Startup Statistics 2026: Failure Rates, Funding and Building With AI

Last updated: 19 October 2026

Every number on this page links to the organization that collected it: government statistics agencies, research firms and the original studies. We checked each figure against its source and say what it actually measures, because several of the most-quoted startup statistics are misread or have no data behind them.

1. 22.1% of new US businesses close within their first year. 51.4% are still open after five years and 34.7% after ten. (U.S. Bureau of Labor Statistics)
2. No dataset supports "90% of startups fail". In US government data, closures only reach about 87% after three decades. (BLS)
3. 70% of VC-backed startups that shut down since 2023 ran out of money. The root causes were poor product-market fit (43%), bad timing (29%) and weak unit economics (19%). (CB Insights, Mar 2026)
4. The median failed VC-backed startup shut down 22 months after its last fundraise. (CB Insights, Mar 2026)
5. Only 24% of US startups that raised a $1M+ seed round in 2023 have raised again or exited so far. Through 2020 the rate was 55% or higher. (Crunchbase, May 2026)
6. Startups raised $679 billion worldwide in Q1–Q3 2026, the most ever for the first three quarters of a year. (Crunchbase, Oct 2026)
7. AI startups took 82.7% of US venture deal value in 2026 through Q3, but only about 45% of deals. (PitchBook-NVCA, Oct 2026)
8. The median US seed round was about $3 million in 2025, three times its 2018 size. (Crunchbase, May 2026)
9. The average US pre-seed SAFE or note hit a record $276,000 in Q2 2026. (Carta, Aug 2026)
10. 80% of product features are rarely or never used, in Pendo's data on its customers' products. (Pendo, 2019)
11. Experienced developers took 19% longer with AI tools in METR's 2025 trial, while believing they had been 20% faster. (METR, Jul 2025)
12. 42% of companies abandoned most of their AI initiatives before production, up from 17% a year earlier. (S&P Global Market Intelligence, May 2025)

About one in five new US businesses closes within a year, half within five years and two in three within ten. For venture-backed startups, Harvard's Shikhar Ghosh found about three in four don't return their investors' money.

• 77.9% of new private-sector businesses that opened in the year to March 2024 were still operating in March 2025. (BLS)
• 51.4% of those that opened in the year to March 2020 were still operating five years later.
• 34.7% of those that opened in the year to March 2015 were still operating in March 2025.
• In the Information sector, which includes software publishers, fewer survive: 71.6% after one year, 45.7% after five and 30.0% after ten. (BLS)
• Averaged over 1994 to 2022, 67.7% of new US employer businesses survived two years, 49.2% five years, 33.9% ten years and 25.5% fifteen years. (SBA Office of Advocacy, Feb 2026)
• About three-quarters of US venture-backed firms don't return investors' capital, in Ghosh's study of more than 2,000 companies funded from 2004 to 2010. (The Wall Street Journal, Sep 2012)

How to read these: BLS and SBA count business locations with employees, across all industries. They are the best public data on survival, but they aren't limited to tech startups.

No public dataset shows it. In US government data, about two in three new businesses close within ten years. Closures only approach 90% after about three decades.

• Of the businesses that opened in the year to March 1994, 12.6% were still operating in March 2025, so about 87% had closed after 31 years. (BLS)
• The best-known early written source is Startup Genome's 2011 report, which says "more than 90% of startups fail" without citing data, and admits "we don't yet have longitudinal data". (Startup Genome, 2011)
• Ghosh's answer depends on the definition: 30–40% fail if failure means liquidation, 70–80% if it means missing the projected return, and 90–95% if it means falling short of projections. (Harvard Business School Working Knowledge, Mar 2011)

Most failed startups run out of money, but that's how they end, not why. The causes underneath are usually poor product-market fit, bad timing and weak unit economics.

• Of 385 VC-backed startups that shut down since 2023, 70% ran out of capital. The root causes were poor product-market fit (43%), bad timing or market conditions (29%) and unsustainable unit economics (19%). Startups could have several reasons, so the shares add up to more than 100%. (CB Insights, Mar 2026)
• About two-thirds of the startups that failed on product-market fit were early-stage companies that never found a market.
• In a 2026 survey of 200 US tech founders, 81% said their company had pivoted from its original idea, and 42% wished they had pivoted or changed their business model sooner. (Wilbur Labs, Apr 2026)
• In the same survey, 25% blamed running out of money as the main cause of failure, down from 38% in 2023. Technology or product issues (44%), outside factors (31%) and hiring mistakes (30%) were cited more often.
• Startup Genome's 2011 study of about 3,200 high-growth internet startups found 70% had scaled too early. It concluded that "74% of high growth internet startups fail due to premature scaling", and that 93% of startups that scale too early never pass $100,000 in monthly revenue. (Startup Genome, 2011)

Often misquoted: "42% of startups fail because there's no market need" comes from an older CB Insights analysis of 101 post-mortems, not the 2026 study above.

Testing demand before the money runs out is the job of an MVP. See why most MVPs fail and how to validate a SaaS idea.

The median failed startup shuts down 22 months after its last raise, and the gap between seed and Series A is now more than two years.

• The median time from last fundraise to shutdown is 22 months. Nearly a quarter of failed startups had gone more than three years without raising. (CB Insights, Mar 2026)
• The 431 VC-backed startups that shut down since 2023 had raised $17.5 billion between them. The median company raised $11 million.
• Nearly 50,000 VC-backed startups haven't raised funding since the start of 2023.
• US startups now take more than two years to raise a Series A after a $1M+ seed round. (Crunchbase, May 2026)
• Only 24% of companies that raised a $1M+ seed round in 2023 have raised again or exited, and 16% of the 2024 group so far. The 2024 group is still young, so its rate will rise.

Americans filed about 5.7 million applications to start a business in 2025, the most in any year since the Census series began in 2004 (our totals of Census monthly data).

• 5,671,836 business applications were filed in 2025, and 1,708,842 of them were "high-propensity" applications likely to become employers. These are our totals of the Census Bureau's monthly figures; the previous high was 5,469,302 in 2023. (U.S. Census Bureau)
• January to September 2026 brought 4,927,840 applications, about 15% more than the same months of 2025 (our calculation from Census monthly data). September 2026 alone had 534,947, seasonally adjusted. (U.S. Census Bureau, Oct 2026)
• Census projects that 29,022 of September 2026's applications will become businesses with payroll within four quarters.
• In 2025, 0.36% of American adults (360 in every 100,000) started a new business each month, up from 0.33% in 2024. (Kauffman Foundation, May 2026)
• There are 36,207,130 small businesses in the US, and 82.3% of them have no employees. (SBA Office of Advocacy, Feb 2026)
• In 2023, 1.3 million US business locations opened for the first time and about 1.2 million closed for good.

How to read these: an application is a request for a tax ID, not an operating business, and Census's 2025 covers 53 weeks (2023 had 52). Per week, 2025 is still the highest.

2026 is already a record year: $679 billion worldwide in the first three quarters.

• Startups worldwide raised $679 billion in Q1–Q3 2026, the highest amount for the first three quarters of any year. (Crunchbase, Oct 2026)
• Q3 2026 alone brought $159 billion, down 25% from Q2 but up 53% on Q3 2025, with close to 6,000 startups funded.
• In the first half of 2026, OpenAI and Anthropic alone took $217 billion, 43% of all startup funding. (Crunchbase, Jul 2026)
• All of 2025 brought $425 billion, up 30% on 2024 (Crunchbase, Jan 2026). Crunchbase has since revised 2025 to $440 billion.
• US startups raised $515.8 billion across an estimated 14,199 deals in Q1–Q3 2026, already about 44% more than 2021's full-year record. (PitchBook-NVCA, Oct 2026)
• US first-time financings reached $34.3 billion across an estimated 8,625 deals through Q3 2026, already more than 2021's full-year record of 8,507 deals.

How to read these: Crunchbase and PitchBook count deals differently, so their totals don't match. KPMG's $560.4 billion for H1 2026 uses PitchBook data (KPMG, Jul 2026). Funding totals are also revised as late rounds come in, so quote each number with its report date.

Most of it. AI took about half of global venture funding in 2025 and more than 80% of US deal value in 2026 so far.

• About half of global venture funding in 2025, $211 billion, went to AI companies, up 85% on 2024. (Crunchbase, Jan 2026)
• AI took 64% of global venture funding in Q3 2026 ($102 billion), after 80% in Q1 and more than 70% in Q2. (Crunchbase, Oct 2026)
• AI took 82.7% of US venture deal value in 2026 through Q3, the highest share PitchBook has recorded, but about 45% of deals by count. (PitchBook-NVCA, Oct 2026)
• AI companies took 49% of US pre-seed dollars on Carta in H1 2026, about the same as 2025's 50%. (Carta, Aug 2026)
• Global private investment in AI reached $344.7 billion in 2025, up 127.5%. Generative AI companies took $170.9 billion of it. (Stanford AI Index 2026)
• US private AI investment was about $285.9 billion in 2025, 23.1 times China's $12.4 billion.
• 1,953 AI companies were newly funded in the US in 2025, against 172 in the UK and 161 in China.

The median US seed round is about $3 million, and PitchBook's median seed valuation is $18.3 million before the investment.

• The median US seed round was about $3 million in 2025, three times its 2018 size. (Crunchbase, May 2026)
• The median US seed valuation before investment (pre-money) is $18.3 million in 2026 so far, up from $16.0 million in 2025. The median seed deal fell to $3.0 million from $3.4 million. (PitchBook-NVCA, Oct 2026)
• On Carta, the median seed valuation after investment (post-money) reached a record $24 million in Q4 2025, and the median priced seed round was $4.5 million. (Carta, Mar 2026)
• US seed funding totaled $19.4 billion in 2025, and 51% of it went into rounds of $10 million or more. (Crunchbase, Mar 2026)
• Global seed funding was $13 billion in Q3 2026, after $12 billion in each of Q1 and Q2. (Crunchbase, Oct 2026)
• In 2025, the top 10% of US startups on Carta that closed a round raised about half of all capital, and the bottom 50% raised 14%. (Carta, Mar 2026)

How to read these: each dataset defines "seed" differently, and PitchBook reports pre-money valuations while Carta reports post-money. Compare numbers only within one source.

The average pre-seed SAFE or convertible note is about $276,000, and almost all pre-seed rounds now use SAFEs.

• US startups on Carta raised $3.19 billion across more than 11,500 pre-seed SAFEs and convertible notes in Q2 2026. The average instrument was a record $276,000, up 27% on a year earlier. (Carta, Aug 2026)
• 93% of pre-seed rounds on Carta in Q2 2026 were SAFEs, 91% of those were post-money SAFEs, and 94% of post-money SAFEs in H1 2026 had a valuation cap. (Carta, Aug 2026)
• Median valuation caps ranged from $10 million for SAFEs under $250,000 to $35 million for SAFEs of $2.5 million or more.

How to read these: $276,000 is the average per SAFE or note, not per round. A round often stacks several instruments.

UAE startups took about two-thirds of MENA venture funding in H1 2026, after Saudi Arabia led in 2025.

• MENA startups raised $1.35 billion across 214 deals in H1 2026, down 22% on a year earlier. UAE startups took $895 million, 66% of it. (MAGNiTT, Jul 2026)
• MENA startup funding grew 74% to a record $3.8 billion across 688 deals in 2025. (MAGNiTT, Jan 2026)
• UAE startups raised $1.58 billion across 231 deals in 2025, up 67%, second only to Saudi Arabia's $1.72 billion. (MAGNiTT for Saudi Venture Capital, Jan 2026)
• Counting debt, MENA startups raised a record $7.5 billion in 2025, $4 billion of it debt. (Wamda, Jan 2026)
• Counting debt, UAE startups raised $1.2 billion across 83 deals in H1 2026, about 70% of MENA's $1.7 billion. (Wamda, Jul 2026)
• The Dubai Chamber of Digital Economy helped 1,690 digital startups set up or expand in Dubai in 2025, up 39.7%. About 15% were AI businesses. (Dubai Media Office, Jan 2026)

How to read these: MAGNiTT counts equity only and Wamda includes debt, so their totals differ a lot. The Dubai figure counts startups the chamber supported, most of them global companies; it isn't a count of startups founded in Dubai.

Building in the UAE? See our Dubai product studio.

Most aren't. In Pendo's data, 80% of features are rarely or never used.

• 80% of features are rarely or never used (24% never, 56% rarely), and 12% of features generate 80% of daily usage. (Pendo, 2019)
• In the median product, 6.4% of features drive 80% of user clicks. In the top 10% of products it's 15.6%. (Pendo)

Often misquoted: "80% of features are never used". Pendo's own chart shows 24% never used and 56% rarely used.

That's the case for cutting scope without cutting the outcome.

Big projects fail far more often than small ones. Large IT projects run 45% over budget on average.

• The Standish Group rated 29% of software projects successful in 2015 (on time, on budget, with a satisfactory result). 52% were challenged and 19% failed. (Standish Group CHAOS Report 2015)
• 61% of small software projects succeeded, against 6% of the largest. 43% of the largest projects failed, against 7% of small ones.
• 39% of agile projects succeeded, against 11% of waterfall projects.
• Large IT projects run 45% over budget and 7% over time on average, while delivering 56% less value than predicted. (McKinsey and University of Oxford, 2012)
• 17% of IT projects go so badly they can threaten the company's existence.
• Each extra year a large IT project is scheduled to run adds 15% to its cost overrun.

How to read these: the 2015 CHAOS report is the latest edition Standish has published for free. "Challenged" isn't the same as "failed", so "71% of projects fail" is a misreading.

It depends on the task and the developer. A controlled GitHub test found a 55% speed-up, but a randomized trial with experienced developers found they were 19% slower.

• Developers using GitHub Copilot finished a set coding task 55% faster: 1 hour 11 minutes on average, against 2 hours 41 minutes without it. (GitHub, Sep 2022)
• Experienced open-source developers took 19% longer when allowed AI tools, though they expected a 24% speed-up and afterward believed they had been 20% faster. (METR, Jul 2025)
• METR's late-2025 follow-up pointed toward a speed-up (18% less time for returning developers, 4% for new ones), but the margins of error for both include zero and METR calls the data unreliable. (METR, Feb 2026)
• Google's DORA 2024 report estimated that each 25% increase in AI adoption came with a 1.5% drop in delivery throughput and a 7.2% drop in delivery stability. (Google DORA, Oct 2024)
• DORA's 2025 research found AI adoption now improves delivery throughput, reversing 2024, but still increases instability. 90% of respondents use AI at work, and 30% have little or no trust in AI-generated code. (Google DORA, Sep 2025)
• 83% of developers use at least one AI tool at work, and 66% use AI coding assistants or agents. (Stack Overflow Developer Survey 2026)
• 69% say "plausible but wrong code" is the AI mistake that cost them the most time in the past year. (Stack Overflow 2026)
• Only 9.6% use AI-generated answers as they are; 76.5% run the code locally first. (Stack Overflow 2026)

• PostgreSQL is the most-used database: 57.9% of developers worked extensively with it in the past year, and it's also the most admired (62.3%). (Stack Overflow 2026)
• Node.js (46.4%) and React (41.5%) are the most-used web technologies. Next.js is used by 20.6%. (Stack Overflow 2026)

• The median annual wage for US software developers was $135,980 in May 2025. (BLS Occupational Outlook Handbook)
• Employment of software developers is projected to grow 10% from 2025 to 2035, against 3% for all jobs. (BLS)

It depends who's counting and what "fail" means. Estimates range from Gartner's prediction that 30% of generative AI projects would be abandoned after proof of concept, to S&P Global's finding that 46% of proofs of concept are scrapped.

• 42% of companies abandoned most of their AI initiatives before production, up from 17% a year earlier. On average, 46% of proofs of concept were scrapped before broad adoption. (S&P Global Market Intelligence, May 2025)
• MIT NANDA's 2025 report says 95% of organizations are getting zero return from generative AI, and just 5% of integrated AI pilots are extracting millions in value. (MIT NANDA, Jul 2025)
• In the same report, 60% of organizations evaluated enterprise AI tools, 20% reached a pilot and 5% reached production.
• AI tools from outside partners reached deployment about 67% of the time, against about 33% for tools built in-house.
• In July 2024, Gartner predicted at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025. (Gartner, Jul 2024)
• In June 2025, Gartner predicted over 40% of agentic AI projects will be canceled by the end of 2027, and estimated only about 130 of the thousands of "agentic AI" vendors are real. (Gartner, Jun 2025)
• CEOs say only 25% of their AI initiatives delivered the expected return, and 16% have scaled across the company. (IBM Institute for Business Value, May 2025)
• In BCG's 2026 study, nearly half of companies now get meaningful value from AI: 7.5% are "future-built" and 41% are scaling. In 2024, 74% had yet to show tangible value. (BCG, Sep 2026; BCG, Oct 2024)

Often misquoted: "MIT found 95% of AI projects fail". MIT measured organizations with no measurable profit impact from generative AI, based on interviews with 52 organizations, a survey of 153 senior leaders and a review of 300+ public AI initiatives. "RAND found 80% of AI projects fail" is also wrong: RAND repeated an earlier estimate; its own research was 65 interviews (RAND, Aug 2024).

Planning an AI product? See how to build an AI MVP and evaluating LLM features before shipping.

About a quarter of US employer businesses use AI, and nearly nine in ten organizations in McKinsey's global survey do.

• 23.8% of US employer businesses used AI in at least one business function in the previous two weeks, and 27.7% expect to within six months. (U.S. Census Bureau BTOS, Oct 2026)
• 89% of respondents to McKinsey's 2026 global survey say their organization regularly uses AI in at least one function, and 44% say it's scaling across the company. (McKinsey, Aug 2026)
• 37% attribute at least some profit (EBIT) impact to AI. Only about 6% attribute 5% or more of EBIT to it.
• 32% say their organization decided against buying at least one software product because it could build the feature in-house with AI coding tools.
• Companies spent $37 billion on generative AI in 2025, up from $11.5 billion in 2024. (Menlo Ventures, Dec 2025)
• 76% of enterprise AI use cases are now bought rather than built in-house, up from 53% in 2024.
• Gartner forecasts worldwide AI spending of $2.7 trillion in 2026, up 49.5%. (Gartner, Sep 2026)

We looked for independent data and found none. The MVP cost and timeline figures online come from agencies' own estimates, with no published method or sample, so we haven't included them.

Our MVP cost guide explains what drives the price, our MVP development service lists fixed prices, and our case studies show what we've built.

• Original sources only. Every figure comes from the organization that collected the data: government agencies, research firms and the original studies. We didn't use stats roundups.
• Checked against the source. On 19 October 2026 we opened each source and confirmed the number and wording. Where a page has moved or is paywalled, we link the Internet Archive copy.
• Our own math is labeled. Where we added or subtracted figures (for example, yearly Census totals), we say so.
• Updated as sources update. Funding data is revised as late rounds are reported, so each number carries its report date.

If you quote a number, please cite the original source we link to, and this page: