What the YC record actually says

Companies funded per year

each column is one year, shaded by batch — YC went from two batches a year to four. Numbers above the bars are that year’s total.

Sector mix, year by year

share of each year’s batch — B2B went from 31% in 2007 to 58% in 2026. Numbers in the bars are B2B’s share, in percent.

• B2B
• Consumer
• Healthcare
• Fintech
• Other

How each year’s cohort turned out

older cohorts have had time to exit or die; recent years read as almost entirely active because they have not shaken out yet. Numbers in the bars are the share now inactive, in percent.

• Active
• Public
• Acquired
• Inactive

What is getting crowded, and what is emptying out

Share of each era’s batches carrying a tag — biggest movers first. Useful for reading whether your space is filling up or being abandoned.

AI is now a bigger share of YC than SaaS ever was. It went from 2.7% to 26.4% of batches. Marketplace and e-commerce have nearly emptied out. If your idea is in a rising column you have competition; in a falling one, ask whether the market left or nobody has cracked it yet.

What the people who got in actually have

Pedigree is not one thing, so it is counted as separate markers across 1,019 fully-checked profiles. Verified is what a full professional profile confirms; mentioned is what the founder’s own YC bio reveals. The gap between the two columns is the reason bios alone cannot be trusted here.

93% of the founders who got in carry at least one of these markers. Genuinely credential-free founders are about 5%. The markers stack — the median accepted founder has several, and roughly half are already repeat founders. If you are deciding whether to borrow money on the expectation of getting in, that is the number to weigh, alongside an acceptance rate near 1%. The counterweight, from the tables further down, is equally real: once you are in, none of this predicts whether you build a top company. It is a filter on who gets through the door, not on who wins afterwards.

How many of the people who got in went to a top school

Per batch, newest first. Two measures of the same thing: mentioned counts founders whose YC bio names a top-tier school, and verified checks full professional profiles. Verified only exists for the batches that have been enriched — and it runs 17.5 points higher, because a short bio simply leaves a lot out. Read the mentioned column as a floor, never as the rate.

About 56% of the people who got in went to a top-tier school. Which means roughly 44% did not. That is the honest picture from 1,019 verified profiles, not the 39% the bios alone suggest. Pedigree is common here — more common than founder folklore admits — so anyone telling you it does not matter is not reading the same data. But it is not a gate: a large minority get in without it, and the tables above show that once you are in, it barely predicts whether you survive and does not predict a top company at all. If you are weighing real money on this, weigh it against a roughly 1-in-100 acceptance rate first, and treat your CV as the smaller variable.

Does founder background predict anything

Read from YC’s own founder bios across 2,446 companies in the 2015–2022 batches. Bios are self-written, so a miss means “not mentioned”, never “did not attend” — treat these as signals founders chose to lead with.

Background barely moves the needle — and the more data you use, the less it moves. Every signal here shifts outcomes by only a few points. Run the same test on the 2007–2019 batches and the gaps look three to four times bigger: naming big tech went with +10.8 points more exits there versus about +1 here. That is either an effect that needs a decade to show, or noise in a smaller sample — either way, the larger and more recent the window, the more the advantage disappears. Nothing in a founder’s CV predicts a top company.

Solo, pair, or three

The trade is not what founder folklore says. Solo founders shut down least often (20.7%) but also exit least (14.5%); three-founder teams die more (23.1%) and exit far more (20.4%). More founders buys variance, not safety — and it does not predict outliers: top-company rate barely moves.

A cofounder buys you upside, not safety. Every founder you add raises both the death rate and the exit rate. If you want the best odds of still existing, go solo; if you want the best odds of a real exit, don’t. Nobody should read this as “solo is safer, therefore better” — it is a different bet, not a better one.

Does being in San Francisco help

Both more death and more exits: SF companies shut down at 28% versus 19.2% elsewhere in the US, and exit at 20.4% versus 15.4%. The Bay Area raises the spread, it does not raise the floor.

San Francisco is a variance machine. It raises your chance of dying and your chance of exiting, by roughly the same 5–9 points. Being outside the Bay Area is the lower-variance path, not the losing one — and Canada quietly has the lowest shut-down rate of any group here.

Which sectors actually survive

Fintech companies shut down at 16.5%; Consumer at 37.6% — roughly 2.3× the rate. Sector is not destiny, but it is the single biggest split in this data.

• shut down
• acquired or public

Your sector is the biggest single lever in this data. Consumer kills 2.3× as often as the safest sector and produces the fewest top companies. B2B has the best combination of survival and exits. Fintech is the outlier: lowest-but-one death rate and the highest top-company rate at 1.6%. The absolute rates are floors — these batches are still young — but the ordering is the same on every window tested back to 2007.

How big the survivors get

Headcount of companies from the 2015–2022 batches that are still operating. Median is 18 people; the 90th percentile is 127. Most surviving YC companies are small businesses, not rockets.

• 1–311.0%210
• 4–1025.7%492
• 11–5039.7%761
• 51–20018.2%348
• 200+5.4%104

The median YC survivor is a 30-person company. Only 5.4% get past 200 people. The batch photo is full of companies that worked without becoming famous — which is a more realistic target than the one the headlines describe.

What happens to a batch over time

each row is one cohort, oldest first. Of the 2007 batch, 50% are gone and 40.6% exited; the 2022 batch is at 13% gone. Read down to see how long the shake-out actually takes.

Failure shows up slowly — a healthy-looking recent batch tells you nothing. The newest cohorts read as almost entirely alive because nothing has had time to happen. Real separation starts around year four and cohorts only settle near year ten, by which point roughly a third have shut down and a third have exited. Read this table downward, not across.