Coal overtook gas in Bangladeshi generation in July, despite years of investment in LNG import infrastructure. The terminals work. The plants work. The one variable the country did not control turned out to be the one that decided whether any of it would run.
Coal generated 3.92 terawatt hours of Bangladeshi electricity in July, overtaking gas at 3.66 terawatt hours for the first time, despite substantial investment in liquefied natural gas import infrastructure. High international LNG prices have made imported gas difficult for a price-sensitive electricity system to absorb, pushing generation towards cheaper coal even as renewables expand.
The terminals were built, the regasification works, the pipelines connect and the gas plants are commissioned. Nothing in the chain has failed in an engineering sense, and the chain is nonetheless producing less electricity than the coal plants it was meant to displace.
Bangladesh assembled a sequence that reads as a coherent plan: import terminal, long-term supply contracts, gas-fired generation, electricity at a lower carbon intensity than coal. Each element was financeable, each had a clear engineering specification, and the overall logic of gas as a transitional fuel was endorsed by most international energy institutions for two decades.
The sequence contained one input the country could not control, and it was the input that determines whether the plants dispatch. Imported LNG is priced in a global market responding to European storage policy, Asian winter demand, Atlantic basin supply disruptions and the state of Russian pipeline flows, none of which is responsive to Bangladeshi affordability.
When that price rises above what a subsidised, price-sensitive electricity system can pass through to consumers, the gas plant does not run. The capacity payment is still owed, the terminal charge is still owed, the debt service is still owed, and the electricity comes from coal.
That is a different failure mode from the one infrastructure appraisal is built to detect. Feasibility studies test whether an asset works, whether it can be built on time and whether demand exists. They are considerably weaker at testing whether the asset will be dispatched under a price path nobody controls, because that requires a view on a global commodity market over a twenty-five year horizon.
There is an argument for renewables in a price-sensitive importing economy that rarely gets made in those terms, and it is stronger than the climate case in a country whose own emissions are marginal.
Shimon Awerbuch and Martin Berger applied mean-variance portfolio theory to electricity planning and reached a conclusion that is counterintuitive until it is stated: fixed-cost technologies such as renewables must be part of any efficient generating portfolio. Earlier work by Awerbuch had shown that adding wind, photovoltaics and other fixed-cost renewables to a portfolio of conventional generation reduces overall portfolio cost and risk, even where their stand-alone generating cost is higher.
The mechanism is ordinary portfolio reasoning applied to generation. A technology whose cost is almost entirely sunk at construction has no fuel price variance at all, so adding it to a portfolio dominated by fuel-price-exposed assets reduces the variance of the whole portfolio. A planner comparing technologies one at a time on levelised cost cannot see that effect, because it only exists at the level of the mix.
For an importing country the implication is specific. Solar and wind are not primarily a decarbonisation choice. They are a twenty-five year hedge against a price set in Rotterdam and Tokyo, purchased up front, and that hedge has a value which does not appear anywhere in a levelised cost comparison.
The same reasoning cuts the other way on coal, which is why Bangladesh's substitution is not a return to safety. Coal is also imported and also internationally priced. The system has moved from one exposure to a cheaper one, not from an exposure to none.
Price exposure should be a line in the appraisal, not a sensitivity annex. The question is not whether the project survives a ten per cent fuel price increase. It is what share of national generation becomes undispatchable at each price level, and who pays the fixed charges in that state.
Take-or-pay terms deserve the scrutiny that construction risk receives. A terminal charge payable whether or not gas flows converts a commodity price problem into a sovereign balance sheet problem, which is the same structure we described in the Kenyan capacity payments that keep electricity expensive in a system that is 93 per cent renewable.
Portfolio variance belongs in the planning model. Most national generation plans optimise expected cost. Optimising expected cost subject to a variance constraint produces visibly different mixes, and the method has existed for twenty years.
And the counterfactual for a transitional fuel should be stated at the start. If gas is justified as a bridge, the appraisal should say what the bridge is to, how long it is, and what happens if the fuel becomes unaffordable before the far bank is reached. Bangladesh is on that bridge now.
Dispatch against capacity, by plant and by month. The gap between installed gas capacity and gas generation is the direct measure of stranding, and it is published or derivable in most systems.
Fixed charges paid on undispatched capacity. Terminal fees, capacity payments and debt service attributable to plant that did not run, expressed per unit of electricity actually delivered. That figure is almost never compiled and it is the real cost of the exposure.
Who absorbed the difference. Whether it landed on the utility's balance sheet, the fiscal account, the tariff or the consumer through load shedding. The question of which party ends up holding the volatility is the one that determines the social consequence, and the answer is rarely the party that signed the contract.
And the renewable counterfactual, costed properly. What the same capital would have bought in solar, wind and storage, evaluated on portfolio variance rather than on levelised cost alone. That is a retrospective calculation, it is uncomfortable, and it is the only way the next decision gets made differently.
The Lab works on this in energy and water systems and on what a system-level decision means for the households at the end of it through measuring change.
If you are appraising generation infrastructure whose economics depend on a price set somewhere else, tell us what you need to know.
If this reads close to a decision you are weighing, we would like to hear about it. A short note is enough to start.
The Lab is an independent research team, not an advocacy organisation. We do not campaign, endorse products, or accept commissions where a finding is expected in advance. Where a partner's own theory of change is what needs testing, that is a legitimate engagement, and the Lab tests it honestly.