Distilled
A Dutch joint venture between two banks was funded with limited-recourse loans from a Luxembourg lender, the interest matched exactly to the dividends on hybrid preference shares it acquired. Luxembourg taxed that interest at a higher rate than the Netherlands. The Supreme Court held the deduction failed anyway: a compensating levy protects only within the anti-base-erosion structure, and the artificiality here sat outside it.
The facts
The taxpayer was a Dutch joint venture company, the vehicle for a co-operation between a Netherlands bank and a French bank. A Luxembourg subsidiary of the French bank held 95 per cent of its nominal share capital; the Dutch bank held 5 per cent, together with a very large share premium. The structure was implemented in two tranches, in August 2006 and March 2007, and unwound in September 2009.
The Luxembourg shareholder lent to the taxpayer on three facilities: a super senior loan of 75 million and two limited-recourse loans of 410 million and 5 million, with a further limited-recourse loan of 325 million in the second tranche, whose limited-recourse provisions were identical to the first. Recourse was capped: repayment of principal and all accrued interest was at all times limited to what the taxpayer had received on a holding of preference shares in a second Luxembourg company — proceeds on redemption or sale, plus dividends.
The taxpayer applied the borrowed funds to acquire ordinary shares and all the preference shares in that second Luxembourg company, taking 15 per cent of its nominal capital. They were equity legally and commercially, but debt for Luxembourg tax — dividends deductible there, exempt in the Netherlands under the participation exemption.
Every leg was rate-matched. The preference dividend was fixed at 4.176 per cent in the first tranche and 4.108 per cent in the second; the bond portfolios at both levels were swapped from floating into precisely those same fixed rates; and the limited-recourse interest was calibrated to the preference dividends. The shareholders' agreement went further still, adjusting the Dutch bank's dividend by reference to the Dutch tax saving the structure produced — a contractual sharing of the benefit.
The litigation
Gerechtshof Amsterdam · 30 September 2021 · authority prevails
The Court of Appeal held the taxpayer had failed the business-reasons rebuttal outright — no commercial reason was shown for interposing it between the two Luxembourg companies. It then held the anti-base-erosion provision denied the interest from 1 January 2008 to the end of the following year, and that abuse of law denied it for the preceding quarter. It accepted that the taxpayer had satisfied the compensating-levy limb — Luxembourg did tax the lender's interest income — but held that the counter-rebuttal available to the Inspector applied from 1 January 2008.
Hoge Raad · 1 May 2026 · cassation dismissed
Four grounds were advanced. The first, on the scope of the dispute, failed — and the reasoning deserves attention. The Inspector had confined his counter-rebuttal argument to the later year; the Court of Appeal raised of its own motion at the hearing that the rule had entered into force on 1 January 2008, and allowed him to extend his case into the earlier year. That was upheld, because his original narrower stance rested not on a deliberate choice but on a manifest error of law. The remaining three grounds also failed: that the counter-rebuttal cannot apply where the foreign rate exceeds the Dutch rate; that abuse of law could not reach the earlier quarter; and that the outcome breached freedom of establishment. The Court followed the Advocate General, who had concluded for dismissal almost four years earlier.
ajiho commentary
Arm's length pricing does not protect an artificial structure
For a financial transactions practice this is the passage that matters. Departure from arm's length terms is not a condition of applying either the anti-base-erosion rule or abuse of law. And whether an act forming part of an avoidance construction was on arm's length terms is no usable criterion — because a wholly artificial arrangement can consist of a set of legal acts each of which is, individually, on arm's length terms. Two qualifications matter before this is cited. The words are the Court of Appeal's, which the Supreme Court held it had not misapplied, rather than the Supreme Court's own formulation. And the criterion is unusable for a particular purpose — distinguishing legislation that combats wholly artificial arrangements from legislation of wider scope, which is a point about classifying rules under EU law. It is not a general statement that pricing is irrelevant. It remains striking that every rate in this structure matched every other rate to three decimal places, and that the precision was evidence rather than defence.
The compensating levy now has two holes in it
The taxpayer's position looked strong on its face. It contended the interest was taxed in Luxembourg at 29.63 and 28.59 per cent against a Dutch rate of 25.5 — more tax paid abroad than saved at home. The Court did not adopt those figures; it answered at the level of principle, and the answer was that they made no difference. Be precise about what was decided. Inside the anti-base-erosion structure the levy still worked, and for calendar 2007 it protected the deduction. But from 1 January 2008 the counter-rebuttal, introduced to cut back the absolute safe harbour of 2007, applies wherever the levy on the interest income is 10 per cent or higher — so no foreign rate, however high, is a shield. And abuse of law does not stop at the levy where the artificiality sits outside the structure. Two defined carve-outs rather than a general demotion — but between them, "the interest is properly taxed somewhere at a proper rate" is no longer an answer.
The perimeter is the point
The most transferable piece of reasoning is the delimitation. The anti-base-erosion structure is bounded on one side by the debt owed to the related party and on the other by the legal act performed with the funds. Everything else is outside it — and artificiality outside the perimeter can defeat a deduction that survives inside it. Here the loans and the preference share acquisition sat within; the joint venture relationships between the two banks and the taxpayer, and the acquisition of the bond portfolio, sat outside. That is where the artificiality was found, and it is the move that allowed abuse of law to reach a structure the statutory rule had let through.
Limited recourse mattered — but not for characterisation
Worth being precise, because the instinct is to reach for debt-equity re-characterisation. No such issue arose: it was common ground that these were loans. The limited-recourse feature mattered evidentially. It was the mechanism coupling repayment and interest to the preference share proceeds, and it fed the findings of artificiality and of stripped-out risk. Combined with the swaps neutralising the bond portfolios, the structure had been engineered to bear no risk at all — and engineered risk-lessness reads as design, not prudence. One further caution: the hybrid character of the preference shares made the structure more attractive to the French side, but the Court of Appeal expressly did not rest the motive finding on it, and the taxpayer's attempt to attack that finding failed as a misreading of the judgment. The avoidance intent found was the taxpayer's own, because it was its profit that bore the interest.
And a drafting warning in the shareholders' agreement
The Dutch bank's dividend was adjusted by reference to the Dutch tax saving the structure generated. That clause did nothing but share the benefit between the parties — and it was treated as close to dispositive evidence that the design was tax-driven. A commercial agreement that prices the tax outcome is a document that proves the tax outcome was the point.
What this means for your business
On benchmarking as a defence. A file showing every leg priced at arm's length answers the pricing question and nothing else. Where a structure may be challenged as artificial, the pricing file is not the defence — and if every rate ties exactly to every other rate, it may be part of the problem.
On compensating levies. Do not build a Dutch position on the foreign rate being high. The test is a floor, not a comparison, and clearing it comfortably provides no additional protection.
On limited recourse. Capping recovery to the proceeds of a specific asset is a legitimate commercial technique, but it links the two sides of a structure in a way an authority can read as coupling. If the feature exists for a genuine reason, evidence that reason at the time.
On commercial agreements. Any clause that shares, allocates or adjusts for a tax benefit is discoverable and will be read as a statement of purpose. If the arrangement has a commercial rationale, the contract should show it.
ajiho will continue to monitor Dutch case law on the anti-base-erosion rules and abuse of law in financing structures, and cover further decisions in future FT Cases.
Case reference
[X] B.V. v Staatssecretaris van Financiën, Hoge Raad der Nederlanden, Belastingkamer, zaaknummer 21/04746, ECLI:NL:HR:2026:736, judgment of 1 May 2026. On appeal from Gerechtshof Amsterdam, 30 September 2021, nrs. 19/00872 and 19/00873, ECLI:NL:GHAMS:2021:2832; at first instance Rechtbank Noord-Holland, HAA 16/1410 and HAA 16/1411. Conclusion of Advocate General P.J. Wattel, 24 May 2022, ECLI:NL:PHR:2022:507, followed. Vice-President M.E. van Hilten presiding, with M.A. Fierstra, E.F. Faase, P.A.G.M. Cools and F.G.F. Peters. Financial years 2007/2008 and 2008/2009. The judgment is anonymised. Primary source: official judgment.
No currency symbol appears in the text as extracted; amounts are stated as rounded figures without a unit, and euro is inferred from context. The first-instance decision is identified only by case number — neither its date nor its outcome is recorded. Two sub-limbs of the first ground were disposed of without reasons under article 81 of the Judiciary Organisation Act; the remaining grounds were fully reasoned. The fraus legis framework is drawn from a line of recent Supreme Court authority rather than developed afresh here.