Case Overview (2019, European Union)
In Joined Cases T-760/15 and T-636/16 (EU:T:2019:669), the EU General Court annulled the European Commission’s State aid decision targeting a Dutch advance pricing arrangement (APA) for Starbucks’ Netherlands roasting entity. The Court accepted that an arm’s length benchmark can be used as a tool to assess advantage where national tax law aims to tax group companies as if they had dealt at market prices, but held the Commission failed to prove an economic advantage.
Key Facts
- Starbucks Manufacturing Emea BV (SMBV) operated Starbucks’ coffee-roasting activities in the Netherlands, supplying the EMEA region.
- The Dutch tax authorities concluded an APA on 28 April 2008, valid from 1 October 2007 to 31 December 2017.
- The APA set a cost-based mark-up for SMBV’s production/distribution activities and endorsed a deductible royalty paid by SMBV to Alki LP (UK) for roasting-related IP/know-how, calculated as residual profit (i.e., SMBV’s operating profit minus the routine return).
- In Decision (EU) 2017/502 (adopted 21 Oct 2015), the Commission found the APA granted unlawful and incompatible State aid and ordered recovery. The Commission publicly estimated recovery at €20–30 million for Starbucks (exact amount to be computed by the Netherlands).
The Dispute
The legal question was whether the APA reduced SMBV’s taxable profit compared with the “normal” application of Netherlands corporate income tax (and therefore conferred an advantage under Article 107(1) TFEU), in a way that was selective.
The Commission’s core technical criticisms were that the APA/TP report:
- endorsed the TNMM (in a cost-plus style application) rather than a more direct approach (notably a CUP analysis for the royalty),
- did not properly analyse the royalty (and the Commission argued the arm’s length royalty should be zero), and
- accepted elements of the structure (including issues the Commission raised on green coffee bean pricing) that allegedly shifted profits out of the Netherlands.
The Netherlands and Starbucks argued (among other points) that methodological imperfections are not enough: the Commission had to show, on the evidence available at the time, that any errors actually produced a non–arm’s length outcome that lowered SMBV’s Dutch tax base.
Holding
The General Court annulled the Commission decision.
Key points from the Court’s reasoning:
- Arm’s length as an “advantage” testing tool (not a standalone EU tax code). Where national law does not distinguish between integrated and standalone companies for corporate income tax purposes, it is intended to tax group-company profits as if they arose from market-priced transactions. In that setting, the Commission may use an arm’s length benchmark as a tool to compare the taxpayer’s burden under the ruling with the burden under normal taxation.
- Burden of proof: methodological flaws ≠ advantage. The Court stressed that mere non-compliance with methodological requirements (e.g., disputing TNMM selection, gaps in analysing the royalty) does not automatically show a reduced tax burden. The Commission needed to demonstrate that the identified issues prevented a reliable approximation of an arm’s length outcome and led to a lower tax base.
- Royalty “should be zero” not proven. Based on the functional analysis and the Commission’s comparables discussion, the Court held the Commission did not prove that an arm’s length royalty would be nil (or otherwise that the royalty endorsed by the APA conferred an advantage).
- No hindsight / green coffee beans. The Court noted that the price of green coffee beans was outside the APA’s scope and, in any event, the Commission could not rely on post-APA developments to establish advantage.
Takeaways for Practitioners
- Outcome focus matters. In State aid-style challenges, it is often decisive whether the authority can show the ruling’s method leads to a materially understated tax base, not merely that the TP report is imperfect.
- Residual-profit royalties are high-risk. If royalties are set as residual profit, ensure the file clearly supports (i) the IP being licensed, (ii) who controls DEMPE-relevant functions/risks, and (iii) why the royalty level is arm’s length.
- Be explicit about scope. If a key cost or transaction (e.g., raw materials pricing) is outside the APA, document that boundary and keep contemporaneous support to avoid “gap-filling” later.
- Contemporaneous evidence beats hindsight. Preserve the information set available when the APA was agreed, including method selection rationale and comparables/adjustments support.
Related Content
For practical guidance on building defensible files that can withstand both tax authority and broader regulatory scrutiny, see our Transfer Pricing Documentation Guide. For the conceptual foundation underlying the Court’s benchmarking discussion, revisit the Arm's Length Principle.
Frequently Asked Questions
Q1. What measure was challenged?
A Dutch APA for Starbucks Manufacturing Emea BV (effective 2007–2017) setting a routine return and endorsing a deductible royalty to Alki LP calculated as residual profit.
Q2. Did the Court accept an arm’s length benchmark in State aid cases?
Yes, as a tool to assess whether a ruling reduces the tax burden compared with “normal” taxation where the domestic corporate tax system aims to approximate market outcomes for group companies.
Q3. Why did the Netherlands and Starbucks win?
Because the Commission did not demonstrate that the alleged methodological errors led to an economic advantage (i.e., a lower Dutch tax base) and did not prove the royalty should have been zero.
Q4. Does this judgment mean tax rulings cannot be State aid?
No. It mainly underlines the Commission’s evidentiary burden: it must prove an actual advantage, not just flawed TP reasoning.
Q5. How large was the recovery the Commission sought?
The Commission publicly estimated recovery at €20–30 million for Starbucks (with the exact amount to be computed by the Netherlands under the Commission’s methodology).
