Home / Insights / Onshore or Nearshore Software Development in 2027: What the Figures Say
Insights

Onshore or Nearshore Software Development in 2027: What the Figures Say

Onshore Dutch development is competitive per unit of output: Polish software cost per employee rose to 63% of the Dutch level, and measured against value added the hourly gap largely disappears. Dutch software firms also use AI more.

Summarize with AI Prompt copied. Paste it into the chat

Nearshore software development is no longer the automatic saving it was. On our analysis of Eurostat data, Polish personnel cost per software employee rose from 49.7% to 63.1% of the Dutch level between 2021 and 2024, while Dutch software firms raised productivity by 22.1% against a 17.6% rise in cost. Measured per euro of output, most of the hourly gap disappears. AI-assisted development now works on the output side of that sum, and Dutch software firms use it more than Polish ones.

This piece is for the IT manager, IT director or CIO at a Dutch or Flemish organisation of 250 to 5,000 FTE who decides, before the 2027 budget closes, where the next system gets built. Crux Digits designs and builds AI systems and the software around them, alongside your ICT partner. We are an onshore supplier ourselves, so we have kept every number checkable against the study.

Is nearshore software development still less expensive than building in the Netherlands?

Per hour yes, per unit of output mostly not: Polish employee cost per hour is 41.7% below the Dutch level, but output differs more.

In 2024 a person working in Dutch computer programming (NACE 62.01) produced €108,440 of value added, sixth of 27 member states and 34.3% above the EU figure. Polish value added per person stayed at about 35% of the Dutch level. Put cost against output and Polish personnel cost per euro of value added lands between 0.98 and 1.80 times the Dutch figure. The range is wide because 51.8% of the people working in Polish software are not employees, so the answer depends on how the self-employed are counted.

Romania sits at 0.96 to 1.02 times the Dutch level, close to parity. Germany, often treated as the safe nearby alternative, comes out at 1.18: Dutch software staff produce 14.5% more value added per head at 3.1% lower personnel cost per employee. Belgium is the one neighbour that beats the Netherlands per unit of output, at 0.90.

That is the good news inside the data. Dutch software firms became more productive faster than they became more expensive, and the onshore option is more competitive than an hourly comparison suggests.

How should you compare an onshore and a nearshore supplier?

Divide the cost ratio by the productivity ratio: below 1 the other country costs less per unit of output, above 1 it costs more.

This is the rule of thumb from our software productivity league, and it works for any pair of countries. Value added per euro of personnel cost gives the same picture as one number per country: in 2024 Belgium scored 1.41, the Netherlands 1.28, the EU 1.15 and Poland 0.71, last of 27.

Two refinements matter for a buyer of your size. First, compare against the right size class. In Polish software firms with 250 or more people, value added per person was €61,360, almost three times the €21,340 of firms with fewer than ten. Large Dutch firms produced €126,380. The supplier you would actually contract sits in the large-firm figure, not in the national average.

Second, remember what the data is not. Value added per person is not a price: it includes the supplier's operating surplus, and a contract adds margin, travel and management on top. The ratio shows where the structural advantage sits. It does not predict what a quote will say.

What will the cost gap look like in 2027?

On the 2022 to 2024 pace, Polish cost per software employee reaches 78.0% of the Dutch level in 2027.

That is a labelled extrapolation on three published points, not a forecast and not a Eurostat figure. Polish cost per employee grew 14.3% a year from 2021 to 2024, against 5.5% in the Netherlands. The same arithmetic puts Romania at 51.9% and Portugal at 61.0% of the Dutch level in 2027.

Pull quote from Crux Digits: Compare software suppliers per euro of output, not per hour: that is where the nearshore gap disappears.

The movement is not all one way, and that is encouraging. Between 2022 and 2024 Portuguese software productivity rose 26.2% against cost growth of 14.0%, and Romanian productivity 22.1% against 19.6%. Those two markets, like the Netherlands, are getting more productive faster than they get more expensive. Poland is the market where cost is outrunning output.

If these trends hold, 2027 is a good year to be a buyer. The Dutch market keeps turning cost growth into more output, two nearshore markets are improving on the same terms, and the comparison that matters, output per euro, is easier to make than it was. For a contract signed now that runs into 2028, the third year is the one to price.

How does AI-assisted development change the onshore versus nearshore equation?

AI works on the output side of the ratio, and Dutch software firms use it more: 71.5% in 2025 against 42.0% in Poland.

Across EU software and IT service firms (NACE 62 and 63), AI use rose from 28.2% in 2021 to 65.7% in 2025. The share using AI to generate text, speech or code went from 6.7% to 45.5%, and in the Netherlands from 10.0% to 52.7%. Countries whose software firms use more AI tend to produce more value added per person: the rank correlation is 0.56 across 24 member states, and 0.72 without the four small economies where profit booking distorts the figures. Correlation does not prove that AI causes the gain, because productive firms may simply adopt it sooner. For a buyer both readings point the same way.

The productivity research is real and still maturing. In a controlled experiment, developers with GitHub Copilot finished a set task, an HTTP server in JavaScript, 55.8% faster than a control group. METR found the opposite in early 2025: 16 experienced open-source developers took 19% longer on 246 issues in their own repositories when AI tools were allowed. In a February 2026 update METR marked those results out of date, estimated an 18% reduction in task time for returning developers, and wrote that developers are likely more sped up now, while calling its own data very weak evidence of the size. Google's 2025 DORA report, a survey of nearly 5,000 technology professionals, found that 90% use AI at work and that AI adoption has a positive relationship with delivery throughput and a negative one with delivery stability.

Our reading, not the data's: AI makes writing code faster and leaves specification, review, testing and integration as the work that decides the outcome. DORA describes AI as an amplifier of what a team already does well or badly. That work sits close to your organisation: the integration with AFAS, Exact or SAP, the change request through your change advisory board, the audit trail your security officer signs off. As code generation speeds up, that close-in work becomes a larger share of every project. That favours onshore delivery for integration-heavy work, and it favours any supplier, onshore or nearshore, that can show how AI is used in its delivery.

What should an IT manager do in the next twelve months?

Five moves, each tied to a finding: buy output, price year three, test AI in delivery, keep integration close, and benchmark Belgium.

  1. Ask for output, not hours. The hourly gap with Poland is 41.7%, the per-unit gap between 0.98 and 1.80 times the Dutch level. Ask every bidder for a fixed price per milestone with acceptance criteria, and use the rule of thumb to check the structural position behind each bid.
  2. Price the third year. Polish cost per employee grew 14.3% a year, and the extrapolation reaches 78.0% of the Dutch level in 2027. Write an indexation clause with a named index and a cap into any multi-year nearshore contract, and model year three in the business case.
  3. Make AI in delivery a selection criterion. Dutch software firms lead Polish ones on AI use by 29.5 points, and AI use tracks value added per person. Ask each supplier where AI generates code and tests, who reviews that output, and how it reports throughput and stability, the two delivery measures DORA tracks. A supplier that cannot answer is pricing last year's productivity.
  4. Keep integration, change and audit work close, and send bounded work where the ratio favours it. Romania and Portugal sit near parity and are improving faster than their cost. Well-specified modules with clear acceptance tests travel well; integrations with core systems and anything inside your audit scope do not.
  5. Benchmark Belgium. Belgian software firms produce €140,060 of value added per person, fourth in the EU, and per unit of output cost 10% less than Dutch ones. For Flemish organisations that is the first comparison; for Dutch ones it is a nearby alternative worth pricing.

Where does Crux Digits fit?

Crux Digits designs and builds AI systems and the software around them in the Netherlands and Flanders, alongside your ICT partner or IT team.

We hand over code you own. We do not run infrastructure, hosting or a helpdesk, and we do not replace your ICT partner. We are an onshore supplier with an interest in this question, which is why the argument above rests on Eurostat figures and published research you can check, not on our word.

For how we approach that work, see custom software development.

Last updated: 28 October 2026.

Need custom software built?

Software built around your process and data, with AI where it changes the outcome. Something working every week, and the code is yours.

Custom software development in the Netherlands →

Frequently asked questions

Does AI-assisted coding actually make developers faster?

On bounded tasks, yes: in a controlled experiment developers with GitHub Copilot built an HTTP server in JavaScript 55.8% faster than a control group. On real work in large, familiar codebases the evidence is mixed. METR measured a 19% slowdown in early 2025, then in February 2026 called that out of date and estimated an 18% reduction in task time for returning developers, while calling its own evidence weak. The gain depends on the task and the team.

How do I put AI use into a software development tender?

Ask three things and score the answers. Where in delivery does AI generate code and tests, who reviews that output before it reaches your environment, and how does the supplier report throughput and stability, the two delivery measures the DORA research tracks. DORA found that AI adoption goes with higher throughput but lower stability, so a supplier that shows both numbers is showing whether its testing keeps up with its speed.

Is Belgium a less expensive place to build software than the Netherlands?

Per unit of output, yes. On our analysis of Eurostat data for 2024, Belgian software firms produce 29.2% more value added per person than Dutch ones at 16.8% higher personnel cost per employee, so per unit of output Belgium costs about 10% less. Belgium ranks fourth in the EU on value added per person employed in computer programming. These are industry figures, not supplier quotes, so a tender still decides the price.

Is the 78% figure for Poland in 2027 a forecast?

No. It is a labelled linear extrapolation in our study, not a forecast and not a Eurostat figure. It takes the share of Polish personnel cost per employee relative to the Dutch level in 2022 and 2024 and extends that pace by three years: share in 2027 equals share in 2024 plus three times the two-year change divided by two. It shows the direction if nothing changes, and it will be wrong if wage growth slows.
Our AI services AI consultancy AI automation AI agents AI implementation Pricing

Want any of this applied to your business?

We turn these concepts into working tools: grounded, safe and measurable. Start with a free consultation.

Book a free consultation →