SBI loan write-offs: RTI shows 14% recovery

Social media discussions on SBI loan write-offs have picked up after RTI-linked figures were widely shared alongside NCLT recovery numbers and small-borrower recovery comparisons. Much of the debate is about what “write-off” means in bank reporting, and how it differs from losses taken in insolvency resolutions. Posts also highlighted that recoveries from written-off accounts can span multiple years, which can make single-year comparisons misleading. Another theme is the perceived gap between recovery outcomes for large borrowers and smaller borrowers, even though the datasets cited often belong to different banks. The core SBI-specific RTI figure being discussed is a recovery of around 14% against technically or prudentially written-off large-borrower loans over a long period. A separate set of RTI-linked numbers focuses on NCLT and similar resolution forums, where the bank’s recovery is compared with its claim amount. While the numbers are being shared in the same threads, the accounting categories are not identical and should not be combined. That nuance is central to understanding what these figures do and do not say.

Why SBI loan write-offs are trending again

The current discussion centres on an RTI response cited in a Moneylife report and amplified across Reddit and other platforms. In the posts, SBI is described as having technically or prudentially written off a large sum of loans linked to borrowers owing more than Rs 100 crore each. The same RTI-linked dataset is also used to discuss how much was recovered against those write-offs. Separately, another RTI-cited set of figures is being used to discuss NCLT and similar resolution forums, where recoveries are compared with the bank’s claims. Several users are also sharing a comparison from Central Bank of India data that suggests small borrowers have higher recovery rates than large borrowers. That comparison is not SBI data, but it is being used in SBI-focused threads to argue about fairness and enforcement. Posts repeatedly point out that the “real story” is the amount recovered against the written-off amount in the books, not the original borrowed amount. The online argument is less about whether write-offs occur and more about the recovery effectiveness and transparency.

What the RTI says on large-borrower write-offs

As per the RTI response cited in the context, SBI technically or prudentially wrote off Rs 1,51,857 crore between FY2016-17 and FY2025-26 for large borrowers owing more than Rs 100 crore. Against these write-offs, the bank recovered Rs 20,838 crore, which is described as around 14% of the relevant sum. This specific metric is being interpreted online as a low recovery rate for large-borrower written-off loans. Several posts also link the figure to the idea that SBI recovered “around Rs 1 lakh crore less” than amounts claimed in certain resolutions, though that claim gap is discussed more directly in the NCLT dataset. The RTI-linked “14% recovery” number is repeatedly cited as the headline takeaway for the large-borrower write-off pool. Some users also argue that the figure represents the recovery against written-off amounts recorded in the books, not the gross defaulted exposure. The distinction matters because write-offs can be accounting actions, while recovery actions can continue after a write-off. Here is a consolidated snapshot of the main figures being circulated.

NCLT resolutions: claims vs recoveries

A second cluster of posts focuses on recoveries through the National Company Law Tribunal (NCLT) and similar resolution forums over about nine years. According to the RTI-cited figures in the context, SBI sent 309 loan accounts to NCLT or similar forums between FY2017-18 and FY2025-26. In these cases, SBI’s total claim amount is stated as Rs 1,49,895 crore. The amount recovered through resolution plans is cited as Rs 49,727 crore. The difference between the claim and the recovery is described online as roughly Rs 1 lakh crore lower recovery than claimed. The context explicitly states this difference is about Rs 1,00,168 crore, and calls it a “haircut” of about 67% of the total claim. These are the figures driving the “big haircut” narrative in social media threads. Importantly, this set of numbers refers to resolution outcomes relative to claims, not recoveries relative to book write-offs.

Write-off recoveries vs resolution haircuts - not the same

Several posts and reposts mix write-off recovery percentages with insolvency haircuts, but the context itself flags that this is not a clean comparison. The RTI-linked write-off number refers to “technical” or “prudential” write-offs in the bank’s books, and recoveries reported against those written-off loans. The NCLT data refers to claims filed and the realised amount under resolution plans. The context clearly states that resolution haircuts and loan write-offs are different accounting categories and should not be combined while calculating losses or recoveries. This matters because a bank can take a haircut in a resolution plan while also having separate write-off treatment for the same or related exposures. It also matters because recoveries from written-off accounts can be booked later, after the accounting write-off, depending on the mechanism used. Online discussions often treat “write-off” as “waiver,” but the numbers being shared are framed as recoveries continuing against written-off accounts. Without aligning the base metric, percentage comparisons can be misleading even when each number is accurate within its own category. The safest reading of the trending posts is that they reflect multiple lenses on stressed-loan outcomes, not one unified recovery ratio.

Small vs large borrower recoveries - the Central Bank example

A widely shared comparison in the threads comes from Central Bank of India’s own data for the last six financial years, as presented in the context. It states that when a small borrower defaults, the bank recovered about 74% of dues, while for a large borrower it recovered about 14.5%. The underlying figures shared say Central Bank of India wrote off Rs 6,774.23 crore for borrowers with dues of less than Rs 1 crore and recovered Rs 5,004.28 crore. For large borrowers with dues of more than Rs 100 crore, it wrote off Rs 26,701.55 crore and recovered Rs 3,874.41 crore. That leaves more than 85% of the written-off amount in the large-borrower category unrecovered as of the reported period, as stated in the context. These figures are being used to argue that recovery mechanisms are tougher on retail and weaker on large corporates. However, this comparison is not SBI-specific, even though it is circulating in SBI-focused discussions. What it does show, based on the posted numbers, is how sharply recovery outcomes can diverge by borrower size in a bank’s write-off dataset.

SBI yearly write-offs and recoveries: recent trend

Another set of figures circulating in the same discussion focuses on SBI’s annual write-offs and recoveries in recent years. The context says SBI wrote off Rs 19,666 crore of bad loans in 2021-22, which increased to Rs 24,061 crore the next year, and then declined to Rs 17,803 crore by 2025-26. It also states that the amount recovered from written-off loans has increased every year, from Rs 24,739 crore in 2021-22 to Rs 42,889 crore in 2025-26. The Ministry is cited as specifying that recoveries can pertain to loans written off in a particular year or in previous years. That note is crucial because it explains why recoveries in a year can exceed write-offs in the same year. The context also says the annual ratio of recoveries to write-offs has improved from 21.4% in 2021-22 to 60.8% in 2025-26. These figures are being used online to argue that recovery actions have become more effective over time. At the same time, the RTI-cited long-period recovery rate of around 14% for a specific large-borrower write-off pool is prompting questions about older legacy accounts.

How recoveries happen: IBC, OTS, ARCs, asset sales

The debate is also pulling in a separate disclosure from SBI’s annual report about recoveries from written-off loan accounts in FY26. According to the context, SBI recovered Rs 10,054 crore from written-off loan accounts during FY26, up from Rs 8,002 crore in the previous financial year. The same passage says the bank used multiple bad-loan resolution mechanisms, including the Insolvency and Bankruptcy Code (IBC), one-time settlements and asset sales to reconstruction companies. It also notes that SBI continued to carry a sizeable stock of legacy stressed assets on its books. The context further says that during FY26, SBI wrote off fresh loans worth Rs 17,803 crore. These disclosures are being cited to argue that write-offs and recoveries can run in parallel, and that write-offs do not necessarily mean recovery efforts stop. They are also being used to highlight that different reporting lines can produce different “recovery” numbers for the same year. Because the context provides no reconciliation between the FY26 recovery numbers (Rs 10,054 crore versus Rs 42,889 crore), posts are largely treating them as separate disclosures with different scopes.

What investors and depositors are debating online

The dominant investor-facing question in the threads is what these numbers imply for accountability and future credit discipline. Many posts focus on the optics that recoveries look stronger in small-borrower buckets and weaker in large-borrower buckets, even when the dataset being cited is from another bank. Others focus on the NCLT claim-versus-recovery gap and treat it as evidence of large haircuts in insolvency resolutions. A more technical set of commenters emphasises that technical or prudential write-offs are accounting actions and do not automatically mean a loan is forgiven. Some users point to the improving annual recovery-to-write-off ratio figures cited for 2021-22 through 2025-26 to argue that recoveries are trending better in recent years. Others respond that the long-period 14% recovery against a large-borrower write-off pool remains a concern regardless of recent-year improvements. The context also includes non-English posts framing the issue as a large loss figure linked to NCLT-mediated resolutions, reflecting the political sensitivity of the topic. Overall, the conversation is less about one single statistic and more about how different recovery channels, definitions, and time windows create very different headline numbers.