Why does Pakistan still run out of power?
On paper, the country has almost twice the power plants it needs. The lights still go out, on bad evenings for hours. Here is where it all goes, in six short chapters.
0MW
power plants, installed
0MW
most power ever needed at once
Chapter 1
Where the power goes
Most of that power can't show up when it matters. Step through the busiest evening of the year, 11pm on 20 June 2025, and watch the usable power shrink. Every drop is the model's own arithmetic.
48,316MW left
Start with everything
48,316 MW of power plants, every one in the country. The most power ever needed at once was 29,201 MW. On paper, that is plenty.
From the engine's per-fuel availability factors on the mapped fleet; the solar zero is measured (NPCC, all top-10 FY25 demand hours). Steps add up exactly to the model's 27,528 MW available.
Chapter 2
Paid even when switched off
Most power plants here are paid just for being available, even when they are switched off. These are called capacity payments, and you pay them in every bill. That is why the idle oil plants from Chapter 1 still cost money.
Then the money runs out. Distribution companies don't collect enough (Chapter 4 shows why), so they can't fully pay the power buyer, who can't pay the generators, who can't buy fuel, so plants that could run sometimes don't. The pile-up is called circular debt.
People
pay bills, or don't
DISCOs
collect too little
Power buyer
can't pay in full
Generators
invoices pile up
Fuel suppliers
want cash up front
Money should flow left to right, and power back the other way. The first arrow is where it breaks.
Rs 0bn
circular debt at the end of 2023-24
Rs 0bn
at the end of 2024-25, after a big government paydown
Rs 0bn
added to the debt in one year by grid losses alone
Debt figures: NEPRA State of Industry Report 2025. Loss contribution: Pakistan Electricity Review 2025, p.19. The reports don't give one clean figure for yearly capacity payments, so this page doesn't quote one.
Chapter 3
The power is in the wrong place
Much of the power is made in the south: Thar coal, Karachi's plants, the wind corridor. But most of the people, and most of the demand, are up north in Punjab. Switch between the two maps and you can see the gap.
Loading map…
But the wires aren't the problem
You'd expect the long haul from south to north to be the bottleneck. It isn't, at least not in the evening. This project measured the lines between provinces against the power that needs to move, and the busiest one runs at just 24.7% of what it can carry. The rest are lower. At the peak, the problem isn't moving the power. It's that there isn't enough power to move.
Model finding: required transfers (zonal balance) vs measured PSS/E line limits, FY25 evening peak.
Chapter 4
Power no one pays for
Between the power station and a paying customer, electricity leaks away: old wires, theft, and bills that never get collected. How much depends a lot on where you live. These are the measured loss rates for every distribution company.
Losses measured (NEPRA SIR 2025). The what-if uses those rates and each company's reported sales: bringing SEPCO, QESCO, PESCO and HESCO down to IESCO's 8.6% saves the energy shown, roughly half the evening gap, without building anything.
Chapter 5
The problem is at night
Rooftop solar now quietly covers the daytime, so the hours the grid strains hardest have slid into the evening. 47 of the year's 50 busiest hours fall between 7pm and 1am. Here is the peak day itself: national demand on top, and underneath, to the same scale, everything the grid's solar made.
Measured: NPCC hourly demand and grid-solar output, 20 June 2025 (the FY25 peak day). Rooftop solar is invisible to the grid operator and shows up here only as lower daytime demand.
So adding more solar does nothing for the evening shortfall, because the shortfall starts after sunset. Only storage or firm power moves it. Don't take this page's word for it.
Try it in the simulatorChapter 6
When the whole grid fails
The nationwide blackouts of January 2021 and January 2023 were a different problem from the evening squeeze. They happened in winter, at low demand, when the grid had plenty of power but lost its balance: the system split apart and shut itself down within minutes. A different failure, with a different fix, and outside what this project's model covers.
The 2023 blackout shows the imbalance best. Breaking took about four minutes. Putting the country back together took more than twenty hours.
Both bars to the same time scale. Source: NEPRA inquiry report, 23 January 2023.
Figures on this page: engine outputs and availability factors from this project's model; measured series from NPCC; losses, circular debt and transformer loading from NEPRA's State of Industry Report 2025; loss cost from the Pakistan Electricity Review 2025. Where a report does not give a number, this page says so instead of guessing one.