Cloud in seven firms out of ten, ICT staff in three

Whether Dutch companies can leave the American cloud is usually framed as a procurement question: which European provider exists, what does it cost, and how long does the current contract run. That is the wrong question. The constraint is staffing.

In 2024, 71.0% of Dutch enterprises with 10 or more employees bought paid cloud services. In the same year, 29.7% of that same group employed an ICT specialist. Seven in ten bought cloud services without employing anyone with ICT specialist skills.

The headline adoption figure hides the problem, because it counts a company buying email the same as a company running its production database in someone else's data centre. For an exit, the second layer is what matters.

In 2025, 55.4% of Dutch enterprises with 10 or more employees hosted a database or their files with a cloud provider. Across the European Union the figure was 24.0%. Only Denmark was higher, at 55.5%. On cloud security software the Netherlands ranks fourth of the 27 member states at 55.7%, on development platforms fourth at 24.3%, and on bought-in computing power sixth at 26.3%.

On the headline measure the Netherlands is seventh of 27, at 68.5% against a European average of 52.7%. Dutch firms are not unusual in buying cloud. They are unusual in how much company data they place in it.

Size shifts the picture without removing it. Among Dutch enterprises with 250 or more employees, 90.8% bought paid cloud services in 2025 and 84.3% employed an ICT specialist in 2024, a gap of 6.5 points. Among those with 10 to 49 employees the figures were 64.5% and 21.8%, a gap of 42.7 points. The exposure runs deepest in the firms with the fewest ICT staff of their own.

The Netherlands is not short of ICT people in absolute terms. In 2025 the country employed 708,800 ICT specialists, 7.2% of everyone in work, against 5.0% across the European Union. Only Sweden and Finland score higher. A decade earlier the figure was 414,400.

The problem is in the hiring. In 2024, 15.3% of Dutch enterprises with 10 or more employees recruited or tried to recruit an ICT specialist, against 9.55% across the European Union. Of those that tried, 63.1% were left with a vacancy they could not fill. The European figure was 57.5%.

The reason cited most often was not money. Lack of applications was named by 8.60% of all Dutch enterprises, the second highest share among the 27 countries that publish the breakdown, behind Malta at 10.4%. Applicants lacking the right qualifications followed at 7.77%, lack of relevant work experience at 6.23%, and salary expectations being too high came last at 5.95%.

What happens when firms try anyway has been measured. The Netherlands Authority for Consumers and Markets surveyed 420 business cloud users. Of those, 172 had at some point tried to switch provider: 120 succeeded and 52 did not, so 30.2% of attempts failed. Among the firms that did switch, 52.5% used an intermediary, 53.3% said the provider they were leaving made it difficult, and 50.9% said the move delayed or otherwise obstructed their business processes. Just over half moved to a European provider.

That 52.5% is the key figure. Where a migration happened, the ICT capacity was more often bought in than supplied from inside.

Three things argue against the reading above, and they belong in it.

First, the ICT labour market has loosened rather than tightened. The job vacancy rate in the Dutch telecommunications, computer programming, consulting and computing infrastructure sector fell from 8.2% in the second quarter of 2022 to 5.1% in the second quarter of 2026. Across the whole Dutch economy it fell from 5.1% to 4.1% over the same period. ICT is still tighter than average, but by a narrower margin than it was: 1.6 times the national rate in 2022, 1.2 times in 2026. A shortage of vacancies is not the constraint. Vacancies that never fill are.

Second, Dutch recruitment difficulty has eased. The share of recruiting firms left with an unfilled vacancy fell from 70.4% in 2022 to 63.1% in 2024. In 2020 it was 70.1%. So 2024 is the best of the three waves, not a deterioration.

Third, the switching survey was in the field between 3 December 2024 and 13 January 2025, and the regulator says so itself: geopolitical developments since then could have influenced the results. The figures describe how hard switching was before firms had a political reason to want it. Whether demand has risen since is not something this data can answer.

There is also a wobble in the cloud series itself. Dutch adoption stood at 71.0% in 2024 and 68.5% in 2025, the first fall in the series. A single wave is not a trend, and it is reported here rather than smoothed away.

Companies do not publish their cloud dependence, so the only Dutch organisation whose position has been counted from the inside is central government.

In January 2025 the Netherlands Court of Audit reported 1,588 cloud services in use across the ministries. Of those, 700 were public cloud, 477 private or hybrid, and for 411 services, 26% of the total, the ministry could not say which form it was. Of the 700 public cloud services, 126 were classified as material to a primary task such as tax collection or issuing visas. For 84 of those 126, that is 67%, no risk assessment had been made before the decision to use public cloud. More than half of the material public cloud services are bought from Amazon, Microsoft and Google.

The market structure behind that is not new. In its 2022 market study the Netherlands Authority for Consumers and Markets found that Microsoft Azure and Amazon Web Services each hold 35 to 40% of the infrastructure and platform layers, in the Netherlands and in Europe, and that switching barriers are organisational, legal and financial at once.

• Cost an exit in people, not licences. Moving a hosted database or a development platform is an engineering project with a duration, and the Dutch firms that completed one mostly bought that capacity in.
• The constraint is applicants, not budget. In 2024 more Dutch firms named the absence of applications than named salary expectations. Raising the offer does not create a pipeline.
• The gap is widest at small firms. Between 10 and 49 employees, 64.5% buy cloud and 21.8% employ an ICT specialist. For an agency, that is the segment with the widest gap between dependence and in-house staff.
• The cooler market is an opening, not a solution. The ICT vacancy rate is at its lowest since 2022, so candidates are easier to reach than they were, and still 63.1% of recruiting firms did not fill the role.
• Treat a stated exit plan as a multi-year hiring commitment or a multi-year contracting one. This data supports no middle option.

Enterprise figures come from the Eurostat survey on ICT usage in enterprises, covering enterprises with 10 or more persons employed in NACE Rev. 2 activities C10 to S951 excluding the financial sector. Cloud adoption is from isoc_cicce_use, reference years 2021, 2023, 2024 and 2025. Enterprises employing ICT specialists is from isoc_ske_itspen2 and isoc_ske_itspe, reference year 2024. Recruitment and hard-to-fill vacancies are from isoc_ske_itrcrn2, reference year 2024.

Two reference years are unavoidably mixed. The cloud survey published both a 2024 and a 2025 wave; the ICT specialist and recruitment modules are biennial and stop at 2024, with no 2025 point. Where a Dutch cloud figure sits next to a Dutch staffing figure, both are taken from 2024. Where a Dutch cloud figure is compared across countries, 2025 is used, because that is the only year with full coverage.

All 27 member states report the 2025 cloud indicators. France is absent from the 2024 hard-to-fill vacancy indicator, and Ireland and Italy do not publish the breakdown of reasons, so that ranking covers 27 reporting countries rather than the full set.

Vacancy rates are from Eurostat jvs_q_r21, unadjusted, NACE Rev. 2.1. Under that classification the ICT sector is section K: telecommunication, computer programming, consulting, computing infrastructure and other information service activities. This is not the same boundary as section J under NACE Rev. 2, where publishing and broadcasting sat alongside computing services, and figures under the two classifications should not be read as one continuous series. Second quarters are compared with second quarters throughout.

ICT specialist headcounts are from Eurostat isoc_sks_itspt, reference years 2015, 2020 and 2025.

Switching figures are from the Data Act Cloud study by the Netherlands Authority for Consumers and Markets, reference ACM/25/197283, dated 8 April 2026. The sample is 420 business cloud users recruited through a panel and distributed roughly to the Dutch spread of firms by size: 10 at large enterprises, 90 at medium, 120 at small and 200 at micro enterprises. Fieldwork ran from 3 December 2024 to 13 January 2025. The report notes that geopolitical developments since then could have influenced the results, and that caveat applies to every use of those figures here. Market shares are from the same regulator's 2022 cloud market study, reference ACM/INT/440323.

Government figures are from the Netherlands Court of Audit report on central government cloud use, published on 15 January 2025.

One source was examined and left out. The experimental Eurostat statistics on online job advertisements, isoc_sk_oja1 and isoc_sk_oja3, report a 74.6% year-on-year fall in Dutch ICT job advertisements in the first quarter of 2026 and a 55.1% fall in the fourth quarter of 2025. Those movements cannot be reconciled with a vacancy rate that stayed flat at 5.1 to 5.3% across the same quarters. The series is therefore treated as affected by a change in collection, and no claim here rests on it.

Eurostat data is published under CC BY 4.0.