Datary Technologies
Datary TechnologiesThe odds

Building got easier.
Being right didn’t.

Most startups don’t close because their founders didn’t work hard enough. They close because they built something too few people needed, at a cost that didn’t add up, at a moment that wasn’t ready. AI has made building faster than ever. It hasn’t changed what decides the outcome.

OUT OF EVERY 100
  1. 100new businesses open their doors.All U.S. private-sector businesses
  2. 20have closed within a year.
  3. 49have closed within five years. About half.
  4. 65have closed within ten. Nearly two in three.
  5. 62of the 4,369 venture-tracked U.S. startups founded in 2018 had shut down seven years later. 32 were still running, 5 had been acquired and 1 had become a unicorn.Startups on Carta, as of 2025

“Nine in ten” is the number you hear most. The honest answer depends on what you call failure, and when you count. Among venture-backed companies, about three in four don’t return their investors’ capital. Fewer than two in a hundred become unicorns.

Why they fail

Running out of money is how it ends. It is rarely why.

CB Insights studied 385 startups that shut down since 2023. Seven in ten ran out of capital. Underneath, the same causes keep returning: a product the market didn’t need, the wrong moment, and economics that never worked.

Companies could give more than one reason, so the total is above 100%.

The AI era

More people can build. The odds haven’t moved.

AI coding tools can turn a weekend into a working product. That is real progress, and it is why more people are starting companies than ever. But the shutdown data still tells the same story. What changed is how quickly an untested idea can be built, launched and abandoned.

5.67M

new business applications in the U.S. in 2025, the most on record and 58% more than in 2019.

U.S. Census Bureau
36%

of new U.S. startups had a single founder in the first half of 2025, up from 24% in 2019.

Carta
1 in 4

of Y Combinator’s startups in early 2025 had 95% of their code written by AI, according to its CEO.

Y Combinator
21%

of AI-app subscribers are still paying a year later. For other apps it is 31%.

RevenueCat, 2026
40%

median gross revenue retention for AI-native companies in 2025. Tools under $50 a month kept 23%.

ChartMogul

When anyone can build in a weekend, building stops being the advantage. Knowing what deserves to be built becomes the advantage.

Before you build

Find the signal before you scale.

In the Startup Genome study of more than 3,200 high-growth startups, premature scaling was the most common pattern behind failure: growing the team, the spend or the product before the market had answered.

  1. 01Talk to the people who have the problem. Ask what they do about it today.
  2. 02Look for commitment, not compliments: a pilot, a deposit, a signed letter.
  3. 03Price it early. Check the unit economics before you grow them.
  4. 04Decide in advance which result would make you stop.
Sources and notes

Figures are from the United States, where the most complete data exists. “Closed” in government data counts every reason a business stops, including retirement or a sale.

  1. U.S. Bureau of Labor Statistics, Business Employment Dynamics: of establishments born in March 2013, 79.6% were operating a year later, 50.6% after five years and 34.7% after ten (2024). Closures here include every reason a business stops, not only failure.
  2. Carta data shared by Peter Walker, then Carta’s head of insights (February 2025): of 4,369 U.S. startups founded in 2018, 61.9% had closed, 5.5% had been acquired, 1.3% were private unicorns and 0.34% had gone public; the rest were still operating.
  3. Shikhar Ghosh, Harvard Business School, reported in The Wall Street Journal (“The Venture Capital Secret: 3 Out of 4 Start-Ups Fail”, September 2012): about three-quarters of venture-backed U.S. firms don’t return investors’ capital, from 2,000+ companies funded, generally with at least $1M, from 2004 to 2010.
  4. CB Insights, “The top 9 reasons startups fail” (March 2026): 431 VC-backed shutdowns since 2023; reasons identified for 385. Companies can cite more than one reason, so the total exceeds 100%.
  5. U.S. Census Bureau, Business Formation Statistics (via FRED, series BABATOTALNSAUS): 5,672,436 business applications in 2025 against 3,599,590 in 2019. Applications are not all startups, but they show how many people are starting something.
  6. Carta and Solo Founders, Solo Founders Report 2025: the share of new startups with a solo founder rose from 23.7% (2019) to 36.3% (first half of 2025).
  7. Garry Tan, CEO of Y Combinator, speaking to CNBC (March 2025): “for about a quarter of the current YC startups, 95% of the code was written by AI.”
  8. RevenueCat, State of Subscription Apps 2026 (March 2026): annual subscriber retention of 21.1% for AI apps against 30.7% for other apps; AI apps churn about 30% faster.
  9. ChartMogul, “The AI churn wave” (December 2025): median gross revenue retention of 40% for about 200 AI-native companies with at least $250k in annual revenue; 23% for AI products under $50 a month.
  10. SimpleClosure, State of Startup Shutdowns 2025: AI companies made up 15.9% of shutdowns; thin application-layer “wrappers” saw the sharpest correction.
  11. Startup Genome Report Extra on Premature Scaling (2011), from about 3,200 high-growth internet startups: “74% of high growth internet startups fail due to premature scaling.”