Ask what a nearshore developer costs in Europe and you'll get a number. Ask twice and you'll get a different one. The published rate tables for this market disagree with each other by a factor of three on the same country in the same year, and almost none of them say where their figures came from.

That's not sloppiness at the edges. It's the structure of the category.

What does a nearshore software developer cost in Europe in 2026?

There's no single answer, because "rate" names four different measurements. Employer labor cost in the ICT sector ranged from €24.00 an hour in Bulgaria to €71.70 in Ireland in 2025, per Eurostat. Vendor billing rates for senior engineers across Europe ran $64–76 an hour, per Accelerance. Those two numbers measure different things and neither is "the rate".

The gap between them isn't margin. Employer labor cost is what a company in Bucharest spends to keep a person employed for an hour. A vendor billing rate is what a client pays a supplier for an hour of delivered work, and it carries recruitment, management, quality assurance, unbilled bench time, holiday, sick leave, office, equipment, and profit. Comparing them is a category error, and it's the single most common mistake in this market.

Below, we've separated the four layers, named the source for each, and said what each one includes. We haven't published a rate of our own, because we'd be adding a thirteenth unsourced number to a market that already has twelve.

Why do published nearshore rates disagree so wildly?

Because most of them aren't measurements. They're estimates, and several are estimates of each other. Poland is the best-covered country in European nearshore content, which makes it the clearest illustration: five sources, all current, all presented as the cost of a Polish developer, spanning a factor of three from bottom to top.

Only the first row is a measurement. The rest are estimates, and two of them are estimates of each other.

The Brainhub page states its method in one sentence: "We've based our research on software quality rankings by HackerRank, TopCoder, and Pentalog, and companies' rates information available on Clutch." Clutch publishes bands, not point figures. Somewhere between the band and the published range, $50–99 became $45–70, and the conversion isn't shown.

Devico goes further in the wrong direction. Its Hungary figure is, by its own account, "calculated based on the average monthly salaries in the main hubs, like Budapest", linking to Levels.fyi, with Glassdoor used the same way for Moldova. That's a vendor billing rate derived from employee salary data, with no stated assumption about utilization, overhead or margin. Those are the three things that separate the two numbers. Its Polish figures carry no methodology statement at all.

Then there's the sanity check nobody ran. Index.dev publishes Hungary at $9–11 an hour for a full-time developer. Hungary's ICT employer labor cost in 2025 was €24.40 an hour, per Eurostat. A published rate roughly a third of the measured employer cost isn't a cheap market. It's a broken figure.

The four numbers everyone calls "the rate"

Four distinct measurements circulate under one label, and telling them apart is most of the work of reading a rate quote. Each answers a different question, each is published by a different kind of organization, and each includes and excludes different costs. Here they are side by side, with the question each one actually answers.

The two layers most often confused are the first and the last, and the confusion runs in one predictable direction: an employer cost figure gets quoted at you as though it were a price. Eurostat's €28.20 for Portugal is not what anyone will invoice you. It's what a Portuguese employer pays to have someone at a desk for an hour, before anything that makes that hour into shipped software.

What employers actually pay: the only harmonized number in Europe

Eurostat's labor cost series is the only figure in this article that's comparable across countries by construction rather than by hope. It's compiled to a common legal standard by every member state's statistics office, published free, and documented in full. For the ICT sector in 2025:

Source: Eurostat lc_lci_lev, lcstruct D1_D4_MD5, EUR per hour. ICT column is NACE Rev.2 Section J; whole-economy column is aggregate B-S_X_O from the same dataset, with the EU, Netherlands, Bulgaria, Romania and Hungary figures also stated in Eurostat's March 2026 release. Netherlands ICT 2025 is flagged confidential, so the 2024 figure is shown and the premium is left blank. Premiums are our arithmetic on Eurostat's published values.

Two things in that table are worth more than the ranking. The first is Serbia at €28.20, identical to Portugal, from Eurostat's own series, with no third-party estimate involved. The second is the premium column. Romania's ICT sector costs 1.91 times its whole economy and Bulgaria's twice, against 1.32 in Spain. The cheap-country story that gets told from whole-economy numbers doesn't survive contact with the sector figures, because ICT is where the wage convergence has already happened.

What the Eurostat number includes, in Eurostat's own words: "Wage and salary costs include direct remuneration before deduction of employees' social contributions, bonuses, and allowances paid by an employer in cash or in kind to an employee in return for work done, payments to employees' saving schemes, payments for days not worked and remuneration in kind such as food, drink, fuel, company cars, etc. Non-wage costs include the employers' social contributions plus employment taxes regarded as labour costs minus subsidies, intended to refund part or all of the employer's cost of direct remuneration."

Two details matter when you use it. The denominator is hours actually worked, not hours paid, so holiday and sick leave sit in the numerator and out of the denominator. And bonuses and payments in kind are included, which is why this number comes out higher than any salary figure you'll see for the same country and year.

Section J isn't the software sector, and it doesn't overstate Eastern Europe

Eurostat's Section J covers "Information and communication", which means software sits alongside telecoms, publishing, film and broadcasting in the same figure. The obvious worry is that the bundle distorts the software number, and that quoting Section J flatters cheap markets by mixing in low-paid media and telecom work. It does distort it. Just not in that direction.

The Labour Cost Survey publishes a narrower division, J62 "Computer programming, consultancy and related activities". Comparing J62 against Section J in the same reference year:

Source: Eurostat lc_ncost_r2, NACE J62, D01 total labour costs excluding apprentices, EUR per hour, enterprises with 10 or more employees, reference year 2020. Section J column from lc_lci_lev for the same year, all enterprise sizes.

At EU level the two are almost the same. In exactly the markets nearshore content is about, Section J runs low: programming costs 18.5% more than the sector average in Romania, 20.3% more in Bulgaria, 28.7% more in Serbia. In Western Europe the effect reverses, mildly. Which means the standard move of quoting Section J as a proxy for software understates Central and Eastern Europe and slightly overstates Ireland and Spain, narrowing the apparent gap between them.

Three limits on that table, and they're real. The Labour Cost Survey runs every four years and 2020 is the most recent published round. The 2024 round is due, but it wasn't in the database when we checked on 24 August 2026. The J62 column is restricted to enterprises with 10 or more employees while the Section J column carries no size restriction. And D01 excludes apprentices where the annual series includes them. So the two columns aren't perfectly like-for-like, and the percentages should be read as a direction rather than a precise differential. We've published it anyway, because a directionally clear finding from the official source beats silence, and because we'd rather show the caveat than hide the number.

What developers are paid, country by country

Developer pay is the layer you'll most want and the layer where cross-country comparison genuinely breaks down. Here's the best official figure per country, taken from statistics offices rather than vendor surveys, because official series disclose their methodology and are free to check.

Now the part that stops you putting that table in a spreadsheet and sorting it.

Some of those figures are net and some are gross. Romania and Ukraine above are net, because that's the convention their sources publish in. Poland, Bulgaria, Romania and Ukraine habitually report IT salaries net, and it's the most common cause of a bad cross-country comparison. Read a net figure as gross and you'll understate employer cost by roughly a third.

Payment counts differ. Portugal and Spain pay 14 monthly salaries a year, not 12. Multiplying a Portuguese monthly figure by twelve produces a number nobody receives.

Contract mix differs structurally. Senior IT work runs largely through B2B and self-employment in Poland, Romania and Ukraine, and largely through employment contracts in Portugal, Ireland and the Netherlands. Ukraine's sector is built on the private-entrepreneur model, which is why an employer-cost figure comparable to Eurostat's doesn't really exist there, and why Ukraine is absent from Eurostat's labor cost series entirely.

Sector isn't occupation. Most of those rows are NACE Section J, which includes the telecoms account manager and the broadcast engineer alongside the Back-End developer. Czechia is the exception and it's instructive: the Czech statistics office publishes software developers as a distinct category at 113,200 CZK a month, against roughly 99,800 CZK for ICT specialists as a whole. That occupational split is why Czechia has the most trustworthy pay data of any country in this set.

Portugal shows the sector-versus-occupation gap most clearly. INE's official figure for Section J is €2,084 a month. The Landing.Jobs and Damia Tech Talent Trends Report 2026 puts the average gross tech salary at €53,671 a year, up 0.9% on the previous year and up roughly 74% since 2019. Both are right. INE counts every employee in the sector from administrative staff up; the report surveys self-selecting technology professionals on a jobs platform. The gap between those two numbers is the finding, not an error to reconcile.

What freelancers charge, and what that number leaves out

Freelance rate data for Europe is better measured than any other layer and covers almost none of the countries you'd ask about. That's not a coincidence. The platforms with real transaction data operate in markets where day rates are the norm, and those aren't the nearshore markets.

YunoJuno publishes the strongest dataset in the category: more than 182,000 contractor, booking and rate data points across 2024 and 2025, taken from platform transactions rather than a survey, with outliers removed. Software engineering came out at £533 a day in the UK and $88 an hour in the US, published as a single aggregate. The report notes that role and skill-level splits appear "only where there is sufficient and statistically meaningful volume of data", and software engineering didn't clear that bar for a seniority breakdown. Neither market is a European nearshore market.

Freelancermap's Freelancer-Kompass 2026 surveyed around 5,400 freelancers between 17 November 2025 and 8 February 2026, cross-validated against data from over 340,000 platform users, and published its own sample skew rather than burying it. Average rate across all disciplines: €103 an hour. Coverage is Germany, Austria and Switzerland.

In the Netherlands, HeadFirst Group and Intelligence Group reported ICT freelance rates passing €100 an hour in 2025, up 4.7% year on year, with enterprise architects at €127. The figures are credible and the report is behind a form with no methodology on the landing page.

For Portugal, the most specific published figures come from Lemon.io's rate calculator: $40–45 an hour for a senior developer, $44–53 for a lead or architect, sourced to "actual contracts with 2,400+ vetted developers, not self-reported surveys that skew high" and 248 vetted Portuguese developers. Use it with one caveat that the page never resolves: it doesn't say whether that rate is what the developer receives or what the client pays. The platform is a marketplace, and the difference between those two numbers is its business model.

That ambiguity is the norm. Of every source in this article, only Malt states plainly what its rate covers: the freelancer's own rate, excluding Malt's service fees, travel and taxes. And Malt's 2026 Tech Trends report contains no rate figures at all.

What vendors bill, and why nobody publishes it properly

Vendor billing rates are the number you actually need, and the layer with the weakest published data. Every source that names European countries either declines to give a figure, gives a band without a sample, or discloses a sample small enough to invalidate the figure.

Accelerance publishes the most usable European figure: junior $31–39 and senior $64–76 an hour, down 4.4% year on year. It's a single Europe-wide range: no Western Europe figure, no Central and Eastern Europe figure, no country broken out. And the sample is stated twice, differently: the blog post says "our survey of 60 software development partners from around the world", while the guide's landing page says it draws on "data from over 100 firms across North America, Latin America, Europe, and Asia". Either way, respondents are Accelerance's own certified partner network, and 60 to 100 firms across four continents leaves a per-country sample of roughly nothing.

Clutch publishes bands by country, sourced "through its first-party reviews submitted by verified software development clients", with no sample size, no company count per country and no median. Of the countries in this article it covers Poland at $50–99, Spain at $25–49 and Ukraine at $25–49. Portugal, Romania, Czechia, Estonia, Ireland and the Netherlands have no cell. The United Kingdom's cell reads "Unknown".

Two sources deserve credit for disclosure that undermines their own numbers. Techreviewer surveyed 127 providers in January and February 2026 and published its geographic composition: Portugal 3.1%, Poland 3.9%, Ukraine 5.5%. That's roughly four Portuguese companies and five Polish ones. To its credit, Techreviewer publishes no country-level rate. GoodFirms surveyed 100-plus companies in September and October 2025, disclosed that Western Europe was 11.4% of respondents, and published a distribution rather than country averages: 56.3% of firms bill $20–50 an hour, 12.5% bill $50–100, 10.9% bill $100–250.

The advisories that hold the real data don't publish it. Everest Group's Pricing Index covers Poland among six locations and shows no figures publicly. ISG's Rate Card Benchmark is a paid service claiming the most comprehensive rate card database in the world and publishing none of it. Deloitte's Global Outsourcing Survey is free, but the most recent edition is 2024 and it carries no rate benchmarks at all.

And almost nobody says what the rate includes. Clutch, GoodFirms, Techreviewer, Lemon.io, Devico, nCube, Brainhub and YunoJuno all publish rate figures without stating whether project management, quality assurance, overhead or margin sit inside the number. For a vendor rate that omission is decisive: two suppliers quoting the same hourly figure with different inclusions aren't quoting the same price.

Which sources disclose a methodology, and which just assert numbers

We checked 33 published sources across the four layers, and the useful way to sort them isn't by the numbers they give but by whether you can verify what you're reading. Four tiers, plus a fifth for the organizations that hold the best data in the market and publish none of it:

One entry in the second tier deserves singling out. DOU.ua's summer 2026 developer salary report surveyed 4,541 developers in May and June 2026, published its field dates, published its sample composition, and published the raw data as a CSV on GitHub. It's a free Ukrainian community site, and it's the most verifiable rate publication we found anywhere in Europe. The outsourcing advisories charging for benchmark access don't come close.

The correlation nobody in this market wants to talk about

Across all 33 sources, methodological quality and country coverage run in opposite directions. The better a source's method, the less likely it covers the countries you're asking about. The more countries a source names, the less it discloses about where the numbers came from. That's the structural finding, and it explains why this question never gets a good answer.

The sources with genuine transaction data, YunoJuno and Freelancermap, cover the UK, US, Germany, Austria and Switzerland, and not one of the nearshore markets. The sources that name Portugal, Romania or Estonia are the ones with no sample at all. In between sit Techreviewer and GoodFirms, which have real samples, disclose them, and are then honest enough not to publish country figures their sample can't support.

So when a page gives you a confident hourly rate for eight European countries, the confidence is the tell. Nobody has that data. What they have is Clutch's bands, somebody's salary aggregator, and a spreadsheet.

Eurostat is the exception, and it's the exception because it isn't trying to sell anything.

How to read a nearshore rate quote

Five questions turn an unusable number into a comparable one, and they work as well on a supplier's proposal as on a published rate table. None of them require you to know the market. They only require the person quoting to say what their number measures. If they can't answer three of the five, you don't have a price yet.

1. Which of the four layers is this? Employer labor cost, developer pay, freelance rate, or vendor billing rate. If the source doesn't say, assume it doesn't know.

2. Gross or net? Ask explicitly for Poland, Romania, Bulgaria and Ukraine, where net is the publishing convention.

3. What's inside the rate? Project management, quality assurance, delivery management, holiday cover, recruitment, bench time. Two identical hourly figures with different inclusions are different prices.

4. What's the utilization assumption? A rate is meaningless without the hours behind it. Billed hours, hours worked and hours paid are three numbers, and the gap between them is where cost hides.

5. What's the sample? How many companies, in which country, collected when. A rate table with no sample is an opinion in a table.

Why a rate card can't answer the question you're actually asking

If you're comparing an hourly rate against building a team, the rate isn't the number you need, and picking the cheapest hour is how teams end up with the most expensive year. A rate prices an hour of someone's attention. What you're buying is a team that ships, and the distance between those two things is where most of the money goes.

Rate times hours doesn't give you the cost of a team. It leaves out the six to nine months before a self-built hub produces steady output, the recruitment you run twice because the first hire didn't stay, the management time that shows up on your side of the ledger, and the difference between hours billed and features shipped. Those are the numbers that decide whether an engagement worked, and none of them appear on a rate card.

We publish one figure and we'll repeat it here with its context rather than as a headline: a Tech Hub in Portugal on the Build-Operate-Transfer model starts at €400k a year on a 12-month minimum, covering the team and the operational support around it, with the pilot phase running at a minimum 30% off our standard rates. That's a total, not a rate, because a total is the thing you can actually compare. We've broken down what sits inside it in how much it costs to set up a Tech Hub in Portugal, and the country-choice question separately in best countries to build a Tech Hub in Europe. For the models either side of it, Staff Augmentation versus Managed Services versus Build-Operate-Transfer sets out the tradeoffs.

The honest summary of this whole category is short. There's one harmonized cross-country dataset for what employing an engineer in Europe costs, it's free, and it's Eurostat. Everything above that layer is either a small self-selected sample or somebody's arithmetic on somebody else's arithmetic. Anyone quoting you a precise figure for eight countries is telling you more about their sourcing than about the market. But once you can name which of the four numbers you're being shown, you can ask the only question that matters, which is what it costs to get the thing built.