In 2025, 19.95% of EU enterprises with ten or more employees used at least one AI technology. In Romania it was 5.21%, the lowest share in the EU. The gap is mostly about size: 55.0% of large EU firms use AI, against 17.0% of small ones. Of the firms that considered AI and stopped, 70.3% cite a lack of expertise.
All the figures here are Eurostat's, from its annual survey of ICT use in enterprises, and three details matter before reading them. The survey covers firms with ten or more employees. It leaves out the financial sector. And "uses AI" means the firm reports using at least one of several AI technologies, so the figure is self-reported and says nothing about how deeply the tool is used.
How many European companies use AI in 2025?
One in five, and the share is rising fast. EU adoption went from 13.48% of enterprises in 2024 to 19.95% in 2025 (Eurostat). The leaders are in the north: 42.03% in Denmark, 35.04% in Sweden, 33.21% in the Netherlands and 25.97% in Germany. At the other end, Eurostat's own report puts it plainly: "the lowest adoption rates were recorded in Romania (5.2%), Poland (8.4%), Bulgaria (8.6%) and Greece (8.9%)" (Eurostat, KS-01-26-009).
Why does company size matter so much for AI adoption?
Because the gap between countries is smaller than the gap between a large company and a small one. Across the EU, 55.0% of large enterprises used AI in 2025, against 30.4% of medium-sized ones and 17.0% of small ones (Eurostat, KS-01-26-009). Romania has the same shape at a lower level.
| Enterprise size, 2025 | Romania | EU-27 |
|---|---|---|
| Small (10 to 49 employees) | 4.1% | 17.0% |
| Medium (50 to 249) | 7.8% | 30.4% |
| Large (250 or more) | 20.8% | 55.0% |
| All enterprises (10 or more) | 5.2% | 20.0% |
Source: Eurostat, KS-01-26-009, Table 1 and page 14. Read the last column against the first row and something stands out: Romania's large companies use AI at only slightly above the rate of the EU's small ones. Romania's national figure is low mostly because Romania's economy is made of small firms, and Romanian small firms barely use AI at all.
Size matters for a mundane reason. A company of 300 people can afford someone whose job includes working out where AI fits. A company of 30 usually cannot, and the barrier data below says that this missing person is exactly what holds firms back.
Which sectors use AI least in Romania?
Every sector Eurostat publishes sits well below the EU average, including the one you would expect to lead.
| Sector, 2025 (enterprises with 10+ employees) | Romania | EU-27 |
|---|---|---|
| IT services (programming, consultancy, information services) | 34.2% | 65.7% |
| Telecommunications | 13.9% | 39.8% |
| Administrative and support services | 5.0% | 19.9% |
| Manufacturing | 3.2% | 17.3% |
| Retail trade | 1.7% | 15.5% |
Source: Eurostat, isoc_eb_ain2. The first row is the surprise. Romania's IT services firms are its most AI-heavy sector, and still the lowest in the region: 42.0% in Poland, 44.7% in Bulgaria, 51.2% in Hungary and 64.1% in Czechia use AI, against 34.2% in Romania. For the people side of that sector, see Romania as an IT delivery location in 2026. At the other end, fewer than two Romanian retailers in a hundred report using any AI technology.
Is AI adoption in Romania growing?
Yes, from a low base, and mostly at the top. Romanian adoption rose from 3.07% of enterprises in 2024 to 5.21% in 2025. Among large Romanian firms it nearly doubled, from 11.26% to 20.75%, while large EU firms went from 41.17% to 55.03% (Eurostat). The movement is real. It is happening in the companies that already had the people to make it happen.
What stops companies from using AI?
Eurostat asks the firms that considered AI and did not adopt it why. 70.3% named a lack of relevant expertise, the most common reason by a wide margin (Eurostat, KS-01-26-009, Table 7). Unclear legal consequences and data-protection concerns came next, each cited by about half. Cost came only sixth, at 38.4%, and 17.8% said AI was not useful for their business.
The denominator matters. These are shares of the firms that looked at AI, not of all firms. Measured against all enterprises, the share that does not use AI for lack of expertise was 10.77% in Germany and 7.81% in the Netherlands in 2025 (Eurostat). So the honest reading is narrower than "Europe lacks AI skills". Among the companies that got as far as asking the question, most stopped because nobody inside could answer it.
Do European companies build AI themselves or buy it?
Large companies increasingly bring someone in. In 2024, 18.69% of large EU enterprises had AI systems developed or modified by external providers (Eurostat). Whether buying works better than building is a harder question, and the evidence on it is thinner than vendors suggest; I went through it in what Klarna's AI U-turn teaches.
What does this mean for a small company in Romania?
The data points at the person, not at the budget. A firm that stalls on AI is, by its own account, usually missing someone who can tell which process is worth automating and what a good result would look like. Cost is a real constraint, and it is the sixth one.
That changes the first step. Before choosing a tool, it is worth borrowing the expertise for one process, measured before and after: a guide to that is in how to implement AI in your company, and the question of which process to start with is in what is worth automating. If the answer involves outside help, how to choose an AI consultant and what an AI consultant costs cover the buying side.
A country at 5% is usually described as behind. Read by company size, Romania looks less like a country that rejected AI and more like one where the large firms have started and the small ones are still waiting for someone who can show them where.
If you want an outside view on where AI would pay in your company: vladtudor.com/consulting.