What Lukoil means for the refusal to deal doctrines
The Court of Justice delivered its judgment in Lukoil as 2025 was coming to an end. It was the second case in the year that engaged head on with the conditions set out in Magill and Bronner.
It may be true that this judgment is not as consequential as the one in Android Auto for the present and future of the refusal to deal doctrines. This said, Lukoil marks a subtle but clear move away from the traditional rationale underpinning the application of these doctrines.
What makes the legal development even more interesting is that it appears to expand the scope of application of Magill and Bronner, where as Android Auto decisively reduced their scope.
Traditionally (and as recently as 2021, when the Court ruled in Slovak Telekom), the Magill and Bronner doctrines were justified on two grounds. Compelling a firm to deal with third parties interferes with fundamental rights and must therefore be confined to exceptional circumstances. What is more, it can be expected to negatively affect the undertaking’s incentives to invest and innovate.
This original rationale no longer reflects the reality of the case law. In Android Auto, the Court held that requiring a firm to deal with third parties with which it has chosen not to deal is not necessarily confined to the refusal to deal doctrines. Magill and Bronner will only be relevant, moving forward, when compelling the sharing of an asset would ‘fundamentally alter‘ the dominant undertaking’s ‘economic model‘.
Accordingly, a firm running a partially open platform can no longer invoke the refusal to deal doctrines by virtue of the fact that it has chosen to open other markets, not necessarily concerned by the dispute, to third parties.
Lukoil raises a different set of issues. One of the questions asked by the Bulgarian Administrativen sad Sofia-oblast was, in essence, whether the refusal to deal doctrines are applicable where the development of the infrastructure has been supported by means of State aid.
The answer would be clear if one were to look at the traditional rationale behind Magill and Bronner. Where State aid is involved in the roll-out of the infrastructure, the traditional concerns with firms’ incentives to invest and innovate are not present.
The award of public funding is an alternative, and mutually exclusive, approach to deal with the incentive-related issues. To the extent that it is, one could reasonably claim that public funding rules out, by definition, the application of Magill and Bronner.
One could, in fact, go further and argue that the award of State aid does not just exclude the application of the refusal to deal doctrines, but requires that the subsidised infrastructure be shared with third parties.
This is, after all, the (very reasonable) position taken by the European Commission in its Guidelines on State aid for broadband networks, where it holds that ‘[e]ffective wholesale access for third parties to the funded networks is an indispensable condition of any State aid measure‘.
The Court of Justice, however, did not follow the logic of the traditional case law. According to the judgment, the applicability of Magill and Bronner does not hinge on whether the infrastructure was built with public support, but whether the owner of the infrastructure acquired it ‘at a price and under conditions resulting from a competitive procedure‘.
As a result of this choice, a dominant firm controlling an asset can validly invoke Magill and/or Bronner even when its incentives to invest and to innovate would not have been negatively affected by an obligation to deal with third parties.
For the same reason, it is inevitable to conclude that the scope of the refusal to deal doctrines has been expanded to comprise scenarios where it would not have been applicable, at least in light with the twofold rationale provided by the Court in Slovak Telekom.
The conclusion seems clear: the scope of refusal to deal doctrines seems to have shrunk in digital scenarios, whereas it has expanded in the realm of public utilities (telecommunications, energy, gas, rail), where the incumbent often exploits infrastructures rolled out with public support.(if not outright ownership).
In a sense, the case law appears to have moved in line with enforcement priorities.
Chillin'Competition 
Dear Pablo, I do notfollow you on this one.
What the Court has correctly argued here is that for the incentives to invest in an infrastructure it makes no difference whether the dominant firm has build the infrastructure itself or whether is has acquired the infrastucture at a competitive market price. This makes perfect economic sense, both as regards the incentive to build respectively acquire the infrastructure as for the incentives to maintain/expand/improve the infrastructure.
The incentive question would be different in case the building of the infrastructure was done with public funds or supported by state aid and the subsequent privatisation and sale of the infrastructure was not done at a competitive price (‘not done at a price and under conditions resulting from a competitive procedure’). In that case the firm acquiring the infrastucture would effectively be subsidised. If it would subsequently argue that having to give access would undermine its incentives to invest, those arguments should indeed be taken with a pince of salt.
Assuming that Lukoil bought the infrastructure at a competitive price and did not receive subsidies later, questions of fact on which the Court did not have to rule, the judgment makes perfect sense and does not, as claimed by you, expand the scope of the refusal to deal doctrines. Basically, it shouldn’t matter, for the application of the law, whether the dominant firm bought the infrastructure at a competitive price from a state/authority or from another private firm.
Finally, but this is a different point, I also do not agree that with Android Auto the Court has shrunk the scope of refusal to deal doctrines in digital scenarios. The scope was shrunk with Teliasonera and Slovak Telekom. Those two judgments made clear that the Bronner conditions (only) apply in case of a denovo refusal and not in case the dominant firm is or has been giving access (as, for instance, in case of termination of supply and margin squeeze). Android Auto is just following that logic. This may in practice be particularly relevant for the digital context, but that is simply because those platforms start to be open and accessible and the termination or squeezing starts later when the platform has become dominant and it is money time.
Luc Peeperkorn
5 February 2026 at 9:10 pm
Thanks, Luc!
Just to say that Android Auto was very much about a de novo refusal (and rightly analysed by the Italian authority as such). Which goes to illustrate that it has shrunk the scope of the doctrines
Pablo Ibanez Colomo
6 February 2026 at 6:58 am
Pablo, thanks for the quick reaction. However, I again beg to differ.
In Android Auto the judgment makes clear that Google offered interoperability “for entire categories of apps in the form of templates for each interoperability solution”. When Enel X Italia asked for an interoperability template for its type of app, Google refused to provide or develop such a template, arguing it made such templates only available for multimedia and messaging apps.
My interpretation is that the Court did not see this as a denovo refusal (and I would think rightly so). In the words of the Court: “where that platform has not been developed by the undertaking in a dominant position solely for the needs of its own business.”
So the only possible shrinking of Bronner I see is that it doesn’t matter – for the Bronner conditions not to apply – whether access is refused while access was given before for exactly the same use of the input or for similar/other uses of the requested input.
Luc Peeperkorn
6 February 2026 at 12:08 pm
Thanks again, Luc!
It is sufficient to take a look at the facts in Magill and Bronner (para 8) to realise the shift in the case law. As in Android Auto, the dominant firms in Magill and Bronner were dealing with rivals in other markets unconcerned by the refusal at issue in the case (in Magill, the listings were licensed to newspapers prior to the refusal ; in Bronner, the firm printed and distributed another newspapers when the access request was put).
In spite of this fact, indispensability was deemed necessary in both. Had Magill and Bronner been examined under the approach introduced in Android Auto, they would have been decided differently (neither the copyright – in Magill – nor the distribution platform – in Bronner – had been developed ‘solely for the needs of [the dominant firms’] own business’).
It is something I discuss in this paper (on which your comments would be very much welcome): https://ssrn.com/abstract=5648051
Best wishes
Pablo Ibanez Colomo
6 February 2026 at 1:35 pm
Disclosure: I work in-house at Google and Android Auto is one of my cases.
Pablo refers to one element in the Android Auto CJEU ruling which leaves me perplexed. The Court says that the strict Bronner criteria are not applicable to situations where an infrastructure/platform owner already deals with 3rd parties. Yet this was exactly what the infrastructure owner Mediaprint did in the Bronner case: in addition to distributing their own newspaper they also distributed a rival newspaper Wirtschaftsblatt. One of Bronner’s claims was therefore discrimination between Bronner’s newspaper Der Standard – which Mediaprint refused to distribute – and the rival newspaper Wirtschaftsblatt which Mediaprint already distributed.
If the Court had applied the new Android Auto law to Bronner, Mediaprint would have abused dominance by refusing Bronner’s access to its platform which it had already opened to a 3rd party. It hadn’t developed the platform solely for its own needs.
Regarding the “open to 3rd parties” element: Android Auto had developed templates for two categories of apps which were the most widely used at the time of Enel’s request to develop a template for an EV charging app. EV charging apps at the time could theoretically be used by 0.04% of electric car drivers in Italy (many who had an iPhone and could not use Android Auto). As it took 10 months for the Android Auto team to develop a template for EV charging apps, one can question how “open to 3rd parties” the platform was at the time. There are hundreds of possible app categories each requiring their own template to reduce driver distraction.
Compare this with Bronner. In Bronner Mediaprint had an existing distribution network for newspapers in Austria. Bronner asked for their newspaper to be distributed through that network in Austria. Mediaprint could legally refuse this access.
In Android Auto everyone had free access to the two templates Google had developed. Enel requested access to something which did not yet exist. In Bronner this would have meant Bronner asking Mediaprint to distribute, as an example, large parcels instead of newspapers. If Mediaprint had said that their network is not equipped to handle large parcels and that they have to invest in different types of vehicles and distribution centers to manage that, they would have been required to do so under Android Auto. Even if the demand for large parcels had been under 0.04% of the volume they distribute.
Tero
18 February 2026 at 12:52 pm
Dear Tero, like Pablo in his reply to me, you are referring back to the facts of the Bronner case to argue that also in Bronner the (supposedly) dominant firm Mediaprint had not developed its distribution activities solely for its own needs, as it not only distributed the newspapers it published itself but also a competing newspaper. However, it is relevant to note that this newspaper, while not published by Mediaprint, was not only distributed but also printed by Mediaprint, while Bronner requested access to only the home delivery part of Mediaprint’s distribution activities.
While I agree it is useful to go back to the facts of old cases, I see important differences with Android Auto, even more if one takes into account the underlying principles of the case law.
In Bronner the Court was trying to preserve the incentive to invest, also by dominant firms, in infrastructures. The fact that Mediaprint offered the combination of printing and full distribution services to another rival newspaper was not considered discriminatory, as Bronner did not want access to Mediaprint’s printing services combined with full distribution services. The Court considered that a refusal to give access to Mediaprint’s home delivery system would only be abusive if that access could be deemed indispensable to compete on the Austrian newspaper market. In all other circumstances the Court considered it would be better for competition if competitors would keep the incentive to invest in their own infrastructures instead of relying on each others’ investments.
With Teliasonera and Slovak Telekom the Court made clear that the strict Bronner conditions do not apply in case the dominant firm was already giving access to the requesting firm or other competitors on the downstream market. The underlying logic seems to be that it did not consider the risk of negative effects on the incentives to invest in infrastructures to be as strong in such ‘termination and margin squeeze’ cases, the previous supply indicating there was apparently a commercial logic for the dominant firm to give access.
The question now is how the incentives issue applies in the digital platform cases. My understanding is that the Court considers these platforms as infrastructures which are created with the intention and expectation to be open, i.e. to allow suppliers and buyers (in Android Auto app producers and app users) to meet on whatever downstream markets that may develop. To preserve and protect the incentives to invest it is consistent in that context for the Court not to apply the Bronner conditions, as these incentives are based on the platform being open and accessible. So, to revert the case law, I guess a platform will have to convincingly argue that when investment decisions to create and expand the platform were taken, this was not done to create an open platform with open downstream markets but to create a platform to be used exclusively by itself on (certain) downstream markets.
Luc Peeperkorn
24 February 2026 at 4:59 pm