In the evening roundup of the world press, there was a line that would usually pass unnoticed: ministers from five countries—Germany, Austria, Greece, Denmark, and the Netherlands—had agreed on “guiding principles” for deportation hubs outside the EU, expected an understanding with one unnamed partner country “around the New Year,” and promised a launch in 2027. In retellings, this was presented as a shift in migration policy from a humanitarian framework to a budgetary one.
The political side of this story can be picked apart from either angle. The budgetary side cannot. Because a project with eight months to go before its supposed launch has no country, no price tag, and no source of funding from the EU budget. Previous attempts at the same thing, though, have all three. So the question is not “is this good or bad?” but something duller and harsher: how much did it cost the state to stop handling the matter at home, and how many people ultimately left.
Spoiler: more than at home, and almost no one.
The Italy–Albania scheme is the most thoroughly documented example, because it had time to happen, fail, reinvent itself, and undergo a financial audit. The protocol was signed on 6 November 2023 and ratified on 21 February 2024. Adult men from “safe countries,” intercepted at sea, were to be transported by ship to Albanian centres, have their claims examined within 28 days, and most often be sent back. The government announced a figure of €670 million over five years; the parliamentary opposition said it was almost a billion.
Construction at an abandoned airbase near Gjadër was assigned to the engineering corps of the airbase in Bari on 4 April 2024, with a deadline of 223 days and a reference to “exceptional necessity and urgency.” Eight days later, when it emerged that the ground was unstable and thousands of concrete piles would have to be driven into it, while construction was running into the peak of Albania’s tourist season, contractors had to pay workers €150 a day—against an average salary in Albanian construction of around €700 a month, or thirty a day.
The financial trail is tangled to precisely the extent the law allows. Initially, €39.2 million was earmarked for construction. Ten days later, a clause in the recovery-plan law raised the cap to €65 million; in May 2024 the protocol was amended again—to €73.5 million. Contracts were ultimately signed for €74.2 million, and by March 2025 €61.3 million had been transferred. The money came from everywhere except the “migration” line item: €15.8 million from a fund for “unavoidable needs,” usually spent on climate emergencies; €3.8 million from the Education Ministry, €3.8 million from the Culture Ministry, €3.2 million from the Health Ministry.
The comparison that wipes out all the rest of the arithmetic is the cost of a single place. The centre in Modica, Sicily, opened in 2023 as a model for the Albanian project, cost €1.65 million for 256 places—€6,445 per bed. The Albanian centres, with 1,024 places—€72,461 per bed. Eleven times more expensive.
The protocol’s emergency clauses made it possible to bypass public-procurement rules: every contract was awarded through direct negotiations, without an open tender, and the signatories were bound by confidentiality agreements limiting contact with the press. The biggest beneficiary was a Tirana construction company that received €36.4 million—almost as much as it had earned from all Albanian public contracts since 2019 combined. One hotel in Shëngjin received €352,630 for housing construction workers at €110 per person per day; another, up to €9 million for feeding and housing 295 Italian police officers.
The seismic assessment of the site was given to Bari Polytechnic and the Institute for Seismic Risk in Pavia, with 30 days allotted for it. A professor of urban planning from Brescia commented on the deadline in a single sentence: “Thirty days seems too short.” An engineer from Milan, who asked to remain anonymous, said a project of this scale normally requires six months. Italy had already had a precedent: in 2019, part of a centre in Potenza had to be demolished and rebuilt because of soil liquefaction.
Now for the result. According to Italy’s Interior Ministry, by July 2026 685 people had been transferred to Gjadër, of whom 101 were repatriated—14.7%. The rest returned to Italy: they applied for asylum, were deemed medically vulnerable, or were released by court order. Directly from Albania, bypassing Italy, five Egyptians were deported over the entire period, in May 2025.
The centre, meanwhile, was almost always half-empty. On 29 June 2026, 26 people were physically there; in a later update, 83. In February 2026, the TAI monitoring group recorded around 90 people and called it a record since opening. Before that, transfers came in batches of roughly ten people, with around twenty present at any one time. Out of the fourteen and a half thousand places envisioned by the scheme, fewer than a hundred were used.
The cost of maintaining the building itself does not depend on how many people are in it. Running the Albanian facilities in 2025 cost around €50 million, while the multi-year financial plan provides for roughly €671.6 million through 2028. Keeping one bed in Gjadër for two months while the centre is half-empty costs around €1,500—which, according to ActionAid’s calculations, is roughly equal to the entire annual cost of a bed in Modica.
A separate line is the legal scaffolding around the procedure. Human-rights advocates who visited the centre described how some people received no formal transfer order, although courts had already found the absence of one unlawful; how selection criteria were opaque; how transfers involved the use of force without an individual assessment of necessity. It later emerged that in cases where the vulnerability commission did convene, the person was returned to Italy as unfit for detention. Italy’s Court of Auditors received a 60-page complaint alleging waste of public funds, while the National Anti-Corruption Authority received a notice of suspected irregularities in the €133 million tender to operate the centres.
There is an earlier scale for comparison, too. In 2023, 10% of people issued with removal orders were repatriated from Italian temporary detention centres—2,987 out of 28,347. From 2014 to 2023, 50,000 people passed through them, and they were occupied just over halfway. The average cost of repatriation set by Italy’s Interior Ministry for 2026 is €3,820 per person, as a weighted average across 6,172 actual returns in the previous year. The trip to Albania and back does not fit into that calculation in any way.
The British version of the same idea is the most candid in its documentation, because the Home Office published the breakdown itself. By June 2024, the Rwanda scheme had consumed £715 million. Of that, £290 million went to Rwanda: £270 million for an economic development and integration fund, regardless of the number of people transferred, plus £20 million in advance for operating costs. Each person transferred was separately entitled to a five-year integration package—£150,874, with the first year costing £45,262 and the last £15,087. Plus £20,000 per person into that same fund and a one-off £120 million once the threshold of 300 transfers was reached.
Over more than two years of operation, four volunteers and not a single asylum seeker went to Rwanda. The first scheduled flight was cancelled in June 2022, a few hours before take-off, by an ECHR ruling. In November 2023, the UK Supreme Court found the policy unlawful. In the summer of 2024, the new government scrapped it, Rwanda presented a bill for £50 million, and the dispute went to the Permanent Court of Arbitration in The Hague. In May 2026, the arbitrators ruled for the UK: the written exchange between the parties after the cancellation was treated as confirmation that no payments would be made.
The Danish story is interesting for something else: the way a law can exist without ever being used. Bill L226 was introduced in February 2021 and passed on 3 June that year by 70 votes to 24. Its premise was more radical than Britain’s: to shut down the processing of claims on Danish territory altogether, and send a person recognised as a refugee not to Denmark but to an unnamed host country. One detail passed almost unnoticed at the time: not a single country had agreed to take part before the law was passed, and no partners were named. The countries politicians mentioned as possible candidates—Morocco, Tunisia, Algeria, Jordan, Libya, Egypt—all refused once they learned they had already been named.
Since then, the project’s format has shrunk in four moves: in 2016 they promised “huge refugee cities with hospitals, schools, universities and farms,” then it became “camps for asylum seekers,” in 2018 “reception centres,” and in 2021 the minister was already talking about an “experimental mini-centre.” In 2026, Denmark once again gathered ministers in Copenhagen, and EUobserver accompanied its report with the phrase “despite a history of failed projects.”
The Australian model has neither deadlines nor any prospect of 2027—it has been operating since 2013, and it comes with a long cost history. Keeping one person on Nauru costs the Australian budget nearly A$22 million a year. For comparison: detention in a centre on Australian territory costs A$429,000, while an asylum seeker living in the community while their case is considered costs A$3,962.
The offshore-processing system cost an average of at least A$1 billion a year, reaching A$1.49 billion in 2017–18. Spending exceeded budget allocations every year: A$483 million was planned for 2022–23; A$611 million was spent. The Refugee Council of Australia counted A$12 billion from July 2012 to June 2024. By a 2015 estimate, Australia spent twice as much to keep an asylum seeker in detention as the United States, Canada, or European countries—around $655 per person per day.
The UN, in its report on outsourcing migration management, writes that thousands of people, including children, were held for years in arbitrary detention in inhuman conditions, and that this led to at least twelve deaths. The return rate in this system has been counted too: of the thousands sent to Nauru and Manus Island, nobody ultimately made it to Australia.
The United States has taken the same logic to an industrial scale: agreements with more than 35 countries, including South Sudan, and, by human-rights groups’ count, more than 23,000 people taken to countries with which they have no ties at all. Liberia accepted the first group of 20 people under a deal for 1,200. A joint statement by Human Rights First and Refugees International puts it briefly: the agreements treat people as bargaining chips. The irony is that European agencies look at these statistics as a benchmark for a catch-up model.
The main argument made by opponents of outsourcing is usually not about money. It is that a state can build a building anywhere, but responsibility for decisions taken inside that building remains with it. And that is not an abstract moral argument—it is the position of courts, built up through specific cases.
In 2009, Italian ships intercepted three vessels in the Mediterranean carrying around two hundred people, transferred them onto military ships, and returned them to Libya. In Hirsi Jamaa and Others v. Italy, the Grand Chamber of the ECHR in 2012 set out the point around which the entire field has turned ever since: a vessel on the high seas is subject to the exclusive jurisdiction of the flag state; therefore, the applicants were within Italy’s jurisdiction under Article 1 of the Convention.
The Court extended Article 4 of Protocol No. 4 on collective expulsions beyond national territory too—otherwise, a substantial part of modern migration routes would fall outside its scope. Its wording on the maritime context reads like a warning sent fourteen years ahead: the special nature of the sea cannot justify an area outside the law where people are not protected by any legal system.
There is a ruling in the opposite direction, and it is useful to keep it beside this one. In 2020, the same Grand Chamber, by 16 votes to 5, found no violation in the case of two people immediately returned to Morocco after trying to climb the Melilla fence. The argument was that the applicants could have used official border-crossing points but did not. In other words, the line between a violation and no violation runs through whether the person had an alternative. A hub project, where a person is taken by force, abolishes that alternative by design.
Then comes the European procedure that is now deciding everything. On 23 April 2026, Advocate General of the Court of Justice of the EU Nicholas Emiliou delivered an opinion in Case C-414/25, arising precisely from the Italian scheme: EU law does not prevent a member state from creating a detention centre for return procedures outside its territory. But the state remains bound by every safeguard: the right to a lawyer, language assistance, contact with family and authorities, full protection for minors and vulnerable people, access to healthcare and education, the right to a judge and prompt judicial review. A separate point states that an asylum seeker’s right to remain in the member state while their case is being examined does not entitle them to be returned to the territory of that state.
Then comes the place where sources diverge, and it should not be smoothed over. On 4 September, a German publication wrote that “the legal assessment by the CJEU in April concluded that the Italian centres in Albania do not violate European law.” In reality, it was a non-binding opinion from the Advocate General, and a conditional one at that: the centres are lawful provided people’s rights are fully respected. The Court itself had not yet delivered a judgment at the time of publication. Reuters made the qualification honestly that same day: the opinion does not bind the judges, though they usually follow it, and no date for the verdict has been announced.
The legal framework has finally appeared. In June 2026, the European Parliament approved the returns regulation: 418 votes for, 218 against, 30 abstentions. The document permits detention for up to 24 months, with a possible extension of up to six months; searches of premises, personal searches and the confiscation of electronic devices; electronic monitoring and bail as alternatives. Hubs in third countries are permitted under agreements concluded by the member state, not Brussels; the mechanism excludes unaccompanied minors, and before applying an agreement, a country must notify the Commission and the others.
And here, it becomes clear what exactly this construction has bought. Twenty-four months of detention is not a return procedure. It is the time it takes for the question “what do we do with this person?” to acquire an administrative answer. The right to notify Brussels about a hub is a way of removing any responsibility from the Commission for choosing the partner country. The obligation to take human rights into account in an agreement is a rule without a verification procedure: the text contains no criteria for who exactly decides that a third country respects human rights.
On 28 September, the Commission confirmed that it would consider hub projects once they were sufficiently developed, but did not say either what EU money it was prepared to put into construction, management, and operation, or on what legal basis that could be done. So the political decision has been made, the legal framework is ready, but nobody has yet been assigned to pay or answer for it. A letter from nineteen states, initiated by Denmark and Italy, called for “concrete results... based on solutions with third countries, as soon as possible.” In July, the Council of Europe Commissioner for Human Rights, Michael O’Flaherty, sent letters to the signatories warning of a “human rights black hole” and the risk of repeating the worst mistakes of the past.
At the Munich meeting on 26 September, progress, by observers’ accounts, was minimal: eighteen countries and the Commission, no Spain; the main outcomes were restated commitments and expanded powers for the border agency. Over a year, the list of potential partners had narrowed from twelve to two: Libya and Egypt were dropped, leaving Rwanda and Uganda, both described by German journalists with the same adjective—autocratically governed. A Rwandan spokesperson confirmed on 6 August that talks were under way, but that there was no agreement yet. The existing transit centre in Gashora, cited as infrastructure for the future hub, was built in 2019 for people stranded in Libya.
What seems worth keeping in mind here. The conversation began with a line about five countries and 2027, and ends with the Commission saying it will assess projects when they mature and talk about money later. Between those two points lie three countries, nine hundred and fifteen million in two currencies, one hundred and one people out of six hundred and eighty-five, and €3,820—the sum at which the state officially values one deportation. Moving a person to another jurisdiction has not become cheaper anywhere—it has become possible to stretch out the bill and call it policy.