Two studies by the economic analysis firm Asterès, commissioned by Cigref — the association of large French companies and public institutions — attempt to put a figure on something that usually stays an impression: what it costs Europe to depend on American cloud and software.
The numbers are large, and worth looking at closely. Not least because, as we shall see, they are not all of the same kind.
First study: how much leaves
La dépendance technologique aux softwares & cloud services américains (April 2025, 33 pages) estimates that:
- European Union companies buy €264 billion a year of cloud and software services that benefit the American economy;
- American suppliers hold around 83% of the European cloud-software market — €330 billion for the Union, €54 billion for France alone;
- those €264 billion amount to roughly 1.5% of EU GDP, and to one and a half times the European Union’s budget.
The most striking comparison is in the study itself: EU energy imports in 2024 came to €360 billion. Europe’s digital bill, Asterès writes, is “of an amount comparable to its energy bill”.
On the American side, those purchases generate $323 billion of value added — 1.1% of US GDP — and 1,943,000 jobs, or 1.2% of total US employment.
And the reverse? The study tries to quantify it: if the Union managed to produce 15% of what it currently buys from the United States by 2035, the result would be 463,000 additional jobs in Europe.
Second study: how fast it grows
A year later, De la dépendance technologique à la captation économique (May 2026, 34 pages) shifts from volume to price, and the picture worsens:
- cloud-software costs rose +8.7% a year over the past three years, and organisations expect +12% a year over the next five;
- the largest increase faced at contract renewal averages 51% — with extreme cases reported at 100%, and up to 200–300%;
- the average excess cost for 2026–2030 is €140 billion a year at constant volumes;
- of which €93 billion would leave the continent permanently instead of funding the local economy. The study notes this is the same order of magnitude as the annual digital budget required by the Draghi plan;
- the overall estimated effect is a loss of €107 billion of value added per year, 0.6 points of GDP, and 1.4 million jobs by 2030.
One detail deserves to be pulled out, because it bears directly on how artificial intelligence is being sold: the increases are justified by new AI features, imposed in 40% of cases, while around 60% of respondents judge the related productivity gains to be theoretical in the short and medium term.
What these numbers are not
Here a clarification is needed that almost nobody makes — and that the study itself makes honestly.
The €264 billion is not a measurement: it is an estimate extrapolated from interviews with six Chief Information Officers of large companies, who reported cloud-software spending of around 2.2% of revenue, with 83% going to American suppliers. The French results were then extended to the Union in proportion to GDP.
Asterès explains why it had to proceed this way: trade data is unusable because of Ireland's overwhelming weight in the flows — of $225 billion of European surplus in IT services in 2022, $194 billion was Irish — and foreign direct investment profit data is too imprecise.
Something else should be said: Cigref is an interested party. It is the association of large French companies and public institutions, and it has a legitimate but evident interest in pushing this subject onto the European agenda. That does not disqualify the study — the methodology is stated and the limits are written down — but anyone citing it should know.
Finally, a distinction that matters particularly to us: the 2025 study measures companies, not public administrations. The 2026 one broadens to “organisations”. The perimeter is therefore adjacent to ours, not identical.
Why we keep building our own measurement
One could object: if Asterès has already done the sums, what is our second instrument for?
Because the two measurements answer the same question from opposite directions, and need each other.
Asterès measures from above: a sample of interviews, a macroeconomic extrapolation, a continental figure. It is the only way to get a European order of magnitude, and it carries the limit of any extrapolation — six interlocutors are not a census.
What we are building measures from below: public contracts actually awarded, body by body, year by year, from open data, with the method published and the figures reproducible by anyone. It will never produce a European figure. It will produce an Italian one that is verifiable.
Where they have the order of magnitude and a sample of six, we would have the precise figure for Italian public administration. We are not competing with these studies: we are their consistency check.
And there is a political reason why the second measurement matters as much as the first. A continental number is impressive and unassailable precisely because it is abstract: nobody feels named. A per-body number — this municipality, this health authority, this amount, this supplier — is far smaller, and far harder to ignore for the person who signs it.
The state of progress of that work is in the roadmap.
Sources
- Asterès for Cigref, La dépendance technologique aux softwares & cloud services américains, April 2025
- Asterès for Cigref, De la dépendance technologique à la captation économique, May 2026
- Cigref page on the second study
Both PDFs are archived in our repository with their hashes, so they remain verifiable even if the source pages change.