Letterbox Companies: How the Dutch Tax Treaty Network Became a Global Controversy, Then Reformed

The Dutch letterbox company was not an accident of lax regulation; it was a deliberate product, designed in the 2000s, marketed by trust offices and defended by successive governments until the political cost exceeded the tax revenue. At its peak the system allowed multinationals from Google to Nike to route profits through Dutch special financial institutions that had no employees, no offices beyond a brass plate on the Herengracht and no purpose beyond exploiting the gap between the Dutch treaty network and the tax codes of Bermuda, Ireland and Luxembourg. The Tax Justice Network still ranked the Netherlands seventh in its 2024 Corporate Tax Haven Index. Yet between 2018 and 2024 the Netherlands introduced a set of reforms that measurably redirected flows, closed the routes used in the most famous cases and changed what it means to incorporate in Amsterdam. This article traces the structure, the scandals and the reforms, and asks what a founder or CFO should expect from the Dutch system in 2026. It is part of the Netherlands Company Stories hub.

How did the Netherlands become a conduit country?

The Netherlands became a conduit country because it combined three features that few other jurisdictions offered together: no withholding tax on outbound interest and royalties, a participation exemption that made dividends from subsidiaries tax-free, and a tax treaty network of more than a hundred agreements that reduced foreign withholding taxes on money flowing in. A multinational could therefore route a royalty from a German subsidiary to a Dutch BV at a reduced or zero German rate under the treaty, and then pay it onward from the Dutch BV to a Bermuda affiliate with no Dutch tax at all.

This was not a historical accident. Under State Secretary Joop Wijn between 2003 and 2005, the Netherlands liberalised its rules after lobbying by US tax advisers, and in July 2005 a provision designed to curb tax avoidance by American companies was abolished. The ruling practice, under which the tax authority gave advance certainty about transfer pricing and the tax treatment of a structure, made the Netherlands predictable in a way that Luxembourg, Ireland and Switzerland competed to match. A 2017 analysis published in Nature ranked the Netherlands as the largest of five global “conduit offshore financial centres”, ahead of the United Kingdom, Ireland, Singapore and Switzerland, together routing almost half of global corporate tax avoidance flows into 24 “sink” jurisdictions.

The scale was enormous relative to the real economy. Economist Ewald Engelen estimated in 2013 that roughly €12 trillion passed through Dutch special financial institutions each year, earning the state about €1.5 billion in tax. DNB, the central bank, tracked these “bijzondere financiële instellingen” separately from the real economy because their balance sheets, running into trillions of euros, would otherwise have distorted every national statistic. For the Netherlands the arrangement generated revenue, legal and accounting employment and a reputation among multinationals; for the countries whose tax base leaked through Amsterdam, it generated resentment.

What exactly did trust offices do?

Trust offices, or trustkantoren, are licensed companies that provide directors, a registered address, bookkeeping and compliance to legal entities that have no staff of their own, and they were the operational layer of the letterbox industry. A multinational wanting a Dutch conduit did not rent an office or hire a finance team; it engaged a trust office such as Intertrust or TMF, which supplied a Dutch-resident director, filed the accounts and handled the paperwork for thousands of such entities from a single building. The directors were employees of the trust office, often serving on the boards of hundreds of companies simultaneously.

That model is what gave the Netherlands the term “letterbox company”: a mailbox on an Amsterdam canal, a trust-office director and a set of intercompany agreements. The companies were real in law and empty in substance. The Dutch argument for decades was that substance was irrelevant because the treaties and the participation exemption applied to legal residence, not to employment.

Supervision has tightened in stages. The Act on the Supervision of Trust Offices was overhauled in 2018 (Wtt 2018), raising licensing, integrity and client due-diligence requirements and putting DNB in charge of enforcement. DNB maintains a public register and withdraws licences; it continues to impose penalties, for example a fine on International Trust Services BV on 24 November 2025 for failing to commission a required audit. The industry consolidated and shrank, and the Dutch arms of the largest trust groups have moved toward corporate services for companies with real operations rather than pure conduit work.

What did the Starbucks and Nike cases reveal?

The Starbucks and Nike cases revealed that the Dutch advance ruling practice could be challenged as illegal state aid, and even where the Netherlands ultimately won in court, the publicity changed the politics. The European Commission opened an in-depth investigation into a Dutch ruling for Starbucks Manufacturing EMEA in June 2014, alongside the Apple-Ireland and Fiat-Luxembourg cases. On 21 October 2015 it ruled that the Netherlands had granted Starbucks a selective advantage through a transfer pricing agreement that understated the taxable profit of its Amsterdam roasting operation, and ordered recovery of up to €30 million. The decision also exposed the Dutch BV-CV structure, in which a Dutch limited partnership was treated as transparent in the Netherlands and as a corporation in the United States, leaving income taxed nowhere.

The Netherlands and Starbucks appealed, and in September 2019 the EU General Court annulled the Commission’s decision, finding that it had not demonstrated that the ruling conferred an advantage. The legal victory did not restore the reputation. In January 2019 the Commission opened a separate investigation into Dutch rulings for Nike European Operations Netherlands and Converse Netherlands, concerning royalties paid for the use of Nike’s trademarks to entities that were not taxed in the Netherlands. The Paradise Papers had already shown how Nike moved ownership of the Swoosh trademark to a Bermudan subsidiary that charged royalties to the Hilversum European headquarters, with US tax court filings citing royalties of US$3.86 billion for 2010 to 2012, and how a later reorganisation shifted the intellectual property to Nike Innovate CV, a Dutch limited partnership whose non-resident partners owed no Dutch tax.

The third case was IKEA. In December 2017 the Commission opened an investigation into rulings for Inter IKEA Systems, the Dutch entity that collected franchise fees from IKEA stores worldwide. The three cases together covered coffee, shoes and furniture and three different mechanisms, transfer pricing, royalties and franchise fees, but one common feature: a Dutch entity with a ruling that allowed profits to pass through at a very low effective rate. The headquarters exodus story describes how, in the same years, the dividend tax that supposedly protected the Dutch base was driving Dutch multinationals abroad.

How did the Google “Dutch Sandwich” work, and why did it end?

The Dutch Sandwich was a structure in which an Irish-registered company paid royalties to a Dutch BV, which paid them on to a second Irish company tax-resident in Bermuda, so that profits earned across Europe ended in a jurisdiction with no corporate tax having passed through Ireland at 12.5 per cent on a thin margin and through the Netherlands at zero. The Dutch layer existed because Irish law charged withholding tax on royalties paid directly to Bermuda but not on royalties paid to an EU member state, and Dutch law charged no withholding on royalties paid to Bermuda. Google was the best-known user; the Dutch entity, Google Netherlands Holdings BV, was reported to route well over €15 billion a year at its peak, and Dutch shell companies using the structure accounted for roughly 10 per cent of all income reported by such companies in the Netherlands.

The structure ended for two reasons. Ireland closed the “Double Irish” to new entrants in 2015 and to existing users by the end of 2020, after EU and US pressure. Google announced that it would stop using the Bermuda route from 2019 and 2020, moving its intellectual property to the United States following the 2017 US tax reform. NOS and the Dutch finance ministry attribute much of the fall in Dutch outflows to tax havens between 2019 and 2020 to this single decision rather than to Dutch policy.

The Dutch response was the conditional withholding tax, announced for 2021. It targets exactly the Google scenario: interest and royalties paid from a Dutch entity to an affiliate in a jurisdiction on the Dutch list of low-tax countries, or in abusive situations, are taxed at the Dutch corporate rate of 25.8 per cent. Multinationals moved about €22 billion of royalties and interest through the Netherlands in 2016 to avoid tax, according to estimates cited in Dutch parliamentary debate, and the withholding tax was designed to make that flow uneconomic.

What reforms were introduced between 2018 and 2024?

The reform programme ran under State Secretary Menno Snel from 2017 and continued under his successors, including Marnix van Rij, and it combined Dutch initiatives with EU directives and OECD agreements. The main elements, in order of effect, were:

The conditional withholding tax on interest and royalties, in force from 1 January 2021, at the top corporate rate (25.8 per cent from 2022), applying to payments to affiliated entities in jurisdictions with a statutory rate below 9 per cent or on the EU list of non-cooperative jurisdictions, and to abusive arrangements. The extension of that tax to dividends from 1 January 2024, closing the route by which dividends could be paid to low-tax affiliates free of the ordinary 15 per cent dividend tax. The EU anti-tax-avoidance directives, implemented from 2019 and 2020, introducing interest limitation, controlled foreign company rules and anti-hybrid rules, the last of which neutralised the CV-BV structure that Starbucks and Nike had used. Stricter substance requirements for rulings from 2019, under which a Dutch entity seeking advance certainty must have real economic activity in the Netherlands, which removed the trust-office model from the ruling practice. The Wtt 2018 trust-office regime, and later a prohibition on trust offices serving clients from Russia after 2022. The OECD Pillar Two global minimum tax of 15 per cent, implemented in the Netherlands from 2024, which limits the benefit of routing profits to any jurisdiction with an effective rate below 15 per cent.

The measurable result, reported by Van Rij to parliament in 2023, was that outflows from the Netherlands to listed low-tax jurisdictions fell from €38.5 billion in 2019 to about €6 billion in 2020 and stayed there in 2022. The government presents this as proof the withholding tax works; critics note that the fall was already complete in 2020, a year before the tax took effect, and coincided with Google’s exit.

Where did the money go after the reforms?

The money did not disappear; a large part of it was rerouted to jurisdictions that are not on the Dutch low-tax list, above all the United Kingdom. Ministry of Finance and DNB data show that Dutch companies typically sent €20 billion to €25 billion a year to UK counterparts before 2022, and that in 2022 this roughly tripled, with about €50 billion of additional transfers from some 180 Dutch entities, while about €17 billion flowed the other way. The ministry said it did not know the reason. Leiden University tax professor Jan van de Streek suggested that the UK, which does not tax dividends paid to subsidiaries in no-tax countries, may be serving as an intermediary for profits ultimately bound for Bermuda, and that some groups may have rerouted cash to avoid the Dutch withholding tax. Shell’s 2021 move of its headquarters to London also contributed.

The Tax Justice Network’s figures tell the same story from the outside. Its 2023 analysis attributed 17 per cent of US$301 billion of annual global corporate tax avoidance to the Netherlands, up from 6.3 per cent two years earlier, though the Dutch affiliate’s coordinator Arnold Merkies cautioned that the rise reflected better data rather than more avoidance. The network’s November 2024 State of Tax Justice report found that the Netherlands, UK, Luxembourg and Switzerland together accounted for a third of global corporate tax losses, with multinationals channelling profits into the Netherlands at effective rates below 10 per cent. For every dollar of tax the Netherlands collects from such structures, the report estimated, other countries lose more than five.

Domestic avoidance has also adapted. NRC reported in April 2024 that Heineken heiress Charlene de Carvalho-Heineken, whose dividends had flowed through a Luxembourg letterbox to Jersey at a 2.5 per cent rate until the 2024 dividend withholding rules closed that route, had shifted to repayments of share capital, which are not subject to Dutch withholding tax. The same month NRC showed that chemicals group OCI had returned about €5 billion to shareholders through capital reductions, avoiding roughly €750 million of dividend tax, a route also used by AkzoNobel, TomTom and Ahold. The finance ministry said it was investigating. Each closed route produces a new one, and the Heineken family story shows how the largest Dutch fortunes have managed that evolution.

Is the Netherlands still a tax haven?

By the measures used by its critics, yes, but a smaller and more conditional one than in 2015. The Tax Justice Network ranked the Netherlands seventh in its 2024 Corporate Tax Haven Index, behind the British Virgin Islands, Cayman, Bermuda, Switzerland, Singapore and Hong Kong, and Oxfam Novib continues to campaign for a UN tax convention on the grounds that the Dutch system benefits a small group of multinationals at the expense of poorer countries. By the measures used by the government, the conduit function has been largely dismantled: outflows to listed low-tax jurisdictions are down more than 80 per cent and the rulings practice now requires real activity.

The honest answer lies in the definition. The Netherlands never had a low headline corporate tax rate; at 25.8 per cent it is above the EU average. Its role was as a conduit, not a sink, and the reforms have specifically targeted conduit behaviour toward jurisdictions the Netherlands itself defines as low-tax. What the reforms have not touched is the attractiveness of the Dutch holding company for groups with real operations, nor the flows to jurisdictions such as the UK, Ireland or Luxembourg that are not on the list. The participation exemption remains, which is why Prosus, Stellantis and Exor remain Dutch NVs and why the Prosus story is one of inbound domicile rather than outbound avoidance.

A second factor is the global minimum tax. With Pillar Two in force from 2024 in the Netherlands and across the EU, the marginal benefit of routing profits to any jurisdiction below 15 per cent has shrunk for large groups, which reduces the value of the Dutch conduit function independently of Dutch law. The remaining arbitrage is in the gap between 15 per cent and the rates of high-tax countries such as Germany and France, which is where the UK flows may be heading.

What can founders and investors learn from the letterbox era?

The first lesson is that a tax product built on treaty arbitrage has a political half-life. The Dutch conduit model generated perhaps €1.5 billion a year in revenue for the state and a few thousand jobs in trust offices, law firms and accountancies, against reputational damage that the Starbucks, Nike and Google headlines made impossible to ignore. Once the Netherlands was named alongside Bermuda in the Paradise Papers, defending the model cost more in European and domestic politics than it earned.

The second lesson concerns the difference between law and substance. For two decades Dutch policy held that legal residence was enough. The EU directives, the OECD and finally the Dutch legislature concluded that substance was what mattered. Founders structuring a group today should expect every jurisdiction to ask where people work and decisions are made, and should design the structure around that reality rather than around the treaty map.

The third lesson is that reforms redirect flows rather than stop them. The €38.5 billion that left for tax havens in 2019 became €6 billion, but the €50 billion extra that went to the UK in 2022 shows the water finding a new channel. Investors in Dutch-domiciled groups should read the cash flow statement for where distributions go and under which exemption, as the OCI and Heineken examples illustrate.

Finally, the Netherlands remains a good place for a holding company with real operations, which is a different thing from being a good place for a letterbox. The Dutch chip cluster around ASML and NXP, the cooperatives in agriculture and the Amsterdam financial sector described in the ING story all use Dutch structures for ordinary commercial reasons. The reforms have made the distinction between those businesses and the conduit entities clearer, which is probably the most durable outcome of the whole episode. Readers comparing conduit jurisdictions will find the British equivalent discussed in the UK Company Stories hub.

Frequently Asked Questions