IP · Tax

IP holding and royalty structures after BEPS

Cyprus, Ireland and the Netherlands compared through the modified nexus fraction, plus the DEMPE question that decides whether the royalty is respected at all.

8 min readUpdated February 2026Director-written
Structure diagram
IP holding company inside an operating group

The structure only works if the IP company genuinely develops and controls the asset — the nexus fraction and DEMPE functions decide the outcome.

Tier 1 — Group
Group parent
Consolidating entity
Pillar Two applies above €750m consolidated revenue.
Tier 2 — IP
IP holding company
Cyprus / Ireland / Netherlands
Owns the patents and copyrighted software, employs or directs the R&D, bears the development risk.
R&D team
In-house or unrelated contractors
Qualifying expenditure — related-party outsourcing and acquisition cost do not qualify.
Tier 3 — Users
OpCo — market A
Licensee
Withholding on royalties reduced by treaty or the EU Interest and Royalties Directive.
OpCo — market B
Licensee
Third-party licensees
External revenue
How value moves
Royalties up
Operating companies deduct arm's-length royalties; the IP company receives them into a concessionary regime.
Nexus fraction
Relief applies only to the proportion matching qualifying R&D spend, with a 30% uplift cap — not to gross royalty income.
R&D spend down
The IP company funds and controls development. Related-party outsourcing sits in the denominator and dilutes the benefit.
Pillar Two
For in-scope groups a sub-15% effective rate draws top-up tax, partially mitigated by the substance-based income exclusion.
Effective rates: Cyprus as low as 2.5% on nexus-adjusted qualifying profit, Netherlands innovation box 9%, Irish KDB 10%. All are nexus-restricted; headline figures are not applied to gross income.

The nexus fraction in one paragraph

Since BEPS Action 5, an IP box benefit is only available in proportion to the research the claimant actually did. Take qualifying expenditure — R&D the company performed itself, plus R&D outsourced to unrelated parties — and divide it by total expenditure including acquired IP cost and related-party outsourcing. A 30% uplift on the numerator is permitted, capped at total expenditure. That fraction is applied to the IP income; only that portion enters the concessionary rate.

So a company that buys finished IP and licenses it out gets essentially nothing from an IP box. A company that built the asset with its own team gets close to the headline rate. Marketing that quotes a flat 2.5% is quoting the best case.

DEMPE decides whether the royalty stands

The IP box is the second question. The first is whether the royalty is respected at all. Under the transfer pricing rules from BEPS Actions 8 to 10, returns follow the functions performed, assets used and risks assumed in developing, enhancing, maintaining, protecting and exploiting the IP.

A company that holds legal title but has no people making development decisions and no capacity to bear the risk will be treated as a conduit, and the return will be reallocated to where the DEMPE functions sit. Legal ownership on its own buys nothing.

Choosing between Cyprus, Ireland and the Netherlands

Cyprus is the lowest effective rate and the cheapest to run, and suits software and patent-adjacent IP developed by a modest in-house team. Ireland suits groups already operating there with substantial R&D headcount and grant support. The Netherlands offers 9% with an exceptional treaty network and a mature ruling practice, which matters where withholding tax on inbound royalties is the binding constraint.

  • Rate matters least where withholding tax on the inbound royalty is high — treaty coverage is worth more than a lower box rate.
  • For groups above €750m consolidated revenue, Pillar Two claws back much of the difference, partly offset by the substance-based income exclusion.
  • Migrating existing IP triggers exit taxation in the country it leaves. That cost is usually the deciding number.
Where these structures fail
  • Applying the headline box rate to gross royalty income instead of the nexus-adjusted portion.
  • Putting acquired IP or related-party outsourced R&D in the numerator of the nexus fraction.
  • No per-asset tracking of R&D expenditure, so the fraction cannot be evidenced on audit.
  • Legal ownership without DEMPE functions, leading to reallocation of the return under transfer pricing rules.
  • Pillar Two top-up tax ignored, so the modelled benefit never materialises for a large group.
  • Exit tax on migrating existing IP discovered after the decision rather than before.

Seen in practice