Taking Stock of the Seed Stage<br>Argument map<br>Modepublic essay<br>Statusmodel based estimate<br>DateJune 12, 2026
Decision question<br>How many seed-stage companies are alive at any point in time, and is that stock rising or falling?
Source trail<br>Graduation benchmarks, aggregate Seed and Series A deal counts, modeled cohort graduation curves, inferred exits, and a stock-flow identity.
Argument mapModepublic essay<br>Statusmodel based estimate<br>DateJune 12, 2026
Decision question<br>How many seed-stage companies are alive at any point in time, and is that stock rising or falling?
Source trail<br>Graduation benchmarks, aggregate Seed and Series A deal counts, modeled cohort graduation curves, inferred exits, and a stock-flow identity.
Four years ago I set out to correct a long-standing misnomer in how founders and investors talk about startup valuations. Turns out, it was a thornier issue than I realized, and a proper fix required solving a few puzzles first. This is the first of two essays building up to that answer.
Here’s the first insight: the population of active Seed-stage startups is shrinking. Not necessarily the funding (in aggregate dollars or the average round size) but rather the number of active Seed companies.
Despite the AI boom, exits now outpace new Seed financings, and no one seems to have noticed.
Join the Syndicate<br>Long-form essays on the economics of tech, with the data to back it up. All signal, no noise.<br>Read by 5,000+ founders, engineers, and investors.<br>Subscribe_
Missing data
To make that claim, I needed a measure of the total number of seed stage companies that are “alive” at any moment. This is surprisingly difficult to locate, as no one tracks this.
This is understandable – companies announce their initial Seed financings, they might publicly celebrate their Series A funding, but they typically don't announce shut downs (ironic, given it's the most common outcome). They mostly just die off (“quiet quitting” as they say). When startups do announce an abrupt exit, databases like PitchBook don’t always record this, and even when they do you can't extract the data at an aggregate, "macro" level.
This creates a missing data problem. One can get around this, but it requires making assumptions about the way the world works that allow you to plausibly "guess" what the missing data looks like.
In this case, I needed to make assumptions about the graduation and “death” / exit rate of Seed stage companies. Knowing this, I could estimate the count of live Seed startups by adding up all the surviving companies across past and present Seed cohorts, estimated with those graduation and exit rates.
Observed flow
New Seed deals
Seed financings tell us which companies enter the population.
Partially observed
Graduations
Series A rounds reveal when Seed companies leave the Seed stage successfully.
Inferred flow
Exits and deaths
Shutdowns and small acquisitions are usually quiet, so they have to be modeled.
Graduation statistics
Thankfully, there is some data out there on graduation rates that I can leverage in multiple ways.
I’ve previously discussed the shifting winds around the graduation of Seed stage companies to Series A, particularly after the 2021 boom period:
... the numerous startups that raised Seed funding during the boom are now facing a massive bust... It's safe to say the graduation or survival rate from Seed to A has fallen by more than half. -- The Series A Bust
This has been well-documented by multiple sources:
The percentage of companies who make it from seed to Series A within two years fell by a lot for the 2021 seed cohort.
27.5% of companies that raised a seed round in 2019 made it to Series A within 2 years.
Only 17.6% of companies who raised their seed in 2021 have "graduated" to the next round. – Peter Walker (Carta)
It’s hard to believe today, but at some point in the not-too-distant past over 50% of Seed stage companies “graduated” to Series A. That’s unthinkable these days, as most recent Seed cohorts are trending well below that:
I fit S-curves to various benchmarks of Seed conversion to Series A. That part was easy enough (and yes I realize they don’t start right at zero, this was hard to enforce for various technical reasons):
However, I also needed two more things:
something to inform the earlier Seed cohorts that Carta and other sources don’t publish, and
some way to ground this in the Series A data, most of which represents Seed companies graduating to the next stage.
In other words, the number of graduations our model estimates should correspond to the quantity of Series A deals getting done.
I assume that all Series As “come from” a Seed company that was part of a prior cohort. This is not exactly true: some companies go straight to Series A as their first round of funding. This was more common in the distant past than it is today. So I’m cheating here slightly to keep things simple, but I think it’s fine.
I embedded this in a larger statistical...