IP migrations after the OECD transparency reforms

The classic IE–NL–LU stack and why it broke

For the better part of two decades, the default international IP structure for a US, UK or European group followed a familiar path. A principal company in Ireland (12.5% headline rate, with the Knowledge Development Box for qualifying IP), a Dutch holding to manage royalty flows under the EU Interest and Royalties Directive, and a Luxembourg finance company to round out the leverage. Layered onto that, a Bermudan or Cayman entity owned the underlying IP itself, and a cost-sharing arrangement (or a cost contribution arrangement under the OECD Guidelines) fed development cost up the chain.

The model was elegant in 2008. By 2026 it has been hit by four consecutive policy waves — the original BEPS modified nexus approach, the EU's Anti-Tax Avoidance Directives (ATAD I and II), Pillar Two, and a cluster of transparency rules (DAC6, DAC8, EU public CbCR) that make the structure visible to regulators, journalists and competitors well before any tax authority opens an audit. The result is not that the stack is illegal. It is that the steady-state ETR is no longer demonstrably below the rate a single onshore principal could achieve, while the operational and reputational drag is materially higher.

We have advised on roughly forty IP repatriations and reorganizations since 2023. This essay is the consolidated version of what is actually working — and what is being abandoned.

The four waves that closed the gap

To understand what changed, it is worth being specific about which rule did what. The popular narrative collapses everything into “BEPS”; the practical reality is more interesting.

1. Modified nexus approach (BEPS Action 5, in force since 2016)

The modified nexus approach restricted preferential IP regimes (patent boxes, IP boxes, the Knowledge Development Box) to the proportion of income attributable to qualifying expenditureincurred by the taxpayer itself. The corollary: if the IP was acquired or its development was outsourced to a related party, the preferential rate did not apply to that fraction. Most IP-rich groups quietly shifted into compliant nexus arithmetic between 2018 and 2021. This was the wave that ended “migrate finished IP to a low-tax jurisdiction and let the rate fall to 6.25%”.

2. Anti-Tax Avoidance Directive (ATAD I & II)

ATAD I's controlled foreign company (CFC) rules, transposed by every EU member state by 2019, brought passive income from low-taxed subsidiaries into the parent's tax base unless genuine economic activity could be demonstrated. ATAD II (2022) neutralized hybrid mismatches that had previously enabled double-non-taxation of royalty flows through US disregarded entities. Together, these directives cut the principal value of a Bermudan or Cayman tier sitting between an EU operating company and an Irish principal.

3. Pillar Two (in scope from fiscal 2024 / 2025 for most groups)

The 15% global minimum tax does the most damage to a stack that was already running close to 15% — which most modern IP stacks were, after the modified nexus approach. The marginal benefit of the structure is now compressed into the gap between the jurisdiction's effective rate and 15%. For groups operating at a 13% jurisdictional ETR, a fully modeled top-up of 200 basis points is no longer enough to justify the operational complexity.

4. Transparency: DAC6, DAC8, public CbCR

DAC6 (in force since 2020) requires reporting of cross-border arrangements that bear hallmarks of aggressive tax planning. DAC8 (in force from 2026) extends mandatory automatic information exchange to crypto-asset transfers — relevant for a small but growing subset of IP-adjacent intangibles. EU public CbCR (first reporting cycles in 2025–2026) makes country-level revenue, profit and tax visible in a public register. The three instruments do not change the tax cost; they change the political cost. CFOs whose 2024 sustainability reports flagged an Irish ETR of 6% have been asked to explain it on earnings calls. Many would rather not.

Onshoring: when it actually makes sense

Bringing IP “onshore” — meaning into the operating jurisdiction or the headquarters jurisdiction — is the most reported response, but it is not always the right one. Onshoring makes sense when three conditions hold simultaneously:

1. The R&D activity already sits in the destination jurisdiction. Onshoring IP into a jurisdiction whose modified nexus ratio is genuinely high is the cleanest position. Onshoring into a jurisdiction where the IP would have to be re-developed to qualify is theatre.
2. The destination jurisdiction has a usable preferential regime. The Italian Patent Box, the Spanish IP regime, the Belgian innovation income deduction, the French IP regime and the UK Patent Box are still valuable — typically delivering an effective rate of 8–11% on qualifying income. The KIDD in Ireland and the Patent Box in the Netherlands remain workable but require disciplined nexus tracking.
3. The exit cost is bearable.An IP migration triggers an exit charge in the source jurisdiction, valued on an arm's-length basis. For mature IP with strong forward cash flows, the exit charge can exceed five years of the projected tax saving. Run the present-value math before you commission the legal work.

Where these conditions hold, onshoring is the most defensible and the most boring answer. It removes the structure from the DAC6 / public CbCR conversation, simplifies transfer pricing, and absorbs cleanly into the Pillar Two computation.

The cost contribution arrangement (CCA) revival

For groups whose R&D is genuinely distributed across two or three jurisdictions, the cleanest answer in 2026 is not a single principal — it is a cost contribution arrangement under Chapter VIII of the OECD Transfer Pricing Guidelines. The CCA was somewhat out of fashion through the 2010s because the centralized-principal model was easier to defend in a single audit. With Pillar Two narrowing the rate-arbitrage benefit and transfer-pricing audits increasingly focused on DEMPE (development, enhancement, maintenance, protection and exploitation) functions, distributed ownership has become structurally more defensible.

A workable 2026 CCA has four characteristics:

• Genuine, costed contributions from each participant— typically R&D headcount with documented projects, not just cash.
• Anticipated benefits aligned with contributions— typically expressed as a regional licensing right rather than a global one.
• Buy-in / buy-out mechanicsfor entrants and leavers, valued on an arm's-length basis at the point of the transaction.
• Annual benchmarking against the actual commercial outcome of the arrangement, with adjustments.

Where a single-principal model collapses the IP into one tax jurisdiction and creates a single audit point of failure, a well-documented CCA distributes both ownership and audit risk across the participants. For groups whose R&D and commercial footprint genuinely span the EU, the US and the UK, this is the structure most often surviving 2026 reviews.

Substance: what regulators are looking at in 2026

The substance question has moved from “does the entity have employees?” to a more textured analysis. In our recent experience with EU and Irish revenue authorities, the core questions are:

1. Decision-making. Are the people who can meaningfully approve a major licensing decision physically and legally located in the jurisdiction? Board minutes that show unanimous approval after a fifteen-minute meeting do not pass scrutiny in 2026; they did in 2018.
2. Risk control.Are the people who control the risks that could destroy IP value (legal disputes, cyber loss, product liability) employed by the entity, with a mandate, a budget and a P&L responsibility?
3. Functional cohesion. Does the bundle of functions, assets and risks at the entity hang together as a coherent business unit, or has the entity been engineered to the minimum bar of substance?
4. Documentary trail. Are decisions evidenced in contemporaneous documents created in the jurisdiction, by people resident there?

The shift is from a checkbox view of substance to a narrative-and-evidence view. The Cadbury-Schweppes line of EU jurisprudence remains good law for what cannot be denied; the practical question is now what can be sustained at audit without protracted litigation.

Worked example: relocation from Bermuda to a Dutch cooperative

Consider a software group with $400m of annual royalty income, IP historically held in Bermuda, R&D performed in the Netherlands and the United States, and a commercial footprint across Europe. The Bermudan tier delivers a near-0% rate; under Pillar Two, the group's consolidated ETR is being topped up by ~15 percentage points on the Bermuda profit, and the income is being collected either via a Dutch QDMTT proxy (where the income is allocated under the IIR) or via a UTPR adjustment in operating jurisdictions.

The repatriation analysis we typically run on this profile looks at three options:

• Option A: Dutch coop with Innovation Box.Migrate the IP to a Dutch cooperative; rely on the Innovation Box (effective rate ~9% on qualifying income subject to modified nexus). Exit charge in Bermuda is symbolic (no corporate tax). ATAD II hybrid issues are eliminated. Pillar Two top-up shrinks materially. Public CbCR profile is much cleaner.
• Option B: US repatriation.Bring the IP to the US under the FDII regime (effective rate ~13.125% on foreign-derived intangible income). Useful where the US parent already exists; less useful where the group's center of gravity is European.
• Option C: CCA between NL and US.Establish a CCA between the Dutch coop and the US parent, with each participant owning regional rights commensurate with its R&D contribution. Most operationally defensible; most documentation-intensive.

In four of the last five projects on this profile, Option C has won. The combination of post-Pillar-Two ETR (12.5–13.5% in steady state), transfer pricing defensibility, and the elimination of Bermuda from the public CbCR has been worth the additional documentation cost.

Myths to discard before you start

• “We just need to add staff to the Bermuda entity.” This was a viable answer in 2017. In 2026, hiring two product managers in Hamilton does not fix the modified nexus ratio, the public CbCR exposure, or the ATAD II issues. It increases cost without changing the outcome.
• “Pillar Two means rate doesn't matter anymore.” Pillar Two compresses the rate-arb benefit; it does not eliminate it. A 9% Innovation Box is still better than a 25.8% Dutch headline rate even after any UTPR allocation, because the SBIE and the local QDMTT absorb a substantial share.
• “The exit charge is the dealbreaker.”For mature IP, sometimes. For IP whose value is concentrated in the next five years of cash flow, the exit charge often clears in two to three years of the post-restructuring tax saving. Always run the present-value math before assuming the conclusion.

Action checklist

1. Build a current-state map: where is the IP held, where is the R&D done, where is the income earned, what is the jurisdictional ETR, and what is the Pillar Two top-up by jurisdiction?
2. Quantify the public CbCR exposure for the next two reporting cycles. If a journalist or a competitor opened the disclosure tomorrow, what would they see?
3. Identify the modified nexus ratio jurisdiction-by-jurisdiction. Where it is below 1.0, decide whether to remediate or accept the dilution.
4. Model three restructuring options against the steady-state ETR and the transition cost (exit charge, internal reorganization, advisor fees). Choose on the five-year present value, not the year-one rate.
5. Pre-clear with the destination tax authority where possible. APAs and unilateral rulings in the EU are slow, but they are cheaper than a contested audit four years later.