Startup failure rate in India, what the closure data really shows

The startup failure rate in India has no single figure. The Ministry of Corporate Affairs categorised 6,789 of 2,12,283 DPIIT-recognised startups as closed (dissolved or struck off) on 31 January 2026. Tracxn lists 55,233 deadpooled companies among 180,227 tracked, a different measure. Neither is a failure rate.

Is it true that 90% of startups fail?

The 90% figure is not an Indian closure count. A failure rate needs a definition of failure and a base to divide by, and the two published counts differ on both.

The Ministry of Corporate Affairs (MCA) counts a recognised startup as closed only when it is dissolved or struck off, and its base is the startups DPIIT has recognised. Tracxn counts a company as deadpooled under its own classification, and its base is every Indian tech company it tracks, funded or not. The MCA count does not cover a startup that stopped operating but stayed on the register, and neither count covers a founder who never registered an entity.

Both are cumulative totals on a date, so they cannot say how many startups fail in a year. Read as what they are, they describe formal closures and a tracked database, not the share of startups that fail.

What does the Ministry of Corporate Affairs count, and how many recognised startups have closed?

The figures come from written replies to the Lok Sabha that the Press Information Bureau (PIB) published. As on 31 January 2026, DPIIT had recognised 2,12,283 startups, and MCA categorised 6,789 of them as closed (dissolved or struck off), according to the reply of 17 March 2026. The reply of 2 December 2025 gave 1,97,692 recognised startups as on 31 October 2025, with 6,385 categorised as closed in MCA data shared on 11 November 2025. The Ministry told the Lok Sabha that the Government has not observed any increase in startup closures.

Recognition itself runs under notification G.S.R. 127(E) dated 19 February 2019, and the base for these counts is the set of startups recognised under it.

Of the 6,789 closed, 2,950 had at least one woman director or partner, against 1,02,054 of the 2,12,283 recognised.

Two limits apply. Closed means dissolved or struck off, so a startup that has stopped trading but is still on the register does not appear.

Recognition is also cumulative and rests on what a startup declares when it applies, so the base is startups that applied, not every startup that began. Both counts are higher in the later snapshot, which is what a cumulative total does, and two dates three months apart cannot show a yearly trend. What recognition does and does not give is covered in who can get DPIIT startup recognition.

The table sets each measure beside its definition.

Do scheme-backed startups close less often?

As on 31 January 2026 the Fund of Funds had supported 1,382 startups, of which 17 were categorised as closed by MCA. SISFS had supported 3,311, of which 26 were closed, and CGSS had supported 281, of which 1 was closed. Each scheme's closed count is a smaller share of its supported startups than the 6,789 of 2,12,283 is of all recognised startups.

That does not show that support prevents closure. Supported startups are a selected group, chosen by a fund, an incubator or a lender, and the counts say nothing about founders who never received support. Closed again means dissolved or struck off only.

What do Tracxn's deadpool count and the funding gap add?

Tracxn's all-time India Tech statistics in its FY 2025-26 report list 180,227 companies, of which 18,044 are funded, with $182 billion in total funding, 86 unicorns, 307 soonicorns, 1,552 minicorns and 55,233 deadpooled companies. Deadpooled counts all tracked companies, including unfunded ones, and the unicorn count follows Tracxn's own definition, which differs from government and media tallies.

Because MCA and Tracxn use different bases and different labels, the two numbers do not reconcile and should not be netted or averaged. The 55,233 is also a total over the whole dataset, not a rate for any period.

On the funding side, PIB's release of 17 April 2026 says more than 55,200 startups were recognised in FY 2025-26, up 51.6% from 36,400+, the highest in a single year since Startup India launched on 16 January 2016.

In the same year Tracxn recorded 459 first-time funded companies in India, down 32% from 678, and 218 in H1 2026, down 31%. The two counts come from different publishers with different bases, so the second is not a conversion rate from the first.

Recognised, funded and still operating are three different states, and the MCA closure count speaks only to the first. Funding in 2026 is covered in how startups get funding in India.

How should a founder read a run of failed attempts?

None of these figures counts attempts. They count entities. A founder who tests and drops several ideas without registering a company appears in none of them, while one who registers a company for each idea and closes it appears as a closure each time.

Recognition counts run far above first-time funding counts, though from different bases. A run of failed attempts therefore sits mostly in the stretch before a company is registered or funded.

The blog's guide to picking and validating a business idea and its post on a Founders Night about validating a startup idea take up the testing step. When a company should be registered at all is set out in when to incorporate a startup in India.

Where does the Founders Cohort fit alongside this?

The Leap by Masai Founders Cohort is a programme for aspiring founders that moves from an in-person start in Goa through an online build phase to an in-person pitch event in Bangalore for teams that meet the benchmarks. It covers problem discovery, building a live MVP, talking to users, finding distribution, pricing and investor pitching, with sessions led by founders and practitioners. It promises no funding, and only teams that meet the benchmarks reach the pitch event.

Frequently asked questions

Is it true that 90% of startups fail?

The 90% figure is not an Indian closure count. MCA categorised 6,789 of 2,12,283 DPIIT-recognised startups as closed (dissolved or struck off) as on 31 January 2026, and Tracxn lists 55,233 deadpooled companies among 180,227 tracked. The two measure different things, and neither is a failure rate.

How many startups fail in India and why?

The counts that exist are closures, not failures: 6,789 of 2,12,283 recognised startups closed as on 31 January 2026, and 55,233 deadpooled among 180,227 tracked by Tracxn. No reason for closure is recorded in either count, and a startup can close for many causes.

How many startups fail every year?

Both counts are totals on a date, not yearly figures. MCA's figures are cumulative snapshots: 6,385 closed of 1,97,692 recognised on 31 October 2025 and 6,789 of 2,12,283 on 31 January 2026. Two dates three months apart cannot show an annual rate, and the Ministry told the Lok Sabha that the Government has not observed any increase in startup closures.

Do government-backed startups fail less often?

The closed counts are smaller as a share of supported startups: 17 of 1,382 for the Fund of Funds, 26 of 3,311 for SISFS and 1 of 281 for CGSS, as on 31 January 2026. Supported startups are a selected group, so the counts do not show that support prevents closure or say anything about founders who never received it.

What is the difference between MCA closures and Tracxn's deadpool?

MCA counts recognised startups that are dissolved or struck off, out of 2,12,283 recognised on 31 January 2026. Tracxn's deadpool is its own classification, covering 55,233 of the 180,227 companies it tracks, funded or not. The bases and labels differ, so the figures should be read separately and not combined.

Next step

The Founders Cohort, Leap by Masai page sets out the curriculum and how to apply.