Crunchbase finds 2 US AI startups for every European one. Dealroom finds 1.
Published 28 September 2026
Yesterday we counted 4,067 funded AI startups in San Francisco and 4,147 in the 27 EU countries in our copy of Crunchbase (San Francisco has as many AI startups as the entire EU). The objection we expected first: Crunchbase is an American database and misses European companies. So we put its counts next to those of Dealroom, a startup database run from Amsterdam, and next to the national AI startup maps of Germany and France, whose startups we then looked up in Crunchbase one by one.
Crunchbase vs Dealroom: AI startup counts in the US and Europe
- For AI startups founded 2015 to 2020, Crunchbase lists 934 a year in the US and 476 in Europe, 2 to 1. Dealroom counts about 900 new VC-backed AI companies a year on each side, 1 to 1.
- In the Bay Area, Crunchbase lists 6,246 VC-backed AI startups, 1.38 times Dealroom's 4,529. In New York City it lists 2,376, 1.33 times Dealroom's 1,786.
- In the ten EU countries of Dealroom's country table, Crunchbase lists 2,850 AI startups across all sectors. Dealroom counts 3,248 in ten industrial sectors alone.
- Germany's national map lists 935 AI startups founded since 2015, funded or not. Crunchbase has 800 funded ones of that age, and 287 in Berlin against the map's 283.
- France Digitale's map lists 1,114 French AI startups, 67% of them funded. Crunchbase has 617 funded ones.
- Matched company by company, 79% of the startups on the German map and 75% of those on the French map are in Crunchbase. Only 52% and 38% carry its AI category.
- The sets overlap much less than the counts suggest. 41% of Crunchbase's 800 funded German AI startups are on the German map, and 46% of its 617 French ones are on the French map.
The same database on both sides of the Atlantic
Dealroom's March 2026 report on European AI, written with Prosus, draws a funnel for the US and Europe that starts at about 900 new VC-backed AI companies a year on each side. Crunchbase agrees on the US, with 934 funded AI startups a year founded 2015 to 2020. For Europe it has 476.
Other founding-year windows give similar ratios: 2.1 US AI startups for every European one for 2013 to 2022, and 2.3 for 2015 to 2024. Europe here is the 27 EU countries plus the United Kingdom, Switzerland, Norway, Iceland and Liechtenstein. Dealroom's report does not say which countries its Europe covers.
Dealroom's own counts, place by place
| Place | Dealroom | Crunchbase | Crunchbase as % of Dealroom |
|---|---|---|---|
| Bay Area | 4,529 | 6,246 | 138% |
| New York City | 1,786 | 2,376 | 133% |
| Germany | 765 | 803 | 105% |
| Italy | 255 | 240 | 94% |
| Ireland | 122 | 109 | 89% |
| Austria | 103 | 90 | 87% |
| Finland | 156 | 132 | 85% |
| France | 694 | 579 | 83% |
| Spain | 392 | 327 | 83% |
| Sweden | 263 | 218 | 83% |
| Netherlands | 379 | 271 | 72% |
| Portugal | 119 | 81 | 68% |
The two US rows compare like with like. Dealroom's AI guide counts VC-backed AI startups in all sectors, and so does our Crunchbase column. The EU rows tilt the comparison in Crunchbase's favour. Dealroom's European counts come from a report for the EU's Apply AI strategy (data as of October 2025), which only counts AI startups in ten industrial sectors such as health, defence, energy and manufacturing. Our Crunchbase column still holds every sector, and it is below Dealroom's in nine of the ten countries anyway.
When we keep only the Crunchbase startups whose categories fall in those ten sectors, Crunchbase has 1,229 against Dealroom's 3,248 in the ten countries, or 38%. The mapping from Crunchbase categories to Dealroom's sectors is ours.
The national maps
| Map | Startups on the map | Crunchbase, funded | Crunchbase as % of map |
|---|---|---|---|
| Germany, appliedAI 2025 | 935 | 800 | 86% |
| Berlin | 283 | 287 | 101% |
| Munich | 200 | 145 | 72% |
| Hamburg | 71 | 52 | 73% |
| France, France Digitale 2026 | 1,114 | 617 | 55% |
We expected the national maps, built by local associations, to show the biggest gap to Crunchbase. The gap they show is smaller than Dealroom's. appliedAI's German AI Startup Landscape counts startups founded since 2015 with at least two full-time staff and AI at the core of the product, whether or not they have raised money. Crunchbase has 800 funded German AI startups of that age that are still operating. Berlin matches almost exactly. Munich is where Crunchbase falls short, with 145 against 200. The first version of this post said 118: Crunchbase spells the city Munich for some companies and München for others, and we had counted only the first spelling.
France Digitale's 2026 map lists 1,114 French AI startups, and 67% of them have raised money, about 746. Crunchbase has 617 funded French AI startups that are still operating, 83% of the funded part of the map. The summary we read gives no founding-year cutoff for the French map, so the Crunchbase count takes every founding year.
Matching the maps company by company
The counts could not tell us whether Crunchbase lacks a startup or holds it without the AI category, so we matched the lists one company at a time. appliedAI's online list had 933 entries on 28 September 2026 and France Digitale's dataset had 1,137. Some companies appear twice under two names, such as Apheris and Apheris AI, so we merged entries that share a website. That left 900 German and 1,120 French startups. We looked each one up in our full copy of Crunchbase, which also holds companies without funding and outside the AI category.
| Place | Startups on the list | In Crunchbase | With the AI category | With a funding round | In our Crunchbase count |
|---|---|---|---|---|---|
| Germany, appliedAI | 900 | 79% | 52% | 59% | 37% |
| Berlin | 271 | 85% | 53% | 71% | 39% |
| Munich | 186 | 80% | 56% | 64% | 44% |
| Hamburg | 68 | 81% | 53% | 56% | 32% |
| France, France Digitale | 1,120 | 75% | 38% | 55% | 25% |
| Ile-de-France | 718 | 79% | 42% | 60% | 29% |
We expected most of the gap to be startups that Crunchbase holds under another category. That is the largest group in both countries. Of the 900 German startups, 192 (21%) are not in Crunchbase at all, 236 (26%) are in it without the AI category, and 118 (13%) have the AI category but no funding round. For the 1,120 French startups the same groups hold 280 (25%), 412 (37%) and 130 (12%).
France Digitale's data also records each startup's latest round. For the 645 startups it lists with a seed round or later, Crunchbase does better: 89% are in it and 75% have at least one funding round there. It gives the AI category to 42% of them.
What surprised us is how little the maps and Crunchbase's AI category overlap. Only 327 (41%) of the 800 funded German AI startups in Crunchbase are on appliedAI's list, and 281 (46%) of the 617 French ones are on France Digitale's. In Berlin, where the counts matched almost exactly at 287 and 283, the two sets share 109 companies. Some of the Crunchbase startups that are missing from the maps may fail appliedAI's rules, such as two full-time staff and AI at the core of the product; we did not check them one by one. Munich startups on the list are in Crunchbase about as often as Berlin ones, 80% against 85%.
What this does to San Francisco vs the EU
Dealroom and the national maps point in different directions. By Dealroom's count, Crunchbase sees Europe at about half the rate it sees the US. If Crunchbase covers San Francisco like the rest of the US and the EU like the rest of Europe, the EU's 4,147 AI startups stand for roughly 8,100 at San Francisco's rate, about twice San Francisco's 4,067. By the German and French maps, Crunchbase's funded counts reach 86% and 55% of everything the maps list, funded or not.
The company-by-company match shows both: Crunchbase lacks some of the maps' startups and holds others without the AI category, and the second group is larger. 26% of the German list and 37% of the French list are in Crunchbase without the category, against 21% and 25% that are not in Crunchbase at all. Our San Francisco count uses the same category.
Does Crunchbase tag US startups as AI more often?
If it did, the San Francisco vs EU comparison would lean toward San Francisco. The two national maps cannot answer this, so we matched one list that covers both sides: the Y Combinator companies YC tags as AI (Artificial Intelligence, Generative AI, Machine Learning and similar tags), read from YC's public directory on 28 September 2026, matched to Crunchbase by website domain.
| YC AI companies | On YC's list | In Crunchbase | With the AI category, of those in Crunchbase |
|---|---|---|---|
| US | 1,714 | 1,307 | 944 (72%, 95% interval 70 to 75%) |
| EU and UK | 166 | 140 | 101 (72%, 64 to 79%) |
| EU | 89 | 80 | 60 (75%, 65 to 83%) |
| UK | 77 | 60 | 41 (68%, 56 to 79%) |
We expected the US share to be higher. It is the same, within the precision that 140 European companies allow. Older companies carry the category less often on both sides, so we also split by YC batch: before 2023, 57% of the US companies and 53% of the European ones have it (487 and 47 in Crunchbase); from 2023 on, 81% and 82% (820 and 93).
On the national maps the share is lower: of the startups we found in Crunchbase, 66% of the German ones and 51% of the French ones carry the category. With one list on both sides the US and Europe come out level, so we read the lower shares on the maps as coming from which companies the maps include. Within YC's companies we found no sign that the category favours US startups.
Method
We started from the same copy of Crunchbase as the San Francisco vs EU post, read on 27 September 2026: organizations with a funding total or at least one funding round, in Crunchbase's Artificial Intelligence (AI) category group, for-profit, not public (last round not a post-IPO round), duplicates removed, with a headquarters country. On top of that, each comparison uses its own filter.
- Per year: startups counted by founding year, US against Europe as defined above.
- Against Dealroom's counts: founded 1990 or later, Dealroom's own cutoff. Dealroom counts VC-backed startups. Our copy of Crunchbase holds only each company's last round, so we drop companies whose last round was a grant, debt, non-equity assistance, product crowdfunding, an ICO or a secondary sale. A company that raised a seed round and later a grant drops out too.
- Bay Area: Crunchbase's San Francisco Bay Area location group, plus a list of Bay Area city names for California records that lack the group. New York City: the Greater New York Area group plus the five boroughs.
- Same sectors: Crunchbase category groups Health Care, Biotechnology, Government and Military, Privacy and Security, Manufacturing, Energy, Sustainability, Natural Resources, Agriculture and Farming, Food and Beverage, Transportation, Messaging and Telecommunications, Media and Entertainment, Music and Audio, Video, Content and Publishing and Gaming, plus the categories Robotics, Construction, Industrial, Industrial Automation, Industrial Engineering, Mechanical Engineering and Civil Engineering.
- Against the national maps: still operating (not closed, not acquired), with funding, founded 2015 or later for Germany and in any year for France. Munich counts Crunchbase records with the city Munich or München.
- Matching the maps: each startup's website domain against the domain and website of every Crunchbase organization, in any country, funded or not. Domains used by more than three Crunchbase organizations, and social or site-builder hosts, were not used. If no domain matched, we took the exact company name, with accents, punctuation and legal forms such as GmbH or SAS removed, among Crunchbase organizations in the same country. 679 German and 775 French startups matched by domain, 29 and 65 by name. We checked 20 name-only matches by hand: 19 were clearly the same company and one was unclear. A looser pass over the unmatched startups (the same website name with another ending, or names without words like AI or Labs) found 19 more candidates, 9 of them the same company, which would add less than one percentage point to the found share. "In our Crunchbase count" applies filter 5 to the matched organization.
- YC companies: every company in YC's public directory on 28 September 2026 with one of the tags Artificial Intelligence, Generative AI, AI, AI Assistant, Machine Learning, Deep Learning, Computer Vision, NLP, AIOps, Conversational AI or AI-Enhanced Learning. US and Europe by YC's region field, EU and UK by the location string. Matched by registrable website domain only, on both sides; a domain held by more than three Crunchbase organizations was not used. Intervals are Wilson 95% intervals.
The benchmark numbers come from these reports:
- Dealroom and Prosus, The State of AI in Europe: The Invisible Giant, March
- The US vs Europe funnel starts at "~900" new VC-backed AI companies per year on each side. A note ties its conversion rates to seed rounds from 2015 to 2020; the report does not say which years the 900 covers.
- Dealroom for the EU's Innovation Radar Bridge project, Startups driving Europe's AI transformation, November 2025, data as of October 2025. Key metrics by country, ten countries shown and 17 not shown; over 4,100 VC-backed startups in the Apply AI sectors across the EU.
- Dealroom, Artificial Intelligence guide, read on 28 September 2026: 4,529 VC-backed AI startups in the Bay Area and 1,786 in New York City.
- appliedAI Institute for Europe, German AI Startup Landscape 2025, 21 October 2025: 935 startups; Berlin 283, Munich 200, Hamburg 71.
- France Digitale, Mapping 2026 des startups françaises de l'IA, 10 February 2026: 1,114 startups, 67% funded, 63% in Ile-de-France, as reported by Maddyness.
- appliedAI Institute for Europe, the startup list on its AI Startup Landscape (Deutschland) page, read on 28 September 2026: 933 entries, 282 in Berlin, 194 in Munich, 71 in Hamburg.
- France Digitale, the dataset behind its Mapping IA 2026 page, read on 28 September 2026: 1,137 entries with each startup's website, region and latest round.
Data
The files below are under CC BY 4.0.
- coverage.csv, Dealroom and Crunchbase counts for the Bay Area, New York City and ten EU countries, with each benchmark's definition
- per_year.csv, funded AI startups in Crunchbase by founding year, US, Europe and EU27, 2013 to 2025
- national_maps.csv, map counts and Crunchbase counts for Germany, three German cities and France
- map_matching.csv, for each place on the two maps: startups on the list, how many we found in Crunchbase, with the AI category, with funding and in our count, why the others are outside our count, and how many of Crunchbase's own count are on the list