Distilled
A French property holding company borrowed EUR 32 million from its Luxembourg parent at 4 per cent, and left the rate unchanged when it extended the loan eight years later. The Court accepted its Moody's-based credit rating but rejected its comparables — mostly dollar bonds, mostly lower-rated, all issued before the amendment date. One comparable survived, and one comparable is not a range.
The facts
SAS Le Trema Holding France heads a French tax-consolidated group. Its sole subsidiary held a single asset — an office building at Asnières-sur-Seine, let to a United States consumer goods multinational. The structure was ultimately held, indirectly, by the sovereign wealth fund of the Sultanate of Oman.
On 3 August 2006 the company borrowed EUR 32,000,000 from The European Acquisition Company 2 SARL, a Luxembourg entity and its sole shareholder, to fund the same-day acquisition of the property company. An amendment that December fixed the rate at 4 per cent. The loan was due to mature in August 2014; by an amendment of 19 November 2014, expressed to take effect retroactively from 1 August, maturity was extended to 2019 and the rate was left unchanged — the same figure set eight years earlier, in a materially different rate environment.
The company deducted interest of EUR 1,448,919 and EUR 1,430,538 in the years ended 31 August 2015 and 2016. The authority found no relevant justification of the rate and capped deductibility at the safe-harbour rate under Article 39, 1, 3° CGI — 2.3 and 2.12 per cent respectively — reintegrating EUR 613,779 and EUR 664,457.
The company defended the 4 per cent with two EY reports applying the comparable uncontrolled price method. A Moody's REIT and commercial property scorecard produced a rating of B1, equivalent to B+. A search on Thomson Reuters Eikon for non-financial B1 to B3 issuers, with three to seven years left to run at 1 August 2014, returned ten bonds, nine of them dollar-denominated. After adjustment these gave an arm's length interval of 4.42 to 9.51 per cent with a median of 5.98 — placing the 4 per cent below the range the company's own study had produced. A supplementary study narrowed to real estate but widened the rating and repayment criteria, returning sixteen bonds, fourteen in dollars, of which only two were adjusted and priced. The company also produced an indicative quotation obtained from Société Générale in January 2019 for terms the bank said it could have offered in August 2014.
The litigation
Tribunal administratif de Paris · 13 March 2024 · authority prevails
The company's claim was rejected in full.
Cour administrative d'appel de Paris · 16 January 2026 · authority prevails on the rate
Two procedural grounds were dismissed. On the substance, the Court set out the standard: the rate is what independent financial institutions would have been likely to grant, taking account of the borrower's own characteristics and in particular its risk profile. It cannot be the rate the company would have had to offer subscribers had it issued bonds, given the difference in nature between a bank loan and a bond issue. The burden sits on the borrower, which may discharge it by any means.
The rating survived — but for want of a valid challenge rather than by endorsement: each of the authority's four attacks on it failed for lack of precise evidence. The comparable set did not survive, though not on the ground the authority had led with.
ajiho commentary
The reference is a bank loan — but bond evidence is admissible
The orthodox French position, that the benchmark is what a bank would lend rather than what bondholders would demand, came with a useful qualification. The borrower may rely on loans to non-financial companies rated close to itself even where those companies sit in heterogeneous sectors, because rating systems already account for sector. It may also use bond yields where, on the facts, bonds are a realistic alternative to an intra-group loan. The authority attacked the sample on both sector and size and got nowhere.
You can win the rating and still lose the case
The authority attacked the credit analysis on several fronts and lost on every one. The strong tenant covenant did not remove risk — the tenant could vacate or renegotiate, and did renegotiate downwards in 2017. Rising values in the Asnières market, coupled with the fact that the loan was secured by mortgage security, were said to reduce credit risk; but the authority produced nothing precise to show that EY's scoring of the asset quality sub-criterion at Ba, on the valuation reports consulted, was too generous. And the implicit support argument failed for want of any commitment. So the rating stood — not because the scorecard was endorsed, but because every challenge to it failed evidentially. The company established it was a B1 credit and lost anyway. Rating and comparables are two separate pieces of work, and only one of them landed.
A single comparable is not a range
The authority's lead attack — heterogeneous sectors, differing sizes — was dismissed outright. What killed the set was three independent defects. Nine of ten bonds were dollar-denominated. Bonds issued in 2013 and early 2014 were not shown to reflect conditions at the November 2014 amendment date. And eight of the ten came from lower-rated issuers with no corresponding adjustment, while the supplementary study adjusted in the opposite direction for two higher-rated bonds. One comparable survived — a Gecina euro bond adjusted from 1.77 to 4.69 per cent — and a single comparable is not capable of determining a range. The analysis was not held to be false. It was held to be insufficient, which is a different and more easily fatal thing.
The currency adjustment must be evidenced, not asserted
Building a euro benchmark from a dollar sample and adjusting for the cross-currency basis is common practice where the euro universe is thin at the relevant rating. The Court did not hold the technique impermissible — it held that the company had not established the adjustment was sufficient. That is a burden point rather than a prohibition, but the effect is identical for any file that asserts the adjustment rather than showing its workings.
Extending a loan resets the pricing date
By extending maturity to 2019 the lender must be regarded as having reassessed the risk of its commitment, and the fact that formally it is the same loan renewed is irrelevant. The Court then turned the point into evidence against the company: carrying forward a rate identical to one set eight years earlier, in a different financial context, told against it. Note the symmetry, because it is the trap. The same logic that moved the testing date to 2014 — helpful to the company on the rating — is what made a comparable set drawn from 2013 and early 2014 issuances too stale to support it. Any amendment touching tenor is a repricing event, and the benchmarking has to be repriced with it.
Implicit support needs an anchor
Unusually, it was the authority arguing implicit support — contending that indirect ownership by Oman's sovereign wealth fund limited default risk and should lift the rating. The Court rejected it: it was common ground there was no guarantee, and although such support might be envisaged by an investor and a fortiori by the lending shareholder, absent any commitment the authority showed nothing to make it an objective parameter. Useful authority for resisting an uplift — and worth reading carefully by anyone relying on implicit support in the other direction, because the reasoning is symmetrical.
What this means for your business
On refinancing and extensions. Treat any amendment touching maturity as a repricing event. Either reprice to market or document contemporaneously why the existing rate still holds in the conditions prevailing at the amendment date. Rolling a rate forward unchanged is affirmative evidence against you.
On the currency of your comparable set. Where the sample is predominantly in another currency, the swap adjustment must be evidenced. If the euro universe at your rating is genuinely thin, show the search that established it rather than reaching silently for the dollar market.
On sample size and rating dispersion. A set that collapses to one or two usable observations after proper adjustment has not produced a range. Screen with the adjustments already in mind — a sample sitting mostly below the tested rating, used unadjusted, will not survive.
On bank quotations. Obtain them at the time, not in anticipation of a dispute, and press for terms the bank would actually have transacted. An ex post indication built on figures you supplied proved nothing here, even read alongside the benchmarking studies.
ajiho will continue to monitor French administrative court decisions on Article 212 CGI and intra-group interest rates, and cover further decisions in future FT Cases.
Case reference
SAS Le Trema Holding France v ministre de l'économie, des finances et de l'industrie, Cour administrative d'appel de Paris, 2ème chambre, N° 24PA02156, 16 January 2026. Inédit au recueil Lebon. Appeal from Tribunal administratif de Paris, jugement n° 2110247 of 13 March 2024. Presiding: Mme Fombeur, présidente de la cour, with Mme Vidal, présidente de chambre; rapporteur M. Alexandre Segretain; rapporteur public M. Perroy. Counsel for the company: Me Arras. Primary source: official judgment.
Note on the source text: the published judgment gives the date of the sole accepted comparable, a Gecina euro bond, as 30 July 2024. That is not an artefact of the scan — the characters are clean — but it cannot be right: both EY studies are dated 2018 and the screen ran to issuances between 1 August 2013 and 1 August 2014. The intended date is 30 July 2014. Cite it as 2014, noting that the published text reads 2024.