Layer 2 · In depth
The banking crisis
The facts of Bangladesh's banking crisis - how many loans are bad, where they are concentrated, the 2026 bank buy-back law and its repeal, who pays, and why trust in the central bank matters.
On this page
This page holds the facts. How to repair the banks (the “bank deal”, the order of steps, the trade-offs) is on Strategy and deals; the loop that links banks, politics and credit is on The loops. The wider economy is on The economy in late 2026.
In one paragraph
About one-third of outstanding bank lending, by value, is classified as non-performing, and a large share of that sits with about 5,775 borrowers (about half the reported stock by value in September 2025) and a handful of banks. Banks have little capital left, so they lend little to anyone else. Private credit grew only 4.5%, the slowest in 33 years. Someone has to pay for the losses, and in 2026 Parliament repealed a clause that let former owners buy their banks back, which one analyst (Zia Hassan) had judged the least bad fiscal option. The bill is still unpaid. The value of the currency rests on trust in the central bank, so how the clean-up is run matters as much as what it costs.
How big is the problem
- Bad loans. Non-performing loans (NPLs) were 32.78% of all loans, Tk 6.07 trillion, at end-June 2026, the highest in the world on IMF comparisons.1 One source cited in the note says the Tk 606,555 crore figure is 27 times the 2009 level and that the jump reflects a 2025 switch to the international 90-days-overdue rule, which ended years of understated figures.2 Officially reported NPLs were about Tk 1.7 trillion in mid-2024 (unverified; no primary source located).
- By type of bank. The World Bank cites 33.2% overall, Islamic banks at 58.9%, state banks at 43.2%, and system capital adequacy of -2.6%.3 Foreign banks stand at 6.5%.4
- Wider “distressed” loans. Including rescheduled and written-off loans, distressed assets are near Tk 11 lakh crore, about 60% of all loans (59.73% at end-2025, up from Tk 756,553 crore a year earlier).52
- Losses. The sector made a record net loss of Tk 136,666 crore in 2025, against a profit of Tk 12,158 crore in 2024. Net interest income turned negative (Tk -12,537 crore): banks paid depositors more than they earned on loans. The ten worst banks lost Tk 154,745 crore between them. A few banks still made money, including BRAC Bank, City Bank and Pubali Bank.5 (Unverified: from a summary whose source codes cannot be traced.)
- Solvency, not liquidity. Deposits were still growing about 12% a year to March 2026, so Moody’s calls this a solvency crisis, not a liquidity crisis.5
- Credit freeze. Private credit grew 4.5%, the lowest in 33 years, while credit to government grew 30.4%.3
Capital adequacy and the NPL ratio
System capital adequacy is -2.6% (World Bank), -2.64% at end-2025 against +3.08% a year earlier (unverified: no primary source located), and -3.8% (Bangladesh Bank Financial Stability Report; unverified). The dates and definitions differ, but all are well below the required 12.5%. The NPL ratio is 32.78% (Bangladesh Bank), 33.2% (World Bank) and 30.6% at December 2025 (classified loans, Tk 557,217 crore). See Disputed figures.
Where the bad loans are
Bad loans are not spread across the economy.
| Fact | Figure | Source |
|---|---|---|
| Default held by a small group of borrowers | about Tk 3.04 trillion with 5,775 borrowers at September 2025, roughly half the reported stock by value, alongside many smaller defaulters | Bonik Barta6 |
| Each of six state banks’ top-20 defaulters, combined | Tk 92,627 crore, 63% of all defaulted loans at those six banks; 1% recovered. The lists are per bank, and any overlap between them is not stated, so this is not a count of 20 borrowers | Daily Star7 |
| Loans rescheduled before default, in one year | Tk 2.68 trillion, up 99% | bdnews248 |
| Longest repayment concession | up to 15 years for loans of Tk 10bn or more | Prothom Alo9 |
| Small versus large defaults | small defaults falling, large ones piling up | Bonik Barta10 |
| Bank credit reaching rural areas | 7.4%, and the stock is shrinking | Bonik Barta11 |
| Worst banks’ share of defaults | “15 worst banks hold 85%” | unverified: no primary source located |
Rescheduling and incentives
Our reading is that each new rescheduling window can teach borrowers that waiting costs less than paying. This assumes that repayment is not otherwise enforced; it is developed in Root causes.
What the concentration implies
The figures suggest that small defaults are falling while a few very large ones pile up, and that about half of the classified stock by value sits with about 5,775 borrowers. A clean-up would therefore need to deal with those borrowers and a small number of banks, as well as many smaller defaulters. This is our reading of the figures above.
The merged Islamic banks
Five Islamic banks linked, in press reports, to the S Alam Group were merged into one institution, called Sammilito Islami Bank (earlier United Islami Bank in one summary). The constituent banks had NPLs of about 84%.12 Zia Hassan’s arithmetic for the merged bank:13
- Deposits of Tk 142,000 crore belonging to 7.5 million depositors.
- Loans of Tk 193,000 crore, about 90% non-performing.
- Net asset value negative and equity zero.
- Tk 35,000 crore needed in the interim period alone and more than Tk 100,000 crore in future.
- The Finance Minister told Parliament that the state has already paid more than Tk 80,000 crore and will need another Tk 100,000 crore.
- If the interim approach continued, the cost could be a quarter of government revenue, or nearly 2% of GDP.
Separately, a larger political and financial inquiry exists. The 2024 Economic White Paper estimated that $234bn left the country between 2009 and 2023, about $16bn a year through trade mis-invoicing and informal transfers, and that $14-24bn of $60bn in public procurement was lost to kickbacks.14 The government reports Tk 70,446 crore of assets frozen and six verdicts.15 It is negotiating with ten countries to recover funds.16
The White Paper’s $234bn
This is a committee estimate “built on certain assumptions”, and its method is contested.17 Critics also say the night-lights method behind the paper’s larger GDP corrections should not be taken at face value.18 A committee estimate is not a court finding. We report it as an estimate. Press accounts differ on how many banks the group controlled (5 and 7 are both reported); that is separate from the five Islamic banks that were merged, and we state only the merger of five.
The buy-back law: passed in April, repealed in September
The question the state faced was who pays for the merged banks’ losses. Four options were on the table: keep paying, liquidate, sell, or hand the banks back to their former owners under strict conditions.13
- April 2026Added
- The interim ordinance becomes an Act. Section 18(a) lets former owners reclaim a bank by repaying 7.5% of the state's money up front and the rest within two years
- The central bank reportedly had not known the clause was coming
- May 2026Argued
- Zia Hassan argues it is the least bad fiscal option
- Opposition parties, bank-owner associations and bankers' associations object
- One critic says owners could pay the 7.5% and borrow the rest from the banking system
- June 2026Withdrawn
- The government retreats
- 9 September 2026Repealed
- Parliament repeals the clause
Details:
- The clause was added by the government when the interim ordinance became an Act in April, and it allowed former owners to reclaim a bank by paying 7.5% of the state’s money up front and the rest within two years.1920 One account of the terms adds 10% simple interest, repayment of all state money, new capital, settlement with depositors and creditors, and payment of taxes.5
- Opposition parties said it was written for specific groups. The associations of bank owners and of bankers also objected.21 A critic said owners could pay the 7.5% and borrow the rest from the banking system.22
- The central bank reportedly had not known the clause was coming.23
- The government retreated in June and Parliament repealed the clause on 9 September.24 The bank resolution amendment bill was passed.25
- Part of the cost of the merged bank has also been met by converting the largest deposits into shares (used at Sammilito Islami Bank).26
What the episode shows
A two-thirds majority backed the clause, and it was reversed within five months after public, opposition and industry objection. That is evidence that the public cares about who bears the cost. The repeal did not pay the bill. Lessons for the design of any settlement are on Strategy and deals.
Who pays, and how much
| Estimate | Figure | Source |
|---|---|---|
| Cost of clean-up | about 1% of GDP a year for several years | Jyoti Rahman; central bank governor put it at Tk 60,000 crore a year27 |
| Finance Minister’s figure | $40bn over several years | Cited by Jyoti Rahman27 |
| Full recapitalisation | about 10% of GDP | attributed to Moody’s (unverified: no primary source located) |
| Another proposal | roughly 3-5% of GDP, amortised over 10-15 years | a proposed roadmap (a proposal, not a forecast) |
| How many banks survive | 10 to 12 of 61 | Jyoti Rahman28 |
The ranges differ because they cover different time spans and what they count (a one-off recapitalisation or an annual flow). They are our reading, not a settled number. Jyoti Rahman adds that money already moved abroad is “extremely difficult to recover”, so part of the loss is permanent.28
Trust in the central bank
Zia Hassan argues that the value of a currency is trust. Remittances rose from about $24bn to $33bn a year and reserves from $18bn to $30bn after the former governor, Ahsan Mansur, was believed when he said he would hold interest rates and run a transparent exchange rate. A successor seen as the government’s choice, he argues, loses that belief, so even correct decisions are read as favours to allies. He reports that a New York banker advising five global banks said he would now tell them not to raise credit lines. The figures and the account are his own, and the article gives no primary sources.29 CPD separately criticised the abrupt removal of the governor.30 The current governor is named in one summary as Mostaqur Rahman.5
Other reported 2026 events:
- In mid-2026 the central bank issued a monetary policy statement and almost at once a circular capping lending spreads, which one analyst reads as contradicting it.31 One source cited in the note reports June 2026 lending-rate caps and extended rescheduling, and warns that these risk reviving forbearance.2
- A claim that a November 2026 policy let court-stayed defaulters access stimulus funds appears in one input without a source. We could not verify it.
How should the central bank behave?
Jyoti Rahman wants tight money and warns that rate cuts prop up over-borrowed firms at the cost of inflation. Zia Hassan describes a balance-sheet recession in which lower rates and case-by-case rescheduling are needed, but only from a central bank that people trust. They agree that credibility is the binding constraint.
Courts and recovery
Money-loan courts held 72,540 cases worth Tk 1.78 trillion in 2023.32 Mediation targets for defaulted loans were set at a 1% pilot.33 Court capacity is on State and institutions.
Who writes the rules
Politics and banking overlap. About 60% of MPs are businesspeople and half carry loans or liabilities (see What the numbers miss). This is a description of incentives, not of any individual.
Gaps
- Bank-by-bank losses have not been published in the inputs.
- The fiscal cost of the 2024-26 restructuring is not known (an open question; unverified).
- What fraction of NPLs is recoverable is not known.
-
Prothom Alo, NPL at 32.78%, Tk 6.07 trillion. ↩
-
The Statesman (its ref 6) and the Daily Star (ref 7). ↩↩↩
-
Prothom Alo on the World Bank Bangladesh Development Update, https://en.prothomalo.com/business/local/qgylemzmn1. ↩↩
-
Unverified: no primary source located. ↩
-
Bonik Barta, Tk 3 trillion defaults, 5,775 borrowers. ↩
-
Daily Star, Top 20 loan defaulters: 1% recovered at six state banks. ↩
-
bdnews24, rescheduled loans. ↩
-
Prothom Alo, concessions of up to 15 years. ↩
-
Bonik Barta, small defaults falling. ↩
-
Bonik Barta, rural credit. ↩
-
Daily Star, Islamic 4th-gen banks buckle under NPL and liquidity crises; bdnews24, report. ↩
-
Zia Hassan, The Bank Bailout Trap, Counterpoint, May 2026. ↩↩
-
Economic White Paper (December 2024), via Prothom Alo and a further report that cites the White Paper PDF. ↩
-
BSS, assets frozen and verdicts. ↩
-
Prothom Alo, asset recovery and the $234bn estimate. ↩
-
TBS, Dented but still a minor miracle: a comment on the White Paper. ↩
-
Daily Star, Bill placed in Parliament to abolish provision allowing former owners to reclaim troubled banks. ↩
-
Daily Star, Cabinet approves amendment to bar former owners from regaining weak banks. ↩
-
bdnews24, objections to Section 18(a). ↩
-
Prothom Alo, central bank not informed of the clause. ↩
-
Dhaka Tribune, Parliament passes bank resolution amendment bill. ↩
-
Dhaka Tribune, Bank Resolution Act: cabinet repeals controversial clause. ↩
-
Jyoti Rahman, A State of Economic Emergency, Counterpoint, 20 October 2026. ↩↩
-
Prothom Alo, interview with Jyoti Rahman: the decision to reform the banking sector is a political one, 2 August 2026. ↩↩
-
Zia Hassan, Why the New Governor Won’t Work, Counterpoint, 27 February 2026. ↩
-
CPD, Six Months of the Government, 24 August 2026. ↩
-
Unverified: no primary source located. ↩
-
Financial Express, loans trapped in Artha Rin courts. ↩
-
Financial Express, Bangladesh Bank guidelines for appointing mediators. ↩
Go deeper
- Key statisticsData & sourcesThe important figures from the research, grouped by theme in tables, each with its value, date, source and link. Where sources disagree, see the disputed-figures ledger.
- Disputed figuresData & sourcesThe ledger of figures where our sources disagree, with both values, the likely reason and how this site handles each, plus a list of known gaps.
- SourcesData & sourcesThe consolidated bibliography of every external source cited in our research, grouped by type and deduplicated, with the links as given.
Summarised in
- Where Bangladesh stands Overview
Where this comes from
Synthesised from 7 research reports, checked against the sources cited above and reviewed before publication. How we work · All sources.