Why Startups Fail: AI Analysis of 1091+ Post-Mortems
The data behind every startup death. Five verdicts, four interactive charts, and six recurrent risk indicators observed in failed companies — derived from the largest open post-mortem corpus on the web.
- 1091+post-mortems
- 7ymedian lifespan
- $5.7Bavg mega burn
The post-mortems behind the verdicts below.
6 recurrent risk indicators (frequency in failed companies)
How often each indicator appears in the post-mortems in our corpus (0–10). Frequency in a failure-only sample is not a calibrated predictive probability.
The 5 verdicts from 1091+ failures
Every chart and signal below resolves to one of these five conclusions. Read them, then go deeper.
42%
Cash isn't the killer — demand is
Across 280+ post-mortems, "no market need" beats "ran out of cash" 2:1 as the root cause.
Read: Quibi$2.8B
Mega-rounds accelerate failure
The average $1B+ failure raised $2.8B before collapsing. Capital buys speed, not product-market fit.
Read: WeWork7y
Median lifespan is 7 years
In our corpus the median time from founding to shutdown is 7 years. Most cash-funded failures wind down between years 3–7.
See the timeline73%
Funded ≠ safe (in our corpus)
Among $50M+ failures we track, roughly 73% collapsed with clearly broken unit economics on record. This is a share of failures we observed, not a base failure rate for well-funded startups.
Read: FTX5x
Governance debt is fatal
Missing CFO, weak board, related-party deals — every $10B+ failure in our corpus shares the same pattern.
Read: Theranos1091+
AI-analysed post-mortems in our corpus
14%
are mega-failures (>$1B raised)
7y
median lifespan, founding → shutdown
$5.7B
avg capital destroyed per mega failure
Live values computed from the IdeaProof failure database, last updated May 2026.
The top 12 reasons startups fail — with our examples
CB Insights' post-mortem study established the canonical taxonomy of startup death. Below: each reason cross-referenced with a case study from our corpus.
| Reason | Share | What it looks like | Case study |
|---|---|---|---|
| No market need |
|
Built a category nobody asked for. | Quibi |
| Ran out of cash |
|
Burn outpaced any plausible path to profitability. | WeWork |
| Wrong team |
|
Charismatic founders, zero scientific rigor. | Theranos |
| Got outcompeted |
|
Better product, weaker distribution than Spotify. | Rdio |
| Pricing / cost issues |
|
Unit economics inverted from day one. | MoviePass |
| Poor product |
|
Shipped a glitchy EV nobody trusted. | Fisker |
| Bad business model |
|
Hardware company funded like a SaaS. | Jawbone |
| Poor marketing |
|
Lost the narrative to Apple Watch. | Pebble |
| Ignored customers |
|
Built ops nobody wanted to scale. | Beepi |
| Mistimed product |
|
Right idea, 20 years too early. | Webvan |
| Pivot gone bad |
|
Strategic whiplash burned trust and cash. | Jumia |
| Burnout / disharmony |
|
Compliance shortcuts torched the leadership team. | Zenefits |
Share percentages reflect the joint frequency in CB Insights' post-mortem study and our internal corpus. Reasons overlap — most failures cite 2–3.
Interactive failure analysis
Four lenses on the same corpus: by reason, by funding stage, by lifespan, by industry.
Failure reason distribution
Share of cited primary cause across the corpus.
- Capital & funding
- Competition
- Execution
- External shocks
- Governance
- Market & demand
- Unit economics
Failures by funding stage
Most failures cluster at $10–200M — the "growth trap".
Failure by startup age
Years from incorporation to shutdown.
- Failures
Industry distribution
Top 10 industries by failure count in the corpus.
Funding raised vs. years-to-fail
Each dot is one failed startup. No correlation between capital and survival.
Funding ≠ survival
If raising more money kept startups alive, the cloud would slope upward. It doesn't.
- Mega-rounds buy time, not viability. $1B+ failures lasted only ~1.4 years longer on average than $10M failures.
- The $50–200M zone is the danger zone. Enough capital to over-hire, too little to survive a downturn.
- Capital amplifies the underlying bet. Good bets compound; bad bets implode faster.
The 5-step kill chain
Almost every failure in our corpus walks the same five stages — usually in this order.
Misread the market
Confuse interest with demand. Skip discovery, build on conviction.
e.g. Quibi
Force growth
Hire ahead of revenue, buy users that never come back.
e.g. Jawbone
Burn the runway
CAC > LTV, burn multiple >3x. Cash converts into headcount, not value.
e.g. WeWork
Lose the team
Top performers leave first. Politics and rationalizations remain.
e.g. Zenefits
Wind down
A fire-sale, a Chapter 11 filing, or a quiet "acqui-hire" exit.
e.g. Pebble
Stage × root cause heatmap
Where each failure mode concentrates by funding bucket.
| Stage \ Cause | Market | Cash | Team | Competition | Product |
|---|---|---|---|---|---|
| <$10M |
90
|
24
|
6
|
19
|
137
|
| $10-50M |
57
|
15
|
8
|
20
|
97
|
| $50-200M |
47
|
19
|
15
|
14
|
127
|
| $200M-1B |
64
|
17
|
19
|
16
|
126
|
| $1B+ |
24
|
14
|
17
|
6
|
93
|
Darker = more failures. Market and cash dominate every funding stage; team risk concentrates at the top.
Recurrent risk indicators observed in failed companies
Patterns that appear repeatedly across the 1091+ post-mortems in our corpus. Because the corpus contains only failed companies, these are diagnostic hypotheses to stress-test — not calibrated predictive probabilities. Computing a true failure rate would require a matched sample of survivors and standard precision/recall metrics.
Negative unit economics at Series B
CAC persistently exceeding LTV after PMF claims is one of the most frequently recorded patterns in failed Series B+ companies in our corpus.
Founder-market fit mismatch
Non-technical founders building deep-tech, or consumer founders doing enterprise.
Burn multiple > 3x
Spending $3+ to generate each $1 of new ARR is unsustainable beyond Series A.
No revenue after 18 months
Pre-revenue past 18 months with a team of 20+ signals fundamental demand issues.
Single customer concentration > 40%
One customer providing 40%+ of revenue creates existential risk.
Governance red flags
Missing CFO, board cash-outs, or related-party transactions — see FTX, Theranos, WeWork.
When Metrics Lie: Startup Failures
Growth without unit economics is a ticking time bomb. These startups proved it.
Rivian (Value Destruction)
Rivian IPO'd at $150B — briefly worth more than Ford and GM. The stock fell 90% as production couldn't match hype.
Cruise (Robotaxi)
Autonomy is a decade-plus, capital-black-hole problem. When a single incident and a cover-up destroy regulator trust, no amount of parent-company cash can restart the flywheel.
Byju's
Growth at all costs through aggressive M&A and high customer acquisition without sustainable unit economics is a recipe for disaster, especially when product-market fit is superficial.
WM Motor
Capital-intensive hardware businesses require a massive minimum viable scale and aligned capital structure, which WM Motor failed to achieve despite significant funding.
How we built this analysis
Corpus. 1091+ publicly documented startup failures from 2000 to 2026, sourced from SEC filings, Chapter 11 dockets, press releases, founder post-mortems, and tier-1 financial press. We exclude failures < $1M raised and any company still operating in any form.
Categorization. Each failure is tagged with a primary reason (CB Insights taxonomy), funding raised, peak valuation, lifespan, and geography. AI-assisted clustering surfaces secondary patterns (governance, timing, GTM).
Sources. CB Insights post-mortem study, Failory cemetery, Crunchbase, PitchBook public data, SEC EDGAR, TechCrunch, Bloomberg, WSJ, The Information, and primary-source founder essays.
Limits. Our corpus skews toward US/EU venture-funded failures with public documentation. Bootstrapped failures and emerging-market shutdowns are under-represented. Percentages should be read as directional, not statistical.
Frequently asked questions
What's the #1 reason startups fail?+
Building something nobody wants. Across CB Insights' canonical post-mortem study (~150 companies) and our 280+ case studies, 'no market need' is the single most cited cause — present in roughly 42% of the CB Insights sample. Cash, team, and competition are downstream of this root cause.
Do well-funded startups fail less often?+
Not automatically. Our corpus tracks failures only, so we cannot compute a base failure rate for well-funded startups from it — that would require a matched sample of survivors. What our records do show is that among $50M+ failures, the vast majority (~73% in our corpus) collapsed with clearly broken unit economics on record. WeWork, Quibi, Fisker and FTX collectively raised over $20B and still collapsed.
How long do failed startups typically last?+
Median lifespan in our corpus is about 4 years from incorporation to shutdown. Roughly 1 in 5 fail within 24 months; another third die between years 5–10 after a "zombie" phase where they keep raising but stop growing.
Which industries have the highest startup failure rate?+
In our corpus, consumer hardware, food delivery, and clean-energy hardware show the highest density of $100M+ failures. By 10-year failure rate, food (90%), retail (88%) and construction (85%) lead. AI/ML has the lowest 1-year failure rate (33%) but the highest average capital destroyed per failure.
Are AI startups failing more in 2025–2026?+
Yes, but in a specific way. Foundation-model wrappers without proprietary data or distribution are failing fastest. Vertical AI companies with real workflows are surviving. Expect a sharp 2026 cohort of "AI-native" failures driven by margin compression and OpenAI/Anthropic platform risk.
Can you actually predict startup failure with AI?+
No — not in the strict statistical sense. Because our corpus contains only failed companies, we cannot compute precision, recall, ROC-AUC or a proper predictive accuracy. What we can report are recurrent risk indicators observed in failed companies — patterns like burn multiple >3x, CAC>LTV at Series B, founder-market mismatch, and >40% revenue concentration. They are diagnostic hypotheses to stress-test, not calibrated predictions.