500 kVA, with a grid-connected consumer able to source up to 30% of its energy directly, and a consumer not connected to the grid up to 100%. We have not read the statutory instrument itself, so treat those thresholds as a law firm's reading rather than as verified text and confirm with ERA. What is not in doubt is that any such arrangement rests on a credible demand profile — which is something you either measure or estimate.
Power interruptions in Uganda — common questions
Is Uganda short of electricity?
No. Reliable capacity is around 1,550 MW against a February 2026 system peak of roughly 1,337 MW, and Uganda exported about 270 MW to Kenya, Rwanda and Tanzania in the same month, with Kenya seeking more. Karuma's 600 MW came fully online in 2024. The constraint is in transmission and distribution reliability, not in generation.
Have outages got worse since UEDCL took over?
According to UEDCL. In August 2026 its acting managing director told Parliament's Public Accounts Committee (Central) that the number of outages had almost doubled since the changeover from UMEME, and that restoration time had risen from about 12 hours to nearly 20. Separately, EPRC's Business Climate Index for April–June 2025, covering 1,152 firms, found around four in ten reporting operational harm from the transition, with more outages, unstable voltage and degraded power quality the leading issues.
What is the industrial electricity tariff in Uganda?
ERA's approved rates for July–September 2026 are UGX 308.1/kWh for large industrial and UGX 207.7/kWh for extra-large industrial, with medium industrial at 363.8 and commercial at 562.1. ERA reviews and republishes tariffs quarterly, so verify the current schedule before relying on these figures for a business case.
What does an outage cost our factory?
It depends on your contribution margin per hour on the affected line, your process's tolerance for interruption, your restart profile and your standby fuel rate — which is why a figure quoted without seeing your plant is worth nothing. What we can do is measure the components: interruption duration, changeover gap, restart ramp, generator hours and fuel, and the production actually lost. One line and one month usually settles the argument in either direction.
We already have generators. What does monitoring add?
Three things a generator cannot tell you on its own: whether it starts reliably, discovered in testing rather than during the next outage; what each hour actually costs in litres per kWh, which is the number that decides what stays on standby; and how much production the interruption cost regardless of whether the generator caught it, since the changeover gap and the restart ramp happen either way. Fuel visibility alone pays for the instrumentation on a surprising number of sites.
Can voltage instability really damage equipment?
Yes, and it is the part of the problem that hides best because nothing stops. Sags, swells and repeated restarts accumulate stress on motors, variable-speed drives and control electronics, and the eventual failure is usually recorded as a component fault months later. Logging supply quality at the intake is what lets you connect a pattern of failures to a pattern of events instead of treating each one as bad luck.
What does this cost to set up?
Determined by how many intake points and lines you want covered, whether your switchgear and gensets can be read electronically, and how much of your production data already exists in a PLC. We quote a written cost band after a remote assessment — panel photographs and a genset nameplate are usually enough to start. Sites typically begin with the intake and one line, then extend once the first month's data has made the case.
Stop estimating what the grid costs you.
Tell an engineer what you run and how often you think you lose supply. One month of data usually replaces the argument with a number.