In May 2025 Pakistan's finance ministry announced that the government had allocated 2,000 MW of electricity in the first phase of a national initiative to power bitcoin mining and AI data centres. The pitch was simple. Pakistan had more generating capacity than it could use, especially in winter, and it was paying capacity charges on idle plants regardless. Selling some of that surplus to energy-hungry computing facilities at a price above marginal cost would raise utilisation, bring in foreign exchange and spread fixed costs over more units.
Seventeen months later the plan has not been implemented at the proposed rates, and the conditions that justified it have shifted. The International Monetary Fund has not agreed to concessional tariffs. And with QatarEnergy's force majeure on deliveries to Pakistan extended to 6 November because of the disruption around the Strait of Hormuz, the cost of producing an extra unit of electricity this winter is far higher than the price the plan envisaged.
What the government proposed
The idea predates the 2,000 MW announcement. Business Recorder reported in July 2025 that in September 2024 the Power Division proposed a six-month incremental consumption package for October to March at marginal cost, about Rs23 per kWh, based on the previous year's usage. The IMF approved only a three-month version. In November 2024 the Power Division proposed a targeted package at Rs22 to Rs23 per kWh for energy-intensive users, including copper and aluminium melting, data centres and crypto mining. The IMF rejected it, saying it resembled sector-specific tax holidays.
Dunya News reported that the government planned to offer electricity at up to Rs24 per unit for bitcoin mining, against a surplus it put at 7,000 MW, of which 2,000 MW would be allocated to crypto mining. It said the government had presented the plan to the Fund three times without success, and that the IMF had argued that past experience with cheap industrial power had not produced positive results.
In July 2025 the Power Division and the IMF denied reports that the Fund had rejected the plan outright. A Power Division spokesperson told Arab News that the secretary had said the government was still negotiating and hoped to reach a solution. The IMF representative said both sides were still in talks and stressed the importance of a level playing field. As of September 2026, TechJuice reported, citing energy ministry sources, that Pakistan had still not secured IMF approval and expected to raise the proposal again in talks with the Fund.
What winter power now costs
The case for the plan assumed that surplus electricity could be produced cheaply. That depends on the fuel at the margin. When domestic gas, hydropower or coal is available, the incremental cost of running an idle plant can be modest. This winter, the marginal fuel is imported and expensive.
Arab News reported on 2 October that an official at the Independent System and Market Operator said at least 800 MW was being generated from furnace oil between 5 pm and 1 am, out of 1,400 MW of available furnace oil capacity, to limit imports of regasified LNG. Shankar Talreja of Topline Securities told the paper that furnace oil generation cost had been over Rs40 per kWh, almost the same as RLNG in the previous two months. Figures presented at NEPRA's fuel adjustment hearing on 29 September, reported by The Nation, put the cost of imported LNG generation at Rs45.92 per unit in August and furnace oil at Rs45.25.
Against those numbers, a tariff of Rs22 to Rs24 per unit does not cover the fuel cost of the marginal plant, let alone its capacity payment. If a bitcoin mine or data centre adds load in the evening peak this winter, it is served by furnace oil or LNG, and selling that power at half its fuel cost would be a subsidy paid by other consumers or the budget.
Capacity surplus is not energy surplus
The confusion at the heart of the debate is between capacity and energy. Pakistan does have far more installed capacity than it uses on most days, and it pays for that capacity through fixed charges whether the plants run or not. That is a real cost and a real argument for finding new demand. But surplus capacity only translates into cheap energy if the fuel to run it is cheap and available. Gas-fired plants built around imported LNG have no cheap fuel when LNG cargoes are cancelled or diverted.
A better-designed offer would distinguish between hours and seasons. Demand that runs when hydropower is abundant, or that can shut down during the evening peak and during fuel shortages, could be priced near the true marginal cost of those hours, which is sometimes low. Demand that runs flat around the clock, as bitcoin mining and most AI training do, would be priced at the average marginal cost across all hours, including the expensive ones. Interruptibility would be a condition, not an option.
Why the IMF objects
The IMF's stated concern is distortion. Targeted tariffs below cost for favoured sectors shift costs onto other consumers, and Pakistan's power sector already carries a heavy circular debt. In the same July 2025 Senate committee session, Business Recorder reported, senators debated the government's agreement with scheduled banks to reduce a circular debt stock of Rs1.275 trillion, and the Power Secretary said the cost would be recovered through a debt servicing surcharge of Rs3.23 per kWh expected to continue for five to six years. Adding a new class of subsidised consumers would cut against that effort.
The case that remains
None of this means the idea is worthless. Pakistan has idle plants and fixed costs to recover, and data centres could be valuable customers if they pay prices that reflect when they consume. The country also has a strategic interest in domestic computing capacity. But the winter of 2026 shows why a flat concessional tariff is the wrong instrument. A grid that is burning furnace oil at Rs45 a unit to cover the evening peak does not have cheap power to sell, whatever its installed capacity says.