Furnace oil is the costliest power on Bangladesh’s grid
A fleet of 5,634 MW burns imported oil that India, Vietnam and China barely touch. At half load it costs up to $6bn a year.
The rest of the picture is LNG, coal, solar, Rooppur and Indian power, and which of them can take the oil fleet’s place.
5,634 MW of furnace oil plants · run at a 50% plant factor · cost per year
$5.8-6.3 billion
# 24.7 TWh a year, about a quarter of the 96 TWh the grid made in FY2023-24
# $3.9-4.4bn is imported fuel: 4.8 Mt at 200 ml per kWh
# $1.9bn is capacity payments, owed whether the plants run or not
# Tk 29-31 per kWh, against Tk 13.20 for coal and Tk 7.09 for gas on BPDB’s FY2025 books My estimate.
Capacity: Daily Star, April 2026. Fuel prices: BERC Tk 108.78 a litre (1 October) and Singapore HSFO $806 a tonne (9 October).
Capacity payments scaled from IEEFA’s Tk 9.5 per kWh. At last year’s pace of about 33% the bill is nearer $4.5-4.8bn.
Bangladesh has 5,634 MW of furnace oil plants. Run at half load they cost up to $6bn a year, and almost none of its garment rivals lean on oil this way.
Furnace oil is the most expensive way Bangladesh makes electricity. BPDB’s FY2025 books put liquid fuel at Tk 27.39 a kWh, against Tk 13.20 for coal and Tk 7.09 for gas. The country has 5,634 MW of furnace oil plants installed, plus 768 MW of diesel units, and IEEFA counts oil as 10.7% of generation in FY2025.
Fig. 1 · Oil supplies 12% of Bangladesh’s power and 1% or less in India, Vietnam and China
Oil’s share of electricity generation, per cent. Bangladesh is highlighted.
Bangladesh, Sri Lanka, Cambodia, China, Indonesia: Ember via Our World in Data, 2025 (Indonesia 2024). India, Pakistan, Vietnam: IEEFA, FY2024-25 basis. *Ember’s country page lists Pakistan at 9%, which conflicts with IEEFA’s 0.6%. I use IEEFA because it covers India and Vietnam on one basis, but treat Pakistan as unsettled.
The gap with rivals is the point. Bangladesh’s oil share is 12% on Ember’s 2025 data and 10.7% on IEEFA’s FY2025 count; either way it is many times the 1% or less seen in India, Vietnam and China. Pakistan is at 0.6% on IEEFA’s count, though Ember lists 9%. Sri Lanka, at 11%, is the one comparator that leans on oil as Bangladesh does.
The bill depends on how hard the plants run.
What that share costs depends on how hard the fleet runs. Last year’s pace, about 33% of capacity, used about 3.2 Mt of fuel.
At half load the fleet would burn 4.8 Mt and make 24.7 TWh, roughly a quarter of the 96 TWh the grid produced in FY2023-24. PDB’s summer plan assumed 3,500 MW of average output from oil, and its September plan nearly 4,000 MW.
Fig. 2 · At half load the oil fleet costs up to $6bn a year, and a summer-plan load costs about $7bn
Capacity: The Daily Star, 7 April 2026. Fuel: 200 ml per kWh, density 0.98 kg per litre, Tk 122.7 per dollar. Capacity payments: IEEFA’s Tk 9.5 per kWh at about 50% load, held fixed at Tk 41.6m per MW a year. Summer and September plans: PDB, as reported by The Daily Star and Prothom Alo.
Fuel is the larger half of the bill, not all of it. Capacity payments of about $1.9bn are owed whether the plants run or not, and the scenarios hold them fixed.
That is why the cost per kWh falls as load rises, and also why leaving plants idle does not bring the bill to zero.
The price they pay keeps moving, and the plants need more fuel than the jetties can unload.
Fig. 3 · The regulated furnace oil price is up 55% since April, after peaking at Tk 113.54 in May
BERC price for power plants and other consumers, taka per litre, by effective date in 2026.
BERC decisions reported by The Financial Express (12 April and 18 May), UNB (1 July) and bdnews24 (6 September and 1 October). The price includes taxes and margins.
Price is only one limit. BPC’s furnace oil imports of 1.74 Mt between January 2025 and March 2026 came 57% from Singapore and 39% from Malaysia, none from the Gulf, and the Dolphin jetties can handle about 100,000 t a month.
A fleet running at the summer-plan load needs more fuel than that route moves easily.
Oil is one line in a fuel and power import bill of roughly $9bn a year.
Oil is one line in a larger bill. LNG cost $3.8bn in 2025 for 109 cargoes, coal imports run 16-20 Mt a year against 0.7 Mt mined at home, and Adani’s power cost about $1bn in FY2023-24.
On these estimates the grid sends roughly $9bn a year abroad for fuel and power, close to CPD’s figure of nearly $10bn for primary fuel.
Fig. 4 · Roughly $9bn a year leaves the country to fuel the grid, with furnace oil now the fastest-growing line
Annual import cost of fuel and power for electricity, US$ billion. Hatched bars are my estimates; the whisker shows the range. Coal imports of 16-20 Mt compare with 0.7 Mt mined at home.
LNG: IEEFA, 2025. Coal: 16-20 Mt at $85-120/t. Oil: 3.2 Mt, scaled by BERC’s price rise from Tk 70.10 to Tk 108.78 per litre (rough).
India: Adani payments in FY2023-24 only. CPD has put the annual primary-fuel need for power near $10bn.
A price spike lands directly on that bill, and LNG is where it bites hardest.
Each $1 per MMBtu adds about $317m a year to the LNG bill at 2025 volumes.
A move from $12 to $20 adds roughly $2.5bn, and Q2 2026 spot prices were near the top of that range.
Fig. 5 · At 2025 volumes, an LNG price of $20 adds $2.5bn to the bill
Annual LNG import cost at a given price per MMBtu, holding volume at the 2025 level, US$ billion. IEEFA put the extra subsidy at $1.07bn for April-June 2026 alone at about $20.
My arithmetic: IEEFA’s $3.8bn at about $12 implies roughly 317 million MMBtu a year. Excludes regasification and terminal charges.
Gas still carries most of the exposure.
Gas carries most of the exposure because it makes most of the power: 64% of generation, against 22% for coal, 12% for oil and diesel and 3% for solar and hydro.
Oil’s share is small, but its import dependence, price volatility and chokepoint risk all score high, so it adds about 10 points to the index. Two-thirds of 2025 LNG crossed Hormuz.
Fig. 6 · Gas carries two-thirds of the system’s price and supply risk
Share of generation (Ember, 2025) and my 1-5 scoring on import dependence, price volatility and chokepoint risk.
Contribution = share × average score ÷ 5. This is an editorial index, not an official one.
System exposure index: 74 out of 100. Imported electricity from India (about 15.6% of 2025 supply) sits outside these shares and adds its own payment and political risk. Generation shares: Ember via Our World in Data, 2025.
Coal is the cheaper way to replace oil.
Coal is the nearest substitute. At Tk 13.20 a kWh on BPDB’s books it costs less than half of oil power, and the Matarbari tariff is Tk 8.45.
On those costs every TWh moved from oil to coal saves about Tk 14bn, or $116m (my arithmetic). That is the case for running the idle coal capacity before any oil plant is dispatched.
The catch is supply: Payra and Matarbari stopped in 2026 when coal supply lapsed, so the units help only if coal and letters of credit are arranged ahead of time.
Solar and nuclear add supply without a fuel bill.
Solar is the fastest source Bangladesh can add. Rooftop capacity is 418 MW on the official count and 667 MW on IEEFA’s, and about 1,000 MW if small units are included, against a 2030 target of 5,500 MW. Grid-scale solar and wind add another 859 MW.
Fig. 7 · Rooftop solar needs to grow roughly five to eight times by 2030
Installed solar capacity and government targets, MW. Hatched bars are targets or estimates. At a 21% capacity factor, 5,500 MW of rooftop solar makes about 10 TWh a year and saves about $1bn of imported fuel (IEEFA: $180,000 per MW).
Official and IEEFA rooftop counts, June 2026. Targets: National Renewable Energy Development Strategy 2026-2030. The 1,000 MW figure is IEEFA’s estimate including units below 150 kW.
Pakistan cut its fuel import bill with solar. Japan and South Korea leaned on nuclear and coal.
Pakistan still buys spot LNG for evenings, and some contracted cargoes went surplus. Japan’s renewable share doubled from 6% to about 12% between 2017 and 2023. Bangladesh has agreed to buy 117 US LNG cargoes for 2026-2038. Sources: Dawn, IEEFA, S&P Global.
Rooppur adds 2,400 MW, but its grid date keeps slipping.
Fuel loading in Unit 1 began on 28 April. The tariff is not fixed, and the grid needs reserve for a 1,200 MW unit trip. Sources: The Business Standard, Daily Observer, Rosatom.
Indian power is far cheaper than oil, though not cheaper than domestic coal.
Indian power sits in the middle of the cost ladder.
In FY2024 NVVN averaged Tk 8.07 a kWh, PTC Tk 9.28 and Sembcorp Tk 10.42, all cheaper than system-average coal, while Adani’s Godda plant averaged Tk 14.87.
Every one of them is well under the Tk 27.39 that liquid fuel cost in FY2025, so Indian power helps most when it displaces oil-fired hours.
Fig. 8 · Indian power beats oil by a wide margin but not domestic coal
Cost of electricity by source, taka per kWh. Years differ by source, so read the ladder as a guide. Hatched = estimate.
BPDB FY2025 via The Daily Star (gas, coal, liquid fuel); Financial Express, FY2024 payments (Indian suppliers, Matarbari); LNG-fired gas breakeven $0.12/kWh from BigMint at Tk 122.7 per dollar.
Put the levers side by side and they could offset $3.7-5bn a year.
No single source closes the gap. Rooftop solar at target, Rooppur at full output, another 1 GW of Indian power and captive-power efficiency could together offset about $3.7-5bn a year on my estimates, roughly the size of the furnace oil bill at half load.
Fig. 9 · Four levers could offset $3.7bn to $5bn of the yearly import bill
Approximate annual fuel-bill offset at full delivery, US$ billion. All figures are my estimates built from the sources above.
Rooftop: 5.5 GW × $180k per MW (IEEFA). Rooppur: 19 TWh × $0.075-0.12 per kWh. India: 1 GW at 80% load, replacing Tk 27.39 oil power with Tk 8.07-14.87 supply. Efficiency: IEEFA’s 50.18 Bcf, more than 15% of 2025 LNG, valued at 15% of $3.8bn.
What would make the shift happen:
1. Make BPC the single importer of fuel for power stations. Reports describe private plants importing most of their own furnace oil on their own letters of credit, and I found no approval step on those imports.
A bill of up to $6bn a year is large enough to strain the dollar reserves if many owners import at once, and many plants are held by politically connected owners.
BPC already supplies part of the demand: its share of private plants’ needs rose from 12% in FY2021-22 to 44% in FY2023-24, after the 2022 dollar crisis.
Making it the sole buyer of all power-station fuel would put the volume under one set of letters of credit and bulk tenders, with BPDB approving monthly volumes and publishing plant-wise offtake.
It needs storage and jetty room, since BPC normally handles 75,000-100,000 t a month for power plants, and a rule that BPDB takes what it orders: in January 2025 it took 18,434 t against a demand of 65,577 t. This is the author’s proposal; plant ownership was not independently verified.
2. Run idle coal capacity before oil. Dispatch available coal units ahead of oil-fired plants, and back the coal units with firm supply: longer contracts, stocks and letters of credit opened in advance.
At Tk 13.20 against Tk 27.39 a kWh, each TWh shifted from oil to coal saves about Tk 14bn.
3. Contract LNG for flexibility. Hold back from locking in long-term volumes beyond what a solar-rich grid will burn. Seek price caps, collars and annual reviews, and test the new terminal plans against the solar path.
4. Extend the solar tax cut and clear the remaining barriers. The NBR has already cut the combined tax on solar equipment from 17% to 1% for 180 days from September 2026, after exempting duties in June.
Extend the cut beyond the 180 days so importers can plan, speed up net metering and open household finance, as Pakistan’s zero-rated import regime did.
5. Get the grid ready for Rooppur. Fix the tariff, carry enough reserve for a 1,200 MW unit trip and add transmission before the reactor reaches full power.
6. Buy more from India on better terms. Target oil-displacing hours, diversify beyond one seller, add Nepali and Bhutanese hydropower, and settle payment terms to avoid the arrears that cut supply.
7. Squeeze waste before buying fuel. Cut LNG use in captive industrial power and gas system losses. IEEFA sees 50 Bcf in savings from captive efficiency alone.
8. Keep a fuel buffer. Sign longer coal contracts with solid suppliers, hold stocks and keep drilling at home. Payra and Matarbari stopped in 2026 when coal supply lapsed, and a tender for 26 offshore blocks opened in May.
