What the big stalled-project collapses looked like in the filings

Every risk regime in finance is a memorial to the specific failure that preceded it. India's real estate disclosure regime has two names on the plaque: Amrapali and Jaypee Infratech. Their collapses were reconstructed in courtrooms through forensic audits and years of hearings, and the reconstruction reads, item for item, like the specification for today's register: certified progress, ring-fenced collections, quarterly disclosure, public extension records. This piece replays the documented record as a filings timeline, then asks the only question that matters for a desk: what would these names have printed on the register, quarter by quarter, and what is printing there now.

Amrapali: the anatomy of diversion, on the record#

The documented sequence, drawn from the Supreme Court's judgment in Bikram Chatterji v. Union of India (23 July 2019) and the court-ordered forensic audit it relied on. Through the early 2010s the group presold aggressively across Noida and Greater Noida, collections flowing in years ahead of construction. The audit found buyer money diverted at scale away from the projects it was collected for, into other companies, other uses, other assets. Construction slowed asymmetrically: sales offices stayed open while sites quieted. By the time buyers organised and the matter reached the Court, more than 42,000 homebuyers were waiting for flats the group had not delivered. The Court cancelled the group's RERA registrations, and handed completion to NBCC under court supervision; by 2025 NBCC reported roughly 25,000 flats completed (proceedings reported by The Print).

Translate that anatomy into register artifacts, and it is a checklist of what MahaRERA now demands: withdrawals from the separate account allowed only against a CA's certificate of the eligible amount (Form 3), physical progress certified by architect and engineer (Forms 1 and 2), the 70 percent account, and, for projects registered from 1 July 2024, the three-account plumbing that executes the split at the bank. The diversion that took a forensic audit and two years of hearings to establish would, under the current regime, surface as a quarterly divergence between certified work and certified money, the exact reconciliation the tranche-monitoring piece turns into a lender's procedure.

Jaypee: the slow-motion version#

Jaypee Infratech failed differently: not primarily diversion but leverage against a delivery machine that stopped delivering. The public timeline: IDBI Bank's application under section 7 of the Insolvency and Bankruptcy Code in August 2017 on a Rs 526.11 crore default; the Supreme Court giving homebuyers a representative at the creditors' meetings within weeks (Chitra Sharma, September 2017); then five and a half years of resolution rounds until the NCLT approved Suraksha's plan on 7 March 2023, covering more than 20,000 undelivered homes. The NCLAT upheld the plan on 24 May 2024, with the homes due within 40 months from August 2024.

The doctrine the matter forced arrived in between. An ordinance of 6 June 2018 made allottees financial creditors under the Code, with a vote on the committee of creditors, and the Supreme Court upheld the change in Pioneer Urban Land and Infrastructure v Union of India on 9 August 2019.

The register translation: years of promised completion dates moving in public would today print as a serial extension record; the widening gap between sold inventory and finished floors would print as the progress-versus-time comparison MahaRERA's 2023 grading design proposed to score; and the group-wide nature of the stress, one balance sheet feeding many projects, is exactly why underwriting the sponsor group rather than the SPV is this series' first rule. Lenders inside the process learned the cost of the clock: almost seven years passed between admission and the NCLAT's ruling, and whatever the plan's nominal recovery, that much time value repriced it, the arithmetic the IBC piece works through.

The replay: five signals against two collapses#

Set the documented facts against the five-signal framework from the early-warning playbook:

A procedural detail from both matters deserves its own paragraph, because it is replicable diligence rather than history. In each case, the decisive facts were established not by any creditor's monitoring but by court-commissioned reconstruction: forensic auditors in Amrapali tracing collections across group entities, a court receiver marshalling assets, an interim resolution professional in Jaypee assembling the claims map. Every instrument those professionals used, entity-level money trails, project-level progress records, buyer payment schedules, is an instrument a lender could have commissioned privately, years earlier, for a fraction of its eventual cost. The register now pre-assembles the raw material: each entity's own project record, certified money certificates, quarterly progress. The forensic audit that ends a collapse and the filings review that prevents one read the same documents; the only variable is the date.

Both columns would have scored as firm multi-signal clusters quarters before the public break. Neither could have been read that way at the time, because the artifacts did not exist. They exist now, refreshed quarterly for every ongoing project, on a register MahaRERA put at 50,162 projects by May 2025 and that carried 55,995 published filings by 29 September 2026 on ReraGenie's count, each filing the project-level fields the forensic teams had to reconstruct at Amrapali and Jaypee. What was archaeology in 2019 is a subscription to public documents in 2026, and the arithmetic of the difference is measured in years of queue and multiples of recovery. That asymmetry between then and now is the whole argument for filings-based surveillance, and the cleanup bill prices the alternative: SWAMIH's completion portfolio alone stood at 1,01,443 homes with 63,200 completed as of 31 January 2026, according to the Finance Minister's reply in the Lok Sabha in March 2026, and SWAMIH Fund 2, a Rs 15,000 crore fund announced in the February 2025 Union Budget to complete another 1 lakh homes, was being formally launched that month. The stalled stock the rescue economy services is a national dataset in its own right.

What today's watchlist looks like#

The forward application, because the point of a post-mortem is the next patient. Maharashtra's register now publishes, continuously, the population in which the next collapse is incubating: the lapsed and revoked registration lists, the enforcement sweeps (8,212 show-cause notices under RERA section 7 in May 2026, for one missed quarterly deadline), the extension records, and since 2023 an orderly exit route, deregistration under Order 42/2023, whose usage separates promoters who end dead projects honestly from those who let them rot into the lists. A desk can, today, pull every project in its book's catchments and rank them by the five-signal score; the register's own numbers say the tail is not small. Set against the healthy majority of the register, the distribution of stress is knowable in a way it simply was not when Amrapali's buyers were discovering theirs from unfinished balconies.

The worked arithmetic of acting early is the same as ever. Take a Rs 50 crore exposure. Caught at a two-signal cluster and restructured to return the full amount in two years, it is worth about Rs 41.3 crore today at a 10 percent discount rate. Discovered at the NCLT instead, it joins the queue: on the Jaypee clock, 6.8 years from admission to the NCLAT's ruling, and at the 30.56 percent of admitted claims that resolution plans across all sectors had realised by June 2026 (IBBI), it recovers about Rs 15.3 crore, worth about Rs 8.0 crore today, 16 percent of the exposure. The discount rate and the two-year restructuring are assumptions, and the 30.56 percent is an all-sector average; real estate, 22 percent of admitted insolvency cases by December 2024 (IBBI), is not a corner of that system. Between those two outcomes sits nothing but the desk's reading habits.

The analogy is aviation's: crash investigations become checklists, and the checklist only protects the crews that fly it. Indian real estate ran its crash investigations in the Supreme Court and the NCLT from 2017 onwards; the checklist came out as forms, accounts and deadlines. Kavya Menon's desk (illustrative, as ever) keeps a one-slide version of the table above in its credit committee pack, less for information than for institutional memory: every argument to skip the quarterly filings review is an argument that this time differs from the documented record, and the documented record runs to tens of thousands of families, two Supreme Court sagas and a national rescue fund.

ReraGenie exists to make the checklist flyable: every registration's current filing on a free project page, with its extensions, complaints and registration status on the page, the Rs 2,999 project analysis for a single name, and the area market report for a pincode at a flat Rs 2,999. The register keeps only each filing's current state and records every entity separately, so the quarter-on-quarter series and the group map remain the desk's own work. The free monthly city reports keep the headline market numbers current and citable.

The one-line summary#

The great collapses were reconstructed after the fact from exactly the artifacts the register now publishes in advance: read the record the failures paid for, quarterly, or budget for learning it the way their creditors did.