We rebuilt the funding ladder for 6,957 US technology companies from 615,718 SEC Form D filings, then ran a survival model on it. The hypothesis was right: there are hard tipping points after a Series A. They arrive earlier than almost anyone plans for.
A Series A buys 18–24 months of runway, but the Series B hazard peaks at month 21 and has fallen below half its peak by month 36. The clock and the cash run out at the same time — and the clock wins.
All figures from our own reconstruction of SEC Form D filings, 2014–2026. Cohort-level results use Series A vintages 2014–2019, which have 78+ months of observation.
01
Of companies that raise an A-scale round, 30.3% ever raise a genuinely larger follow-on. Under a looser definition it is 39%; under a stricter one, 23%. Crunchbase puts the 2020–21 cohort at 36%; Carta, which sees full cap tables, puts it higher at 40–50%. Our level is conservative — the timing is what matters here.
02
Nothing happens before month 9. The quarterly hazard peaks at 3.7% in the quarter ending month 21, then decays monotonically. Half of all eventual Series B raisers have closed by month 24.
03
A company still B-less at month 36 has a 10.2% chance of ever raising one — a third of its day-one odds. By month 48 it is 5.5%. The hazard never recovers.
04
Of those who never graduate, 65% never file another financing of any size. They do not fail loudly. They stop appearing in the record.
05
Contrary to the standard narrative, median time-to-B among companies who make it within three years has barely moved: 18.5 → 18.3 → 17.2 months across the 2014-17, 2018-20 and 2021-23 eras. What collapsed is how many make it, not how long it takes those who do.
06
Among companies still B-less at month 24, those who had taken a smaller interim round went on to raise a real B 25.8% of the time versus 15.8% for those who had not — a 1.6× lift. Either way, the eventual round lands around month 40, deep in the tail.
So what
The urgency is real, but it is not where founders feel it. The pressure point is not month 24 when cash runs low — it is month 9 to 12, when the metrics that will be diligenced are already being generated. By the time the runway is visibly short, the outcome is largely set.
Conditional probability that a company still without a Series B at month m will ever raise one. Kaplan–Meier estimate, Series A vintages 2014–2019 (n = 3,093).
Share of still-waiting companies that close a Series B in each quarter · Series A vintages 2014–2019
The window opens at month 9, peaks in the quarter ending month 21, and falls below half its peak at month 36. After month 48 the hazard is statistical noise.
Cumulative graduation (purple) against conditional probability of ever graduating (lime)
The two curves cross at roughly month 27 — the moment when a company has a better chance of being already funded than of ever being funded.
Each gate is a point where the conditional odds step down and never recover.
Only 3.9% of the cohort has raised. Nothing looks wrong. But the metrics that will be diligenced in the next six months are being generated right now.
Typical Series A runway is exhausted. Half of everyone who will ever raise a B has already done it. Odds have fallen below the point where the average outcome is a raise.
The hazard has fallen below half its peak. 72% of all eventual graduates are already through. This is the point of no return in the data.
Effectively terminal. What follows is recapitalisation, acquihire, a quiet slide to break-even, or nothing at all.
Conditional odds of ever graduating, by how strictly a Series B is defined
The absolute level moves with the definition; the decay pattern does not. Half-life of the odds is 27 months under both the loose and base definitions, 30 months under the strictest.
Cumulative share of each Series A vintage that had raised a Series B by 12, 24 and 36 months.
Kaplan–Meier cumulative incidence · bars omitted where the vintage is not yet old enough to observe
The 2022 vintage is the worst on record: 5.3% at 24 months, against 24.5% for 2020. 2023 recovered to 11.2% and the 2025 vintage is running at 4.7% by month 12 — the best 12-month rate since 2020.
So what
A 2022 Series A company was roughly four times less likely to reach a Series B in two years than a 2020 one, at the same quality of execution. When you plan a raise, you are partly betting on a market you do not control — which is an argument for getting to the window early rather than optimising into it.
Half of all Series A companies raise something again. Only 30% raise something bigger. The gap between those two curves — about 24 points — is the bridge economy: flat rounds, insider extensions, and structured paper that keeps the lights on without resetting the story.
Of the 2,133 companies in our mature cohorts that never graduated, 1,389 (65%) never filed another financing of any kind. The remaining 744 raised only smaller or flat rounds — a median of one.
Cumulative share of the 2014–2019 Series A cohort
The bridge lift is real but partly selection: a company that can raise a bridge already has insiders willing to fund it. Read it as "insider conviction is the single best observable predictor," not "take a bridge and your odds improve."
Share raising a Series B within 60 months · 2014–2019 vintages
An inverted U. The $8–18M band graduates best; both the under-funded and the over-funded do materially worse.
Companies raising $18–30M at the A graduate at 25.5% — no better than those raising $5–8M, and seven points worse than the $8–18M band.
The mechanism is a bar you set for yourself. A larger A means a higher post-money, which means the step-up required to clear a Series B mark is larger, which means the same execution reads as a flat round. Big A rounds do not buy time; they buy a harder exam.
Geography moves the number less than people assume: 29.2% in California and 31.0% in NY/MA against 25.5% for the rest of the US — a real coastal edge, but worth about five points, not a category difference.
If the median successful Series B closes around month 18–21, every upstream requirement has to land much earlier than founders plan for.
Series B closes. Median for companies that graduate within three years.
Process starts. A Series B takes 4–6 months from first meeting to wire.
The diligence window. Investors want two to three consecutive quarters of a go-to-market motion that repeats.
The engine has to be running. Pipeline, conversion, retention and attribution must already be producing clean, consistent numbers — because these are the quarters that get shown.
Series A closes. The measurement infrastructure needed at month 10 has to be built now; you cannot retrofit four quarters of clean data.
The Series B is not won in the raise. It is won in the four quarters that get shown during the raise — which means the go-to-market engine has to be instrumented and repeatable by month nine. Everything after that is reporting.
Source
Every quarterly Form D structured data set from 2014 Q1 to 2026 Q2, downloaded from SEC EDGAR. Form D is the notice a company must file for a private securities offering, so nearly every priced US venture round leaves one.
Universe
US-domiciled C-corporations filing an equity offering under a technology industry group. Filings were collapsed into 32,717 financing events, with amendments and tranches inside 120 days merged into a single round.
Definition
Form D does not carry round names. A "Series A" is a company's first equity event of $5–30M that is at least a 1.5× step-up on anything prior; a "Series B" is a later event of at least max($12M, 1.3× the A). Median detected A is $9.5M — squarely on the published Series A median.
Model
Companies are observed to 30 June 2026, so recent vintages are right-censored. All cohort figures are censoring-aware and suppressed where a vintage is too young to observe the horizon.
Limit 1
Some rounds never file, or file under a different classification. Validation against known ladders: Ramp's five rounds reconstruct exactly; Vanta's and Writer's earlier rounds are missing. This biases the graduation level down and is why our 30.3% sits below Crunchbase's 36%. It does not bias the timing, which is the finding.
Limit 2
Form D carries no sector detail below the industry group and no revenue, headcount or shutdown data. We cannot separate B2B SaaS from AI-native, and "never raised again" is a proxy for failure, not a measurement of it.
Our estimate against the three most-cited external datasets
| Measure | This study | External benchmark | Source |
|---|---|---|---|
| Series A → B, eventually | 30.3% (23–39% by definition) | ~36% of 2020–21 US Series A cohort | Crunchbase News |
| Series A → B by 24 months, 2018 vintage | 14.6% | 25% for the Q3 2018 vintage | Carta |
| Series A → B by 24 months, 2022 vintage | 5.3% | 9% for the Q3 2022 vintage | Carta |
| Median months A → B among graduates | 23.5 | 28 median / 31 mean (2024) | Crunchbase News |
| Graduation by year 4, 2018–20 vintages | 28.7–31.6% | 40–50%+ | Carta (10,755 startups) |
Our levels run consistently below Carta's because Carta sees complete cap tables while Form D sees only filed offerings, and because our step-up test excludes labelled Series B rounds that were flat or down. Directionally every series agrees, including the collapse of the 2021–22 vintages.
To go deeper
Three gaps a paid data source would close: (1) real round labels and sector tags — PitchBook or Crunchbase Pro would let us split B2B SaaS from AI-native and use actual Series A/B designations rather than size heuristics; (2) outcomes — acquisition, shutdown and going-concern status, so "never raised again" becomes a measured outcome instead of a proxy; (3) performance at the A — ARR, growth and headcount, which would turn this from a timing model into a model of what actually earns the B.