Concept
A holding structure is a company that owns interests in other companies and assets, collects the dividends and capital gains they generate, and manages the group. A well-built holding reduces tax on cross-border dividends, makes a business easier to sell, and helps protect assets — but this only works when there is real substance behind it.
Legally, a holding company creates one specific capability: it converts many separate assets — operating companies, real-estate SPVs, portfolios — into a single entry in one share register, governed by one company law, that can be financed, pledged, inherited or sold as a single object. One example shows the whole mechanism. An investor owns an operating company in Germany and another in Poland. Held directly, every event happens twice: two shareholder agreements, two pledges if a bank lends against the business, two sale processes at exit, two national inheritance procedures. Interposing a holding company — say a Dutch BV or a Cyprus Ltd — collapses this: the investor holds shares in one company, the bank takes one pledge over those shares, a buyer signs one share purchase agreement, and the dividends of both subsidiaries arrive in one place before anything is paid out to the owner.
The price of that convenience is threefold. First, every added tier crosses at least one extra border, and each border applies its own tax tests — a two-tier group answers not one question but eight, examined by different authorities under different facts. Second, the holding must actually be run: a board that meets and decides, records, accounts, its own bank account — a company that exists only on paper fails most of the tests below. Third, the tier is easy to build and expensive to unwind: a redundant holding at exit means either an extra taxable liquidation or a migration with an exit charge. The rest of this guide follows exactly that order: what the structure separates, how money moves through it, which independent tests apply to each link, and what the same flows cost under two contrasting regimes.
Why Ownership, Financing and Management Are Separated
A holding structure deliberately splits three functions that in a directly-held business sit in one person, because the law tests each of them separately.
Ownership is the register question: who holds the shares of what. It determines who receives dividends and sale proceeds, whose death or divorce touches the asset, and at which level a shareholders' agreement or a pledge attaches. Ownership of holdco shares is a different asset from ownership of opco shares — a creditor of the investor reaches the former, not directly the latter.
Financing is the balance-sheet question: what the investor put in and in what legal form. Equity creates shares and a right to dividends when declared; a shareholder loan creates a creditor's claim to interest and repayment on contractual dates. The two feed different tax articles at every border they cross — each state decides for itself whether the instrument is debt or equity — which is why the debt-versus-equity choice is examined below as its own scenario.
Management is the decision question: where and by whom the company's key decisions are genuinely taken. It barely appears in the corporate documents, yet it drives the tests with the largest consequences — corporate tax residence and permanent establishment — and colours beneficial ownership and treaty entitlement.
Structures rarely fail because one function was put in the "wrong" country. They fail when the three drift apart unnoticed: shares held from one country, money advanced from a second, decisions actually taken from a third — and each of the tests below then reads its own facts off a different function.
The Operating Path: Investor → HoldCo → OpCo
The minimal working structure has three levels and two links: the investor holds the holding company; the holding company holds the operating company or asset. Before comparing jurisdictions it pays to see what moves along each link, because every instrument on this path has its own tax treatment — the table maps them. The diagram shows the base construction from the example above: the investor holds the holdco, the holdco holds both operating companies, and dividends and interest travel back up the same two links.
| Link | Instruments going down | Flows coming up | What is tested at this link |
|---|---|---|---|
| Investor → HoldCo | Equity subscription; shareholder loan | Dividends; interest; return of capital; liquidation proceeds; price for holdco shares | Withholding tax of the holdco's state on the way out; the investor's personal income tax; the investor's home CFC rules on holdco's undistributed profit |
| HoldCo → OpCo | Equity participation; intercompany loan | Dividends; interest; repayment of principal | Source-state withholding tax; beneficial ownership and treaty/directive entitlement of the holdco; transfer pricing on the loan; participation exemption at holdco level |
| HoldCo as seller | — | Capital gain on disposal of opco shares | Where the gain is taxable; participation exemption on gains; land-rich clauses of the opco's state |
Read along the equity path first. OpCo earns profit and pays corporate tax where it operates — nothing in the holding structure changes that first layer. What remains is distributed as a dividend: the source state may withhold tax unless a directive or treaty reduces it, and grants that reduction only to a holdco that passes its tests. Arriving at the holdco, the dividend is typically exempt under a participation exemption, so it accumulates untaxed and can be reinvested into other subsidiaries — this deferral at the middle tier is the core fiscal function of a holding. Only when the holdco pays its own dividend does the last layer apply: the holdco state's withholding tax plus the investor's personal tax at home.
The debt path runs differently at every step. Interest on a shareholder or intercompany loan is deductible for the payer, so it comes out of pre-tax profit — but deductibility is capped by interest-limitation rules (in the EU, Article 4 of the Anti-Tax Avoidance Directive 2016/1164 caps exceeding borrowing costs at 30% of tax EBITDA, with a safe harbour of up to €3,000,000 that member states may grant; Germany's Zinsschranke uses the same 30% and €3,000,000 figures), the rate must survive transfer pricing, and interest often carries its own withholding tax under a different treaty article than dividends. Crucially, the participation exemption does not cover interest: what arrives at the holdco as interest is taxable income there. Equity money is taxed late but at more links; debt money moves earlier and cheaper at the opco level but engages transfer pricing, interest limitation and the holdco's own tax base.
Where They Came From
Holdings grew out of European practice. In 1990 the Parent-Subsidiary Directive (Directive 90/435/EEC, recast as Directive 2011/96/EU: Article 5 exempts a subsidiary's distributions from withholding tax where the parent holds at least the 10% that Article 3 requires) brought withholding tax on dividends between associated companies inside the EU to zero, and a whole industry formed around it: the Dutch BV and the Luxembourg SOPARFI became the standard intermediate links between an operating business and its owners. The principle that still governs today took hold at the same time: the relief goes to the company that is the beneficial owner of the income. On 26 February 2019 the Court of Justice of the EU, in the "Danish cases" (T Danmark and Y Denmark, joined cases C-116/16 and C-117/16, on dividends; N Luxembourg 1 and others, joined cases C-115/16, C-118/16, C-119/16 and C-299/16, on interest), denied directive relief to conduit holdings with no economic life of their own and set the benchmark that every European structure now works to.
Eight Independent Tests
The most common design error is to treat "substance" as one exam that the structure either passes or fails. In reality a cross-border holding is examined by eight separate rules. Each has its own examining state, its own decisive facts and its own sanction — and passing one implies nothing about another. The matrix gives the map; the subsections give the mechanism of each test.
| Test | Question it answers | Whose law examines it | What failure costs |
|---|---|---|---|
| 1. Incorporation | Under which company law does the entity exist | State of registration | No tax cost by itself — but it fixes corporate procedure, creditor rights and what can be redomiciled |
| 2. Corporate tax residence | Which state taxes the company's worldwide profit | Every state where it is incorporated or managed | Full taxation of the holdco's profit in an unplanned state; dual residence resolved only through the treaty procedure |
| 3. Permanent establishment | Whether part of the profit is taxable where activity physically happens | The state where people act for the company | Profit attribution, local filing and tax in the state of activity — residence elsewhere does not block it |
| 4. Beneficial ownership | Whether the holdco owns the income or merely passes it on | The source state applying a directive or treaty | Withholding tax at the full domestic source rate |
| 5. Treaty entitlement / PPT | Whether a treaty benefit may be claimed at all | The source state under the treaty's anti-abuse clause | Benefit denied even where beneficial ownership is conceded |
| 6. CFC | Whether the investor's home state taxes the holdco's undistributed profit | The investor's state of residence | Current taxation of the holdco's profit at the investor's own rate, before any distribution |
| 7. Transfer pricing | Whether intra-group prices match what independent parties would agree | Every state a related-party transaction touches | Income adjustment, penalties and — absent a treaty fix — the same profit taxed twice |
| 8. Local substance | Whether the holdco meets its own state's economic-presence rules | The holdco's own jurisdiction | Local penalties, spontaneous exchange of information with the investor's state, loss of local regimes |
Eight examiners, eight sets of facts: the sections that follow show what each one actually reads.
1. Incorporation
Registration answers only one question: which company law governs the entity — its organs, capital rules, insolvency ranking, and whether it can migrate. It is established by a registry act and cannot "fail," but it is routinely mistaken for a tax fact. It is not one: a company incorporated in Cyprus can be a German tax resident, and vice versa. Incorporation matters at exit — some company laws permit outbound redomiciliation with continuity of legal personality, others force a liquidation.
2. Corporate tax residence
Residence decides which state taxes the holdco's worldwide profit, and it is claimed by facts, not filings: most systems treat a company as resident where its central management and control or effective management is located, alongside (or instead of) the place of incorporation — Cyprus, for instance, has historically used management and control, and its 2026 reform added incorporation as a default basis. A holdco incorporated in one state but steered from another is potentially resident in both; modern treaties resolve that dual residence not by an automatic rule but by mutual agreement between the two tax authorities — until they agree, most benefits are suspended. The decisive facts are board composition, where meetings with real deliberation happen, and who prepares and approves key decisions.
Three English decisions draw the line the test actually applies, and they are cited far beyond the UK because every management-and-control system reasons the same way. In Wood v Holden [2006] EWCA Civ 26 a Dutch company run by a corporate director executed a scheme designed by advisers in England; the Court of Appeal held it Dutch-resident, because its board — however briefly it deliberated, and on whatever advice — did take the decision: outsiders influencing a board is not the same as usurping it. In Laerstate BV [2009] UKFTT 209 (TC) the opposite facts produced the opposite result: the UK-resident owner made every decision himself, first as director and then after resigning, while the Dutch director signed what he was sent, so the company was UK-resident throughout. HMRC v Development Securities [2020] EWCA Civ 1705 closed the gap between the two: Jersey boards that met in Jersey and satisfied themselves of the legality of transactions the UK parent had already decided on were held not to be deciding at all, and the companies were UK-resident. The rule that survives is the one the two-managements scenario below applies: residence sits where the board genuinely decides, even when it decides little; it moves to wherever a shareholder or parent decides for the board.
3. Permanent establishment
Even a holdco that keeps its residence can leave a taxable trace elsewhere: a fixed place of business or a dependent agent habitually acting for it gives the state of activity the right to tax the profit attributable to that presence. For holdings the classic trigger is an owner or executive who negotiates, signs or manages from another country — often their home. The consequence is partial: not worldwide taxation, but attribution, registration and filing in a second state. The facts examined — premises, habitual conduct, authority to conclude — are entirely different from the residence facts, and the two tests can fail independently.
4. Beneficial ownership
The source state reducing its withholding tax asks whether the holdco is the real owner of the dividend or interest — or a conduit that receives and passes on. Since the CJEU's 2019 Danish cases, EU source states must deny directive and treaty relief to abusive conduits even without a domestic rule; the indicia the Court listed are factual: income passed on shortly after receipt, negligible taxable margin, no genuine economic activity, contractual or de facto obligations to forward what is received. An office and staff help but do not answer the question if the money demonstrably flows through — see how nominee and beneficial ownership diverge for the ownership side of the same distinction.
Those indicia were a deliberate step beyond the older, narrower standard. In Prévost Car (Federal Court of Appeal of Canada, 2009 FCA 57) a Dutch holdco with no employees, owned by Swedish and British shareholders who had agreed between themselves to distribute at least 80% of profits, was still the beneficial owner of Canadian dividends, because nothing legally obliged the holdco itself to pass them on — on that reading a conduit is a company with no discretion over the funds. A source state applying the Danish-cases standard reads the same facts differently: the obligation may be de facto, and "passed on shortly after receipt" is evidence of it. Which standard a source state applies is a question for its own courts; the two lines are traced on the conduit page.
5. Treaty entitlement and the principal purpose test
Beneficial ownership is not the end of the source state's inquiry. Under the BEPS multilateral instrument, most treaties now carry a principal purpose test: the benefit is denied if obtaining it was one of the principal purposes of the arrangement, unless granting it accords with the object and purpose of the treaty. This is a purpose inquiry, not a presence inquiry — a holdco with premises and payroll still fails if the structure's documented rationale reduces to the treaty rate. Inside the EU the directive carries its own version: since Directive (EU) 2015/121 inserted a general anti-abuse rule into the Parent-Subsidiary Directive, relief is refused to arrangements that are not genuine and were put in place for the main purpose, or one of the main purposes, of obtaining a tax advantage that defeats the directive's object.
The test exists because general anti-avoidance rules did not reliably reach treaty shopping. In Canada v Alta Energy Luxembourg (2021 SCC 49) a Luxembourg holdco with a thin connection to Luxembourg sold shares in a Canadian oil-and-gas company and claimed the Canada–Luxembourg treaty's capital-gains exemption; a 6–3 majority of the Supreme Court of Canada refused to apply Canada's GAAR, holding that the treaty's text made residence the only condition and that an unwritten economic-connection requirement could not be read into it. The majority itself pointed to the MLI's principal purpose test as the instrument written for such facts — which is why a holdco that would have survived on a pre-MLI treaty must now answer the purpose question directly. The mechanics, the burden of proof and the EU GAAR counterpart are dissected in GAAR and the principal purpose test.
6. CFC — the shareholder's own tax law
While the source state examines the holdco, the investor's home state examines the investor. CFC rules attribute a controlled foreign company's undistributed profit to the controlling resident: under the EU's ATAD template the trigger is control above 50% plus foreign tax below half of the domestic equivalent, with an escape for companies carrying on substantive economic activity with staff, equipment, assets and premises. But that carve-out is an EU feature, not a universal one — other regimes tax passive holdco income mechanically, whatever the office looks like: the Russian KIK regime, for instance, has no staff-and-premises escape, and its exemptions turn on the share of passive income and the level of foreign tax instead. Which country's rules bite, and how retained earnings inside a holdco become taxable at home, is the subject of the CFC master guide and, for the EU template, ATAD CFC rules.
7. Transfer pricing
Every priced transaction between holdco and its group — the shareholder loan's interest rate, management fees, guarantees for opco's bank debt — must match what independent parties would have agreed. The test is indifferent to substance and to purpose: a fully staffed holdco charging its subsidiary an off-market rate is adjusted exactly like an empty one. Failure means an income adjustment in one state with penalties, and economic double taxation unless the counterpart state agrees to a corresponding adjustment. In holding structures the loan is the usual battlefield: rate, currency, subordination and whether a third-party lender would have advanced the amount at all.
8. Local substance
Separately from everything above, the holdco's own jurisdiction may impose economic-presence requirements — the model spread through no-tax and low-tax jurisdictions after 2019: adequate employees, premises and expenditure for the income-generating activity, with penalties, striking-off and spontaneous exchange of information to the parent's tax authority for failure. Passing it earns exactly one thing: compliance with local law. It is evidence, not a verdict, in the other seven tests — what the regimes require and where, in economic substance requirements.
What Makes a Jurisdiction Suitable for Holdings
Four things make a holding jurisdiction: a participation exemption (dividends and gains on shareholdings freed from tax), a wide treaty network, low or zero WHT on outbound payments, and a reputation that banks and counterparties trust. The classic set is the Netherlands, Luxembourg, Cyprus, Singapore and the UAE.
Where Each Jurisdiction Fits
There is no universal holding: the choice depends on where the operating business runs, where the dividends go and to whom, and the profile of the ultimate owners. Europe leans toward the Netherlands, Luxembourg and Cyprus; for Asia, Singapore and Hong Kong are the natural fit; the UAE covers the Middle East and part of Africa. What follows is a closer look at the five classic jurisdictions.
The Netherlands and Luxembourg
Both countries are the load-bearing structure of European holdings. The Netherlands runs the deelnemingsvrijstelling: from a 5% shareholding, dividends and capital gains on subsidiaries are fully exempt, and the corporate rate that never touches them is 19% on the first €200,000 of profit and 25.8% above it (the 2023–2026 brackets); the exemption is denied only to low-taxed stakes held as passive portfolio investments — the mechanics are on the Dutch BV holding page. The Luxembourg SOPARFI (see Luxembourg and the SOPARFI page) works under the same directive with sharper thresholds: dividends are exempt from a 10% stake or a €1,200,000 acquisition price, gains from 10% or €6,000,000, each after twelve months' holding, and only where the subsidiary is fully taxable at a rate comparable to Luxembourg's — at least 8% from tax year 2025; Luxembourg's own 15% dividend withholding falls to zero for a qualifying EU or treaty-state parent on the same 10%/€1,200,000/twelve-month footing, and interest paid abroad carries no withholding at all. Fund infrastructure (RAIF, SIF) sits alongside. Withholding tax on dividends inside the EU is usually zero; at the same time the Netherlands introduced a conditional WHT on interest and royalties paid to low-tax jurisdictions and, from 2024, extended it to dividends, closing the transit route to offshore.
Cyprus
The cheapest entry into the EU — the route in detail is on the Cyprus holding page. On 1 January 2026 Cyprus raised its corporate rate from 12.5% to 15% — the price of complying with Pillar Two — but almost everything else stayed in place: zero WHT on outbound dividends, an IP box with an effective rate of around 3%, and the non-dom regime for individuals. The SDC on dividends for domiciled residents was cut from 17% to 5% from 2026, so for many holdings the reform came out close to neutral.
Singapore
The gateway to Asia. Singapore's headline rate is 17%, but its territorial logic and a foreign-sourced income exemption push the effective burden lower: under section 13(8)–(9) of the Income Tax Act a foreign dividend remitted by a resident company is exempt if it was subject to tax abroad, the paying jurisdiction's highest corporate rate is at least 15% when the dividend is received, and the Comptroller is satisfied the exemption benefits the recipient — so a dividend from a zero-tax subsidiary is taxed on remittance. Dividends the Singapore company pays out carry no withholding under the one-tier system. There is no capital gains tax, and section 13W makes that certain for a holdco selling at least 20% of an investee's ordinary shares held for at least 24 months — a safe harbour Budget 2025 made permanent and extended to equity-accounted preference shares from 2026, unlisted property-holding targets excluded. The catch for holdings arrived on 1 January 2024 as section 10L: an entity of a group with members outside Singapore that sells a foreign asset and brings the gain into Singapore is taxed on it unless it had adequate economic substance in Singapore — for a pure equity-holding entity, filings up to date, operations managed and performed in Singapore and adequate people and premises; for any other entity, a fuller test of employees, expenditure and where key decisions are taken. Among the five jurisdictions this is the one place where substance directly switches a gain between exempt and taxable, and a network of more than 90 tax treaties is worth exactly as much as the board, staff and decisions that sit in Singapore — operating details on the Singapore company page.
The UAE
The Middle East hub. There is no personal income tax; corporate tax is 9% on profit above AED 375,000. A UAE holdco's dividends and gains from a participation are exempt under Article 23 of the Corporate Tax Law where it holds at least 5% (or an acquisition cost above AED 4,000,000) for twelve months, the participation is subject to tax at a statutory or effective rate of at least 9%, and no more than half of the participation's assets are themselves non-qualifying stakes; outbound dividends, interest and royalties face a withholding tax that exists in the law but is set at 0%. Free zone companies keep 0% on qualifying income, but the definition of that income has been tightened, and large multinational groups (global revenue of €750 million or more in two of the four preceding years) pay a Domestic Minimum Top-up Tax at 15% for financial years beginning on or after 1 January 2025. What a UAE holdco over an EU operating company actually collects — after the EU source state's tests, not the UAE's — is worked through in UAE holding over an EU operating company. The UAE is quickly building up its substance and economic-presence requirements, so a "paper" structure works less and less well here.
Pillar Two
Global groups with revenue of €750 million or more have been under Pillar Two since 2024: an effective rate of no less than 15% in every jurisdiction. In the EU, 22 of the 27 member states have adopted the rule, and the UAE launched its own DMTT in 2025. In January 2026 the OECD's "side-by-side" compromise took effect — U.S. groups are carved out of the IIR and UTPR rules in exchange for the United States' own minimum system. For holdings the conclusion is a single one: a bare offshore rate no longer confers an advantage, and value has shifted to real, economically substantiated structures. How the income inclusion rule, the undertaxed profits rule and domestic top-up taxes are sequenced is in Pillar Two.
Two Regimes End to End: Netherlands Against Cyprus
A holding jurisdiction cannot be judged by its own rate: the economics are set by the full chain from the state where profit arises to the person who finally receives it. To show this, the same flow is traced through two contrasting regimes on explicit assumptions. Assume: a German operating company, wholly owned for more than a year; a holdco genuinely managed in its own jurisdiction that passes the German beneficial-ownership and anti-abuse tests; an ultimate investor who is an individual, resident in a state that has no tax treaty with the Netherlands or Cyprus, is not on the EU list of non-cooperative jurisdictions, and imposes no personal income tax or CFC charge of its own; a distributed profit of €1,000,000 — a fictional teaching figure; rules as they stand in 2026. Change any assumption and the arithmetic changes with it. The table walks the money through both routes.
| Stage of the flow | Dutch holdco route | Cyprus holdco route | Rule that decides |
|---|---|---|---|
| Germany: dividend leaves OpCo | 0% withholding | 0% withholding | Parent-Subsidiary Directive as implemented in German law: ≥10% held for a year, and the holdco passes Germany's beneficial-ownership / anti-abuse screening; otherwise Germany's domestic 26.375% applies |
| Holdco: dividend received | Exempt — participation exemption from a 5% stake | Exempt — foreign dividends are excluded unless the payer is mostly an investment vehicle taxed at a significantly lower rate | Each state's own participation regime; the holdco's headline CIT rate (25.8% top rate in NL, 15% in Cyprus) never touches the exempt dividend |
| Holdco: dividend paid to the investor | 15% dividend withholding tax — €150,000 | 0% — Cyprus levies no withholding on dividends to non-residents (exceptions: recipients in EU-listed jurisdictions, 17% since April 2025; related companies in low-tax jurisdictions, 5% from 2026) | The holdco state's outbound WHT law; for the Dutch route, the conditional 25.8% rate would replace 15% only if the recipient were a related company in a listed or sub-9% jurisdiction |
| Investor: net received | €850,000 | €1,000,000 | The investor's own residence state — here assumed to tax nothing |
The 15-point difference arises entirely at the last link — the exit withholding — while everything the brochures advertise about the holdco's own tax rate proved irrelevant to the flow. And the comparison cuts both ways: if the holdco failed Germany's beneficial-ownership or purpose screening, both routes would lose €263,750 at the first link, because the source state's test does not care which holdco was interposed. Reading a structure means reading every link, from source to final recipient. One more assumption, moved once: make the investor a company resident in another EU state holding at least 5% of the Dutch holdco, and the Dutch 15% disappears as well — the Netherlands exempts a qualifying EU parent from dividend withholding — so both routes end level at €1,000,000 and the difference must be sought elsewhere: in the parent's own tax on the receipt, and in what each holdco costs to unwind.
Scenarios: One Fact Changes the Result
Each scenario below changes a single fact against the base structure and follows which of the eight tests reacts — this is the fastest way to see their independence.
The same legal shell, managed from two places
Two identical Cyprus companies hold the same kind of German subsidiary. In the first, the board meets in Limassol, prepares decisions with local counsel, and the owner attends as one voice among several. In the second, the same documents exist, but every decision is made by the owner from Berlin and the local directors sign what they receive. The incorporation test gives the same answer for both. The residence test does not: the second company is managed in Germany, becomes a German tax resident by fact, and its worldwide profit falls into German tax; the treaty resolves the resulting dual residence only through mutual agreement between the two authorities, with benefits suspended in the meantime. Its beneficial-ownership position at the German border collapses for the same reason. Identical shells — opposite outcomes, driven by one fact that appears in no corporate document.
Shareholder debt against equity
The investor wants €1,000,000 a year out of the operating level and can route it as dividend on equity or as interest on a shareholder loan through the holdco. The dividend is paid from taxed profit, crosses the border under the dividend article, is exempt at the holdco and waits for the exit withholding. The interest is deductible at the opco — it comes out of pre-tax profit — but four tests wake up that the dividend never met: transfer pricing examines the rate and whether an independent lender would have advanced the principal; interest-limitation rules cap the deduction by reference to EBITDA; the source state applies its interest article, and conditional withholding regimes (the Dutch 25.8% model) watch where the interest goes next; and at the holdco the interest is taxable income, because participation exemptions cover dividends and gains, not interest. Debt is not "cheaper" — it is earlier and differently examined.
Run once with numbers on the Dutch route, on the assumptions of the end-to-end example plus four more: the opco has €1,000,000 of pre-tax profit; its combined German corporate burden is taken as 30% — a rounded teaching figure for 15% corporate tax, the solidarity surcharge and an average municipal trade tax; the shareholder loan's rate passes transfer pricing and the opco's net interest stays below the €3,000,000 safe harbour of the interest barrier; the holdco funded the loan from its own equity, so it has no matching interest expense. All figures are fictional.
| Step | Equity route (dividend) | Debt route (interest) | Rule that decides |
|---|---|---|---|
| Germany: tax on the €1,000,000 at OpCo | €300,000 — profit is taxed before it can be distributed | €0 — interest is deducted before tax | Deductibility of interest, capped by §4h EStG at 30% of EBITDA once net interest reaches €3,000,000 |
| Germany: withholding on the outbound payment | 0% on the €700,000 dividend | 0% — Germany levies no withholding on plain loan interest to non-residents | Directive exemption for dividends (≥10%, one year); no interest withholding except on profit-participating or convertible instruments |
| Netherlands: tax at HoldCo | €0 — participation exemption | €244,400 — interest is ordinary income: 19% on the first €200,000, 25.8% on the remaining €800,000 | Deelnemingsvrijstelling covers dividends and gains, never interest |
| Netherlands: 15% withholding on the onward dividend | €105,000 on €700,000 | €113,340 on €755,600 | Dividend tax on the distribution to the treaty-less individual |
| Investor: net received | €595,000 | €642,260 | — |
Debt wins by €47,260 on these assumptions — and two facts wipe that margin out. If the rate exceeds what an independent lender would charge, Germany treats the excess as a hidden profit distribution (§8(3) KStG): it is added back to the opco's taxable income and taxed as a dividend, with the 26.375% withholding the directive would have spared, because the exemption certificate is never in place for a distribution nobody declared. And if the group's net interest in Germany reaches €3,000,000, the 30%-of-EBITDA cap bites: the non-deductible part is taxed at the opco now and carried forward, while the holdco still pays 25.8% on receiving it — the same euro taxed twice in the same year, a risk the equity route never runs. Whether an instrument is debt or equity at all is decided separately by each state that sees it, which is why the classification question precedes the arithmetic.
Selling shares against selling assets
At exit the same business can leave as a share deal (the holdco sells opco shares) or an asset deal (the opco sells its business, then distributes). In the share deal the gain arises at the holdco and is typically exempt — the Dutch participation exemption covers it from 5%, Cyprus exempts gains on securities altogether, subject to land-rich exceptions where the target's value sits in local real estate. In the asset deal the gain is taxed first at the opco's own corporate rate, and the proceeds still have to travel the whole dividend path afterwards — a second layer the share deal never pays. This is the structural reason sellers push for shares and buyers for assets (the buyer loses the step-up in a share deal and inherits the company's history); the negotiating mechanics live in business exit.
Liquidation and redomiciliation
An obsolete tier can be liquidated or moved. Liquidation turns the holdco's accumulated value into a final distribution that many regimes tax like a dividend above returned capital — in the Netherlands liquidation proceeds in excess of recognised paid-up capital sit inside the dividend-tax base, while Cyprus imposes no withholding on distributions to non-resident shareholders. Redomiciliation — continuing the same legal person under another company law — preserves contracts and avoids a disposal, but the tax side does not travel free: under Article 5 of the EU's ATAD, moving residence or assets out triggers an exit tax on the market value of the transferred assets less their value for tax purposes, payable in instalments over five years only where the destination is another EU or EEA state with recovery assistance in place — how the exit charges of different states are computed and deferred is its own subject — the old state's treaty network is lost prospectively, and accrued exposures (a residence challenge, a pending withholding claim) follow the company rather than dissolving with the move. Whether the company law of the departure state permits continuation at all is an incorporation-test question — the one test that finally bites at the end of the structure's life. Across the five jurisdictions of this guide the answer is written in company law, not tax law:
| Jurisdiction | Can the holdco leave as the same legal person? | Basis and gate | What the move does not remove |
|---|---|---|---|
| Netherlands | Yes, into an EU/EEA company form since 1 September 2023; no direct route to a third country | Act implementing Mobility Directive (EU) 2019/2121; the notary must refuse the pre-conversion certificate where the operation serves unlawful or fraudulent purposes, tax circumvention included | Article 5 exit tax; the Dutch treaty network stops applying prospectively |
| Luxembourg | Yes, including to non-EU/EEA states | Transfer of the registered office abroad with continuity of the legal person — which is why a Dutch company bound for a third country converts into Luxembourg first | Article 5 exit tax on latent gains |
| Cyprus | Yes, in both directions | Continuation provisions of the Companies Law, Cap. 113 (sections 354A and following, added in 2006); Registrar consent after tax, creditor and employee obligations are settled | Article 5 exit tax; a residence challenge already open follows the company |
| Singapore | No — the regime is inward only | Companies Act 1967, Part 10A, since 2017: foreign companies may transfer in (two of: assets above S$10 million, revenue above S$10 million, more than 50 employees; solvent); a Singapore company leaves only by share transfer to a new foreign parent, a scheme, or liquidation | No exit tax as such; section 10L and the badges of trade still test any gain realised on the way out |
| UAE (ADGM and DIFC) | Yes, both ways in the two financial free zones | ADGM and DIFC continuance regimes; DIFC publishes a fee for transferring an incorporation out; companies in other zones and on the mainland follow their own rules | Corporate tax residence and the participation exemption are re-tested by the destination, not carried |
Regulation: What Gets Checked
Modern oversight turns on a single question: is there real economic life behind the holding? General anti-avoidance rules (GAAR) and the principal purpose test in tax treaties allow a benefit to be withdrawn where the main purpose of the structure is tax. On top of that sit controlled foreign company (CFC) rules and ATAD: the profit of a passive holding with no presence can be taxed in the beneficiary's own country. Economic-presence requirements have been tightened almost everywhere, and cross-border arrangements have to be disclosed under the DAC6 hallmarks. The basic test is unchanged since the "Danish cases": the benefit goes to the beneficial owner of the income, and an empty mailbox does not pass it.
Where This Is Heading
After Pillar Two the centre of gravity shifted from the rate to the infrastructure. A holding increasingly works as a framework of order: through it a group holds operating companies, SPVs, funds and trusts in a single structure that is transparent for compliance. The jurisdictions that win are those with real economic fabric — people, banks, courts and contracts; where there is only a low rate, benefits are challenged more and more often. The logic is simple: real presence comes first, tax saving follows, and it holds up exactly as long as genuine activity stands behind the structure.
Q/A
Cyprus has raised its rate to 15% — is a Cyprus holding still worth it?
The sum is not decided by one rate. From 1 January 2026 corporation tax went from 12.5% to 15%, but outbound dividends to non-residents still carry no withholding tax, and SDC for domiciled residents fell from 17% to 5%. A defensive withholding survives only for payments to related companies in low-tax and EU-listed jurisdictions.
The holding is registered and has a local director — is that enough for treaty relief?
No. Since the 2019 "Danish cases" relief goes to the beneficial owner of the income, and the principal purpose test and a GAAR allow it to be withdrawn where the main purpose of the structure is tax. What is examined is where decisions are taken and whether there is an office, people and functions of its own; a letterbox with a nominee board fails that test.
We have a real office and staff — can our home country's CFC rules still tax the holdco's profit?
Yes, depending on whose rules apply. The EU's ATAD template carves out a controlled company carrying on substantive economic activity with staff, equipment, assets and premises — but that escape exists only where the investor's state implemented it, and several regimes attribute passive holding income mechanically, whatever the presence looks like. CFC is a test of the investor's own tax law, and an office in the holdco's state is evidence there, not an answer.
Where is the permanent-establishment risk in a pure holding that does nothing operational?
In its people. A holdco resident in one state creates a taxable presence in another when someone habitually acts for it there — an owner who negotiates and signs from home, an executive running the subsidiaries from a third country. The result is not a change of residence but profit attribution and filing obligations in the state of activity; the facts examined (premises, habitual conduct, authority) are separate from the residence facts, and the two tests fail independently.
ATAD III has been dropped — does that mean substance requirements will ease?
They will not. The EU formally abandoned Unshell in June 2025, but the fight against shells moves into the DAC6 reform, and CFC rules, GAARs and the principal purpose test are all still in place. What disappeared is a separate test, not the scrutiny: relief is denied through the general anti-avoidance rules instead.
Our group turns over less than €750 million — can we ignore Pillar Two?
The GloBE rules, yes: the threshold is revenue of €750 million or more in at least two of the four preceding fiscal years, with a 15% minimum effective rate in each jurisdiction. Below it, CFC rules and ATAD still apply — the profit of a passive holding with no presence can be taxed in the owner's own country at their rate.
Can I still route dividends onward to an offshore company through a Dutch holding?
Not any more. From 2024 the Netherlands extended its conditional withholding tax to dividends paid to affiliated companies in jurisdictions with a rate below 9%, on the EU lists, or in abusive arrangements; the rate equals the top corporate rate — 25.8% in 2026. The transit route is closed at the exit, not at the entrance.
Is a shareholder loan a cheaper way to take money out than dividends?
It is an earlier way, not automatically a cheaper one. Interest is deductible at the operating level, but the rate must survive transfer pricing, the deduction is capped by interest-limitation rules tied to EBITDA, the source state applies its own withholding article, and the holdco pays tax on the interest it receives — participation exemptions cover dividends and gains, not interest. The comparison has to be run through the full chain, the way the debt scenario above does.
Why does the buyer want an asset deal when selling shares is tax-free for us?
Because the exemption is the seller's, not the buyer's. In a share deal the gain sits at the holdco and is typically exempt under a participation regime, but the buyer inherits the company's entire history and loses the step-up in the assets' tax base. In an asset deal the buyer gets clean assets at market cost, while the seller's group pays corporate tax on the gain and then the dividend chain on the proceeds. The two prices differ precisely by these tax positions, which is why the deal form is negotiated, not assumed.
Can we move the holdco to another jurisdiction instead of liquidating it?
Sometimes — and it is two tests, not one. Company law decides whether the entity can continue abroad with the same legal personality at all; tax law charges the move: under ATAD an exit tax applies to unrealised gains when residence or assets leave, the old treaty network stops applying prospectively, and accrued exposures follow the company. Liquidation, by contrast, crystallises a final distribution that many regimes tax like a dividend above returned capital. Which is cheaper depends on where the value and the latent gains sit.
Which of the five holding jurisdictions lets the company leave without liquidating?
Four of them, on different terms. Cyprus allows continuation out in both directions under the Companies Law, Cap. 113, once tax, creditor and employee obligations are settled; Luxembourg allows a transfer of seat abroad with continuity, including to non-EU/EEA states; the Netherlands allows conversion into another EU/EEA form since 1 September 2023 but has no direct route to a third country; ADGM and DIFC allow continuance both ways. Singapore's regime is inward only — a Singapore company leaves by share transfer, scheme or liquidation. In every case the move is a company-law act; the tax on latent gains is charged separately by the departure state.
Singapore has no capital gains tax — so a Singapore holdco's gain on selling a foreign subsidiary is always tax-free?
Not since 1 January 2024. Section 10L taxes a gain from the disposal of a foreign asset when it is received in Singapore by an entity of a group with members outside Singapore, unless the entity had adequate economic substance there — for a pure equity-holding company, filings up to date, operations managed and performed in Singapore, and adequate people and premises. The section 13W safe harbour (20% held for 24 months) does not rescue an entity that fails that test, and a gain kept offshore is untaxed only until it is remitted. Substance, for once, is the switch itself.