Signal — FT Cases

Autoridade Tributária e Aduaneira v Banco 1… — Portuguese branch

Transfer pricing corrects the transaction you entered into. It does not invent the one you did not.

Distilled

A Portuguese bank branch funded its lending book from a sister branch in Madrid and deducted the interest. The authority never disputed the rate. It argued instead that the branch should have held more free capital, computed the shortfall by applying the arm's length principle, and disallowed the corresponding interest. The Supreme Administrative Court dismissed its appeal: the transfer pricing article corrects transactions carried out, not transactions recharacterised.

The facts

The taxpayer is the Portuguese branch of a credit institution headquartered in France, registered for other monetary intermediation and supervised by the Banco de Portugal. In 2006 it obtained medium-term funding from the Madrid branch of the same bank to support its own lending book, together with overnight funding whenever it lacked the liquidity to meet client withdrawals. Three establishments, one legal person: a French bank with branches in Lisbon and Madrid. The judgment records that the branch's client lending was squared by Madrid, which had a central role in its treasury management — in substance, a centralised treasury arrangement.

The branch documented the arrangement using the comparable uncontrolled price method. It recorded that it had obtained no financing from independent parties that year and therefore had no internal comparable, and benchmarked instead against euro-area interbank money market rates. The monthly average rates it paid on medium-term funding tracked those rates, and sat below the rates it earned on its own medium-term lending to third parties and group clients alike.

The authority did not directly contest the amount of interest — it questioned only whether part of it was deductible. What it attacked was the capital structure. Reasoning from the 1984 OECD report on the taxation of multinational banking enterprises, it took the view that an arm's length level of own funds must be determined by reference to what regulators would require given the assets held and risks assumed, and that capital so determined is free capital which cannot bear interest. Branch equity of EUR 4,327,382 represented about 0.314 per cent of total assets of EUR 1,376,979,890 — a ratio it compared with the same ratio at head office. Labelling the exercise an application of the comparable uncontrolled price method, it computed a free capital deficit of EUR 1,108,441 — the same figure serving both as the capital shortfall and as the interest added back to taxable profit.

One fact framed everything that followed. In early 2001 the French head office had asked the Banco de Portugal to cancel and repatriate the capital allocated to the branch — EUR 21,982,721 — and the regulator had not objected.

The litigation

Tribunal Tributário de Lisboa · 21 September 2022 · taxpayer prevails

The challenge was upheld in full.

Supremo Tribunal Administrativo · 6 May 2026 · appeal dismissed

Sitting in conference, the Court dismissed the authority's appeal unanimously, confirmed the judgment below, annulled both the assessment and the compensatory interest, and put costs on the authority.

Net result. The taxpayer prevails. The transfer pricing article in its 2006 wording is a classic transfer pricing rule with a circumscribed and perfectly delimited scope. Neither Portuguese law nor the Portugal–France convention as it then stood contained any rule addressing free capital. The correction of EUR 1,108,441 was unlawful.

ajiho commentary

Correction is not recharacterisation

The central holding, and it travels well beyond Portugal. The transfer pricing article has in view the correction of transactions carried out — not the recharacterisation of a transaction in order then to correct it. What the authority had actually done was disregard part of the credit granted and recharacterise it as a free capital endowment, then price the remainder. That is a different exercise, and it needed a different legal route: a specific anti-abuse rule directed at the situation, or simulation, or the general anti-abuse clause with the procedure that goes with it. None had been invoked. The transfer pricing article could not be stretched to cover the gap, however respectable the international materials pointing that way.

The methodology was itself the evidence

The sharpest observation in the judgment is almost an aside. No free capital was proved to exist. The Court rested on two things read together: the branch had computed no such amount, and the authority had arrived at its figure only by applying the arm's length principle. Behind both sat the 2001 repatriation, cleared with the regulator. The second is the telling one — a number producible only by hypothesising it confirmed that the thing being taxed was notional. Any adjustment whose quantum exists solely as an output of the adjusting methodology invites the same objection.

Why only one treaty was in play

The instinct on these facts is to reach for three treaties, and the court below did exactly that — France as residence state, Portugal as source of the interest, Spain as the state receiving it. The Supreme Administrative Court declined. The income in issue is business profits, obtained by a person resident in France through a permanent establishment in Portugal, so article 7 of the Portugal–France convention governs and the situation is bilateral. It is not the taxation of the interest that is in issue, and that alone could have engaged several conventions. Spain never enters, because the Madrid branch is not a separate person.

There must be an underlying transaction

The article sits within the corporate tax code's provisions on determining taxable income — net income for the period, income and gains, costs and losses, positive and negative equity variations. It follows that there must necessarily be an underlying transaction to correct, and the only one relevant here was the payment of interest. Even on the hypothesis that an endowment had existed it would still fall outside, because a capital endowment is not a gain, a cost or a patrimonial variation.

The rate was never the issue — and that is the lesson

It is worth pausing on what the authority chose not to argue. The branch had documented its funding cost against euro-area interbank rates, recorded that no internal comparable existed, and shown it paid less on its borrowings than it earned on its lending. Faced with a well-documented price, the authority moved the argument to the capital structure behind it — a pattern worth recognising, because the answer to it is legal rather than economic. A benchmarking file, however good, does not address how much equity an entity should have held.

The later commentary could not be read back

The Court did decide the retroactivity question, and against the authority. Later commentary bears mainly on conventions concluded from that date and following the revised wording — and the revised article was not the one governing 2006. Where the underlying provision has itself been amended, commentary on the new text cannot produce an updated reading of the old, because the exercise is not extending a provision to new situations within its spirit but applying it to situations that were avowedly outside it. What the Court expressly did not need to decide was whether the comparable uncontrolled price method, or any other, was the appropriate one.

What the Court did not hold

It did not hold that free capital may bear interest. It accepted the orthodoxy: where an endowment is proven, interest on it is not deductible, and the epithet "free" is justified by exactly that. It also found the branch had in practice been treated as a distinct and separate enterprise in determining the interest payable. And it closed with a caveat that deserves attention — citing the current Guidelines, it observed that taking account of a minimum volume of own funds is beginning to make headway. This is a decision about legal route under 2006 law, not a rejection of free capital theory.

What this means for your business

On branch funding structures. Where a branch is funded from group treasury rather than capitalised, the exposure is real — but the question to ask of any assessment is which rule the authority is actually applying. If the answer is a transfer pricing article being used to reconstruct a capital structure, that is an argument on the law and not merely on the numbers.

On documenting treasury arrangements. What protected this taxpayer on the pricing was ordinary and unglamorous: a documented method, an explicit record that no internal comparable existed, and rates benchmarked to an observable market reference. The authority never seriously attacked any of it, which is why the case turned on structure instead.

On regulatory history. The 2001 repatriation, cleared with the banking regulator, was decisive background. Where capital is deliberately withdrawn from a branch with the regulator's knowledge, that record is worth keeping — it goes directly to whether any endowment existed at all.

On the direction of travel. Do not over-read this. It is a decision about a 2006 year under a treaty and a domestic rule as they then stood, and the Court itself signalled that the minimum-own-funds idea is gaining ground. The free capital concept was not rejected; only the route by which it was imposed, at a time when neither the statute nor the treaty supported it.

ajiho will continue to monitor Portuguese case law on the attribution of profits to branches and on intra-group financing, and cover further decisions in future FT Cases.

Case reference

Autoridade Tributária e Aduaneira v Banco 1… — Sucursal em Portugal, Supremo Tribunal Administrativo, Secção do Contencioso Tributário, Processo 02070/09.7BELRS.SA1, decided 6 May 2026, unanimous. Relator: João Sérgio Ribeiro. On appeal from Tribunal Tributário de Lisboa, judgment of 21 September 2022. Tax year 2006. The judgment is anonymised. Primary source: official judgment.

The header field of the published judgment gives the date as 05/06/2026 in month/day form; the signature line reads 6 May 2026 and the document number embeds 20260506. Cite 6 May 2026. The tables setting out capital, interest and monthly average rates for 2004 to 2006 are reproduced in the published text as images and are not recoverable, so no rate figures are quoted here. The same judgment appears twice in the source folder under different file names.

This article is published by ajiho for general information only. It reflects ajiho’s own analysis of publicly available sources and does not constitute legal, tax or professional advice, nor a substitute for taking it. No client or confidential information is used in any Signal publication. Where the subject is a court or tribunal decision, the summary is ajiho’s reading of the published judgment and does not account for any subsequent appeal or development. You should take specific professional advice before acting on anything set out here.

Have an arrangement that raises the same question? That’s a conversation worth having.