Half of new US businesses are still open after five years, as many as three quarters of venture-backed startups never return investors' cash, and 1,981 US venture-backed companies failed in 2025, the most since at least 2019. The one startup failure rate everybody quotes, "90% fail," traces to a 2011 report that never defined failure. These startup failure rate statistics for 2026 answer what percentage of startups fail by year and industry, how many startups fail each year, why they fail in the newest post-mortem data, and which numbers to retire. Every figure carries its year and source, because most startup failure statistics pages still quote a 2014 stat that has been replaced twice.
Key startup failure statistics 2026
1. 22.1% of new US employer establishments fail within one year, 48.6% within five, 65.3% within ten and 80.5% within twenty (BLS Business Employment Dynamics, data through March 2025). 2. Information, the sector holding software startups, has five-year survival of 45.7% against 51.4% for all industries (BLS, March 2020 cohort). 3. 1,981 US venture-backed companies failed in 2025, up from 1,602 in 2024 and 893 in 2023; 2,345 are expected in 2026 (SVB State of the Markets H2 2026, PitchBook data). 4. 966 startups on Carta shut down in 2024, up 25.6% from 769 in 2023; AngelList winddowns rose 56.2% to 364 (Carta and AngelList via TechCrunch, January 2025). 5. 70% of failed VC-backed startups ran out of capital, 43% had poor product-market fit, 29% hit wrong timing or macro conditions (CB Insights, March 2026, 385 shutdowns since 2023 with identifiable reasons). 6. The famous "42% fail from no market need" is CB Insights' October 2014 figure from 101 post-mortems; its 2021 edition put the equivalent at 35% and the 2026 edition replaced it. 7. As many as 75% of venture-backed companies never return cash to investors and 30 to 40% liquidate with investors losing everything (Shikhar Ghosh, Harvard Business School, 2012, more than 2,000 companies funded 2004 to 2010). 8. About 1% of seed-funded US tech startups become unicorns and 67% end up dead or self-sustaining (CB Insights venture capital funnel, 2008 to 2010 seed cohort, analysis published September 2018). 9. 81% of founders had pivoted, 42% wished they had pivoted sooner and 54% named product-market fit their biggest lesson (Wilbur Labs survey of 200 US tech founders, February 2026). 10. 5,671,836 US business applications were filed in 2025, the most on record, after 5,469,302 in 2023 and 5,224,176 in 2024 (US Census Business Formation Statistics). 11. "More than 90% of startups fail" is an unsourced sentence in the introduction of Startup Genome's 2011 report, which defines neither failure nor a cohort; its measured finding was 74% of high-growth internet startups failing through premature scaling.
What percentage of startups fail?
About 22% of new US employer businesses fail within their first year, 49% within five years, 65% within ten and 80% within twenty, so the startup success rate is 77.9% at one year and 51.4% at five. Those are the survival rates in the US Bureau of Labor Statistics' Business Employment Dynamics Table 7, data through March 2025 (released December 17, 2025), the only longitudinal public dataset that tracks every new US private-sector establishment with employees.
The startup failure rate by year
- 1 year: 22.1% fail (77.9% of establishments born in the year ended March 2024 were still open in March 2025).
- 5 years: 48.6% fail (51.4% survive, March 2020 cohort).
- 10 years: 65.3% fail (34.7% survive, March 2015 cohort).
- 15 years: 74.5% fail (25.5% survive, March 2010 cohort).
- 20 years: 80.5% fail (19.5% survive, March 2005 cohort).
Failure is a slow leak, not a cliff: the first year takes 22%, and no later year takes more. Recession-born cohorts do worse: BLS data by census division show 1-year survival falling to 71.4% in the South Atlantic for businesses born in 2008, against 75.2% nationally for that cohort.
Startup failure rate by industry
Software startups sit in the sector with the worst startup failure rate. For establishments born in the year ended March 2020, the five-year survival rate was 45.7% in Information, 50.8% in Professional, Scientific, and Technical Services, 52.6% in Health Care and Social Assistance and 61.8% in Agriculture, against 51.4% for all industries. The gap opens early: 71.6% of Information establishments born in the year ended March 2024 survived their first year, against 77.9% overall.
Do 9 out of 10 startups really fail?
Only under the loosest definition of failure, and the paper trail behind "90% of startups fail" is one sentence long. It opens the introduction of Startup Genome's 2011 "Why Startups Fail" report: "More than 90% of startups fail, due primarily to self-destruction rather than competition." The report gives no source, cohort window or definition of failure for that sentence; its own measured finding is narrower, that "74% of high growth internet startups fail due to premature scaling," based on about 3,200 self-selected tech startups. Pitch decks have repeated the premise as if it were the finding, the same path the "70% of digital transformations fail" claim took, which we traced on our digital transformation statistics page.
What the measured data support instead:
- All new businesses: 48.6% gone in five years and 80.5% in twenty (BLS), so "90%" is approached only over decades, for everything.
- Venture-backed startups: Harvard Business School's Shikhar Ghosh found that as many as 75% of venture-backed companies never return cash to investors, with 30 to 40% liquidating assets and investors losing all of their money, in a 2012 study of more than 2,000 companies that raised at least $1 million from 2004 to 2010.
- The definition ladder: Ghosh told HBS Working Knowledge in 2011 that the startup failure rate is 30 to 40% if failure means liquidating all assets, 70 to 80% if it means missing the projected return, and 90 to 95% if it means "declaring a projection and then falling short of meeting it."
So "9 out of 10" is true only when failure means missing a forecast.
How many startups fail each year?
How many startups fail each year depends on who counts: 1,981 US venture-backed companies failed in 2025, up from 1,602 in 2024 and 893 in 2023, and SVB expects 2,345 in 2026. That is the PitchBook data behind SVB's State of the Markets H2 2026 report: "an estimated 2,345 VC-backed companies are on pace to fail this year, the highest level in recent history, with startups founded during the ZIRP era now accounting for the bulk of closures." The year-by-year counts are in a16z's chart of the same data: companies founded in 2019 to 2021 make up 34% of the 2026 failures.
Venture-backed failures by year
Other counts measure narrower slices and point the same way:
- 966 startups on Carta shut down in 2024, up 25.6% from 769 in 2023 (Carta data via TechCrunch, January 2025). The count covers US Carta customers that left for bankruptcy or dissolution, and Carta's Peter Walker told TechCrunch "I bet we're missing a good chunk." Enterprise SaaS made up 32% of those shutdowns.
- AngelList recorded 364 winddowns in 2024, up 56.2% from 233 (same TechCrunch report).
- 254 startups on Carta closed in Q1 2024, "the highest quarterly total so far this decade" (Carta, July 2024); Carta has published no newer shutdown count.
- Shutdown services are still busier. SimpleClosure, a dissolution service, told Crunchbase News it closed 2.6 times more companies in Q1 2026 than in Q1 2025; its H1 2026 report gives no absolute count.
The widely quoted "3,200 venture-backed startups folded in 2023" was a global count, per a January 2024 New York Times correction, so it is not used here.
How many startups start
The denominator keeps growing. Americans filed 5,671,836 business applications in 2025, the most in the series, after 5,469,302 in 2023 and 5,224,176 in 2024 (US Census Business Formation Statistics, monthly totals summed, not seasonally adjusted). An application is a request for an Employer Identification Number, not an operating business; 1,708,842 of the 2025 filings were high-propensity applications, the ones Census expects to become employers.
Why do startups fail?
In the newest post-mortem data, 70% of failed venture-backed startups ran out of capital, but the report that measured it calls that the final cause of death, not the root problem; poor product-market fit, at 43%, is the leading root cause. That is CB Insights' "The top 9 reasons startups fail", published March 5, 2026, which analyzed 431 VC-backed companies that shut down since 2023 and found identifiable reasons for 385 of them. Reasons are not mutually exclusive, so the top four sum past 100%:
The post-mortem data
- Ran out of capital: 70%. In the report's words, "it's almost always the final cause of death, not the root problem."
- Poor product-market fit: 43%. The leading root cause.
- Wrong market timing or macro conditions: 29%.
- Unsustainable unit economics: 19%.
The internet still quotes "42% of startups fail because there is no market need." That figure comes from CB Insights' October 2014 analysis of 101 post-mortems, and it has been superseded twice: the 2021 edition, based on 111 post-mortems since 2018, put "ran out of cash or failed to raise" first at 38% and "no market need" second at 35%; the 2026 edition replaced the "no market need" label with product-market fit at 43%. Each edition is a different sample, so the three numbers are not a trend.
What founders say
In Wilbur Labs' 2026 survey of 200 US tech company founders, fielded February 3 to 12, 2026 with Wakefield Research (margin of error 6.9 points), self-reported, 81% of founders had pivoted from their original idea, 42% wished they had pivoted or changed their business model sooner, 54% named product-market fit the most important lesson from failure, and 90% experienced stress or burnout severe enough to consider quitting.
What is the MVP failure rate?
No primary research measures an "MVP failure rate," and that absence is the useful finding. The "30 to 49% of MVPs fail" range circulating in agency blogs traces to a 2014 UXPin post on Speckyboy that attributes it to "nineteen peer-reviewed research studies between 1945 and 2004" on product failure, none of them named and none about MVPs. The academic record on new products is smaller: Castellion and Markham's review in the Journal of Product Innovation Management found the "80% or higher" product failure rate repeated in trade press is a myth, with empirical studies since 1977 putting it at 40% or less. Harvard Business School has since edited the 95% figure out of its Christensen "milkshake marketing" article, which now reads "many of them fail."
The defensible adjacent numbers are the fit figures above: 43% of failed VC-backed startups had poor product-market fit (CB Insights, 2026) and 54% of founders called fit their biggest lesson (Wilbur Labs, 2026). The MVP decision in both is scope, a first version small enough to pivot before the money runs out; we have written separately about what an MVP actually costs.
Methodology: which startup failure rate each source measures
Every figure above is dated and linked to the page where it appears, and what percentage of startups fail depends on which definition the source counts:
- BLS Business Employment Dynamics measures establishments, which are locations rather than firms, and covers businesses subject to state unemployment insurance, so the self-employed are excluded. Data run through March 2025, released December 17, 2025; the March 2026 vintage is scheduled for October 28, 2026 and will move every cohort figure.
- SVB and PitchBook count bankruptcies and failures of US VC-backed companies with known founding years; 2026 is a projection.
- Carta and AngelList count shutdowns and winddowns of their own customers, so both undercount.
- CB Insights reads public post-mortems; each edition is a different sample (101 companies in 2014, 111 in 2021, 385 in 2026), not a time series.
- Harvard's Ghosh study measures failure to return investors' capital, which counts operating "zombies" as failures.
- Census Business Formation Statistics count Employer Identification Number applications; the annual totals here are calendar-year sums of the not-seasonally-adjusted monthly series.
Capital runs out last and fit runs out first, in every dataset that asked. Of all the startup failure statistics above, that is the one to build on: the startups that beat the startup failure rate are the ones whose first version was scoped to find fit before the money found its limit, which is the operating thesis behind our startup development work.