The Paris AI Map: 23 Companies, $35B, and 10 of Them Inside 2km

The Paris AI Map: 23 Companies, $35B, and 10 of Them Inside 2km

I plotted the registered headquarters of every serious AI company in Paris onto one map, expecting a scatter. What came back was a cluster.

23 companies. More than $35B of combined reported value. All of it inside one city, and most of the interesting part inside a twenty minute walk. France did the thing people say only San Francisco can do, and almost nobody outside France is looking at it.

Here is the map, and then the part a map cannot show you: why the density itself is the asset.

The 23 companies, by sector

Frontier and infrastructure. Mistral AI, Europe's frontier lab, reportedly raising at a 20B euro valuation. Kyutai, an open-science lab backed with roughly 300M euros by Xavier Niel, Rodolphe Saade and Eric Schmidt. LightOn, the first generative AI company to list on a European exchange. FlexAI in AI compute infrastructure.

Agents and developer AI. Poolside in AI software engineering, at a reported $12B valuation with Nvidia investing up to $1B. H Company, whose $220M seed was the largest in Europe, now shipping Holo 3 computer-use models. Dust in enterprise AI agents, a $40M Series B co-led by Sequoia and Abstract, with zero churn in 2025. Giskard in AI testing and evals. Mindee in document AI APIs.

Go-to-market and revenue. lemlist, bootstrapped to eight figures of ARR with zero VC money. Folk, rebuilding the CRM AI-native. Contentsquare in experience analytics at $5.6B.

Health and science. Owkin, an AI biotech unicorn. Nabla, clinical AI used by more than 85,000 clinicians including at CVS Health. Aqemia in AI drug discovery, with a Sanofi collaboration worth up to $140M. Bioptimus in biology foundation models.

Commerce and creative. Photoroom in AI photo editing, more than 200M downloads and roughly $94M ARR. Mirakl at $3.5B, now going all in on agentic commerce.

Vertical and security. Shift Technology, an insurance AI unicorn. Akur8 in AI pricing for insurers. Doctrine, France's leading legal AI. Filigran in open-source threat intelligence. Wandercraft in humanoid robotics, building Calvin-40 for Renault's factory floors.

Three things that stood out

The density is absurd

10 of the 23 sit inside a single circle 2km across, between the Louvre and the Grands Boulevards. Poolside and FlexAI share an address at 9 rue des Colonnes. Photoroom and Aqemia are next-door neighbours at 229 and 231 rue Saint-Honore. lemlist and Filigran are 47 metres apart.

Two of Europe's five most valuable AI companies are a 15-minute walk apart

Mistral sits near Les Halles, Poolside near the Bourse. Between them, more than $25B of reported value and 1,100 metres of pavement. For context, Europe's full valuation map spreads 27 companies across seven countries. Two of the largest share a neighbourhood.

The go-to-market layer speaks French

lemlist bootstrapped an outbound platform to eight figures of ARR with no venture money, in a category where the American competitors raised heavily. Folk is rebuilding CRM around AI rather than bolting it on. Contentsquare turned behaviour analytics into a $5.6B company. If you run outbound, part of your stack was probably built inside this map, which is the same conclusion the wider France GTM software map reaches with a longer list.

Why density compounds

A map tells you where companies are. It does not tell you why sitting 47 metres from another AI company is worth anything. Three mechanisms, in the order I would weight them.

Talent recirculates without relocating. An engineer leaving Mistral can join Dust without changing apartment, school or commute. In a dispersed ecosystem, changing company often means changing city, and most people with families will not do that. Density turns every hire into a local hire and every departure into a local rehire, so the talent pool grows instead of leaking.

Infrastructure gets shared before it gets built. GPU access, French employment contracts, EU AI Act interpretation, recruiters who already understand what an evals engineer does. The second company to hit a problem in a dense cluster solves it for a fraction of what the first one paid, because the answer is one coffee away.

Investors show up in person. A partner running diligence can take three meetings in an afternoon on foot. That compresses fundraising cycles, and compressed cycles are part of why H Company could raise a $220M seed in Paris rather than needing to move to raise it.

None of that appears in a valuation table. All of it is why the next Paris company is cheaper to start than the last one.

The subset that matters if you run go-to-market

Most people reading this do not need biology foundation models. Three companies on the map already sit in a lot of outbound stacks.

• lemlist for sending, and the bootstrapped path is the interesting part rather than the product category.
• Folk for CRM, if you want something designed after AI rather than retrofitted with it.
• Contentsquare for behaviour analytics, once you have enough traffic for the data to say anything.

The broader point is that the default answer is no longer American. When we evaluate a stack component or compare our own build against what a cold email agency would run, the European options are genuinely in the shortlist now. The 2026 GTM tools market map has the full field.

What the map leaves out

Hugging Face, Dataiku and Algolia were all born in Paris. None of them appear here, because their headquarters are now in New York and San Francisco. That was the previous generation's pattern: build in France, incorporate and scale in the US. So far the 23 on this map have not done it, and that is the actual news in the picture.

Two caveats on the numbers. Every valuation here is a reported round rather than an audited figure, and at least one of them, Mistral, is a round in progress rather than a closed one. Read the ordering, not the decimals. And headquarters registration is a legal fact rather than an engineering one. It cuts both ways: plenty of the 29 US AI companies look American on paper with their engineering somewhere else entirely.

How Paris compares

New York assembled an application-layer cluster around finance, media and health, sitting close to its buyers. The most valuable US AI startups concentrate an enormous share of value in a handful of model labs. Paris resembles neither. It has one frontier lab carrying a large slice of the value, then a long tail of applied companies in insurance, legal, health, security and go-to-market that would look unremarkable individually and look striking as a group.

Cluster effects are slow and then obvious. Paris is somewhere in the middle of that curve.

Frequently Asked Questions

How many AI companies are headquartered in Paris?

This map has 23 with meaningful scale, covering frontier labs, agents, health, commerce, security and go-to-market, worth more than $35B in combined reported value. It excludes companies founded in Paris that have since moved their headquarters abroad, such as Hugging Face, Dataiku and Algolia.

Which is the most valuable AI company in Paris?

Mistral AI, which was reportedly raising at a 20B euro valuation, with Poolside next at a reported $12B. The two sit about 1,100 metres apart, near Les Halles and the Bourse respectively.

Which Paris AI companies are relevant to a go-to-market team?

lemlist for outbound sending, Folk for AI-native CRM, and Contentsquare for experience analytics. Mindee is worth knowing if you process documents, and Dust if you are deploying internal AI agents.

Are these valuations reliable?

They are reported figures from the latest publicly known rounds, not audited numbers, and one of them is a round still in progress. European rounds are also disclosed less consistently than American ones. Use the ordering and the sector spread, and ignore precision below the nearest billion.

RevenueFlow builds AI-native pipeline systems and you pay per qualified meeting, not a retainer. No paying for activity. You only pay when we book you a qualified sales meeting. See if you qualify.