Andy Garcia · freelance dev
FRENES
Back to the blog
Startups & Finance

2,640 French Startups Grew Revenue 11.6% in 2025 While Funding Fell. Six in Ten Still Lose Money.

October 19, 2026 6 min read

On September 25, 2026, the Banque de France published its annual study of the financial situation of French startups. The headline numbers are good. In 2025, a year when fundraising fell 5 percent, the 2,640 startups in the study grew their revenue 11.6 percent, from €26.6 billion to €29.7 billion. Their aggregate operating loss shrank 9.2 percent, to €3.29 billion, after a 17 percent drop the year before. Headcount rose 2.2 percent, to 120,872 employees. Over the same year, revenue grew 1.5 percent at French SMEs (excluding microenterprises) and 1.2 percent at mid-sized companies.

The Banque de France's press release calls it resilience in the face of tighter venture capital. The study supports that sentence. It also runs to fifteen pages, with four appendices and a stack of footnotes, and that is where a founder learns what a startup that survives a tighter market looks like on paper.

I read it as the CTO of a LegalTech and as an indie developer who ships iOS apps and a strategy game. Here is what I took from it, and what it cannot tell you.

Who is in the sample

The study is built from accounting data the Banque de France collects. It identifies 6,300 innovative companies and keeps the 2,640 with complete tax returns that either had revenue above €750,000 in 2024 or 2025, or raised more than €3 million. The panel is also "cylindered": every company must have closed accounts in both 2024 and 2025. Mission French Tech, for comparison, counts more than 18,000 startups in France.

So the 11.6 percent is the growth of companies that already sell, or have raised large rounds, and were still there to file. The study states its perimeter plainly. Read it as a reading instruction, not a flaw.

Two more details change the picture. The startups in the study are 11 years old on average, and 54.4 percent are at least ten. Only 243, about 9 percent, are under five. And growth is slowing: successive editions, each on its own panel, measured revenue growth of 25.0 percent for 2022, 18.6 for 2023, 12.5 for 2024 and 11.6 for 2025. Still more than seven times the SME rate. But the trend points down.

Profitability is improving, slowly

The loss line is the real news. Net losses were 21.3 percent of revenue in 2022, 16.5 in 2023, 13.3 in 2024 and 12.3 in 2025. The share of startups with a positive operating result rose from 37.5 percent to 39.8.

Read the other side too. 60.2 percent still lose money at the operating level, €4.62 billion between them. The profitable 39.8 percent earn €1.33 billion. The €3.29 billion aggregate loss is the difference.

The table by age is the one I would pin above a founder's desk. Median operating result: minus €872,000 for startups under five years old, minus €346,000 at five to nine, minus €71,000 at ten to fourteen, minus €42,000 at fifteen and over. These are different companies at different ages, not one company followed over time, and the older groups are by definition the ones that lasted. Even so, the median only gets close to break-even after a decade.

The cash line

Total cash fell 3.9 percent, to €11.1 billion. The median startup held 57 days of revenue in cash, down from 69 in 2024. SMEs hold 50 days, mid-sized companies 52.

For the loss-makers, the Banque de France does the runway arithmetic for you: their €8.2 billion of cash covers a median of about 12 months of operating losses. In its box on deeptech, the 200 loss-making companies out of 271 have about 17 months on average, but a median of only 10, because a few cash-rich companies pull the average up.

Twelve months sounds comfortable until you subtract the time it takes to raise. Assume six months from first meeting to money in the bank, a round number of my own and not the study's. The median loss-maker then starts its next round with half of its runway already spoken for.

Debt did the bridging

Founders talk about equity. The balance sheets also say debt. 86.6 percent of the sample carries bank debt, a stable €5.9 billion, about 33 percent of equity. It fell from 22.1 to 19.9 percent of revenue because revenue grew and the debt barely moved.

The fastest-growing line is bonds: up 29.4 percent, to €1.85 billion, used by 272 companies (10.3 percent of the sample, almost the same share as in 2024). The median amount went from €1.0 million to €1.3 million. The Banque de France says what they are for: mainly bridge financing "to support startups between two funding rounds and fill a cash shortfall" (my translation).

Meanwhile 17.9 percent of the sample, 472 companies, has negative equity, and half has less than €1.4 million of equity.

Where the money went

The 5 percent decline comes from EY's barometer: €7.39 billion raised in 2025 across 618 rounds, against €7.77 billion across 723. Mistral AI's €1.7 billion round was so large that without it, EY says, the decline would have reached 26 percent. Rounds under €10 million fell from 560 to 446, down 21 percent. In Extenso, also cited by the Banque de France, counted 37 percent fewer deals. Even the decline depends on who counts.

The 2026 rebound has the same shape. EY's first-half barometer, published July 9, counts €4.6 billion raised, up from €2.68 billion a year earlier, but by 280 companies, 10 percent fewer, with an average ticket above €16 million against under €9 million. The money came back, to fewer companies.

What the study does not show

It cannot say why losses fell. Accounting data does not separate a founder who chose profitability from one who cut costs because the round never came.

It is an aggregate dominated by large companies. In tourism and leisure, mean revenue is €17.3 million and the median €3.7 million.

Year-to-year comparisons mix panels. The Banque de France warns that earlier editions did not use exactly the same sample, and that its 2024 and 2025 failure counts cover different perimeters. For the record: 51 of the 2,165 startups in last year's panel entered court proceedings in 2025, 45 receiverships and 6 liquidations. Their profile: no equity, weak cash.

And it says nothing about the startups below the thresholds or without complete returns. That is most of the 18,000.

What I would do with it

Compute runway the Banque de France way: cash divided by monthly operating loss. Then compare it with 12 months, or 10 if you build deeptech.

Track cash in days of revenue. The median fell from 69 to 57 days in a single year. It warns earlier than a bank balance does.

Talk to a bank while you have revenue to show. In 2022 the Banque de France changed its rating method to take the startup growth model into account, and more than 2,200 startups with revenue above €1.25 million are now rated. A company head can see the rating in the free "Espace personnel dirigeant". The press release says its 20 regional startup correspondents hold 800 rating interviews and 350 support sessions a year, and the study describes the correspondents' advisory meetings as free and confidential.

If a bridge might be needed, price it while the runway is long. The study shows bridges growing. It does not show their terms, so read yours closely.

And count customers as financing. The study's appendix lists "Je choisis la French Tech", a program meant to get private companies and public administrations to buy more from French startups. I have a bias here: I built BidForge, an AI that reads French public tender files for small businesses. A contract won is revenue that no investor has to approve.

One line for AI builders. Software and data has the highest value-added rate in the study, 48.3 percent of revenue. Value added is what remains after purchases from third parties, and a model provider's invoice is one of them. As inference grows, that is the ratio I would watch.

The French startup that came through 2025 is older than its label, sells more, burns less and borrows to bridge. The headline says resilience. The tables say patience.

Sources

A project like this one?

I design and deploy products like this. Let's talk.

Let's talk