Retiring Bangladesh's costliest coal plants could save $1 billion

Summary

The Centre for Policy Dialogue (CPD) finds official forecasts overstate Bangladesh's electricity demand, leaving a large power surplus even as the country revives coal projects such as the 1,200 MW Matarbari Phase 2

Under the lower CPD demand forecast, the model builds no new coal, and the existing fleet is squeezed out after the mid-2030s. It meets the 30% renewable target with 70 GW of solar and 10 GW of storage, and retiring Rampal and Payra saves US$1.0bn and 67 MtCO₂ by 2050, a net saving of about US$15/tCO₂

Halting coal is only half the transition: with gas covering demand when solar is not generating, power-sector emissions fall by about 40% and then plateau, leaving the grid exposed to imported liquefied natural gas (LNG) prices

The cost of Bangladesh's coal overcapacity

Bangladesh has committed to 30% renewable power by 2040, yet it continues to expand coal into a grid with far more capacity than it needs. Installed capacity is about 27 GW against a 2025 peak near 17 GW, creating a reserve margin above 60%, around three times what a reliable system needs.

This surplus is, in part, the product of overstated demand forecasts. The official Integrated Energy and Power Master Plan (IEPMP) expected peak demand near 27 GW by 2030 and over 50 GW by 2041; the CPD finds these numbers rest on inflated GDP assumptions and projects demand rising about 2.4 times to 2050, against the plan's 3.8. Planning to the higher figure meant building plants the grid would not need.

Those idle plants are still paid through Power Purchase Agreement (PPA) capacity charges, which compensate them for being available rather than generating. Stranded capacity costs an estimated US$0.9bn-1.5bn a year, and the newest units are the priciest, at BDT 13.57/kWh for Rampal and BDT 12/kWh for Payra against BDT 8.45/kWh for Matarbari. Even so, in May 2025, the Power Division revived the cancelled 1,200 MW Matarbari Phase 2, four years after Bangladesh pledged no new coal at COP26.

We use Scenario Builder to test whether this policy direction is the lowest-cost option for the country. Specifically, is the existing coal fleet worth running under the CPD demand forecast, and does coal expansion cost more than a path of no new coal and greater renewables deployment?

Model and scenario set-up

We ran a capacity expansion model of Bangladesh's national grid to 2050 in TransitionZero's Scenario Builder. It minimises cost, choosing the cheapest capacity to build and operate each year while meeting demand. As a single national node model, it does not capture transmission or grid-region dynamics between zones.

The scenarios pivot on two published demand paths, both applied to a base of 88.5 TWh of net generation in the 2022–23 financial year. The High path follows the IEPMP (demand rising 3.8 times by 2050); the Lower path follows the CPD forecast (about 2.4 times). Three scenarios use them: Reference (High demand, the coal pipeline built), Policy (Lower demand, no new coal, the renewable target imposed), and Retirement (Policy plus early retirement of Rampal and Payra).

For full details on the input data, please download our documentation here.

Coal’s economic life ends in the mid-2030s

The model builds no new coal in any scenario. The existing fleet still operates through the early 2030s: with wind constrained before 2030 and solar still scaling, coal reaches about 40% utilisation under the CPD demand and up to 70% under the IEPMP demand by 2035.

After 2035, solar capacity passes 38 GW and coal utilisation falls to about 11% by 2040 under the CPD demand, and to about 5% under the IEPMP demand. At those utilisation rates, the fleet runs too little to be economic, consistent with the low output already recorded at Payra. Matarbari Phase 2 would add new coal capacity to a system where coal utilisation is below 15% by 2040.

Meeting the 30% renewable target requires 70 GW of solar and 10GW of storage

Meeting the renewable target requires a large-scale build-out of solar. In the model, capacity rises to 16.8 GW by 2030 and 70 GW by 2050, lifting the renewable share to 21% in 2030 and 30% by 2040, with 10 GW of storage shifting output into the evening peak.

The build falls almost entirely on solar because Bangladesh has little else to draw on: hydro is limited, and wind, particularly the offshore potential in the Bay of Bengal, remains largely untapped, held back by gaps in policy, financing, and technical capacity.

Capping coal lowers emissions, but gas limits the fall

Capping coal lowers power-sector emissions, but gas fills the firm gap and keeps them from falling further.

In the Policy scenario, emissions are about 40% below the Reference path in 2050, 35 against 57 MtCO₂. Most of that 2050 gap reflects the lower CPD demand rather than a cleaner mix: by 2050 the two paths have similar carbon intensity, about 0.16 MtCO₂/TWh. The coal cap does more earlier, roughly halving emissions against the Reference path in 2030 (34 against 77 MtCO₂). Gas supplies 97 TWh by 2050 and is the largest source of the emissions that remain.

Two limitations apply. The figures cover power-sector combustion only; upstream methane in the gas supply chain would reduce the net benefit of moving off coal. And gas is imported as LNG, which exposes the system to global price volatility, as seen since 2022 and around the Strait of Hormuz in 2026. The model's gas prices, from CLEAN, currently include domestic price subsidies. Spot LNG import prices currently sit well above these, with IEEFA estimating the price gap which the government covers through subsidies of over $1 billion USD in just three months of 2026. Continued subsidies for gas would therefore place a huge strain on already stretched fiscal resources. Capping coal lowers emissions but does not decarbonise the system; that requires renewables to expand in place of gas.

Retiring Rampal and Payra removes coal by 2050

With retirement the only difference between Policy and Retirement, the effect is clear: taking Rampal and Payra (2,640 MW of the priciest imported coal) offline from 2030 removes coal entirely by 2050 and cuts cumulative CO₂ by 67 Mt. It lowers system cost by about US$1.0bn in discounted terms over 2024–2050, a net saving over the horizon rather than in every year. Because it both saves money and cuts emissions, the retirement carries a negative abatement cost of about US$15 per tonne: the system does not pay to abate; it saves.

The economics point away from new coal

Across the three scenarios, the model builds no new coal, coal utilisation falls below 15% by 2040, and retiring the priciest plants cuts both cost and emissions. Reviving Matarbari Phase 2 would add capacity the model does not select: even under the Reference demand, coal utilisation falls to about 5% by 2040. The least-cost replacement, however, is gas rather than renewables at current gas prices and capital costs, so capping coal secures the cost savings but not the transition.

This balance shifts with cheaper capital: the concessional climate finance underpinning Bangladesh's conditional Nationally Determined Contribution (NDC) would lower renewable financing costs and move more of the replacement from gas to renewables. With the fleet already in surplus, the open questions are how fast to stop paying for capacity the grid does not use, and how to raise its ability to absorb renewables, which we also examine in Nusa Tenggara.

Test it yourself in Scenario Builder

Every scenario here is open. Reproduce the retirement case, or vary the demand path, in Scenario Builder. The clearest next step is to build this out as a multi-node model, resolving the transmission and grid-region dynamics a single national node cannot, to see what regional effects it surfaces.

Start Modelling

Scenario BuilderScenario Builder