On a sunny afternoon, a Lahore household with newly registered rooftop panels can use a unit of electricity to run a fan or sell it to the grid. Running the fan saves the price of a unit the family would otherwise have bought. Sending it down the wires earns a smaller credit on its bill.
In Mehar, in Sindh’s Dadu district, a panel answers a more basic question. In the 2020 village census, 44 of Mehar’s 75 villages said electricity reached only some residents. Three years later, about 28,000 households in Mehar, more than a quarter of the total, named solar panels as their main source of light.
For the electricity company, it matters what the family would otherwise have done. A panel can replace a sale. It can also light a home the wires never reached, or keep the lights on through a power cut. The revenue at stake is what the family would otherwise have bought and paid for.
Pakistan’s ten state electricity companies sold 11 per cent less power in 2024–25 than three years earlier. Over the same period, the amount generators charged for keeping plants available rose by 86 per cent. The dispute over what to pay rooftop owners turns on who should carry that bill. There is another question for the communities being asked to pay. What service did they get before they bought a panel?
Why the bill survives a fall in sales
Pakistan pays power plants for two things. It pays for the electricity they generate. Under long-term contracts, it also pays to have plants ready when people need them. That second payment, called a capacity charge, can fall due even when a plant stands idle.
When customers buy less grid electricity, energy costs can fall, but existing capacity and network commitments still need financing. Solar supplying previously unmet demand need not reduce grid sales at all. The schematic explains the financing mechanism; it does not measure the net cost of solar to the system or who bears it.
A household pays its local electricity company. The company buys power through the central purchasing system, which settles the generators’ invoices. Federal subsidies help companies charge a common national tariff even when their costs differ. If a rooftop supplies the family fan, it can save the system the cost of generating that electricity. The contract that keeps a plant available for the evening still has to be honoured.
Between 2021–22 and 2024–25, the ten state companies sold about 11.6 billion fewer units of electricity a year. Farms alone bought about 5.1 billion fewer units. Agriculture accounted for roughly a tenth of sales at the start of the period and 44 per cent of the fall. Those figures give the tube well a much larger place in the story than its original share of the electricity market would suggest.
The Islamabad company approved more solar export connections per customer than any other state company over those three years. Its sales fell by less than 1 per cent. In the Quetta company’s territory, where few systems were registered, sales fell by 36 per cent.
Farm sales in the Multan and Quetta companies’ territories each fell by about a billion units in 2024–25 alone. A farmer who replaces an electric pump with a solar one needs less grid power. A farmer whose electric pump stands idle during a cut also buys fewer units. Both appear in the accounts as a loss of sales. Industry slowing down and changes in billing can also move the total.
At the beginning of those three years, generators invoiced about Rs 970 billion in capacity charges. By 2024–25 the figure was Rs 1.8 trillion. For every Rs 100 on generators’ invoices that year, roughly Rs 61 went towards keeping plants available. These invoices cover the central purchasing system, including power it supplies to Karachi’s K-Electric. The sales figures cover the ten state companies.
Much of the capacity bill came from contracts signed before the import boom. It peaked at Rs 1.9 trillion in 2023–24, then eased after a round of renegotiations. The Central Power Purchasing Agency reports a further fall to Rs 1.71 trillion in 2025–26. Even then, the annual charge remained far above what it had been when grid sales peaked.
A given bill is harder to cover when sales fall. But the bill itself can change. The ten state companies were charged Rs 13.73 for capacity per unit they bought in 2024–25, down from Rs 14.49 a year earlier. Some of the cost reaches households through tariffs. Some falls on public budgets through subsidies.
Where customers register to sell power back
Look first at the areas where people register to sell electricity. The Islamabad company approved about 17 connections for every 1,000 customers between 2022–23 and 2024–25. Lahore approved 15 and Gujranwala 13. The figures include businesses as well as homes. Researchers Wajid Islam, Jan Rosenow and Asha Amirali have drawn attention to what these areas have in common. Their electricity companies collect a high share of what they bill.
Why should the pattern vary under a common tariff? Two families on the same street can pay different rates for their next unit because the schedule charges according to consumption. The value of a panel also depends on when someone is home to use its electricity, how often the power goes out and how much it costs the family to finance the installation.
Across the five state companies serving Punjab and Islamabad, annual receipts from households came to roughly Rs 101 to Rs 103 for every Rs 100 billed in 2024–25. Companies can collect old debts alongside current bills. In upper Sindh, the Sukkur company collected about Rs 40 for every Rs 100 billed to households. Service differed too. Sukkur reported an average of 1.4 hours of power cuts a day. At the locations it monitored in the company’s area, the regulator observed more than eight.
Now ask households what lights their homes. In the Quetta and Sukkur companies’ territories, about one household in four named solar panels in the 2023 census. In the Lahore company’s territory, fewer than one in a hundred did. A household can run appliances on rooftop power by day and still use the grid for lighting at night. Another can rely on a small panel for light and have nothing to sell back.
The grid has more income at stake when a customer who pays in full starts buying fewer units. Company averages show where collections have been strongest. They cannot identify which families cut their purchases.
Consider the company serving Khyber Pakhtunkhwa’s merged districts. Half the households in its territory named panels as their main light, the highest share among the companies. Yet it collected Rs 87 for every Rs 100 billed to households and had the fewest new solar export approvals per customer. Families relied heavily on solar lighting while the company continued to collect much of what it billed.
Where people rely on solar for light
The 2020 village census lets us look back to the years before panel imports took off. It asked how many residents had electricity in each village. The 2023 population census asked households what they used for lighting. Read together, the answers show where households relied on solar lighting in places the grid had served unevenly. They cannot tell us when a particular family bought a panel or whether it also had a meter.
In Mehar, all 75 villages reported that at least some residents had electricity. A map that marks a village as electrified wherever anyone has a connection would colour in the whole tehsil. Yet only 14 villages said electricity was available to everyone. Asked what other sources they used in 2020, 54 villages already reported solar. Many had experience with panels well before the import surge.
Mehar is one of 105 tehsils, or areas within districts, where at least half the villages had reported electricity for only some residents or none, and at least a quarter of households later named solar as their main light. About 900,000 households in those areas reported solar lighting. Most of the 105 tehsils are in Balochistan. Others lie in upper Sindh and parts of Khyber Pakhtunkhwa.
On Balochistan’s Makran coast, the picture changes. In Ormara, 17 of the 18 villages had reported electricity for some residents or none in 2020. Solar was already in use in 15 villages. By 2023, however, only 15 per cent of households in the tehsil named it as their main light. A technology can be present in most villages and still be the main source of lighting for relatively few families. The village returns count settlements of different sizes. The household census counts families in town as well as in the surrounding villages.
Across Pakistan, solar lighting was generally more common in tehsils where the grid had reached fewer people. That relationship was much weaker within Balochistan and barely visible within Sindh. Poor access alone cannot explain why some places came to rely on panels more than others.
In Bara, in Khyber district, 43 of the 64 villages said in 2020 that none of their residents had electricity. Solar was already in use somewhere in 56 villages. Three years later, nearly 44,000 of Bara’s 76,000 households named a panel as their main light. A service plan for Bara would start with thousands of homes that already have a way to light a room. What else those families can power, and for how many hours, remains an open question.
What a household bill can tell us
The national household spending survey gives families an extra option when it asks how they light their homes. They can say they use both the grid and solar. Who appears to rely on panels depends on which answer you look at. In 2024–25, 17 per cent of households in the lowest-spending tenth of the population reported solar as their only source of light. In the highest-spending tenth, just 2 per cent did. The pattern ran the other way for the combination of grid and solar, which was about twice as common at the top as at the bottom. Among families in the lowest-spending tenth, nearly 18 per cent reported using neither grid electricity nor solar for light. Even in a country importing panels on this scale, many of the least well-off households were using other sources.
What showed up on their electricity bills? Across the survey, the median bill for the most recent month was Rs 3,000 among households that used only grid electricity for lighting. It was Rs 1,800 among those that used grid electricity and solar together. A difference remained when households with broadly similar living standards were compared. Within the lowest-spending tenth, for instance, the corresponding bills were Rs 1,500 and Rs 1,000.
It is tempting to read the Rs 1,200 gap in the national figures as the monthly saving from a panel. But the survey visited different households, and the two groups live in different places. Consider families in the top three spending tenths. Nearly seven in ten of those using only grid lighting lived in Punjab, against fewer than two in ten of those using both sources. The hours of supply, the appliances a family owns and the amount of electricity it wants can all affect a bill. To estimate the saving on the Lahore fan, we would want to see what happened to the same home’s purchases after its panels went up, alongside homes facing similar changes in prices and supply.
Who pays for the grid
An electricity company can have trouble collecting its bills for very different reasons. Take the Quetta company in 2024–25. For every Rs 100 it billed to industry, it collected Rs 98. For every Rs 100 billed to households, it collected Rs 65. For farms, the figure was Rs 8.
Farm bills alone came to almost Rs 81 billion, against receipts of about Rs 6.5 billion. Across all customers, the Quetta company collected Rs 39 per Rs 100 billed. Anyone applying that figure to homes would conclude that most of the money billed to households went uncollected. In fact, household receipts covered nearly two-thirds of billing. Irrigation has a large bearing on the figure usually quoted for the whole company.
The governments in Islamabad and Quetta had already agreed to convert 27,437 registered agricultural tube-well connections in Balochistan to solar. That is a different transaction from a family buying rooftop panels and asking to sell spare units. The public authorities themselves have chosen to move some pumping away from the grid. The electricity accounts do not record which pumps had been converted when each bill was issued, or how much of the gap between billing and receipts arose under the programme.
In Sindh, the same exercise points elsewhere. For every Rs 100 the Hyderabad company billed to industry it received about Rs 100. From domestic bills it received Rs 48. In the Sukkur company’s area the figures were Rs 97 and Rs 40. Here an inquiry into the collection gap would start with household billing and service.
Companies also receive different shares of the national capacity bill. Their share depends on how much power they draw when demand is at its peak. In 2024–25, the company serving Khyber Pakhtunkhwa’s merged districts was charged nearly Rs 24 for capacity per unit it bought, about twice the Islamabad company’s Rs 12. These are costs to the suppliers. Households face a common national tariff schedule, under which bills vary with customer category and consumption. Moving a household onto solar does not translate into a fixed number of rupees passed to its neighbours.
A house can meet its needs at noon and draw from the grid when the family returns in the evening. If its demand at the busiest hour stays the same, the supplier still needs enough plant and wires to serve it then. A customer who cuts purchases at that hour can help relieve the pressure. Annual sales totals treat a unit saved at noon and one saved at peak time alike.
How much solar is there?
Panel imports give one indication of the scale. TransitionZero estimates that Pakistan had imported about 50 gigawatts of panels by August 2025. Separately, Renewables First counts 6.8 gigawatts registered to sell electricity to the grid by the end of June. Imports can include panels still in stock as well as those destined for large power plants. Subtracting the registered total from imports cannot tell us how much solar people were using at home or on farms.
Satellites offer another view. TransitionZero estimates that panels on homes, businesses and farmland across Pakistan had a combined capacity of 27.2 gigawatts. Open Energy Transition’s earthpv atlas, using later imagery, puts the figure at 34.5 gigawatts. Despite the gap, both put Punjab first for total capacity and Khyber Pakhtunkhwa first among the four provinces for capacity per household. Both also find fewer watts per household in places where the earlier village census reported limited electricity, even though many homes there rely on solar lighting. A small panel can supply a family’s lights and be hard to pick out from space. A large commercial roof can add substantial capacity to an area where few households name solar as their main light.
In the Peshawar company’s territory, new export approvals were modest beside Lahore’s. Yet on estimated solar capacity per household, Peshawar ranks first in one satellite model and second in the other. An average for a company’s territory counts panels on factory roofs as well as homes. A district with a few large installations can rank above one where thousands of families have enough power for a few lights.
Even the national estimate depends on how researchers recognise a panel. Until late October, the earthpv atlas put the total at about 24 gigawatts. Its estimate rose to about 35 when its authors changed how they counted smaller roofs the model identified with varying degrees of confidence. That jump came from a new reading of the images. The panels being reconsidered were already in the images.
What changes on an electricity bill
For a Lahore household signing a new agreement, the February 2026 rules changed what happens to electricity it sends back. Under the old system, exported units could offset units bought from the grid. Under the new one, the company credits exports at a separate price, about Rs 11 per unit when the rules took effect, and charges the applicable retail rate for electricity bought. A family therefore gets more value from a unit it can use while the sun is shining. A washing machine that can run in the afternoon offers another way to use that electricity at home. A battery can move some of it to the evening, but the family has to cover the price of the battery and loses some energy in charging and discharging it.
The date of the agreement matters. Someone with a valid agreement under the previous rules keeps the old rate and billing terms until it expires, unless a material change to the installation alters its maximum electrical output. Two neighbours with panels of the same size can therefore receive different credits for power they send down the same street.
Other changes reach customers with or without panels. In February the government extended monthly fixed charges further down the household tariff schedule. For many customers, the charge now depends on the size of the connection, so buying fewer units does not remove it. A further change is planned. In its May 2026 review with the IMF, the government set a January 2027 target for channelling electricity subsidies to low-income households through the Benazir Income Support Programme.
Each measure reaches people through a different route. The export price matters to someone with an agreement to sell. A fixed charge can reach a pensioner who has never owned a panel. A family relying on a stand-alone system may have no electricity account at all. Changing a line on the bill has little to offer that family until it can get a connection and count on the supply.
The sales figures cannot yet tell us how much of the decline solar caused or who will ultimately pay more to keep power plants available. To answer that, we would need to know which customers added panels, how much electricity they then bought, and how their bills changed.
The new rules tell a Lahore household what it will earn for selling spare electricity on a sunny afternoon. They leave a bigger question open. As customers buy less power from the grid, who should pay to keep power plants ready for the evening?
A fair answer has to take account of the service families can count on in return. In Mehar, 44 of 75 villages reported that electricity reached only some residents. For a family there, the value of keeping a power plant ready is being able to switch on a fan after sunset and know it will run.
Company accounts. The sales comparison covers the ten state distribution companies and excludes K-Electric in Karachi. The generator invoices cover the wider central purchasing system, which also supplies K-Electric. The cross-company charts show K-Electric separately. All rupee figures are in the prices of the year shown. Annual cash receipts may include old debts. A collection percentage measures money received against money billed over the year. Solar registrations are new approvals across all customer categories over three years, expressed per 1,000 customers. NEPRA industry reports · NEPRA performance reports · Central Power Purchasing Agency, 2025–26, section 3.1.1
Village access and household lighting. The 2020 Mouza Census counted rural villages. The 2023 census lighting question covered both rural and urban households. For the national comparison we use 397 tehsils with their own figures from both censuses. For the map we also show places where a figure for the parent area or district had to stand in. Of the 105 tehsils that cross both thresholds, 34 use a figure from a larger area. Counting only the 71 with a local village figure gives about 750,000 solar-lit households. Changing either threshold also changes the count. A lighting answer does not establish that a household is off-grid. Gallup Pakistan previously mapped the census lighting answers by district. Mouza Census 2020 · Census 2023 detailed results, Table 22
Living standards. The 2024–25 Household Integrated Economic Survey distinguishes households that use solar alone for light from those that use solar alongside grid electricity. We applied household survey weights and grouped the population into tenths by consumption per adult equivalent, which allows for differences in household size and composition. Electricity bills are the amounts households reported for the most recent month, including zero bills. We use weighted medians when comparing bills. The survey did not follow the same households before and after installation. Pakistan Bureau of Statistics microdata
Satellite estimates. TransitionZero’s October 2025 model gives 27.2 gigawatts. The 13 November 2026 release of Open Energy Transition’s earthpv atlas gives 34.5 gigawatts, based on imagery from November 2025 to March 2026. They use different imagery and assumptions, and both have limited calibration in Khyber Pakhtunkhwa. Their estimates include commercial roofs and farmland. Neither model identifies which installations are registered. TransitionZero’s data are licensed CC BY-NC and earthpv’s outputs are ODbL. TransitionZero · earthpv atlas
Research and policy. The comparison with registered solar builds on work by Wajid Islam, Jan Rosenow and Asha Amirali in Energy Research and Social Science (2026, article 105016) and by Asha Amirali in Geoforum (2026, article 104738). Renewables First’s Pakistan Electricity Review 2026 supplies the registered capacity figure. The fixed-charge change is set out in notification SRO 279(I)/2026. Assembly reply on Balochistan tube wells · February rules for solar exports · April amendment for existing agreements · IMF country report, May 2026
Data and methods downloads
- Figure 2 data: sales by year, ten companies (CSV)
- Figure 2 data: sales by company and category, FY16 to FY25 (CSV)
- Figure 3 data: generator invoices, FY19 to FY25 (CSV)
- Figures 4 and 5 data: billing, collection, shedding, registrations and census lighting by company (CSV)
- Figures 6 and 7 data: village electricity 2020 and solar lighting 2023 by tehsil (CSV)
- Figure 8 data: recovery by consumer category, FY25 (CSV)
- Figure 9 data: capacity transfer charges by company, FY24 and FY25 (CSV)
- Figures 2 to 9 as drawn: the numbers behind the interactive charts (JSON)
- Figure 6 geometry: simplified tehsil outlines from the 2023 census frame (JSON)
- Method note data: lighting by population tenth, HIES 2024-25 (CSV)
- Method note data: census lighting by district (CSV)
- NEPRA State of Industry Report 2025 ↗
- NEPRA Performance Evaluation Report FY2024-25 ↗
- Census 2023, Table 22 (PBS) ↗
- Mouza Census 2020 (PBS) ↗
- HIES 2024-25 microdata (PBS) ↗
- TransitionZero distributed solar layer ↗
- Open Energy Transition earthpv atlas ↗
- Source and method notes (Markdown)