Fund Tax Architecture: Who Pays, Where and On What

Fund tax structure

Fund tax architecture is the design discipline that decides who is the taxpayer at each link of the chain through which investment income travels from a portfolio asset to an investor's account. It is not a choice of a low-tax domicile: the domicile answers only the vehicle-level question, and the vehicle is one of four levels. A fund's legal structure allocates cash — who receives which distribution and when. Tax architecture determines different parameters: the character of income (dividend, interest, business profit, capital gain), the timing of its recognition (allocation years can precede cash years), and the identity of the person liable (the vehicle, the investor, or an interposed company). Cash allocation and tax allocation regularly diverge: a limited partner in a US partnership can owe tax on allocated profit in a year when the waterfall pays them nothing, which is why the waterfall mechanics and tax architecture concern two different aspects of the same fund.

The chain has a fixed shape. Income arises at an asset in a source country — a portfolio company pays a dividend, a borrower pays interest, a property is sold. At the first level, the source country decides whether to tax the payment at source and whether a treaty reduces that tax. The income then passes through the fund vehicle — and possibly through a feeder (an entry vehicle for one investor group) or a blocker (an interposed corporation that changes who the taxpayer is). At the second level, the vehicle's domicile determines its tax classification: transparent, opaque or exempt. Finally the income reaches the LP, whose residence country applies tax rules at the third level, including anti-deferral regimes that can tax the LP before any cash arrives. A parallel, shorter chain runs from the fund to the manager and carry recipients: the management fee is service income where the people work, and carried interest is a profit share whose character depends on the recipient's country.

Three trade-offs in fund tax structuring: first, transparency versus shielding: a transparent partnership gives investors one level of taxation and preserved income character, but exposes them to source-country filings and to income types their own status cannot absorb; an opaque corporation shields them and destroys both the single layer and the character. Second, treaty access versus simplicity: interposing a treaty-resident holding company can reduce source withholding, but only if the company passes beneficial-ownership and principal-purpose scrutiny — an empty conduit fails the test and still entails costs. Third, one architecture versus many investors: the same portfolio produces different tax outcomes for a domestic taxable investor, a tax-exempt institution and a foreign investor, so a single vehicle rarely fits all three, and every additional structure (feeder, blocker, parallel fund) provides investor-specific results at the cost of entity-level tax and administration.

The practical test of a finished architecture is that for every material payment in the structure the payer, the recipient, the income character, the withholding applied, the treaty position and the report filed can all be named — before the first capital call.

Four taxpayers in one structure

Every fund structure contains at least four distinct tax positions, and optimising one of them in isolation is the classic design error.

The four levels are legally independent: a change at one level (say, inserting a blocker) shifts liabilities between levels but never makes a level disappear.

Classification and transparency: the first question

Before any rate matters, each relevant country classifies each entity in the chain as transparent (income is attributed to the owners as it arises) or opaque (the entity is itself the taxpayer, and owners are taxed on distributions or disposal). The classification is made independently by the fund's domicile, by each asset country and by each investor country — and they do not have to agree. A limited partnership treated as transparent at home can be treated as a company by an investor's jurisdiction, and the reverse. These mismatches determine who may claim a treaty on a dividend, who must file in the source country, whether a foreign tax credit reaches the person who bears the tax, and whether anti-hybrid rules deny a deduction somewhere in the chain. The mechanics of entity classification — elections where they exist, equity/debt characterisation, hybrid mismatches — constitute a separate discipline. The structure must be tested in every material country, and classification at home does not itself prove its recognition abroad.

Nor is domestic classification permanent once investors are in place. Where associated non-resident entities hold in aggregate 50% or more of the voting rights, capital interests or profit rights in a hybrid entity established in an EU Member State, and their own jurisdictions regard that entity as a taxable person, Article 9a of the anti-tax-avoidance directive — inserted by Council Directive (EU) 2017/952 and applied by Member States from 1 January 2022 — treats the entity as a resident of that Member State and taxes it on income not otherwise taxed under the laws of any jurisdiction.

The trigger depends on the investor register rather than on the fund's own documents: a Luxembourg SCSp can lose its transparency because of who subscribed. Article 9a(2) exempts a collective investment vehicle, defined cumulatively as a fund that is widely held, holds a diversified portfolio of securities and is subject to investor-protection regulation where it is established — three conditions, any one of which a concentrated, closely held closed-ended fund can fail. That is why classification records are kept per investor cohort and refreshed as the register changes, not settled once at formation.

Within one structure the typical pattern is deliberate: the main fund is a transparent partnership (a Delaware LP, a Cayman ELP, a Luxembourg SCSp), so that investors get a single level of taxation and income keeps its character; feeders and blockers are opaque corporations precisely to obtain opacity; asset SPVs are local companies taxed where the asset sits. An exempt fund is a third state: the vehicle is formally a taxpayer, but a statutory regime lifts the tax subject to conditions. The exemption is always limited to a particular level and conditions. A Luxembourg RAIF pays an annual subscription tax of 0.01% of net assets and, under Article 45(1) of the law of 23 July 2016, no other Luxembourg tax at fund level — which says nothing about withholding in asset countries or tax in investor countries. Singapore's 13O/13U regimes exempt specified income from designated investments subject to ongoing conditions, not every receipt of any vehicle. A Cayman fund is tax-neutral at vehicle level and precisely there. In each case the exemption removes one level of taxation out of four.

Tax transparency of an entity (who is taxed on its income) has nothing to do with tax transparency as information exchange — FATCA/CRS reporting applies to transparent and opaque funds alike, and an entity can be fully opaque for income tax while fully reportable for exchange purposes.

Source withholding and treaty access

The source country's tax is collected mostly through withholding: the payer deducts tax before the money leaves. For a fund the operative questions are which rate applies to which payment, and who in the chain is entitled to reduce it. The general mechanics of source taxation are described in the withholding-tax guide. Taking the US as the worked source country: US-source dividends, interest, rents and royalties paid to non-US persons are FDAP income, withheld under chapter 3 rules at a statutory 30% unless a treaty reduces the rate and the recipient documents its claim on the correct W-8 form. The same payment to a US person is not subject to that regime at all; a rate can be assessed only when the recipient class is identified.

For a transparent fund, the treaty question is asked at the level of the partners, not the vehicle: a Cayman partnership has no treaty network, and each investor's own residence and documentation determine relief on income flowing through it. Interposing a treaty-resident holding company changes the claimant — but the claim is justified only if the company is the beneficial owner of the income (it has real power to use it and is not a mere conduit obliged to pass it on) and if the arrangement satisfies the anti-abuse rules: the BEPS Multilateral Instrument has written a treaty-abuse minimum standard, typically the principal purpose test, into most bilateral treaties, so the applicable text is the treaty as modified by both parties' MLI positions, never the headline rate from a database. The PPT analysis is covered separately; the operational point for fund design is that the analysis begins with a residence certificate and ends with substance evidence.

Two operational facts are relevant. Relief can be applied at source (the custodian applies the treaty rate on payment) or by reclaim (statutory rate withheld, refund requested later): a structure whose relief exists only through reclaims the administrator cannot actually run has the statutory rate as its real rate, plus a financing cost for the delay. Withholding requires documentation beyond closing formalities — a fund that cannot produce valid investor forms and intermediary statements at the moment of payment loses the relief its memo assumed.

Case note: the Danish beneficial-ownership judgments

Danish companies paid interest and dividends up to Luxembourg and Cypriot intermediates held by private equity funds; Denmark refused the directive exemptions. On 26 February 2019 the Court of Justice (Grand Chamber) ruled in joined cases C-115/16, C-118/16, C-119/16 and C-299/16 on the Interest and Royalties Directive and in joined cases C-116/16 and C-117/16 on the Parent-Subsidiary Directive.

The analysis is based on four holdings. The exemption is confined to beneficial owners, "the entities which actually benefit from that interest economically and accordingly have the power freely to determine the use to which it is put". Authorities and courts must refuse the benefit where there is abuse "even if there are no domestic or agreement-based provisions providing for such a refusal" — no local anti-abuse rule is needed. Abuse requires objective circumstances in which the purpose of the rules is not achieved despite formal compliance, plus the intention to create those conditions artificially; the indications include "the existence of conduit companies which are without economic justification and the purely formal nature of the structure of the group of companies, the financial arrangements and the loans". And the authority need not identify who the real beneficial owner is in order to refuse the claimant.

A fifth holding concerns fund vehicles directly: a Luxembourg SCA authorised as a SICAR could not be a "company of a Member State" entitled to the exemption if the interest it received was exempt from Luxembourg corporate income tax. The vehicle's own exemption was the fact that disqualified it.

The limits matter as much. These are rulings on EU directives, not on bilateral treaties, and they concern years before the MLI principal purpose test applied to those relations. What they change for design is the standard of proof: a company interposed for directive or treaty access has to receive income it is free to keep and use, its substance has to exist before the payment rather than after the assessment, and relief claimed at the top of a chain can fail on facts about the vehicle at the bottom.

The scope of the rules and the requirements of proof have developed since. In C-228/24 (Nordcurrent group), decided on 3 April 2025, the Court held that the anti-abuse rule of the Parent-Subsidiary Directive is not confined to conduits — a subsidiary carrying on its own activity falls within it too — and that the assessment is tied neither to the date of incorporation nor to the date of payment: an arrangement initially put in place for valid commercial reasons which reflect economic reality has to be regarded as not genuine from a certain point onwards where it was kept in place despite changed circumstances. The same judgment also benefits the taxpayer: characterising the arrangement as not genuine is not enough on its own, and a separate finding of a tax advantage defeating the object and purpose of the Directive is required. Advocate General Kokott's opinion of 21 May 2026 in C-203/25 addresses one further criterion: matching amounts and closeness in time between receipt and onward distribution are, by themselves, neither sufficient nor necessary for a finding of a non-genuine arrangement — while refusal remains possible even where the direct recipient is a confirmed beneficial owner carrying on real activity, if the dividend moves on to the ultimate beneficiary through an artificial arrangement and there is a direct link between its abusive part and the EU-law benefit. That is an opinion, not a judgment: the Court is not bound by it and has not ruled. For the fund → blocker → portfolio chain the practical consequence is that the file on each link is dated and reviewed annually, because the dispute now concerns a period rather than the existence of a rationale. The full treatment is in GAAR and the principal purpose test.

The blocker: functions and costs

A blocker is a corporation deliberately interposed between a transparent fund (or asset) and an investor so that the investor owns shares in a company rather than a share of the underlying activity. Everything a blocker achieves follows from that one substitution. What flows through a partnership to its partners — business income, filing obligations, debt-financed income — stops at the corporation; the investor is left with dividends and share disposals. That is exactly what certain investors need: a US tax-exempt institution uses a blocker so that debt-financed or operating income is taxed to the corporation instead of becoming its own taxable income; a non-US investor uses one so that a US operating investment does not create US tax-filing obligations for them. The blocker as a device is jurisdiction-generic and appears wherever transparent funds meet investors who cannot absorb flow-through income — the typical placements within a Delaware fund structure are described separately.

The cost has a fixed structure, and each element should be priced before the entity is formed. The blocker pays entity-level tax on its income — for a US domestic blocker, federal corporate tax at the current 21% rate on its taxable income, plus any state tax. Cash leaving the blocker can face a second charge: dividends from a US corporation to non-US shareholders are US-source FDAP, withheld at 30% absent treaty relief; a foreign blocker with US effectively connected earnings faces instead the branch profits tax — 30% of the dividend equivalent amount, reduced by some treaties.

Losses inside the blocker do not flow to investors; they wait for the blocker's own future income. Leverage helps only to the extent that interest deductions remain available under the limitation rules. The exit route determines a large part of the economic outcome: selling the blocker's shares, selling the asset inside it and distributing, or liquidating produce different taxes for different parties — which is why blocker exits are designed at acquisition, together with the question of scope: a fund-level blocker taxes the whole portfolio to fix one asset, an asset-level blocker increases the number of entities but confines the leakage to the position that needs it.

Feeders, parallel funds and AIVs

Blockers change the taxpayer; the remaining additional vehicles change the grouping of investors. A feeder fund aggregates one investor cohort and channels it into the master — a corporate feeder doubles as a blocker for that cohort, a partnership feeder stays transparent and only packages reporting, currency or eligibility.

A parallel fund invests side by side with the main fund under an agreed allocation policy instead of through it — used when a cohort needs a genuinely different tax owner, leverage profile or asset perimeter, and workable only with strict mechanics: no cherry-picking of deals, blocker taxes borne solely by the cohort that benefits, arm's-length cross transfers. An AIV (alternative investment vehicle) is a per-deal alternative: one investment that the main fund should not hold directly is placed in a dedicated vehicle with economics equivalent to the LPA's.

The generic rule for all three: each additional entity adds administration, audit, FATCA/CRS classification, banking and directors, so the additional vehicle must be justified by a cohort or asset that actually needs it — a question of measured cost against measured exposure, exactly as with the blocker.

Three LPs, one portfolio

Tax structure must accommodate the differences between investors holding a single portfolio. The scenario is deliberately US-bounded — the conclusions below hold for US assets under US rules and do not transfer to other source countries, where the same investor types meet different regimes. All numbers are a fictional teaching example.

A Delaware LP fund holds three US positions: (A) minority equity in listed US corporations paying dividends; (B) a leveraged US rental property held through a partnership, 60% financed by acquisition debt; (C) an operating logistics business organised as an LLC taxed as a partnership. Three LPs each hold 10%: LP1, a US taxable individual; LP2, a US university endowment (tax-exempt); LP3, a non-US corporate investor with no US operations of its own. In the year examined, each LP is allocated $100k of dividends from A, $100k of net rental income from B and $100k of operating profit from C.

Within one fund, the identical portfolio produces three unrelated tax positions, and each LP's conclusion is self-contained.

LP1 (US taxable). The partnership is the natural form: one level of taxation, character preserved, credits and losses flow through. The costs are timing and paperwork — tax follows the K-1 allocation, which can precede distributions (the cash-versus-tax gap; see waterfall mechanics), plus state filings where the portfolio operates. The one possible structural mistake for LP1 is entering through an offshore corporate feeder: a foreign corporation whose income is mostly passive is a PFIC for US persons under the 75% income / 50% asset tests, and the default regime taxes distributions and gains at top rates with an interest charge unless a QEF or mark-to-market election is available. LP1 belongs in the domestic partnership, not in the offshore feeder built for others.

LP2 (US tax-exempt). Exempt status covers investment income — dividends, interest and capital gains are excluded from UBTI — but the partnership look-through treats LP2 as conducting the fund's businesses itself. Stream C is therefore taxable in full, and stream B is taxable in proportion to the acquisition debt. LP2's decision is quantitative: pay tax on its UBTI share directly, or invest through a corporate blocker through which streams B and C are taxed at the corporate rate while converting LP2's return into dividends. The blocker justifies the 21% corporate tax and associated costs only if the UBTI otherwise generated is large enough; for a portfolio that is mostly stream A, it is pure cost.

The choice is determined by a calculation based on the same fictional numbers. LP2's direct UBTI for the year is stream C in full ($100k) plus the debt-financed share of stream B (60% of $100k = $60k) = $160k; an exempt organisation that is a corporation is taxed on UBTI at corporate rates, so $160k × 21% ≈ $33.6k before the $1,000 specific deduction and any state tax, and LP2 files Form 990-T because gross unrelated income exceeds $1,000.

When B and C are routed through a blocker instead, the corporation is taxed on all of both streams — $200k × 21% = $42k — after which the dividend up to a US exempt shareholder is not UBTI at all. On these facts the blocker costs $8.4k more than paying the tax directly. If one variable changes — acquisition debt on the property from 60% to 100% — the direct charge rises to $42k and the tax charges under the two options become equal.

Below that crossover the blocker is used to eliminate: the 990-T filing, the debt-financed portion of the gain on a later sale, the risk that leverage rises mid-hold, and the effect of §512(a)(6), under which each unrelated trade or business is computed separately, so a loss in one cannot shelter income in another.

LP3 (non-US). Taxation varies with income character, which is why no single answer to "what does a foreign investor pay on a US fund?" exists. Passive streams are a withholding matter: dividends at 30% or the treaty rate; much US-source interest can qualify for the portfolio interest exemption; gains on non-real-property stock are generally outside US tax for a foreign investor without a US business.

Streams B and C have a different tax regime: they make LP3 a participant in a US trade or business, with a US return to file and §1446(a) withholding — 21% of allocated ECTI for corporate partners, 37% for non-corporate — taken on allocation whether or not cash is distributed. Real estate adds FIRPTA, and a secondary sale of the fund interest itself adds 10% withholding under §1446(f).

A blocker interposed between LP3 and streams B–C replaces all of that with one corporate taxpayer — at the blocker price set out above. Whether LP3's home country then taxes the same income again, and whether its own anti-deferral rules attribute the blocker's profit currently, is a question of LP3's residence law, not of US law; a controlling stake in a low-taxed passive blocker is exactly the fact pattern CFC regimes exist for.

Case law and legislation on secondary sales

A Greek mining company held an interest in a US LLC taxed as a partnership and had that interest redeemed. The IRS treated the gain as effectively connected on the strength of Revenue Ruling 91-32; the Tax Court disagreed, and the D.C. Circuit affirmed on 11 June 2019 (No. 17-1268), declining to defer to a ruling resting on "a single unreasoned sentence" and holding that the US-office rule looks to where the redemption itself occurred rather than to where the appreciation had been generated. Congress had already reversed the outcome prospectively: the Tax Cuts and Jobs Act of 2017 §13501(a)(1) enacted §864(c)(8), under which a foreign partner's gain on the sale of an interest in a partnership engaged in a US trade or business is effectively connected to the extent of that partner's distributive share of the gain that would have been effectively connected had the partnership sold all of its assets at fair market value on the date of the sale. The §1446(f) withholding is the collection mechanism implementing that rule.

Two conclusions follow for tax structuring. The tax on a secondary sale is a function of the partnership's asset mix on the sale date rather than of anything the seller did, so a fund that acquires a single US operating business changes the exit position of every foreign LP already inside it. The certification procedure before settlement enforces a rule Congress wrote specifically to overturn a taxpayer win.

Real estate: the US and the United Kingdom

The conclusions above concern the US. If the fund's real estate is located in the United Kingdom, the policy is unchanged — non-residents are taxed on gains from land — while the taxpayer, the collection point and the available alternative all change.

The charge also applies to indirect disposals. Since 6 April 2019 a non-UK resident is taxable not only on a direct disposal of UK land but on an indirect one: the disposal of an interest in an entity that is UK property rich — 75% or more of the market value of its gross qualifying assets deriving from UK land, under TCGA 1992 Sch 1A Part 2 — where the seller also has a substantial indirect interest, a 25% investment held at any point in the two years before the disposal, which HMRC treats as insignificant only where it was held for 10% or less of that window (75 days or fewer). Because the 75% test runs on market value at the disposal date, the fund's valuation record is used for tax calculations and not only for investor reporting.

Collection follows the reverse sequence to FIRPTA. There is no transferee withholding: the non-resident seller reports and pays within 60 days of completion for completions on or after 27 October 2021, and must file even where there is no tax to pay or the disposal produced a loss. The US withholds the money first and refunds against a return; the UK receives the return first and enforces afterwards — which moves the fund's operational problem from cash withheld at source to investor filings the administrator has to obtain.

The UK also offers an instrument the US does not. A UK property rich collective investment vehicle can make an exemption election under paragraph 12 of Schedule 5AAA, whose effect "is to move the point of taxation on gains on UK land from the relevant fund to its investors" — achieved through reporting rather than tax: information on investor disposals in the two years before the election (para 14) and a report for every period of account, delivered within 12 months of the end of that period (para 15), with revocation available to HMRC for serious failures. Where a US structure achieves the investor-level result through a blocker and its corporate tax, a UK structure can achieve it through an election and a reporting duty. The policy is the same, but the cost and failure mode differ — which is why a conclusion from one jurisdiction remains an assumption when applied to another until the local rule has been examined.

The fund under the global minimum tax

The global minimum tax does not add a fifth taxpayer — it checks how much has already been paid at entity level and tops the difference up to 15%. The threshold is the group's, not the fund's: consolidated revenue of EUR 750 000 000 or more in the ultimate parent entity's financial statements in at least two of the four fiscal years immediately preceding the tested year (Article 2(1) of Directive (EU) 2022/2523, which enacts the GloBE Model Rules across the Union). Revenue of excluded entities counts towards that threshold, so exclusion from the tax charge does not reduce revenue counted towards the threshold.

Fund vehicles do get an exclusion, but only at the top of the chain. Article 2(3)(a) takes an investment fund that is an ultimate parent entity, and a real estate investment vehicle in the same position, outside the Directive; Article 2(3)(b) extends the exclusion to an entity at least 95% of whose value is held by such entities and which operates exclusively or almost exclusively to hold assets or invest funds for them. The definition of an investment fund in Article 3(31) is cumulative: pooled assets from a number of investors, some of them non-connected; a defined investment policy; reduced transaction and analytical costs or collective risk-spreading; a design aimed at investment income, gains or protection; investor rights to a return based on contribution; regulation of the fund or its management where it is established or managed; and management by investment fund professionals. None of the seven conditions is optional, which is how a concentrated club vehicle outside any regulatory perimeter misses the definition.

Below the ultimate parent there is no exclusion, and a separate calculation applies instead. An investment entity that is not tax transparent and for which no election has been made computes its effective tax rate separately from the jurisdiction in which it is located: adjusted covered taxes over the group's allocable share of its qualifying income, with several such entities in one jurisdiction combined (Article 41). Two elections eliminate the separate calculation. The Article 42 election treats an investment entity as tax transparent where the owner is taxed on annual fair-value changes in its ownership interest at a rate at or above the minimum rate. The Article 43 election moves the owner onto a taxable distribution method: distributions and deemed distributions enter the owner's qualifying income, and the owner's share of undistributed net qualifying income of the third preceding year becomes the base for top-up tax. The three regimes differ in the place and year of taxation rather than in whether tax arises.

Joint ventures are computed apart from the group. Article 36 defines a joint venture as an entity whose results are reported under the equity method in the ultimate parent's consolidated financial statements where the ultimate parent holds, directly or indirectly, at least 50% of the ownership interest; top-up tax for the joint venture and its affiliates is computed as if they were a separate group with the joint venture as its ultimate parent. For co-investment this is the ordinary characterisation: the fund's equity-method interest in a deal is subject to its own computation.

The OECD package of 5 January 2026 provides a separate option for US-headquartered groups. Under the Side-by-Side Safe Harbour the top-up tax for a jurisdiction is, at the filing constituent entity's election, deemed to be zero for purposes of the IIR and the UTPR where the group's ultimate parent is located in a jurisdiction with a Qualified SbS Regime; the safe harbour is available for fiscal years commencing on or after 1 January 2026, and on the OECD Central Record that jurisdiction is at present the United States alone. A separate paragraph of the package extends the deemed zero to the group's interests in a Joint Venture or JV Subsidiary, whether or not the group also has constituent entities in the same jurisdiction. Two qualifications apply. The safe harbour does not apply to another group's interest in the same joint venture, so two partners in one deal can have different tax outcomes. And where an entity falls outside the joint-venture definition only because it is itself the ultimate parent of an in-scope group, the safe harbour does not apply to it.

What side-by-side does not remove is the qualified domestic minimum top-up tax. Groups electing the safe harbour remain subject to the QDMTT in every QDMTT jurisdiction in which they operate, and the QDMTT everywhere is computed without the pushdown of taxes on controlled foreign companies or foreign branches. For a fund chain that means a portfolio company in a QDMTT jurisdiction pays alongside everyone else, regardless of the number of US investors at higher levels of the structure. The computation itself, the safe harbour timetable and implementation information are in the Pillar Two guide; filing dates, contents and penalties are in global minimum tax reporting.

The manager and carry chain

Manager taxation is analysed separately from investor taxation. The management fee is service income of the management company, taxable where the company is resident and where its people actually perform the functions — a fee booked to an entity in a convenient jurisdiction while the deal team works elsewhere can lead to transfer-pricing reallocation of the profit and can create a permanent establishment where the people negotiating are actually located, and the choice of manager jurisdiction is a licensing and substance decision before it is a tax one.

Fees are also subject to indirect-tax rules, which refer to national definitions: Article 135(1)(g) of the VAT Directive exempts "the management of special investment funds as defined by Member States", so whether the same fee to the same fund family is exempt, taxable or outside the scope turns on which Member State's definition applies and where the service is supplied. Exemption is not always beneficial: an exempt supply carries no input-VAT recovery, so unrecovered VAT on the manager's own costs is a final cost rather than a timing difference.

Carried interest is economically a profit share and legally an allocation through the carry vehicle, and its tax character depends on the recipient's country — there is no universal answer. In the US, carry allocated on an applicable partnership interest keeps long-term capital gain treatment only where the underlying gains meet the more-than-three-year holding period of §1061; gains from shorter holdings are recharacterised as short-term.

The UK moves qualifying carry into an income-tax regime from April 2026, and other jurisdictions range from capital treatment to full employment income — the country-by-country treatment is described in the carried interest guide.

Two design rules apply in every jurisdiction: separate documentation of the three cash flows — fee for services, carry as profit share, GP commitment return as return on invested capital, because commingling them creates a risk of the least favourable characterisation of the combined income; and recognition that a carry vehicle in a neutral jurisdiction does not relocate the services of a partner who works in London or New York — when carry recipients move country mid-fund, the grant, vesting and payment years are each analysed separately for residence.

The full list of money and tax flows

A tax structure can be recorded as a list of payments: every material cash flow with its taxpayer, source treatment and obligation. The payments in the example relate to the US-bounded scenario; a real fund maintains such a list for each source country.

Money and tax-flow information answers the audit question "who pays what where". Each listed flow has a control the administrator must actually be able to operate: the fund is the withholding agent for several flows, and an obligation the data system cannot support means an impending penalty tied to the payment date.

Reporting, tax advances and data

Three obligations apply to the entire structure rather than to any single flow. First, information reporting: every entity in the chain — fund, feeders, blockers, GP, carry vehicle — needs its own FATCA and CRS classification, and the two regimes classify differently, so reusing one classification for both regimes is a systematic error; information exchange is described in the tax transparency guide, and EU-nexus structures add DAC6 hallmark analysis at design time and again at each material amendment.

Second, tax advances: when the fund pays withholding on behalf of a specific partner, LPA drafting treats it as an amount distributed to that partner — the approach reflected in the ILPA Model LPA's tax-distribution and carry provisions — so that one investor's tax never silently reduces another's return; the same logic gives carry holders tax distributions when allocations precede cash, and it only works if the LPA says so before the first such payment.

Third, the data obligation: allocations, withholding by payment, investor status forms with effective dates, debt balances and tax lots must be stored in the administrator's system, because K-1s, Forms 8804/8805, FATCA/CRS files and treaty relief claims are all generated from the same records — an architecture whose reporting depends on data nobody captures has a structural compliance gap.