(iic_2012_001247_1 1.21)

The Essential Facilities Doctrine – What Was Wrong inMicrosoft?** The essential facilities doctrine is designed to oblige dominant undertakingsto make available their important facilities, including intellectual propertyrights, for other undertakings. It requires a delicate balance of, on the onehand, protecting the exclusivity of ownership and on the other hand en-couraging other undertakings’ incentive to innovate. The balance that wasnicely struck in the previous cases nevertheless was abandoned in the Micro-soft judgment. In that case, the General Court made two mistakes: First, itwrongly defined the primary market as client PC operating systems that wasin no accordance with the request of Sun, i. e. the interoperability informa-tion with Windows client PC operating system. This broader market defini-tion made the General Court struggle in justifying the increasing marketshares of Linux products through unnecessarily expanding the scope of‘‘eliminating all the competition’’ from the requesting undertakings to ‘‘elim-inating all effective competition’’. Secondly, the General Court improperlyinterpreted ‘‘new products’’ hindered by a refusal to grant a license as alsoincluding products with ‘‘technical development’’. This arbitrary extensionencroaches upon the very substance of the system of intellectual propertyrights.
The essential facilities doctrine (EFD) is designed to oblige dominant under-takings to supply a product that is necessary to third-party undertakings.1Since an obligation to deal is in principle contrary to the two enshrinedprinciples of free competition, i. e. freedom of contract and exclusivity ofownership,2 its application is always subject to strict requirements. In theEU, it is commonly accepted that the EFD should be assessed under threeconditions within the context of Art. 102 of the Treaty on the Functioning ofthe European Union (Art. 102): (i) the requested product is indispensible tothe product to be supplied by the requesting undertaking; (ii) the denial risks * Dr. jur.; Associate Professor, KoGuan Law School, Shanghai Jiao Tong University, China.
** The author thanks Prof. Peggy Valcke, Prof. Jules Stuyck, Prof. Pierre Larouche, Prof.
Wouter Devroe, Prof. Alexandre de Streel, and Prof. Robert Queck for their valuable com-ments. All mistakes remain the author’s.
1 This article will nevertheless not discuss the situation where a dominant undertaking sup- plies one group of customers while refusing the equivalent request of others. This can bedealt with under the category of discrimination. See, e.g., Case C-27/76, United Brands v.
Commission, [1978] ECR 207; and Case T-301/04, Clearstream v. Commission, [2009]ECR II-3155.
2 Alison Jones & Brenda Sufrin, ‘‘EU Competition Law: Text, Cases and Materials’’ 480 et seq, (4th ed., Oxford, Oxford 2011).
Electronic copy available at: http://ssrn.com/abstract=2025777 eliminating all the competition from the requesting undertaking; and (iii)there are no objective justifications from the requested dominant undertak-ing.3 In addition, when the requested product is an intellectual property right(IPR), or, in other words when it is a case of refusing to grant an IPR license,one more condition is included, i. e. that the denial risks hindering theemergence of a new product.4 These conditions were applied under a delicatebalance in cases preceding the Microsoft judgment.5 However, this balancewas renounced in the Microsoft judgment delivered by the General Court(GC) (previously the Court of First Instance (CFI)). In this case, Microsoftrefused the request of its competitors to disclose the interoperability infor-mation with its Windows operating systems. The GC made several contro-versial interpretations on the EFD, which in the end resulted in Microsoft’sobligation to supply the interoperability information to its competitors.
Given the controversies in Microsoft, this article serves a dual purpose: first,it observes the common practice of the European authorities in the EFDcases prior to Microsoft; secondly, and most importantly, it deliberates onthe mistakes committed by the GC in Microsoft based on those previouscases. In pursuit of these two purposes, the subsequent five parts are dedi-cated respectively to summarizing the application of fives conditions of theEFD in the pre-Microsoft era: (1) the definition of relevant markets; (2) theindispensability test; (3) elimination of competition; (4) prevention of theemergence of a new product; and (5), no objective justifications. The nextpart compares the analyses in Microsoft with those previous cases in order todemonstrate the errors made by the GC. The last part provides some conclu-sions.
The EFD does not disallow all kinds of refusal to deal, but essentially onlyprohibits refusal to deal in the form of leveraging of market power. There-fore, in all EFD cases the first step is to define two relevant markets: onecontaining the primary product, i. e. the product being requested, where therequested undertaking is dominant, and the other comprising the secondaryproduct, i. e. the derivative product that will be supplied by the requestingundertakings on the top of the requested product.6 Although no literaturedenies the existence of the two-market feature in EFD cases, most, if not all,try to neglect the difficulty of defining two such relevant markets. Manyauthors simply refer to the Commission’s notice on the definition of therelevant market.7 However, defining two relevant markets for the purpose ofthe EFD is considerably different from that in other cases. In the following, 3 Guidance on the Commission’s enforcement priorities in applying Art. 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings, 2009 O. J. C 45/7, para. 81.
4 Joined cases C-241/91 P and C-242/91 P, Radio Telefis Eireann (RTE) and Independent Television Publications Ltd (ITP) v. Commission, [1995] ECR I-743, para. 74.
5 Case T-201/04, Microsoft v. Commission, [2007] ECR II-3601.
6 Ibid., at para. 335.
Electronic copy available at: http://ssrn.com/abstract=2025777 it will be discussed how the two relevant markets (the primary market andthe secondary market) should be defined.
As far as the primary product or the requested product is concerned, EFDcases can be divided into two types. First, the primary product has alreadybeen marketed by dominant undertakings for a while, and the refusal tosupply nevertheless takes place where those dominant undertakings decideto monopolize the secondary market and thus disrupts all the supply to thirdparties on the secondary market. Secondly, the primary product has neverbeen supplied to any third-party undertakings and thus always remains partof those dominant undertakings’ final products.
In the first type of cases, the definition of relevant markets is not so differentfrom that in non-EFD cases. It is true that the transactions on the primaryand secondary markets may not exist at the time of antitrust action. How-ever, competition authorities can determine the product and geographicdimension of the two relevant markets according to past transactions.
Furthermore, the requested undertaking cannot deny the existence of such amarket since there was a group of long-standing third-party customers, i. e.
requesting undertakings. The European courts, based on the previous trans-actions, was confronted with no difficulty in defining the primary market ascontaining raw material (aminobutanol) in Commercial Solvents,8 advertis-ing minutes in Telemarketing,9 spare parts for Hugin’s cash registers inHugin,10 and Hilti-compatible cartridge strips in Hulti.11 However, the definition of relevant markets in the second type of cases iscomplicated. Since the primary product has never been supplied to a thirdparty, it is always sold in a bundle with the final product. Consequently, anobligation to supply the primary product is similar to unbundling the finalproduct, with the purpose of creating an intermediate product (i.e. theprimary product) that never existed. The difficulty can be observed from twoperspectives. First, there is no transaction concerning this intermediary pro-duct. Due to no evidence of demand and supply-side substitution, the princi-ples to define relevant markets are not easily applicable here. Secondly, itmay be argued that competition authorities can always define the primarymarket according to the denied request. Nevertheless, a problem may arisewith regard to the ‘‘unbundleability’’ of the requested product. Since theremay be cases where it is commercially or technically complementary to 7 Commission notice on the definition of the relevant market for the purposes of Commu- nity competition law, 1997 O. J. C 372.
8 Jointed Cases C-6-7/73, Commercial Solvents, [1974] ECR 223.
9 Case C-311/84, Centre belge d’études de marché – Télémarketing (CBEM) v. SA Compa- gnie luxembourgeoise de télédiffusion (CLT) and Information publicité Benelux (IPB),[1985] ECR 3261.
10 Case C-22/78, Hugin v. Commission, [1979] ECR 1869.
11 Case C-53/92 P, Hilti v. Commission, [1994] ECR I-667.
maintain such a bundle, it is unwise for competition authorities to define theprimary market solely based on the denied request.
Regarding the definition of the primary product that has never been mar-keted separately, the European Court of Justice (ECJ) in IMS provided twohints that ‘‘it is determinative that two different stages of production may beidentified,’’12 and ‘‘there is an actual demand for them on the part of under-takings which seek to carry on the business for which they are indispensi-ble’’.13 However, the ECJ’s statement is still ambiguous. The first sentenceimplies that competition authorities should focus on the request of thoserequesting undertakings to see whether the requested product can be un-bundled from the final product. However, in a sense, there is nothing thatcannot be unbundled. Therefore, the question of unbundleablity in essenceturns into how far competition authorities can go. No clear implications canbe found within the statement of the ECJ. Moreover, the second sentencerequires competition authorities to examine whether the requesting under-taking has an actual demand for the requested product. The purpose of thisstatement is also uncertain. Does it suggest that the requesting undertaking isobliged to make pre-investment? If so, how sufficient should this pre-invest-ment be? In addition, it is not observed in other cases that the ECJ’s methodwas further clarified either by the European courts or by the EuropeanCommission (Commission).
Nevertheless, in two other cases the question of whether the requestedproduct can be unbundled was discussed in detail, though the ECJ’s sug-gested method was not followed. The first case was FAG.14 FAG (FlughafenFrankfurt/Main AG) owned and operated Frankfurt airport, and it reservedto itself all the services within the airport. This case was filed as abusing adominant position by third parties that were denied access by FAG to airportfacilities in order to provide ground-handling services on the ramp15 in theairport. The ground-handling services on the ramp were always provided byFAG itself and bundled with the provision of airport facilities for the landingand take-off of aircraft. FAG claimed that the two types of services werecomplementary and should not be provided separately. The Commissionhowever considered that they constituted separate markets for three reasons:(1) from the demand side, the services relating to the provision of facilitiesfor the landing and takeoff of aircraft were distinct from and not inter- 12 Case C-418/01, IMS v. NDC, [2004] ECR I-5039, para. 45.
13 Ibid., at para. 44.
14 IV/34.801, FAG/Flughafen Frankfurt/Main AG, 1998 O. J. L 72/30.
15 They consisted of the provision and operation of equipment for the embarkation and dis- embarkation of passengers, transport of passengers between the terminal and the aircraftposition and vice versa, crew transport, loading and unloading of baggage, cargo andmail, transport, sorting and transfer of baggage, transport of cargo and mail on the ramp,cabin cleaning, toilet and water services, push-back/towing of aircraft, provision and ope-ration of equipment to carry out the above activities, fuelling of aircraft, and the transportof catering supplies to and from the aircraft. Ibid., at para. 20.
changeable with ramp-handling services – moreover, airport customers wereusually charged with two separate bills for the two services, which impliedthat it was not necessary to purchase the two services from the same sup-plier; (2) from the supply side, the provision of airport landing and take-offfacilities and the provision of ramp-handling services were significantlydifferent and can indeed be provided by different suppliers;16 and(3) ground-handling services were provided by independent third parties inother airports.17 The second case was Ricoh.18 The applicant of this case, Info-Lad, was amanufacturer of toner for photocopiers. It requested Ricoh, a photocopiermanufacturer, to supply empty toner cartridges for Ricoh’s machines. SinceRicoh only sold filled cartridges and never provided empty ones, the Com-mission was confronted with the question whether there could be a separatemarket for Ricoh’s empty cartridges. It was concluded that there was nosuch a separate market for empty cartridges because first there was noconsumer demand, and secondly as a matter of fact all manufactures ofphotocopiers only supplied filled cartridges.
There are two major differences between the two cases. First, in FAG theservices of providing facilities for the landing and takeoff of aircraft werehorizontal to the ramp-handling services. In other words, they were notupstream-downstream services. In comparison, empty cartridges and filledcartridges in Ricoh were vertically related: empty cartridges can be consid-ered as an upstream production stage while filled cartridges as a downstreamproduction stage. Secondly, in FAG the two services were unbundled in othercountries; whereas, in Ricoh all manufacturers only sold filled cartridges.
The IMS court only held that that ‘‘it is determinative that two differentstages of production may be identified’’, and did not specify those differentstages as only horizontal productions. In addition, there was no evidencethat the vertical relationship between empty cartridges and filled cartridgesplayed a pivotal role in the Commission’s refuse to define empty cartridgesas a separate market in Ricoh. Consequently, one of the most likely justifica-tions to explain the difference between the two cases turns out to be the lackof practice in unbundling empty cartridges.
This article is of the opinion that relying upon the existing unbundlingexperience to decide whether a requested product can constitute a separatemarket strikes an appropriate balance between respecting input holders’ownership and encouraging competition from third-party undertakings. Itremains those requested undertakings’ own business strategies to bundle therequested product with their final products since this may be regarded by therequested undertaking as the best solution to recoup its investment. It is thusintrusive for competition authorities to unbundle any production stage as a 16 Ibid., at para. 65.
17 Ibid., at para. 66.
18 1 Competition Policy Newsletter 35–37 (February 1999).
separate market. If so, competition authorities would substitute themselveswith the requested undertaking to draft business plans. Moreover, defining aseparate market for the requested product that has never been marketedseparately also changes the existing market structure because there was nosuch intermediary market beforehand. This may result in irreparable regula-tory failure, and impair the interests of requested undertakings, as well astheir own incentive to innovate. Last but not least, competition law aims topreserve competition on the market, rather than to create competition.
Relying upon the existing unbundling practice can thus decrease regulatoryrisks to the minimum.
Furthermore, the EFD represents not only the most controversial topic incompetition law but also an overlap between competition law and sector-specific regulation. When applied under relaxed conditions, the EFD can beused to deal with issues that traditionally fall within the scope of sector-specific regulation. For example, the Commission even went one step furtherto define a requested product as a separate market even with no existingunbundling practice in competition-law cases for the purpose of sector-specific regulation. This was the case for access services in the electroniccommunications sector in the 1990s. In the Notice on the application of thecompetition rules to access agreements in the electronic communicationssector, the Commission referred to the possibility to oblige incumbent elec-tronic communications operators to provide their access networks to alter-native operators based on the EFD.19 Apparently, at the very beginning ofliberalizing the electronic communications sector there was no practice ofunbundling access networks from electronic communications networks andservice as a whole. This goes beyond the conclusion made in the previousparagraphs. This extension was criticized by scholars such as Larouche, asbeing inconsistent with the legitimacy model of the EFD,20 and was also oneof the reasons to adopt the current electronic communication regulatoryframework that integrates competition law principles and methodologiesinto electronic communications regulation.21 Under the current electroniccommunications regulatory framework, the Commission is under no pres-sure to employ its competition law power for liberalization purposes.
Furthermore, similar activities can also be observed within the transportmarket22 and the energy market23 where statutory monopolies were pre- 19 Commission Notice on the application of the competition rules to access agreements in the telecommunications sector – frameworks, relevant markets and principles, 1998 O. J.
C 265/2.
20 Pierre Larouche, ‘‘Competition Law and Regulation in European Telecommunications’’ 21 Alexandre de Streel, ‘‘The Integration of Competition Law Principles in the New Euro- pean Regulatory Framework for Electronic Communications’’, 26 World Competi-tion 489–514 (2003).
22 Case COMP/37.685, GVG/FS, 2004 O. J. L 11/17.
23 Case COMP/39.402, RWE Gas Foreclosure, (not reported).
viously present. However, they should be considered as a special applicationof the EFD and has no general impact on other industrial sectors.
The secondary market within the EFD cases comprises the final products tobe supplied by the requesting undertakings. Since the EFD concerns cases ofleverage of market power, there is no need to prove that the requested under-taking enjoys a dominant position on the secondary market, as established inTetra Pak II.24 Nevertheless, it may be disputed whether the final product tobe supplied by the requesting undertakings can constitute a separate marketon its own merits. This question was raised in Commercial Solvents wherethe Commission defined ethambutol as constituting the secondary market.
The requested undertaking nevertheless argued that there was no such aseparate market for ethambutol because it was a part of a larger market foranti-tuberculosis drugs. The ECJ held that it was not necessary for thederivative market to constitute a self-contained market so long as it can bedifferentiated from the primary product.25 It is clear here that the ECJ onlyrequired the requesting undertaking to produce a derivative product thatwas distinct from the primary product possessed by the requested under-taking. There was no further requirement that the requesting undertakingmust produce something different from all the products supplied by therequested undertaking. However, it should be noted that Commercial Sol-vents was a case of refusing to supply tangible facilities. When the refusal todeal involves an IPR, another condition is added: the secondary product ofthe requesting undertaking must be a new product and thus not falling intothe product range of the requested undertaking. Although the new productshould be different from what are currently offered by the requesting under-taking, it does not necessarily constitute a self-contained market in pursuit ofthe principle established in Commercial Solvents. In other words, it is notnecessary for the new product to be a breakthrough innovation in an abso-lute sense.
After having appropriately delineated the primary and secondary markets,and having identified that the requested undertaking enjoys a dominantposition on the primary market, competition authorities should examinewhether the conditions to analyze the EFD are fulfilled. The first conditionrequires an assessment of the relationship between the primary product andthe secondary product; more specifically, whether the primary product isindispensible to the secondary product.
24 Case C-333/94 P, Tetra Pak International SA v. Commission, [1996] ECR I-05951, 25 Commercial Solvents, supra note 8, at para. 22.
According to the case law related to the EFD, the indispensability test entailstwo elements. First, the primary product should be objectively necessary forthe requesting undertaking to carry out their activities on the secondarymarket. Normally the analysis for this element involves less difficulty. Com-mon knowledge about the industrial sector concerned is normally sufficientto allow competition authorities to determine whether the primary productis necessary. This can be observed from many cases, such as the raw materialversus the finished products in Commercial Solvents;26 TV advertising min-utes versus telemarketing in Telemarketing;27 spare parts versus maintenanceand repair services in Hugin;28 airport facilities versus ramp-handling inFAG;29 railway faculties versus international passenger service in GVG;30the copyright over TV programs versus TV magazine publishing weekly TVguides in Magill;31 and the copyright on ‘‘1860 brick structure’’ versusservices analyzing pharmaceutical sales information based on the ‘‘1860brick structure’’ in IMS.32 Furthermore, it is also not difficult to find that theprimary product is not necessary for the secondary product. For example, inLadbroke, the applicant, a Belgian company making books on betting horseraces abroad, requested the Pari Mutuel group to grant the right to broadcasthorse races in France.33 It is obvious that broadcasting horse races is notnecessary, though perhaps helpful, for Ladbroke to carry out its gamblingservices.
Secondly, there should be no economically viable substitutes on the primarymarket.34 This point was first brought forward in Commercial Solvents. Therequested undertaking claimed that there were actually possible substituteson the primary market. However, the ECJ pointed out that those substituteswere either in an experimental stage or only resulted into a modest produc-tion that could only satisfy self-needs.35 Consequently, it was held that thosesubstitutes were only of minor importance, and it was not possible for therequesting undertakings to ‘‘have recourse on an industrial scale to methodsof manufacture of (the requested product) based on the use of different rawmaterials’’.36 It is clear here the ECJ did not aim to find a monopolizedprimary market in an absolute sense, but only referred to no effectivesubstitutes. This argument can also be found in Telemarketing. The re-quested undertaking was a broadcasting company, and was granted an 26 Ibid.
27 Telemarketing, supra note 9.
28 Hugin, supra note 10.
29 FAG, supra note 14.
30 GVG/FS, supra note 22.
31 Magill, supra note 4.
32 IMS, supra note 12.
33 Case T-504/93, Tiercé Ladbroke SA v. Commission, [1997] ECR II-923.
34 Magill, supra note 4, at paras. 52–53; and Case C-7/97, Oscar Bronner v. Mediaprint, [1998] ECR I-7791, at paras. 44–45.
35 Commercial Solvents, supra note 8, at para. 15.
36 Ibid., at para. 16.
exclusive right to run a French-speaking TV station in Belgium. As a matterof fact, there were also other French-speaking TV channels from Francebroadcasting in Belgium. However, since those stations were aimed onlyrarely or not at all at the Belgium public, they were not considered aseffective substitutes.37 This point was further developed in Bronner where the ECJ maintained thatindispensability should be established not only by the fact that there were noeffective substitutes on the primary market,38 but also by the fact that therewere no technical, legal or even economic obstacles to make it impossible, oreven unreasonably difficult, for the requesting undertaking to produce asubstitute, either alone or in cooperation with others on the secondarymarket.39 It should be noted that the possibility of creating a potentialsubstitute referred not to the possibility of a small company, but to ahypothetical company with a comparable size to the requested undertak-ing.40 Consequently, the ECJ implies that the EFD is not designed for theconvenience of undertakings to free ride dominant undertakings, but onlyfor the necessity of survival on the secondary market in situations wherethere are no effective substitutes.
This article applauds the ECJ for its development in Bronner. As mentionedin the preceding section, the EFD concerns an area where competition lawand sector-specific regulation overlap. When the conditions of the EFD areloosely applied, competition law can be stretched to invade the territory ofsector-specific regulation. Therefore, a difficult question is to which extentcompetition law must limit itself in the application of the EFD. Whenobliging an undertaking to supply a product that it does not intend to sell toothers, competition law should not go beyond the situation where (i) it isobjectively necessary for the secondary product to rely on the primaryproduct; and (ii) there are no substitutes, either actual or potential, on therelevant market. It is reasonable that dominant undertakings are under noobligation to provide any of their products to third parties. An obligation todeal should limit its application to what are indispensible for the productionof the requesting undertakings. Moreover, no actual or potential substituteson the relevant market suggest that the requested undertaking enjoys at leasta super-dominant position on the primary market, and most importantly thisstrong position will not easily disappear in a short term by itself. Thisstrongly justifies the intervention of competition law, and at the same timesignificantly lowers the risk of regulatory failure.
37 Telemarketing, supra note 9, at para. 6.
38 Bronner, supra note 8, at para. 43.
39 Ibid., at para. 44.
40 Ibid., at paras. 45–46.
The second condition states that the refusal to supply risks eliminating allthe competition on the requesting undertakings. European courts maintainthis condition as separated from the first condition that the requested pro-duct is indispensible to the derivative product to be produced by the request-ing undertaking. However, it is in practice impossible to distinguish the twoconditions. As supported by some scholars,41 the risk of eliminating allcompetition is an inevitable consequence of the indispensability of therequested product to the derivative product.
As a matter of fact, European courts have never carried out specific analysisto examine whether the refusal to deal resulted in a possibility of eliminatingall competition on the secondary market. In Commercial Solvents, the ECJonly referred to the fact that the sales of the derivative product from therequesting undertaking stopped after the disruptive supply of the requestedproduct.42 In Telemarketing, it was apparent that the requesting undertakingcould not continue to provide its telemarketing service without acquiringbroadcasting minutes from the requested undertaking that was a legal mono-poly on the TV broadcasting market.43 The same can also be observed in theEFD cases involving IPRs such as in Magill and IMS, where the provision ofthe secondary product must necessarily infringe the requested undertaking’scopyrights on the primary markets.
Furthermore, the indispensability test requires that there are no effectivesubstitutes on the primary market. Once this condition has been fulfilled, itis reasonable to imagine that the requesting undertakings cannot acquire thenecessary inputs for their production on the secondary market. It hence goeswithout any doubt that their surviving possibility on the secondary marketdiminishes.
All in all, the second condition of ‘‘eliminating all the competition on thepart of the requesting undertakings’’ indeed becomes a consequence of theindispensability test.
5 Prevention of the Emergence of a New Product There is always a dispute over whether the EFD should apply to IPRs. Amajor reason for such a debate lies in the fact that the purpose of IPRs is togrant a monopoly that is nevertheless balanced with a limited time period. Adiscussion of whether it is appropriate to apply the EFD to IPRs goes beyondthe scope of this article. Nevertheless, as already mentioned, Europeancourts did not neglect the significance of such a temporary monopoly withrespect to innovation. Therefore, a new condition is included in the EFD 41 Net Le, ‘‘What does ’capable of eliminating all competition’ mean?’’, 26 European Com- 42 Commercial Solvents, supra note 8, at para. 25.
43 Telemarketing, supra note 9, at paras. 26–27.
analysis when IRPs are involved, which is that the refusal to grant a licensemust be able to prevent the emergence of a new product.44 This condition first appeared in Magill.45 However, two questions were leftunanswered in that judgment. First, it did not clarify whether this conditionis additional or alternative to the indispensability test. This uncertainty ledto the fault interpretation in the subsequently case Ladbroke that [t]he refusal to supply the applicant could not fall within the prohibition laiddown by Article 86 (now Art. 102) unless it concerned a product or servicewhich was either essential for the exercise of the activity in question, in thatthere was no real or potential substitute, or was a new product whoseintroduction might be prevented, despite specific, constant and regular po-tential demand on the part of consumers.46 [emphasis added] The ‘‘either-or’’ phase implies that the Ladbroke court considered that theprevention of the emergence of a new product was alternative to the indis-pensability test. This interpretation was even taken on by the Commission inits IMS decision where the it stated that ‘‘there is no requirement for a refusalto supply to prevent the emergence of a new product in order to be abu-sive’’.47 This fault interpretation was corrected subsequently by the ECJ inIMS. The court articulated that the prevention of the emergence of a newproduct for which there was potential consumer demand was one of thecumulative conditions alongside with the indispensability test, eliminatingall the competition and no objective justifications.48 The second question left in Magill was whether the new product, or thederivative product, should be different from that supplied by the requestedundertaking. This point was not touched upon by Ladbroke, but was finallyanswered in IMS where the ECJ firmly maintained that a refusal to licensean indispensible IPR was abusive only where the undertaking which re-quested the license did not intend to limit itself essentially to duplicate thegoods or services already offered on the secondary market by the owner ofIPRs, but to produce new goods or services not offered by the owner of IPRsfor which there was potential consumer demand.49 In addition, as alreadypointed out in the second part of this article, it is not necessary for the newproduct to constitute a self-contained product. In other words, it is possiblethat the new product falls into a broader market definition so long as it isnot provided by the requested undertaking.
44 This additional condition is nevertheless question by other authors as unreasonable. See, e.g., Andreas Heinemann, ‘‘Compulsory Licences and Product Integration in EuropeanCompetition Law – Assessment of the European Commission’s Microsoft Decision’’, 36IIC 63–82 (2005.) 45 Magill, supra note 4, at para. 54.
46 Ladbroke, supra note 33, at para. 131.
47 Case COMP D3/38.044, NDC Health/IMS Health, (not reported), para. 180.
48 IMS, supra note 12, at para. 38.
49 Ibid., at para. 49 Since a refusal to supply is governed by Art. 102, the requested undertakingis allowed to justify its refusal. Nevertheless, it should be noted that anobligation to deal is already an exception to the general principle of freedomto contract and exclusivity of ownership. The justification to decline such anobligation is nothing less than an exception to another exception. It is thusunderstandable that not many requested undertakings will succeed in thisjob because the previous analyses already take into account the requestedundertaking’s interests. European courts have only hinted that possible justi-fications may include capacity limits in supply,50 technical or commercialrequirements that make the supply impossible,51 and protecting consumers’interests from being seriously or irreparably damaged.52 However, as amatter of fact, only few requested undertakings could convince the Europeancourts or the Commission at this stage.
The GC’s 2007 Microsoft judgment concerned two abusive activities ofMicrosoft: (i) refusing to provide third-party producers of workgroup serveroperating systems with the interoperability information with Windows clientPC operating systems; and (ii) tying windows media player to its client PCoperating systems for the purpose of excluding other media players. Thesecond behavior is not relevant to this article. The first abuse was filed bySun to the Commission as Microsoft refused to disclose the information andtechnology necessary to allow its workgroup server operating system tointeroperate with the Windows client PC operating system. After investiga-tion, the Commission decided that Microsoft infringed Art. 102 by refusingto supply interoperability information.53 Later, the Commission decisionwas appealed to the GC, and the GC dismissed the application.54 Microsoftinitiated no further action with the ECJ afterwards and the case came to anend. In this judgment the GC ignored the delicate balance struck in theprevious cases from several aspects. This lack of attention stirred fiercedebate in academic circles. In the following, this article aims to make itscontribution to this debate by examining two mistakes committed by theGC: the first is related to the problematic definition of the primary market;and the second concerns the arbitrary extension of the ‘‘new product’’ to alsocover products with ‘‘technical development’’.
50 Commercial Solvents, supra note 8, at para. 28; and Case C-77/77, BP v. Commission, 51 Telemarketing, supra note 9, at para. 26.
52 Case T-184/01 R, IMS v. Commission, [2001] ECR II-3193, para. 148.
53 Case COMP/C-3/37.792, Microsoft, (not reported).
54 Microsoft, supra note 5.
As far as the definition of relevant product markets are concerned, the GC,in accordance with the previous cases, defined two relevant markets. Theprimary market comprised of client PC operating systems and a secondarymarket containing workgroup server operating systems.55 As a piece ofcommon knowledge, Microsoft enjoyed a quasi-monopoly position on theprimary market. Moreover, it was also found to hold a dominant position onthe secondary market.
Nevertheless, the manner to define the primary market in Microsoft isconsiderably different from that in other EFD cases, and remains the firstmistake made by the GC. In the previous cases that supported requestingundertakings, the primary markets were always defined strictly based on therequest of those requesting undertakings. Where the requested products havenever been marketed separately from the final products provided by therequested undertakings, the authorities had to investigate whether the re-quested product can actually be unbundled and thus tailored the definitionof the primary market to the needs of those requesting undertakings.56 Thisanalysis involves a deliberation on the question of whether it is more com-plementary to maintain the bundle than to break it. In all those, cases boththe European courts and the Commission always carried out their analysesby investigating in the first place whether the requested product couldconstitute a separate market. For example, in Bronner the ECJ examinedwhether the home-delivery service could be considered as a product market,though it finally left the answer to the national court.57 In FAG, the Com-mission concluded that it was reasonable to separate the ground-handlingservice on the ramp from the whole products.58 In those IPR-related EFDcases, the problem of unbundling is less important because the secondproduct must be a new product in comparison with the requested IPR. Forinstance, in Magill the primary product was the weekly TV program infor-mation provided by TV stations, while the secondary product was a compre-hensive weekly TV guide;59 in Ladbroke the primary market was the trans-mission of sound and pictures of French horse races and the secondaryproduct was the horse-race gambling service;60 and in IMS the primaryproduct was the copyright held by IMS on its ‘‘1860 brick structure’’whereas the derivative product was the presentation of regional sales data.61 55 Microsoft Decision, supra note 53, at paras. 342 and 401.
56 Some authors even criticize that the Commission may place excessive focus on the request of requesting undertakings in order to enforce the obligation to deal. See, Michael D.
Diathesopoulos, ‘‘The Relation between Essential Facilities Doctrine and Market Defini-tion’’ (2010), available at SSRN: http://ssrn.com/abstract=1732147 (last visited 16 Decem-ber 2011).
57 Bronner, supra note 34, at paras. 32–35.
58 FAG, supra note 14, at paras. 64–68.
59 Case T-69/89, RTE v. Commission, [1991] ECR II-485, para. 62.
60 Ladbroke, supra note 33, at paras. 81–87.
On the other hand, a market definition that is defined broader than therefused request is usually a sign that the requesting undertaking cannotacquire the primary product. An example can be observed in Ricoh.62 Coming back to the Microsoft judgment, what was requested by Sun wasinteroperability information with the Windows client PC operating system.63However, the primary product was defined by the GC as client PC operatingsystems. This broader market definition did not cause trouble for the GC tofind Microsoft’s dominant position on the primary market, which was differ-ent from Ricoh where the Commission could not confirm that Ricoh had adominant position on the market for consumables for all photocopy ma-chines after refusing to define a narrower primary market for consumablesspecifically for Ricoh’s machines.64 However, this market definition resultsin a gap in the GC’s reasoning for the subsequent analyses. By defining abroader primary market the GC neglected the analysis on whether interoper-ability information could be, as a matter of fact, unbundled from the Win-dows client PC operating system. As discussed earlier, the issue regardingunbundleablity serves the first balance in dealing with EFD cases since anobligation to deal cannot be imposed on a product that is better to bebundled to the final product. Furthermore, this article notices no difficultyconfronted by the GC to define the primary product as the interoperabilityinformation. The second part of this article observes that the definition ofthe primary market is usually based on unbundling practices on the relevantmarket. As shown in the judgment, during most of the 1990s, Microsoftactually granted a license relating to the disclosure of portions of the Win-dows source code (versions before Windows 2000) to AT&T, which devel-oped a product capable of enabling interpretability with Windows products.
AT&T then licensed its product to other companies, including Sun.65 Thismay be considered as an unbundling practice and was relied upon by the GCto define the primary market as the market for the required interoperabilityinformation with the Windows client PC operating system.
This broader market definition furthermore led to two consequences. First,the primary market, which should have been tailored to the needs of therequested undertaking, i. e. the interoperability information for the Windowsclient PC operating system in this case, was not defined due to the broadermarket definition. The reason may be the difficulty in defining a proper levelof interoperability. However, the Commission, in order to implement thejudgment, must specify that information in any case. A lack of definition ofthe level of interoperability in the prohibition decision put the Commissionin an awkward position in the implementing stage, where it had to adopt along list of implementing decisions. This has never been observed in other 61 IMS, supra note 12, at para. 23.
62 Ricoh, supra note 18.
63 Microsoft, supra note 5, at para. 2.
64 Ricoh, supra note 18.
65 Microsoft, supra note 5, at para. 429.
cases. Secondly and most importantly, this broader market definition inevi-tably included operators active on the secondary market that did not neces-sarily need the interoperability information with the Windows client PCoperating system. The inclusion of those operators made the GC struggle incarrying out its analysis on the subsequent conditions for the EFD analysis,in particular the indispensability test and elimination of all competition onthe part of requesting undertakings. To make matters worse, this was thevery reason why the GC was obliged to extend the application of those twoconditions, which will be elaborated in following two sections.
In the previous cases, the indispensability test sought for effective substituteson the primary market. In Microsoft, since the primary product was definedas client PC operating systems, the GC should have looked for substitutes forthe Windows client PC operating system. However, what was requested bySun was not a client PC operating system, but the interoperability informa-tion with the Windows client PC operating system. Therefore, when analyz-ing the indispensability test, the GC had to change the subject matter for itssubstitute-seeking analysis to the interoperability information, which madethe analysis of the indispensability test inconsistent with the definition of therelevant markets.
Furthermore, the devastating effect of the broader market definition did notend here. The interoperability information with the Windows client PCoperating system was a matter of degree.66 Different providers of workgroupserver operating systems need various degrees of interoperability. Withoutadding the required level of interoperability into the definition of the pri-mary market, the GC apparently included all providers into its analysis.
Indeed, Microsoft was a quasi-monopoly on the market for client PC operat-ing systems and imposed a de facto standard for workgroup computing.
Thus, most providers of workgroup service operating systems required ahigh level of interoperability with Windows products. Moreover, it wasimpractical to obtain the interoperability information via other methods dueto technical and time limits. This can be observed by the fact that Microsoft,though claiming that there were five alternative methods to achieve inter-operability with the Windows client PC operating system,67 did not denythat none of those methods or solutions made it possible to achieve the highdegree of interoperability required by the Commission.68 Therefore, it maybe established that the interoperability information with the Windows clientPC operating system was in general essential for those undertakings.
However, there were still some providers on the derivative market that didnot need the level of interoperability required by the Commission. This in 66 Ibid., at para. 15867 Ibid., at para. 346.
68 Ibid., at para. 435.
particular concerned Linux products. Without directly acquiring the inter-operability information from Microsoft, Linux products on the secondarymarket nevertheless managed to continue increasing their market shares.69 Itwas true that Linux products were mainly used for tasks such as Webserving, firewall serving and for mission-critical applications,70 and hencecould not offer a full-fledged set of services covering the sharing of filesstored on servers, the sharing of printers and the administration of groupsand users.71 This nevertheless explained why Linux products required lessinteroperability information. Most importantly, it exposed the defect of notdefining the primary market as the required interoperability information.
Since not every piece of information related to interoperability was necessaryto Linux products, the indispensability between the primary and secondaryproducts in Microsoft was not solidly established. Even the GC could notgive a reasonable explanation why without the interoperability informationLinux providers could still increase their market shares. The GC only statedthat ‘‘the growth of Linux products on the workgroup server operatingsystems market was only modest’’.72 Had the primary product been definedas the interoperability information with the Windows client PC operatingsystem at the level specified by the Commission, those Linux products shouldhave been excluded from the analysis of the indispensability test, and thusthe GC’s analysis would be more tenable.
In the previous cases the analysis of elimination of all competition over-lapped the indispensability test. Since there are no other effective substituteson the primary markets, the requesting undertakings cannot carry out theiractivities on the secondary market. However, it is striking in Microsoft thatthe GC replaced the concept of ‘‘eliminating all the competition’’ with‘‘eliminating all effective competition’’.73 Based on such a change, the GCaccepted limited presence of fringe competitors, in this case, Linux produ-cers. In the following paragraphs, it will be analyzed that such a change isunnecessary and the reason is still the problematic definition of the primarymarket.
The very controversy in Microsoft was that when the Commission decisionwas adopted there were still a large number of competitors active on thesecondary market where Microsoft only had around a 60% market share.74Microsoft’s competitors seemed not to be eliminated in a short term. Such amarket situation was not really consistent with the risk of eliminating allcompetition. This may be the reason why the GC extended the scope of this 69 Ibid., at para. 432.
70 Microsoft Decision, supra note 53, at para. 598.
71 Microsoft, supra note 5, at para. 26.
72 Ibid., at para. 432.
73 Ibid., at para. 563.
74 Ibid., at para. 33.
condition from eliminating all competition to eliminating all effective com-petition. Therefore, the following paragraphs focus on the question whetherthose competitors can actually be excluded from the secondary market dueto the refusal to grant a license.
First, those competitors can be divided into two groups: (i) competitors thatdemand a high level of interoperability with Windows products, for exampleSun; and (ii) competitors that only need a low level of interoperability, suchas Linux producers. For the first group of competitors, as appropriatelypointed out by the GC, their presence was largely due to the fact that manyconsumers were still using older versions of Windows products and theinteroperability problem mainly came out when Microsoft released its Win-dows 2000 product line.75 It is foreseeable that those competitors wouldgradually lose their market shares on the derivative market to Microsoftwith more and more consumers updating to Windows 2000 products or evennewer versions. Consequently, it can indeed be concluded that this condition,even according to the standard established in the previous cases, was fulfilledin relation to the first group of competitors.
However, as far as the second group of competitors is concerned, Linuxproducers, as mentioned above, continued to increase their market shares.
There was no evidence that they would be excluded from the secondarymarket very soon, even without the interoperability information. The GCdid not deny this point, but referred to some forecasts that Linux productswould not become effective competition.76 However, the underlying reasonfor the success of Linux products, though ‘‘modest’’ as described by the GC,was due to the fact that the level of interoperability required by the Commis-sion was not indispensible to their production. Therefore, the refusal tolicense did not eliminate their competition.
Accordingly, the extension to eliminating all effective competition can onlyfind its added value with regard to providers that did not need the level ofinteroperability desired by the Commission, and most importantly was notrelevant to the request of the applicant in this case, i. e. Sun. Had the primaryproduct been defined as the interoperability information, the fringe competi-tion from Linux products should have been excluded immediately. Further-more, it was also not necessary for the GC to expand the scope of thiscondition. In addition, once the primary market had been defined appropri-ately, it would have been not necessary for the GC to refer to the fastgrowing speed of Microsoft on the secondary market in order to prove thateffective competition had been gradually weakened.77 This gives a confusingimpression that a dominant undertaking may be punished simply because ofits extraordinary success.
75 Ibid., at para. 429.
76 Ibid., at para. 567.
77 Ibid., at paras. 569–573.
7.4 Preventing the Emergence of a New Product In the previous cases, a refusal to grant an IPR license may be consideredabusive only when it prevents the development of the secondary market tothe detriment of consumers. This is for the purpose of balancing the interestand economic freedom of IPR owners against the interest in protecting freecompetition from other undertakings. Thus, the requesting undertakings,after acquiring such a license, should not limit themselves essentially toduplicating products already offered by the owner of an IPR, but to supply-ing new products that are currently not offered by that owner and for whichthere is potential consumer demand.78 Turning back to Microsoft, it is quite likely that the requested undertaking,Sun, after obtaining the interoperability information, would still supply aworkgroup server operating system similar to Microsoft’s own product.
Based on the analyses in the previous cases, this condition cannot be satis-fied. Nevertheless, the Microsoft judgment changed completely such aninterpretation. Thereby the GC held that [t]he circumstance relating to the appearance of a new product [. . .] cannotbe the only parameter which determines whether a refusal to license anintellectual property right is capable of causing prejudice to consumerswithin the meaning of Article 82(b) EC (now Art. 102). As that provisionstates, such prejudice may arise where there is a limitation not only ofproduction or markets, but also of technical development.79 If it can still be argued that the extension of Microsoft regarding the indis-pensability test and eliminating all competition only contains minor mistakesthat did not invalidate the GC’s whole conclusion, the extension related tothe condition of preventing the emergence of a new product cannot beforgiven. The interpretation in the previous cases struck a proportionatebalance between protecting the IPR owners’ interest and encouraging otherundertakings’ incentive to innovate.80 The purpose of IPRs is to allow theirowners to enjoy monopoly for a limited time period in order to fully exploittheir IPRs. It is probably not reasonable to protect the IPR owner’s interestwhen other undertakings would like to develop a product based on that IPRthat is on the one hand not provided by the IPR owner and on the otherhand is refused to license. Under such a circumstance competition law isjustified to intervene. However, this balance is broken in Microsoft. The GCinterpreted the new product not only in the sense of a new market but also aproduct with technical development. Consequently, different from the pre-vious cases that only protected requesting undertakings’ interest in creating anew market, Microsoft also cherishes differentiation on the derivative mar-ket. This makes the EFD principle encroach into the very substance of the 78 IMS, supra note 12, at paras. 48–49.
79 Microsoft, supra note 5, at para. 647.
80 Kung-Chung Liu, ‘‘Rationalising the Regime of Compulsory Patent Licensing by the Es- sential Facilities Doctrine’’, 39 IIC 757–774 (2008).
IPR rules, i. e. IPR owners’ willingness to exploit their rights solely for theirown benefits.
Furthermore, the concept of ‘‘technical development’’ is very vague. Thismay cause difficulties in some cases. Suppose A holds patent X, and basedon X it introduces a medicine that can cure flu in eight hours. In the firstscenario, B would like to acquire X to produce an effective medicine forcancer. Within both the previous cases and Microsoft, it is very likely that Awould be forced to grant such a license to B. In the second scenario, B desiresto obtain this patent in order to also produce a medicine that neverthelesscan cure flu in two hours. This seems to deliver technical development, andthus according to Microsoft, A might possibly be obliged to license to X.
However, a complicated issue would be raised in the third scenario where Bintends to produce a medicine that can cure flu in seven hours.81 It is unclearin this case whether the GC’s interpretation can lead to the conclusion that Ashould also license to X. Assuming that the answer is yes, IPR holders areput in a dangerous position since in the future they may be forced to licensetheir IPRs even if the requesting undertakings only intend to produce similarproducts with minor modification in competition with them.
Some authors claim that when making such an extension the GC takes intoaccount a specific feature of the software sector, namely that incrementalinnovation is as important as lumpier breakthrough innovations.82 However,this may lead to a sector-specific competition law, and is against the generalunderstanding that competition law serves as an industry-wide regulation.
This article is hence of the opinion that the extension goes far beyond theoriginal purpose of the EFD.
The previous cases prove that once the other conditions have been fulfilled itis very difficult for the requested undertaking to justify their refusal to deal.
Regarding this point, Microsoft did not make any difference. The GC firstdeclined Microsoft’s claim that the technology concerned was covered by theIPR, and thus the refusal was reasonable.83 Consistent with the previouscases, the analysis of the other conditions already takes into account thespecial features of IPRs, and accordingly the protection granted by IPRscannot be regarded as a justification at this stage. Nevertheless, it is interest-ing to notice that the GC also rejected Microsoft’s argument that an obliga-tion to license would impair its incentive to innovate. The GC’s reasons weremainly two fold. First, Microsoft merely put forward vague, general and 81 Bo Vesterdorf, ‘‘Article 82 EC: Where do we stand after the Microsoft judgement?’’, 1 Global Antitrust Review 1–14 (2008).
82 Pierre Larouche, ‘‘The European Microsoft case at the crossroads of competition policy and innovation’’ (2008), available at: http://ssrn.com/abstract=1140165 (last visited16 December 2011).
83 Microsoft, supra note 5, at paras. 690–691.
theoretical arguments on this point.84 Secondly, it was normal practice foroperators to disclose to third parties the interoperability information withtheir products and such disclosure can allow those operators to make theirown products more attractive and therefore more valuable.85 However, itremains the IPR owners’ decisions to decide how to deal with their proper-ties. By referring to the normal practice, the GC intruded into the IPRholders’ business strategies.
Since the EFD represents an exception to the free will of ownership, itsapplication must be subject to strict restrictions. Regarding this point, thecase law in the pre-Microsoft era strikes a delicate balance of, on the onehand, protecting the exclusivity of ownership, and on the other hand incen-tives to innovate from other undertakings. This can be first observed fromthe definition of the primary market where the European authorities focuson the demand of the requesting undertakings to assess whether the re-quested product can be unbundled from the final product supplied by therequested dominant undertakings. Once the primary and secondary marketshave been defined, the European authorities subsequently seek for effectivesubstitutes on the primary market in order to establish the indispensabilityof the primary product to the secondary product. This indispensability test isimplemented in such a strict way that the subsequent analysis of eliminatingall competition from the requesting undertakings becomes a natural conse-quence. At the final step, the requested dominant undertaking is allowed tojustify its refusal, though in practice they seldom succeed at this stage.
Where a refusal to deal involves an IPR, a more delicate balance is reachedby adding a new condition to the above analyses. This condition is that therefusal to grant a license should risk of hindering the introduction of a newproduct. In pre-Microsoft era, this new product is required to be able toconstitute a new market where the requested dominant undertaking is notactive. This requirement on the one hand gives due respect to the will of IPRholders to exploit their rights fully to their own benefit, and on the otherhand encourage other undertakings to bring innovation in addition to theproducts provided by the requested dominant undertaking. This limits theharm to IPR holders and the market to the minimum.
However, the Microsoft judgment brings down all those appropriate bal-ances. The GC first made a technical mistake in defining the primary marketwithout accordance with the need of the requesting undertaking. While therequesting undertaking demands interoperability information with the Win-dows client PC operating system, the primary market is defined as client PCoperating systems. This broader market definition is firstly not consistentwith the subject matter of the subsequent indispensability test, and secondly 84 Ibid., at para. 698.
85 Ibid., at para. 702.
includes competitors, in particular those providing Linux products, thatdepend to a lesser extent on the interoperability information required by theCommission. The consequence is that the GC struggles in justifying in itsanalysis on the increasing market share of Linux products even without theinteroperability information provided by Microsoft by expanding the condi-tion of ‘‘eliminating all the competition’’ to ‘‘eliminating all effective compe-tition’’. This extension would not have been necessary if the primary producthad been defined as the interoperability information with the Windowsclient PC operating system.
The second mistake committed by the GC is related to the new condition forIPR-related EFD cases. While the previous cases establish a proportionatebalance by requiring that the secondary product must constitute a newmarket, Microsoft acknowledges that products with technical developmentcan also satisfy this condition. An obligation to grant a license under suchcircumstances may force IPR holders to support their direct competitorswith their exclusive IPRs. This in essence encroaches upon the very substanceof the IPR system, which creates a temporary legal monopoly over an IPR.
The Microsoft judgment nevertheless gives an impression that an IPR holderhas to change their monopolized IPR into ‘‘open source’’ simply because oftheir exceptional achievements.

Source: http://cclp.sjtu.edu.cn/userfiles/1/files/2012711173851850.pdf