Arbitrating a patent dispute in Europe: the UPC Centre

Thearticle 35 de l’accord on a Unified Patent Court establishes a patent mediation and arbitration centre, with seats in Ljubljana and Lisbon. The same article sets a limit that nothing can circumvent: a patent may not be revoked or limited in mediation or arbitration proceedings.

That limit is structural, and it is poorly understood. It does not prevent the settlement of a dispute that touches incidentally on validity. It prevents an award from producing the effect erga omnes of a revocation. An arbitral tribunal may assess validity as between the parties, in order to set a rate or decide a contractual breach. It cannot remove the patent from the register. That boundary is not a weakness of the system: it is the line separatinginter partes and what belongs to the public law of patents.

In 2025 the centre put three distinct sets of rules out to public consultation: mediation rules, arbitration rules and rules on expert determination. That trilogy deserves attention, because practitioners often remember only the second.

• Mediation does not decide: it seeks agreement. Its value lies in cost and in preserving the commercial relationship, often decisive when the parties are also supplier and customer.
• Arbitration decides, by an award.
• Expert determination is the most underrated tool: it isolates a technical, economic or licensing question, has it decided by an expert, and spares the tribunal (arbitral or state) from having to resolve it in full. On the essentiality of a portfolio or the selection of comparables, it is often the fastest route.

That is the distinctive feature of this centre: it was not designed as a general arbitration institution transposed to patents, but as a mechanism grafted onto a court. Two rules of procedure organise the articulation.

The Rule 11 of the Rules of Procedure allows the judge, at any stage, to explore an amicable solution with the parties and to steer them towards the centre. The parties may jointly request a stay. The Rule 365 allows the Court to confirm a settlement agreement, which gives it enforceable effect in the Contracting Member States.

That judicial confirmation is the crucial point. It explains how a private settlement can produce effects that the award alone could not: a proprietor may undertake to limit its patent, to assign it, not to enforce it or to accept its revocation, and that undertaking may receive the court’s sanction. The prohibition in Article 35 is not circumvented, it is respected by another route.

An arbitral award benefits from the regime of the New York Convention of 10 June 1958 on the recognition and enforcement of foreign arbitral awards, to which more than one hundred and sixty States are party. That is the decisive advantage over a judgment, even a unified one: the territorial reach of the Unified Patent Court stops at the Contracting Member States, that of an award extends to all States party to the convention.

For a global FRAND dispute the asymmetry is not academic. It explains why a worldwide licence is poorly negotiated before a court whose jurisdiction is regional, and why arbitration finds natural ground there.

In Huawei v ZTE (CJEU, 16 July 2015, C-170/13), the Court of Justice made the parties’ conduct the test for assessing injunctive relief. That framework has a consequence that is still poorly measured: proposing arbitration is a negotiating act, and refusing it is another.

A vague proposal to arbitrate (“we are open to arbitration”) is worth nothing. A precise proposal is worth a great deal: designated institution, neutral appointment mechanism, seat, language, governing law, timetable, exact scope of what is submitted, confidentiality regime, cost control, and the fate of documents already exchanged before the court. Refusing such a balanced and workable proposal may be held against the party that refused it when its genuine willingness to contract comes to be assessed.

Symmetrically, an instrumental proposal designed to delay or to displace the debate is assessed in the same way. The rule runs both ways, as does the FRAND obligation.

Before the Unified Patent Court, the cost consists of a fixed fee per action, to which a value-based fee is added where the value of the dispute exceeds a ceiling set at 500,000 euros. A reduction by half is available for micro-enterprises and small enterprises. The structure is therefore largely determined by the stated value of the dispute.

Before the centre, the structure is different: a registration fee scaled to the amount in dispute, then arbitrators’ fees based on time spent, capped in the expedited track. Two practical consequences follow.

First, the cost variable is not the same: before the court, it is the value of the dispute. In arbitration, it is the time consumed, hence the scope of the referral and the number of arbitrators. Second, a narrowly defined arbitration (sole arbitrator, isolated question, expedited track, online hearing) can be markedly less expensive than full litigation, whereas a three-member tribunal on an undefined dispute is not.

The exact amounts in force (court fees of the Unified Patent Court, the centre’s scale) must be checked at the date of the decision: they are set by decisions of the Administrative Committee and revised periodically.

The objection always returns: entrusting these disputes to arbitrators would privatise patent justice. It deserves better than a defensive answer.

What makes justice public is not the nature of the decision-maker but the object of the decision. Deciding the validity of a patent, defining the scope of a claim, assessing the proportionality of an injunction: these questions build case law, bind third parties and must remain public. Determining the rate applicable to a portfolio between two operators, settling a representative sample, handling comparable licences covered by confidentiality: these questions concern only the parties and are ill-suited to a public hearing.

The real risk is not privatisation, it is opacity: if structuring disputes disappeared into confidentiality, the law would cease to develop. That is why the publication of anonymised awards, unless a party objects, is not a cosmetic detail of the arbitration rules: it is the membrane through which the system learns from what is settled outside it.

In the contract. A useful arbitration clause names the institution, fixes the seat, the language and the governing law, provides for the method of appointment, and (this is the point most often missed) delimits what is arbitrable in the light of the prohibition in article 35.

In the licence negotiation. An offer to arbitrate is drafted like a licence offer: dated, precise, workable, and kept. It forms part of the conduct file.

During proceedings. A partial referral (a single issue, a fixed timetable, a circumscribed stay) is almost always preferable to a global referral, which means giving up a forum already seised.

• The mediation and arbitration centre is established byarticle 35 de l’accord of the UPC Agreement. Its seats are Ljubljana and Lisbon.
• A patent cannot be revoked or limited by mediation or arbitration: the award has no effect erga omnes.
• Three sets of rules were put out to consultation in 2025: mediation, arbitration and expert determination.
• Rules 11 and 365 of the Rules of Procedure organise the referral and the judicial confirmation of settlements.
• The convention de New York de 1958 gives the award worldwide enforceability, where the Unified Patent Court stops at the Contracting Member States.
• Proposing or refusing a precise arbitration offer is an element in assessing conduct under Huawei v ZTE.
• The cost variable is the value of the case before the court, and time consumed before the arbitrator.

This article is an original, expanded adaptation of an analysis by Matthieu Dhenne first published on Kluwer Patent Blog.