Skip to content
AnalysisSociety14 November 2026

Is France behind in AI?

Investment, models, computing power, use by businesses and by residents: France alongside its European neighbours, the United States and China

Guillaume PakulaClément ReynaudSébastien PierfedericiAlexandre Veilly By Projet Celsius
Published on 14 November 2026 · 9 min read
Three open bays of a supercomputer, with their stacked compute blades and the red and blue pipes of their liquid cooling

Racks of the Adastra supercomputer, installed by GENCI at CINES in Montpellier. Photo Rambiera, CC BY-SA 4.0, via Wikimedia Commons.

In Europe, France is 2nd for private investment in AI, behind the United Kingdom: $4.4 billion in 2025, ahead of Germany, according to the Stanford AI Index. It lags on use by its companies, 15th of 27 in the European Union with 18% of companies using AI, and on computing power: its best machine is 29th in the world ranking of supercomputers, against 5th for Germany. Worldwide, the United States invested 66 times more than France in 2025.

Contents

No single indicator measures a country's place in AI. We have selected 5, published by sources that compare countries using the same method: money invested, models produced, computing power, use by businesses and use by residents. For each, France is placed next to Germany and the United Kingdom and, for scale, the United States and China.

1.

France through 5 indicators

IndicatorFranceFrance's rankGermanyUnited Kingdom
Private investment in AI, 2025$4.4bn2nd in Europe, 1st in the EU$3.9bn$5.9bn
Notable AI models, cumulative since 2003, our count512nd in Europe, 1st in the EU2984
Supercomputers in the TOP500, June 202621, the best 29th2nd in Europe by number, 6th by cumulative power41, the best 5th11, the best 13th
Businesses with 10 or more employees using AI, 202518 %15th out of 27 in the EU26 %not covered
People aged 16-74 who have used generative AI in the last 3 months, 202537 %16th out of 27 in the EU32 %not covered

Stanford AI Index 2026 (Quid data); Epoch AI database of 14 November 2026; TOP500 of June 2026, our count; Eurostat, surveys 2025. For scale: United States $285.9bn and China $12.4bn of private investment in 2025; 655 and 192 notable models; 161 and 31 supercomputers.

2.

Investment and models

In 2025, private investment in AI reached $4.4 billion in France, according to Quid data reported in Stanford's AI Index: 4th in the world, 2nd in Europe behind the United Kingdom, 1st in the EU ahead of Germany. Per capita, that comes to about $63 per French resident, against $47 per German resident, by our calculation. 84 French AI companies received funding for the first time in 2025, against 92 in Germany and 172 in the United Kingdom. Over 2013-2025, France remains behind the United Kingdom and Germany. These figures do not include the €109 billion announced at the February 2025 summit, which are multi-year commitments, mostly for data centres.

For models, our count from the Epoch AI database gives 51 notable models to which at least one organisation based in France has contributed since 2003, against 84 in the United Kingdom and 29 in Germany. Meta's Paris laboratory, Facebook AI Research, took part in 34 of them and Mistral AI in 7, all published since 2023. In 2024, France published 4, all from Mistral AI, more than any other European country. The count is low across Europe: 1 French model in 2025, against 65 American and 36 Chinese. Our analysis on French AI shows that the nationality of the provider says nothing about the footprint of an answer.

3.

Computing power

In the TOP500 supercomputer ranking of June 2026, France has 21 machines, 2nd in Europe by number behind Germany (41). By cumulative power, 317 PFlop/s, it is 6th in Europe, behind Italy, Germany, Switzerland, the United Kingdom and Finland, by our count. Its most powerful machine, at the CEA, is 29th in the world; Jean Zay, the public AI research machine, is 49th for its part equipped with H100 cards. Germany has the world's 5th machine, JUPITER, at Jülich, and the United Kingdom the 13th, Isambard-AI. Alice Recoque, the exascale machine ordered by EuroHPC for the CEA, is being installed and is due to open in 2027.

Supercomputers are only part of computing: large models are mostly trained in companies' data centres, for which no open source gives the power by country. France has an asset for hosting them, one of the least carbon-intensive electricity supplies in Europe and a net export balance of 92 TWh in 2025, but projects have already reserved nearly 18 GW of grid connection; our analysis on electricity gives the orders of magnitude, and our viewpoint the conditions.

4.

Use by businesses and by residents

Businesses with 10 or more employees using at least one AI technology, 2025 (Eurostat)

Denmark, 1st in the EU42 %Germany26 %European Union20 %France, 15th out of 2718 %Romania, 27th5 %

French businesses use AI less than the EU average: 18% of those with 10 or more employees in 2025, against 20% in the EU and 26% in Germany, ranking 15th out of 27. Denmark, Finland, Sweden, Belgium, Luxembourg and the Netherlands exceed 33%. Residents, on the other hand, are above the average: 37% of those aged 16-74 have used generative AI in the last 3 months, against 33% in the EU and 32% in Germany, ranking 16th. Eurostat surveys people aged 16-74 on the last 3 months; Arcep's Baromètre du numérique, which surveys people aged 12 and over on their use in 2025, finds 48%. Our analysis on French users breaks this use down by age, qualification and occupation.

5.

Our reading

For us at Projet Celsius, the answer depends on the indicator. Compared with its European neighbours, France is 2nd, behind the United Kingdom, for money raised and models, above the average for use by residents, and behind for use in businesses and public computing power. On a world scale, the question applies to Europe as a whole: the United States invests 66 times more than France, and the Epoch AI database counts 2 notable European models in 2025, 1 French and 1 British, against 36 Chinese. These indicators count amounts and uses; they do not say whether that use is worthwhile, or what it costs in electricity.

Point to note

The investment measured by Quid is private only. The counts of models and supercomputers are our own, based on the Epoch AI and TOP500 databases. Eurostat no longer covers the United Kingdom. No open source gives the computing power of data centres by country.

Guillaume PakulaClément ReynaudSébastien PierfedericiAlexandre Veilly
By Projet CelsiusAI Impact, Projet Celsius · 14 November 2026
Write to us →

Sources

7 documents

Every figure was read in its source; the hub's methodology is described in Methodology and sources.

Cite this analysis

AI Impact, Projet Celsius, "Is France behind in AI?", 14 November 2026. https://projetcelsius.com/en/ai/analyses/is-france-behind-in-ai/

A question about a figure?

An error, a source to add, a methodology question, a press request: write to us. Guillaume Pakula, Clément Reynaud, Sébastien Pierfederici and Alexandre Veilly will reply.

Write to us →

Projet Celsius, which runs this hub, also helps companies measure and reduce the footprint of their AI use.

Accounting for AI in your Bilan Carbone® ↗
Projet Celsius, environmental strategy consultancy

We support companies that account for AI in their Bilan Carbone®

Visit Projet Celsius ↗