Y Combinator is the most legible startup cohort in the world: every company is published, dated and searchable. So it is a good place to ask a question the register asks of everything it holds — where is this, actually? — and to see what the answer looks like when you count rather than assume.
Here are 6,631 unique YC companies, geolocated to UN M49 regions. Every level below sums to the same total, which is the only reason the shares are quoted to a decimal place.
6,631
YC companies
Unique, deduplicated
73%
In one country
4,853 in the United States
87
Countries represented
42 of them have two companies or fewer
5.3%
No country on file
352 companies — see below
One country, then everything else#
The Americas take 5,231 of 6,631 — 78.9%. Almost all of that is Northern America: 5,014 companies, 75.6%, of which the United States is 4,853 on its own. Add Canada’s 160 and Bermuda’s single entry and the continent is accounted for above the Rio Grande.
Everything else — every other region, every other country, and the companies with no country at all — is 1,617 companies. 24.4%.
That is not a surprising result. It is worth having the number anyway, because “YC is mostly American” and “three in four YC companies are in one country” are sentences that get treated as equivalent and are not.
Outside Northern America#
Northern Europe leads the rest at 276, and that is mostly the United Kingdom: 218 of it. Southern Asia is 228, of which India is 215 — 94% of the subregion. Latin America and the Caribbean is 217 across thirteen countries, the most evenly spread block on the list.
A few things that only show up at country level:
- The United Kingdom and India are tied for second, at 218 and 215. Nothing else is within a factor of two of them.
- Nigeria and Brazil have the same count, 50 each. Nigeria is 56% of the whole of Africa.
- Singapore has 56 — more than Brazil, and more than Spain and Sweden combined.
- China has 8. Japan has 2. Eastern Asia as a whole is 29, fewer than Colombia.
The fourth-largest region is not a region#
Sort the regions by size and the fourth row is No region on file: 352
companies, 5.3% of the corpus, made up of three different kinds of not-knowing.
These are not the same thing and should not be averaged into one another.
Remote / distributed is an answer — 93 companies that genuinely do not have a
headquarters to record. Unspecified is a company that never published one.
Unknown is a company that published something we could not resolve to a
country.
Only the third is a defect. The other two are the corpus being honest, and folding all three into an “Other” row — or dropping them, which is what most charts of this kind quietly do — would make every share above look better than the evidence supports. Africa and Oceania together are 109 companies. The not-knowing is 352.
We have written about this before: a missing value means the field was never recorded, not that the answer is zero.
The tail is most of the map#
87 countries · 6,279 companies
Regions
Marker area
The shape is legible before any number is: one marker over North America large enough to cover the continent, a thin band across Western Europe and South Asia, and single sightings almost everywhere else. The map is the establishing shot — the counts below are the reading.
Eighty-seven countries appear at least once. Forty-two of them — nearly half the map — hold two companies or fewer, and between them account for 54 companies: 0.8% of the total.
This is the shape of almost every register we keep, and it is the reason a country filter behaves the way it does. Filtering to Portugal returns two companies not because Portugal is quiet but because two is what the evidence supports. The tail is real, thin, and mostly single sightings.
- United States4853
- United Kingdom218
- India215
- Canada160
- Mexico83
- France76
- Germany67
- Singapore56
- Brazil50
- Nigeria50
- Israel35
- Indonesia30
- Colombia29
- Spain21
- Sweden20
- Argentina19
- Australia17
- United Arab Emirates16
- Chile16
- Netherlands15
- Denmark13
- Switzerland13
- Egypt12
- South Korea11
- Pakistan11
- Kenya9
- Norway9
- Ireland8
- China8
- Philippines8
- Malaysia8
- Peru8
- Hong Kong7
- Poland6
- Slovenia6
- Vietnam6
- Austria5
- Panama5
- Belgium5
- Ghana4
- Turkey4
- Saudi Arabia4
- Morocco3
- South Africa3
- Estonia3
- Japan2
- Portugal2
- Senegal2
- New Zealand2
- Greece2
- Ukraine2
- Croatia2
- Romania2
- Bulgaria2
- Georgia2
- Finland2
- Costa Rica2
- Thailand1
- Nepal1
- Bangladesh1
- Bermuda1
- Russia1
- Hungary1
- Latvia1
- Puerto Rico1
- Iceland1
- Czechia1
- Iraq1
- Kyrgyzstan1
- Tanzania1
- Algeria1
- Kazakhstan1
- Italy1
- Lithuania1
- Ivory Coast1
- Zambia1
- Ecuador1
- Cyprus1
- Uganda1
- Bahrain1
- Venezuela1
- Jordan1
- Ethiopia1
- Taiwan1
- Uruguay1
- DR Congo1
- Cayman Islands1
What this does and does not say#
It says where the register currently records these companies. That is a narrower claim than it looks, in four ways.
- It is a location on a record, not a birthplace. A company that went through YC in Mountain View and now operates from Lagos is counted in Nigeria. We hold where a company is, not where it was.
- It describes YC’s selection, not startup geography. An accelerator’s intake reflects who applied, who was reachable, and who was admitted. Read backwards from this to “where startups are” and you will be wrong in an entirely predictable direction.
- It is the companies we hold, not a claim to hold every company YC has ever funded. Coverage is good and it is not complete.
- Names are not identities. Counting companies means having first decided which records are the same company — the part of this that is genuinely hard.
Three quarters of the map is one country. Half the countries on it are a single sighting. Both of those are facts about the evidence before they are facts about the world.