The first map of companies created from zero by, or with, large French groups. Not investments, not acquisitions, not innovation programmes: companies with their own legal entity, their own accounts, and a group behind their creation.
France measures corporate venture capital and startup partnerships every year. Nobody counts the companies that large groups actually create. The Observatoire exists to fill that gap, venture by venture, with a source for every line.
Version 0, 25 September 2026. Figures move as contributions are verified. Three candidates were excluded after review.
What counts as a corporate venture
A company enters the Observatoire if it meets all three conditions:
- Created from zero. The company did not exist before the group decided to build it, alone or with a studio, founders or partners.
- Its own legal entity. A registered company with its own accounts, not a brand, a division or a programme.
- Active or exited. Still operating, sold, merged or reintegrated. Closures are recorded too: they are part of the picture.
Excluded: acquisitions, internal brands, incubations that never incorporated, joint ventures between established groups, and carve-outs of an existing business. A company that grew mainly by buying others comes off the list, however visible it is.
What the first scan shows
Preliminary findings, before the verification round with the groups themselves:
- A few groups do most of the work. EDF Pulse Incubation has spun out seven companies, all of which meet the criteria. Crédit Agricole has six, mostly out of La Fabrique by CA. AXA, through Kamet, has seven candidates, six of them confirmed.
- Most of the CAC 40 does not build. 22 groups of the index have no venture we could identify. Several run a corporate VC fund or an accelerator instead: the invest and partner levers, not the make lever.
- Closures are accelerating. Nine confirmed ventures have closed, five of them in 2025 and 2026: Blank, Moonshot Insurance, Watèa, Apricity and Ma French Bank.
- Profitability is rare, or invisible. Among the 63 confirmed ventures, only two show a profit in their published accounts so far. For most, accounts are confidential or not yet found, which is itself a finding.
- Announced is not built. Some groups communicate on dozens of new activities; few can be matched to a registered company. The gap between the two is one of the things the Observatoire will measure.
Method
Every entry is documented from public sources: company registries and published accounts, group disclosures, and specialist press. Contributions from people inside the groups are welcome and recorded with their origin, then checked against the registry before they count.
The base was built in four passes: ventures known at launch, the CAC 40, large groups outside the CAC 40, and contributions received after the call on LinkedIn. Each company is tagged by origin (intrapreneurship, spin-out of internal expertise, new opportunity through a studio or programme, co-build with an external studio), status, ownership, and whether the CEO came from inside or outside the group.
No figure is published without a source. Where a number is not public, the Observatoire says so rather than estimating it.
Timeline
- September 2026. Version 0: 92 candidates, criteria fixed, contributions open.
- Autumn and winter 2026. Verification with the groups, registry checks on the 26 pending entries, pre-filled profiles sent to venture-building teams.
- Q1 2027. Full publication: dataset, analysis by group and by model, and survival and exit rates.
Contribute
If your group has created a company that belongs here, or if one of the entries is wrong, the fastest way is a short email: venture name, group, year created, ownership, one number, and a source. Every contribution is verified before it counts, and you choose whether to be credited.
Questions
Who is behind the Observatoire?
Honey Lab, a corporate venture building studio founded by Patrick Amiel. Patrick has helped launch ventures with large groups through his previous studio. The Observatoire is independent of any group and not funded by one.
Will the full dataset be public?
Yes, at publication in Q1 2027. Until then, only verified aggregate figures are shown here, because a third of the candidates are still being checked.
Why are some well-known names missing?
Either they grew by acquisition, they are brands without their own legal entity, or we have not found them yet. The criteria are strict on purpose: the point is to count companies, not announcements.
Can a group correct its own entries?
Yes. Each group represented will receive its pre-filled profile before publication and can correct or complete it, with sources.