Distilled
A Norwegian aluminium holding company borrowed NOK 4.7 billion from group treasury with interest rolled up rather than paid. The Tax Appeals Board reclassified the excess debt as equity and increased income by NOK 272 million. The District Court upheld it, rejecting the argument that a below-market margin should offset the adjustment — and expressly declining to follow the Court of Appeal in Orlen while that appeal was pending.
The facts
Norsk Alcoa Holding AS, since merged into Alcoa Norway AS, was the holding company for Alcoa's Norwegian business — primary aluminium smelters at Mosjøen and Lista, and an anode plant at Mosjøen supplying both Mosjøen and the group's Icelandic smelter.
It took two loans from the Swiss branch of a Luxembourg-resident group treasury company. The acquisition loan of September 2005, up to NOK 2,788 million, funded the purchase of the Norwegian business from a Dutch group company for NOK 3,486 million, alongside NOK 698 million of equity. The anode loan of February 2006, up to NOK 1,950 million, funded the Mosjøen investment. Both carried three-month NIBOR plus 152 basis points, both were unsecured, and both had ten-year terms with an option to extend.
Neither loan required amortisation, and neither required interest to be paid. Interest was capitalised and added to principal until supplementary agreements in 2013 switched to current payment. Because the facility limits did not cover capitalised interest, balances grew well beyond them — combined balances rising from NOK 2.9 billion in 2005 to NOK 6.4 billion by 2013. Long-term debt ran between 94.4 and 77.4 per cent of total assets across the assessment years. In 2006 the debt ratio was 80.8 per cent against 37.2 per cent for the closest comparable aluminium producer.
The tax office held the company thinly capitalised for 2008 to 2014 and increased income by NOK 705,153,000. The Tax Appeals Board, reporting almost seven years later, agreed on principle but found borrowing capacity somewhat higher, reducing the increase to NOK 272,545,000. It also agreed with the tax office that the agreed interest terms were within an arm's length range and should not be adjusted.
The litigation
Skatteklagenemnda · 15 November 2024 · authority prevails on principle
The Board upheld thin capitalisation for every year, reducing the assessment by NOK 432,608,000 but leaving the reclassification intact.
Oslo tingrett · 13 February 2026 · State prevails in full
After a seven-day hearing the District Court dismissed the claim entirely. The company was ordered to pay NOK 1,741,571 in costs including VAT, covering 274.75 hours; its own fees were NOK 4,598,558 excluding VAT, for 853.5 hours. An alternative argument on the quantum of the discretionary assessment was withdrawn during closing submissions and was therefore not reviewed.
ajiho commentary
The object of examination is the balance sheet, not the interest bill
The pivot of the judgment. In a thin capitalisation case what is being tested is the taxpayer's capital structure — the split between debt and equity — and not the interest cost produced by loan size multiplied by rate. The court took that from the Commentary to Article 9, and everything else follows from it. The arm's length structure is the intersection between the maximum an independent borrower would take and the amount an independent lender would advance, and that intersection must so far as possible rest on observable market transactions rather than a theoretical maximum capacity derived from loan sizes combined with a rationality assessment. Pricing is not excluded from the exercise — the court accepted that the portion recognised as debt must still be priced — but it is the second step, not the object of examination.
A usable rule for ranking financial ratios
The most transportable part of the reasoning. Debt to EBITDA measures the ability to repay principal; EBITDA to interest cost measures the ability to service it. Neither says anything directly about capital structure. Because capital structure is what is being examined, the ratios that measure it directly — long-term debt to total assets, and the equity ratio — must be given great weight. Note the asymmetry: the court did not demote the coverage ratios so much as decline to treat them as probative of the question actually being asked. It added a practical sanity check: proxy measures must tie back to economic reality, and the capacity the taxpayer argued for would in several years have exceeded the market value of the company itself. The doctrinal hinge for deciding on capital structure alone came from the 1987 OECD thin capitalisation report — a large deviation from observed market gearing may by itself suffice.
Group control over drawdown converted a point-in-time test into an annual one
Both sides accepted the general rule that the arm's length assessment attaches to the date the agreement is entered into. The court nonetheless tested every year from 2008 to 2014. The facilities were revolving in their contractual wording, and it was common ground that they were not operated that way — but the wording, combined with capitalising interest and no covenants of substance, was enough. Group control was treated as the equivalent of a maintenance covenant, and the company would have been in breach of covenants of that type from 2008 onwards. Worth noting that the acquisition loan, viewed on its own at origination, was undisputedly arm's length. It is an aggressive step, carefully reasoned, and if followed it materially undermines the test-at-origination position for intra-group facilities without hard covenants.
A below-market rate bought nothing
The taxpayer's own expert evidence was that 152 basis points sat below market. The tax authorities found the rate within an arm's length range; the court recorded it as common ground and proceeded on that basis rather than finding it independently — and then qualified it, observing that with quarterly compounding the interest-on-interest effect alone could exceed a percentage point, so the rate could not simply be compared with bond coupons paid currently. The company therefore carried a margin its own expert called cheap, and still faced the full volume adjustment — the economically asymmetric outcome its netting argument had been designed to address. The court rejected that argument squarely: the provision authorises increasing income, not increasing deductions, and Norwegian law contains no re-characterisation in the taxpayer's favour. On that reasoning, a below-market intra-group rate provides no shelter whatever against a thin capitalisation adjustment.
And that is precisely the reasoning that did not survive
Six weeks after this judgment the Supreme Court decided Orlen Upstream Norway and held the opposite: the assessment must be net, and where an authority reduces debt to borrowing capacity it must also correct a below-market rate upward in the taxpayer's favour. This court had declined to follow the Court of Appeal in Orlen on that very question — treating its precedential weight as limited because the appeal was pending, but resting its own conclusion on the statutory wording, which it held told decisively against the taxpayer. The Supreme Court then endorsed the approach it had rejected. Read this decision for its treatment of capital structure and financial ratios, which stands. Do not rely on its rejection of net assessment, which does not.
What this means for your business
On capacity analysis. Lead with ratios that measure capital structure directly. A capacity case built on debt to EBITDA and interest cover is answering a different question from the one a Norwegian court will ask, and can be attacked on that basis alone.
On capitalising interest. Where interest rolls up and facility limits do not cover it, balances outgrow the documented facility and the gearing profile drifts year by year. That drift is visible, and it invites exactly the annual testing applied here.
On covenants. The absence of financial covenants, combined with freedom to draw and repay at will, was what converted an origination-date test into an annual one. Terms that look permissive at signing can widen the period under examination.
On peer data. The comparison that carried most weight was entity-level balance sheet data for listed aluminium producers, not loan transaction comparables. In a capital structure case the relevant market evidence may be the peer group's balance sheets rather than its borrowings — which is a different dataset, a different subscription, and a different piece of work from a rate benchmarking exercise.
This decision should be read together with the Supreme Court judgment in Orlen Upstream Norway AS of 26 March 2026, covered separately in this library, which reached the opposite conclusion on the netting question. ajiho will continue to monitor Norwegian case law on skatteloven § 13-1.
Case reference
Alcoa Norway AS v Staten, Oslo tingrett, TOSL-2025-79984 (25-079984TVI-TOSL/05), judgment of 13 February 2026, published at UTV-2026-6. Main hearing 25 November to 3 December 2025. Counsel: advokat Ståle Rønneberg Kristiansen for the company; advokat Goud Helge Homme Fjellheim for the State. Income years 2008 to 2014. First judicial instance. Primary source: official judgment, published on Lovdata, subscription required.
The copy of the judgment used was a scanned document processed by optical character recognition. The loan balance table does not foot in the 2005 column, and the lender's name is rendered inconsistently. The header names the defendant as the State represented by the Tax Appeals Board while the operative conclusion names the State represented by the Tax Administration. Figures should be checked against the published Lovdata text before any are quoted.