Impact of Cloud Computing on Permanent Establishments Under the OECD Model Tax Convention

I. INTRODUCTION

Cloud computing raises very difficult issues for many areas of law, including both domestic and international tax law. This article presents problems and difficulties generated for the international tax regime by business activity based in the cloud. It does that by analyzing the impact of cloud computing on the concept of the permanent establishment (PE) under the Organisation for Economic Co-operation and Development (OECD) Model Tax Convention on Income and on Capital (OECD Model). The article’s primary goal is to highlight how cloud computing renders the existing PE concept under the Model Convention increasingly obsolete, by showing that it is incapable of fulfilling its primary purposes of acting as a threshold and as a profit center. This is demonstrated by assessing how Article 5 and Article 7 of the OECD Model apply to business activities conducted by multinational enterprises (MNEs) in the cloud and to income generated by those activities. The overarching theme of the article consists in highlighting the difficulties generated by cloud computing for different aspects of the PE concept and in analyzing potential responses to them.

The article first elucidates the most salient features of cloud computing and its wide-ranging implications (Part II). It then introduces the PE concept, along with its most important characteristics and explains how and why it has developed into its present form. Over the next two parts the article critiques the existing provisions of the OECD Model that constitute the core of the PE regulation regime — Article 5 and Article 7 — as inadequate in the rapidly evolving economic reality and as increasingly incapable of dealing with the challenges posed by the cloud (Parts III and IV).

Lastly, the article presents and evaluates various reform proposals aimed at remedying the relevant weaknesses of the international tax regime (Part V). It concludes by putting forward a pragmatic solution based on a limited (non-unitary) formulary apportionment method, which, whilst not a perfect remedy, has the benefit of practicability and should be able to deal with some of the most important shortcomings of the existing arrangements.

II. THE RISE OF CLOUD COMPUTING

This Part is aimed at introducing basic cloud computing concepts (Part II.A) and it discusses some of the economic consequences of increasing prevalence of cloud-based services (Part II.B) in the global economy. Part II.C provides an overview of implications that the cloud has for domestic (mostly U.S.) as well as international tax law.

A. Phenomenon Explained

Cloud computing (also referred to as the “cloud”) has somewhat inconspicuously become one of the major developments in the history of computing. ¹Sean Marston, Zhi Li, Subhajyoti Bandyopadhyay, Juheng Zhang and Anand Ghalsasi, Cloud Computing — The Business Perspective, 51 Decision Support Sys. 176, 176 (2011). The main benefits of cloud computing for business consist largely in delivering the functionality of traditional information technology at reduced costs, increased IT efficiency and improved business agility. ²Id. at 177. What is also often emphasized is that the computing power of the cloud is easily scalable and can be accessed on an “as-needed” basis. The National Institute of Standards and Technology defines cloud computing as “a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management and effort or service provider interaction.” ³Peter Mell & Timothy Grance, U.S. Department of Commerce, the NIST Definition of Cloud Computing, Recommendations of the National Institute of Standards and Technology, Special Publication 800-145 (Sept. 2011) (the “NIST Report”), at 2.

The models for providing cloud-based services have traditionally been divided into three categories: Software as a Service (SaaS), Platform as a Service (PaaS) and Infrastructure as a Service (IaaS). SaaS offers ready-made applications for end-users. PaaS is more like an operating system, such as Microsoft Windows Azure, which allows the client to design applications with appropriate programming tools. IaaS is the most open-ended solution, where the end-user has the largest degree of control over the technology in which an organization outsources the equipment used to support operations, including storage, hardware, servers and networking components. ⁴What Is Infrastructure as a Service (IaaS)? http://searchcloudcomputing.techtarget.com/definition/Infrastructure-as-a-Service-IaaS.

But the recent technological developments in cloud infrastructure have resulted in this division no longer being exhaustive. Many providers begin to offer a new cloud service known as integration Platform as a Service (iPaaS). iPaaS is a cloud integration platform, enabling connectivity to SaaS and other cloud services, and it offers a platform for SaaS users and cloud vendors to build and host packaged integration solutions which ensure that the data provided is available in a synchronized fashion across mobile, social and online mechanisms. ⁵iPaaS Integration Platform as a Service, http://www.mulesoft.com/resources/cloudhub/ipaas-integration-platform-as-a-service. Another emerging model which is also becoming increasingly popular is the Business Process as a Service (BPaaS) model. ⁶Business Process as a Service, http://www.gartner.com/it-glossary/business-process-as-a-service-bpaas. It delivers business process outsourcing (BPO) services that are sourced from the cloud and constructed for multi-tenancy. ⁷“Multi-tenancy is a reference to the mode of operation of software where multiple independent instances of one or multiple applications operate in a shared environment. The instances (tenants) are logically isolated, but physically integrated. The degree of logical isolation must be complete, but the degree of physical integration will vary.” Multitenancy Definition, http://www.gartner.com/it-glossary/multitenancy. It is important to note that these categories are not mutually exclusive and different models can be easily combined. ⁸Anne Welsh, Curt Kinsky, Nick Ronan & Mark Klitgaard, Can Clouds Change Shapes? Transfer Pricing Considerations for Cloud Computing, 64 Tax Notes Int’l at 147, 147 (2011). In fact, as discussed in Part II.C the bundling of different rights within one contract for cloud services gives rise to serious income classification problems.

The development of cloud computing has also far-reaching implications for the ways in which we think about and design computer networks. The orthodox view on provision of computing appliances and services has been based on applications being accessed through software and data stored locally with the end-user. Cloud computing has drastically changed this model because now data and applications are steadily migrating to data centers away from individual users and their computers. ⁹Christopher S. Yoo, Cloud Computing: Architectural and Policy Implications, Institute for Law & Economics, Univ. of Penn. Law Sch. Research Paper No. 11-15 (Apr. 30, 2011), at 3. Available at http://ssrn.com/abstract=1824580. Compare, for example, Microsoft Word run locally on a laptop and Google Docs accessed via a web browser. One of the major effects of this change is that the functions that must be performed by end-user machines become much less cumbersome. What is now needed is only a so-called “thin client” capable of furnishing network connectivity and computing power sufficient to run a relatively unsophisticated interface. ¹⁰Id. at 4.

Aside from the unquestionable benefits associated with the cloud, however, there are also certain dangers and challenges generated by the massive migration of data into this new vehicle for data processing. Insofar as privacy and security of stored data are concerned, questions have been raised about increased exposure to privacy invasion and fraud by hackers ¹¹Christopher Soghoian, Caught in the Cloud: Privacy, Encryption, and Government Back Doors in the Web 2.0 Era, 8 J. on Telecomm. and High Tech. Law, 359, 361. For a more positive outlook see Cloud Security Concerns Are Overblown, Experts Say (Feb. 27, 2014) http://www.computerworld.com/s/article/9246632/Cloud_security_concerns_are_overblown_experts_say. and the uncertainty about the scope of protection afforded by the relevant data protection regulations. ¹²See W. Kuan Hon, Christopher Millard & Ian Walden, The Problem of ‘Personal Data’ in Cloud Computing — What Information Is Regulated? The Cloud of Unknowing, Part 1, 4 Int’l Data Privacy Law, at 211 (2011). The issue of accountability of cloud service providers has also attracted significant attention, largely because of the asymmetry of information about the true level of data protection and security offered by those entities. ¹³See Jens Prüfer, How to Govern the Cloud? Characterizing the Optimal Enforcement Institution That Supports Accountability in Cloud Computing, TILEC Discussion Paper DP 2013-022 (Dec. 10, 2013). Available at http://ssrn.com/abstract=2365713. Overall, there is a steadily growing number of areas of law that are being affected by the expansion of cloud-based services. As with many other technological breakthroughs which have significant impact on economic growth, lawmakers must engage in a careful balancing exercise to ensure that appropriate control measures are deployed while preserving the innovation’s capacity to generate economic growth. With this theme in mind, I now turn to discuss the influence that cloud computing has had on the ways in which business is done under this new paradigm.

B. Economics of Cloud Computing: The Business Side

The impact of cloud computing can be observed at both micro- and macroeconomic levels. Starting with the micro level, the most important benefits on the end-user side include lower maintenance costs, slowing down of hardware replacement cycle and obviation of need to maintain server and storage capacity. ¹⁴Yoo, above n. 9, at 3–4. Another benefit is closely related to the scalability of the service. Without the need to maintain fixed storage capacity the end-user incurs lower sunk costs and benefits from increased flexibility, which allows him to use the cloud’s capacity according to idiosyncratic changes in demand. Yet even greater benefits are perhaps derived from the aggregation of demand for computing capacity. The reduction in the variability of demand that results from aggregation helps individual enterprises to achieve higher efficiency of hardware utilization than they would otherwise have achieved individually due to the fact that their traffic would often peak at different times. ¹⁵Id. at 10.

On the macroeconomic level it is predicted that the cloud is likely to extend IT-induced economic growth in mature economies while also contributing positively to growth in emerging economies which have not yet developed advanced IT infrastructure. ¹⁶Marco Iansiti and Gregory L. Richards, Economic Impact of Cloud Computing White Paper, Working Paper (June 30, 2011). Available at http://ssrn.com/abstract=1875893. Professors Iansiti and Richards predict that small and medium-sized business segments are most likely to create new demand for IT investments related to cloud computing. ¹⁷Id. at 21. The reason for this is that the cloud places many IT services within those companies’ financial reach because of its relatively low costs as compared to traditional IT infrastructure. ¹⁸Id. at 23. It is also expected that cloud computing will significantly lower barriers to entry for start-ups and small companies by facilitating acquisition of IT in a way that will quickly create a positive return on investment. ¹⁹Id. at 30.

On the other hand, there are also potential costs associated with widespread deployment of the cloud. Most importantly, because the cloud necessarily involves deploying applications and data over infrastructure provided by a third party, the user might have to incur additional costs associated with ensuring security and compliance. ²⁰Shyam Kumar Doddavula, Ira Agrawal & Vikas Saxena, Cloud Computing Solution Patterns: Infrastructural Solutions, in Cloud Computing, Methods and Practical Approaches, 202 (Zaigham Mahmood ed., 2013). Ensuring compliance is particularly relevant for cloud providers in the context of international taxation, mainly because of high uncertainty surrounding the classification of income generated by means of this technology. Outside of taxation context some of the regulatory measures which might necessitate monitoring of compliance include the Sarbanes-Oxley Act, the Payment Card Industry Data Security Standard and Health Insurance Portability and Accountability Act, ²¹Id. at 214. and also many international regulations such as OECD Guidelines on the Protection of Privacy and Transborder Flows of Personal Data. ²²OECD Guidelines on the Protection of Privacy and Transborder Flows of Personal Data (2013). Available at http://www.oecd.org/sti/ieconomy/2013-oecd-privacy-guidelines.pdf.

C. Cloud Computing — An Overview of Implications for Domestic and International Tax Law

I shall now briefly consider the impact of developments in cloud computing highlighted in the preceding sections on domestic U.S. tax law and the international tax order more generally. Whilst this article focuses on the interaction between the cloud and permanent establishments, the former also affects many other areas of tax law. This broader picture will serve as a helpful background for the more detailed analysis presented in Parts III, IV and V.

The fundamental difficulty which necessitates adjustment of existing tax laws to deal with the challenges posed by a cloud-based economy stems from the interplay between tax law and the cloud, but does not in itself create a new or unfamiliar category of problems. As has been the case in the past, the difficulty with applying old laws to new economic phenomena stems from the growing divergence between legal rules that were conceived and developed in a different era and the supervening economic reality that such rules were not designed to regulate. The sometimes impressive flexibility and adaptability of those rules notwithstanding, they often turn out to be no longer capable of serving the purposes for which they were constructed. For example, in the context of regulation of free speech, the hitherto relatively uncontroversial concept of non-newsworthiness has become very problematic because the Internet has made it impossible to segregate audiences based on immediate relevance of information being revealed to any given community. ²³Geoffrey Stone, Privacy, the First Amendment, and the Internet, in The Offensive Internet, 192 (S. Levmore and M. Nussbaum eds., 2010). As for tax law, some of the most difficult problems generated by cloud computing arise in the context of income classification, transfer pricing and residence rules. These are now considered in turn.

Insofar as income classification is concerned, the divergence between legal rules and economic reality that they are supposed to apply to can be very clearly seen when one looks at U.S.-source rules relating to foreign income for the income tax purposes. ²⁴Internal Revenue Code (the “Code”) Chapter 1, Subchapter N, Part I: §861–§865. All section references herein are to the Code, as amended, or the Treasury regulations thereunder, unless otherwise indicated. Income from sales is generally sourced to the residence of the seller; ²⁵§865. income from personal services is, as a rule, sourced to where those services are performed; ²⁶§861(a)(3). and royalty income is sourced to where the income-generating property is located. ²⁷§861(a)(4). As Professor Reuven Avi-Yonah already pointed out more than 15 years ago, the growing volume of e-commerce applies significant pressure to the distinctions drawn above. ²⁸Reuven S. Avi-Yonah, International Taxation of E-Commerce, 52 Tax L. Rev. 507, 541 (1996-1997). But this pressure has intensified even further with the advent of cloud computing. In order to ascertain the proper government tax take, income has to be classified as falling within one of the categories enumerated above before any further determinations are made. In the case of income derived from transactions involving computer software, classification was already challenging when most purchases involved a transfer effected by downloads. From the perspective of U.S. tax source rules, this was most likely to be the case when payments were contingent upon the sale, productivity or use of the property, which in turn meant that the payments instead of being treated as sale price payments could have potentially been treated as royalties. ²⁹Rufus V. Rhoades and Marshall J. Langer, Rhoades & Langer, U.S. Int’l Tax’n & Tax Treaties §25.12. Classification problems also affect the applicability of the relevant income classification regulations such as the §1.861-18 regulations (hereinafter “U.S. Software Regulations”). The U.S. Software Regulations, drafted in 1998, do not explicitly refer to a situation in which a user can access a computer program without an antecedent physical delivery or download and so the scope of their applicability is becoming increasingly uncertain. ³⁰Welsh et al., above n. 8, at 4. The classification problem is important insofar as approaching provision of SaaS services could be treated as a transfer whereby the cloud would be treated as just another medium for effecting transfer. Alternatively, the Software Regulations could be deemed inapplicable on the basis of there being no transfer; transfer pricing rules on controlled services transactions (§482 Service Regulations) would apply instead. Id. at 4.

In the context of e-commerce generally, the classification process is far from straightforward because the income-generating payments are usually made in return for a bundle of services and rights, but this is particularly problematic when it comes to transactions carried out in the cloud. Here the distinction between different categories of income sources is blurred even further. Consider the following case concerning the need to distinguish between taxable and non-taxable services for the purposes of New York state law. ³¹The case is quoted in: Walter Hellerstein and Jon Sedon, State Taxation of Cloud Computing: A Framework for Analysis, 117 J. of Tax’n 11, 30 (2012). A taxpayer provides a service to ensure safe storage of customers’ data, enabling them to replace the data if the original data is lost. The software provided by the taxpayer would function only in conjunction with the taxpayer’s service. The New York Department of Taxation and Finance ruled that even though the taxpayer was not selling software, it nevertheless was making a taxable use of the software in the course of providing its non-taxable service. ³²Id. at 20.

Similar problems arise in other jurisdictions. For example, under domestic Australian tax law, payments for services rendered are not royalties unless the services are subsidiary and ancillary to the right to use copyright or the supply of know-how. ³³John Walker, Taxing the Cloud: Australia, 70 Tax Notes Int’l 165, 170 (2013). The determination for these purposes, whether services rendered in the cloud are of ancillary character, especially when they are in the form of iPaaS which serves as an integration platform, will be very difficult. The right to use copyrighted software transferred to the taxpayer at an earlier stage and then subsequently integrated via iPaaS will make it very difficult to determine whether the iPaaS service is of ancillary character and whether it should be treated as royalty. In this case such determination will be of large significance because if a payment is classified as a royalty under the Australian tax law but not under the applicable income tax treaty, then no royalty withholding ought to apply to such payment. ³⁴Id. at 170. Given how hard it is to draw the distinction between ancillary and non-ancillary services in such a case there is a high likelihood that Australian tax authorities might take a different approach than tax authorities of other contracting states that have tax treaties with Australia. This, in turn, is likely to lead to costly litigation over the proper classification of income. ³⁵For an overview of problems arising in other jurisdictions, in that case Singapore, see Harvey Koenig & Kah Chuan Ho, Taxing the Borderless Cloud Within the Singapore Border, 18 Asia-Pacific Tax Bulletin, at 5 (2012).

The second area in which cloud computing is particularly troublesome is transfer pricing. OECD identifies transfer pricing treatment of intangibles as “a key pressure area” with respect to its Transfer Pricing Guidelines ³⁶OECD, Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations (July 22, 2010) (hereinafter 2010 OECD Guidelines). and believes that there may be a propensity among MNEs to associate more profit with intangibles and thus transfer the risk within the group, in effect decreasing the proportion of profits associated with “substantive” economic activity. This is particularly troublesome insofar as it gives rise to the phenomenon of Base Erosion and Profit Shifting (hereinafter “BEPS”). ³⁷OECD, Action Plan on Base Erosion and Profit Shifting (2013) (hereinafter “BEPS Action Plan”). BEPS consists principally of arrangements that achieve no or low taxation by shifting profits away from the jurisdictions where the activities creating those profits take place, and whereby due to gaps between different tax systems, and sometimes due to the application of bilateral tax treaties, income from cross-border activities may go untaxed anywhere, or be only unduly lowly taxed. ³⁸Id. at 10. Whilst the BEPS Action Plan does not explicitly mention cloud computing, through its impact on transfer pricing the cloud can facilitate precisely the sorts of activities that have a propensity to increase BEPs, as will be argued in Parts IV and V. ³⁹The discussion of implications of BEPS is deferred to Part V.C.

Last, provisions governing tax residence present another example of rules that are affected by cloud computing. Article 4 of the OECD Model ⁴⁰OECD, Model Tax Convention on Income and on Capital (July 22, 2010) (hereinafter “OECD Model”). contains so-called “tie-breaker” provisions aimed at resolving competing claims of parties to tax treaties with respect to individual and corporate residence. ⁴¹Arthur Cockfield, Walter Hellerstein, Rebecca Millar and Christophe Waerzeggers, Taxing Global Digital Commerce (2013), 168. The relevant corporate test is one of “place of effective management.” In order to deal with the difficulty that arises in case of dispersed management making decisions through, say, video conferencing, the OECD Technical Advisory Group ⁴²OECD Technical Advisory Group on Monitoring the Application of Existing Treaty Norms for the Taxation of Business Profits, The Impact of the Communications Revolution on the Application of the ‘Place of Effective Management’ as a Tie Breaker Rule (Discussion Draft 2011). has put forward a reform proposal which would allow the countries to resolve dual corporate residence on a case-by-case basis. ⁴³Cockfield et al., above n. 41, at 169. The proposed Article 4(3)(b) would read: ⁴⁴Id.

[I]f the state in which its place of effective management is situated cannot be determined or if its place of effective management is in neither State, it shall be deemed to be a resident only of the State [Option A: with which its economic relations are closer] [Option B: in which its business activities are primarily carried on] [Option C: in which its senior executive decisions are primarily taken].

 

In the context of this proposal, it has been claimed that by putting emphasis on substantive economic nexus under Options A and B the proposed Article 4(3)(b) attempts to make it more difficult to manipulate corporate residency insofar as e-commerce is concerned. ⁴⁵Id. But when considered in light of the cloud-based activities this proposition becomes problematic for at least two reasons. First, in case of a cloud service provider it is not entirely clear what should be determinative of Option A’s “closeness” of economic relations. Should it be location of the servers or perhaps location of customers? Given the decentralized structure of the cloud it is easy to imagine an enterprise with multiple servers located in different jurisdictions serving customers in yet another set of locations. Second, the same dilemma is also inherent in Option B, because there might be no “primary” site where either servers or customers are located. It is symptomatic of the rapid pace of technological progress that rules proposed to deal with one set of difficulties generated by e-commerce, such as instantaneous online management, are undermined by another technological development that casts doubt on the rules’ usefulness even before such rules are implemented.

In sum, the implications of cloud computing for international tax order extend far beyond the rules on permanent establishments, and many institutional arrangements require a fundamental rethinking in light of this change. Having sketched out some of the broader implications that the cloud has created for the international tax regime, in the following Parts I focus on the consequences of the cloud for particular provisions of the OECD Model, namely its Articles 5 and 7, which constitute the normative foundation for permanent establishments under this model treaty.

III. PERMANENT ESTABLISHMENTS UNDER ARTICLE 5

The purpose of this Part is twofold. First, it furnishes historical and conceptual background of the permanent establishment as codified in the OECD Model (Part III.A). Second, it analyzes the deployment of the concept as expressed in Article 5 of the OECD Model and looks at whether it is sufficiently flexible to serve its purpose with respect to cloud-based income-generating activities (Parts III.B–III.D). The relevance of the old debates surrounding the institutional design of the permanent establishment will become apparent when I move on to discuss various reform proposals in Part V. Ultimately, the hope is that the historical approach outlined below will serve to strengthen the normative argument in favor of remodeling Article 5 and abandoning the arm’s-length approach, thus ushering in a transition to a PE concept based on principles that not only more accurately reflect economic reality but also prevent, or at least discourage, Base Erosion and Profit-Shifting. What I also aim to achieve with this exposition is to show the inflexibility of the existing PE provisions under the OECD Model against the historical background. If a less narrow definition of the permanent establishment had been implemented when the OECD Model was given its current form, we likely would not have to face some of the difficulties generated by the cloud that we must deal with today. In what follows I will restrict myself to considering the PE definition under the predecessors of Article 5. The model of business profits attribution will be considered in greater detail in Part IV.

A. Evolution of the PE Concept

Traditionally, permanent establishments have been used to identify when the activities of an entity undertaken in a foreign country are substantial enough to allow that country to tax some of the profits connected with those activities. ⁴⁶David J. Shakow, The Taxation of Cloud Computing and Digital Content, 140 Tax Notes 333, 345 (2013). In other words, it is used to determine the right of one contracting state to tax the profits of an enterprise of another contracting state. ⁴⁷OECD, Commentaries on the Articles of the Model Tax Convention (July 22, 2010) (hereinafter 2010 OECD Commentaries), at 92. The principle behind the PE concept is often referred to as the Benefits Principle pursuant to which a taxpayer ought to pay taxes in accordance with the benefit the taxpayer receives from government programs. ⁴⁸Michael Kobetsky, International Taxation of Permanent Establishments: Principles and Policy (2011), 25. Source countries generally rely on the Benefits Principle to assert tax authority over income derived from business activity conducted on their territory by enterprises domiciled in their respective countries of residence. ⁴⁹Reuven S. Avi-Yonah, International Taxation of E-Commerce, 52 Tax L. Rev. 507 (1996–1997), 545. In this context, certain commentators have emphasized the fact that the source country’s primary right to tax active (business) income, as opposed to passive (investment) income, also rests on the Benefits Principle. ⁵⁰Id. at 520.

The existence of a permanent establishment is, however, a necessary but not in itself a sufficient condition for a source country’s assertion of taxing rights. Critically, a mechanism has to exist through which the allocation of taxing rights between treaty countries, as provided for by the Benefits Principle, can be carried out. ⁵¹Kobetsky, above n. 48, at 59. It is via the mechanism of attribution of profits provided for in Article 7 of the OECD Model that such allocation is made. The Benefits Principle in itself is insufficient to provide an answer as to how the income tax base should be divided among the various benefit-providing jurisdictions. ⁵²Avi-Yonah, above n. 49, at 522. Thus the burden of allocation falls on different methods of profit attribution, such as the arm’s-length standard or formulary apportionment that are supposed to provide the requisite and sufficiently precise profit-allocation formula. ⁵³The arm’s-length standard and formulary apportionment concept are discussed in detail in Part V, below. Any analysis of the PE concept would thus be incomplete if it did not encompass the provisions through which the attribution of profits is carried out, a matter to which I turn in Part IV.

It will now be instructive to briefly consider how the modern rules on permanent establishments evolved in the run-up to the adoption of the OECD Model in 1977. The historical perspective will provide some useful insights as to how the PE concept might be amended to deal with the challenges to profit attribution posed by the cloud.

One of the main goals driving the development of international tax order in the early 20th century was to prevent juridical and economic double taxation, i.e., imposition of comparable taxes in more than one state on the same taxpayer in respect of the same subject matter and for the same periods (juridical double taxation), and the taxation of the same income generated by different taxpayers (economic double taxation). ⁵⁴Kobetsky, above n. 48, at 107. To this end, and in order to facilitate financial reconstruction after the First World War, the International Chamber of Commerce requested that the League of Nations undertake necessary steps to prevent such instances of double taxation. ⁵⁵Id. at 111. This is also the time during which the modern PE concept was created, principally in order to protect the interests of capital-exporting countries, the U.S. featuring prominently among them, from taxation in source jurisdictions. ⁵⁶Avi-Yonah, above n. 49, at 535.

The 1927 Report of the League of Nations Committee of Fiscal Experts ⁵⁷Committee of Technical Experts, Double Taxation and Tax Evasion (1927).. served as the basis for a draft bilateral treaty (hereinafter “the 1928 Draft Treaty”) which contained a business profits article (hereinafter the “1928 Business Profits Article”) which was relatively broad in its scope of application. The pertinent part of the 1927 Report concerning the Business Profits Article read as follows: ⁵⁸Id. at 15.

The second paragraph gives a list of the establishments which are considered as permanent; they are: real centres of management, affiliated companies, branches, factories, agencies, warehouses, offices, depots, no matter whether such establishments are used by the traders themselves, by their partners, attorneys, or their other permanent representatives.

 

As can be easily discerned from the foregoing excerpt, the draft bilateral treaty did not contain the requirement that the PE be fixed, as is currently the case under Article 5(1) of the OECD Model, ⁵⁹Article 5(1) contains the following definition: “For the purposes of this Convention, the term ‘permanent establishment’ means a fixed place of business through which the business of an enterprise is wholly or partly carried on.” and so the requisite degree of permanence was much lower. Moreover, while the list of establishments contained in the 1928 Business Profits Article seems exhaustive on its face, the absence of fixedness requirement meant that the concepts such as “office” or “branch” could be interpreted much more broadly than is now possible in relation to the equivalent provisions under Article 5(2). ⁶⁰2010 OECD Commentaries make clear that “[the] place of business must be ‘fixed,’ i.e., it must be established at a distinct place with a certain degree of permanence” and that “the carrying on of the business of the enterprise [must take place] through this fixed place of business” (2010 OECD Commentaries, above n. 47, at 92, ¶2).

The 1928 Draft Treaty was subsequently amended and served as a foundation for the so-called London Model and Mexico Model drafted between 1940 and 1945. The most important difference between the two models for the present purposes is the difference in thresholds for taxation by the source country. The threshold was significantly lower under the Mexico Model, which was prepared with the inputs from capital-importing (developing) participant-states. ⁶¹Kobetsky, above n. 48, at 143. This should not come as a surprise, because the particular design of the PE concept gains particular importance where economic relations between contracting states are asymmetric, as is the case between developed and developing countries. ⁶²Ekkehart Reimer, Stefan Schmid, and Marianne Orell (eds.), Permanent Establishments, A Domestic Taxation, Bilateral Treaty and OECD Perspective (2013), 13. However, the rules based on the higher-threshold London Model ultimately prevailed resulting in a concomitant shift of taxing power towards capital exporting countries. The League of Nations was dissolved in 1946 and its work on bilateral tax models was subsequently assumed by the Organisation for European Economic Co-operation, which subsequently became the OECD. When the 1977 OECD Model was published, it retained the relatively high threshold not too dissimilar to the one under the London Model. The 1977 version of Article 5 has remained unchanged since then. It is against this historical background that I shall now consider various components of Article 5 and evaluate their capacity to capture cloud-based income which is economically connected with the host (source) country.

B. A ‘Fixed Place of Business’

Article 5 defines the permanent establishment as a fixed place of business, through which the business of an enterprise is wholly or partly carried on. ⁶³2010 OECD Commentaries, at 92, ¶2. This definition is then divided into three conditions that must be satisfied in order for a permanent establishment to be created: (1) the existence of a “place of business”; (2) the place of business being “fixed”; and (3) the carrying on of the business of the enterprise through this fixed place of business. ⁶⁴Id. at 92. The “fixed” requirement is often said to be composed of both a spatial component and a temporal component. ⁶⁵Cockfield et al., above n. 41, at 113. The permanence element seems to be further reinforced by the requirement that the business of the enterprise be “carried on [emphasis added] through the fixed place of business,” ⁶⁶Id. at 114. which suggests that the activity has to take place for a non-trivial amount of time. It is also understood that it is the factual circumstances that are determinative of the permanent nature of the establishment rather than the subjective intention or purpose with which the place of business in the host country was initially set up. ⁶⁷Id.

The following conclusion inevitably arises from the foregoing analysis. At the source country level, the fixed nature of permanent establishments is a major obstacle to taxing electronic commerce in general, and cloud-based e-commerce in particular. As was pointed out by Professor Avi-Yonah, most electronic commerce which takes place in the source jurisdiction tends to fall beneath the PE threshold so defined. ⁶⁸Avi-Yonah, above n. 49, at 533–534. This state of affairs would not in itself be so troubling as long as such income would be taxed elsewhere, i.e., in the country of residence of the e-commerce/cloud provider. A problem, however, arises because the production or provision of such services can easily be shifted to tax havens or low-tax-rate jurisdictions, such as Ireland, resulting in a significant lowering, if not complete nullification, of the level of effective taxation of income derived from those sources. ⁶⁹Id. at 534. In the pre-cloud era, this erosion of tax base could to some extent be mitigated by the “server as PE” rule whereby the maintenance of a server in the source jurisdiction would allow its tax authorities to contend that the server constituted a permanent establishment for tax treaty purposes. But with the widespread shift of e-commerce to the cloud it is no longer plausible to count on this rule to effectively deal with the problem because the server need not be located in the source jurisdiction. The requirement of fixedness has also been criticized in the context of much more tangible operations than those taking place in the cloud, it does not make much sense to draw forced distinctions “between oil rigs which are moored to the seabed and swimming rigs.” ⁷⁰Reimer et al., above n. 62, at 197. The force of this insight is all the more potent when applied to e-commerce. The rigid “permanent” component of the PE concept was already becoming obsolete towards the end of the 20th century, and now the advent of the cloud makes all the more necessary the reconceptualization of the degree of permanence required for a PE to be found to exist in the source country.

C. Servers as Permanent Establishments — Still Too Fixed?

Following a meeting of OECD delegates in Ottawa in 1998, ⁷¹OECD Committee on Fiscal Affairs, Electronic Commerce: Taxation Framework Conditions, A Report by the Committee on Fiscal Affairs as Presented to Ministers at the OECD Ministerial Conference “A Borderless World: Realising the Potential of Electronic Commerce” on 8 October 1998. the OECD revised its Commentaries to Article 5 to address specifically the question of e-commerce, including a determination that a server on which a website is stored may constitute a PE permanent establishment. ⁷²2010 OECD Commentaries, above n. 47, at 100, ¶42.2. Whether the server will in fact constitute a PE depends on the following facts. ⁷³Cockfield et al., above n. 41, at 117. First, the taxpayer must have the server at her disposal. Second, the taxpayer must have and must habitually exercise control of the equipment. ⁷⁴2010 OECD Commentaries, above n. 47, at 101, ¶42.5. The Commentary pursues a narrow approach in this regard, stating that virtual control is not sufficient because data do not constitute a “place of business” under Article 5(1) of the OECD Model. ⁷⁵Id. at 110, ¶42.2. Only if the enterprise that carries on e-business has the server on which its e-business is based at its own disposal will the place where that server is located constitute — subject to other requirements of Article 5 — a permanent establishment of that enterprise. This interpretation highlights the divide between the approach taken by the OECD and the economic reality of e-commerce. Within the cloud paradigm, the server should be conceptualized as a mere vessel through which the economic activity is channelled and it does not add any value to whatever goods or services might be so offered. In other words, as to whether the income is attributed to the source jurisdiction, why should the degree of control over such vessel be determinative rather than the degree of control over the substantive content — i.e., the website, the software or some other form of interface? This approach, which puts so much emphasis on the location of the server, is all the more surprising given the OECD’s otherwise strong commitment to the substance-over-form approach towards activities undertaken by taxpayers. More specifically, there is a contrast, if not an outright inconsistency, with the approach taken in Article 5(3) of the OECD Model where any formal split of contracts will not, in itself, require or justify that a heterogeneous set of facilities or activities be separated into several projects for the purposes of that Article. ⁷⁶Ekkehart Reimer, 5 + 7 = Odd, A Plea for More Consistency Between the PE Definition and Profit Allocation Rules in the OECD Model Tax Convention (working paper on file with the author), 3.

To better elucidate the tensions created by the existing fixedness requirement, consider the following example involving the use of the SaaS platform. Enterprise E, domiciled in residence jurisdiction J uses software S which is based on the SaaS platform and is provided by cloud provider P who also owns the server on which the software is run (Hypothetical A). E uses it to serve its customers, 100% of whom are based in (source) jurisdiction T, in which E otherwise has no physical presence. A tax treaty, based on the OECD Model, is in force between countries J and T. Under Hypothetical A, E will most likely be considered not to have a permanent establishment in country T under Article 5 because of lack of control over the server. Now assume (Hypothetical B) that P still owns the server but leases it to E. The underlying economics of E’s operations have not changed, it still serves its customers the same product i.e., software S, and all of E’s customers are still based in T, but now tax authorities of country T might potentially be able to claim, because of the higher degree of control exercised by E, that it does indeed have a PE in country T and that some part ⁷⁷Potentially even all of E’s profits given that all of its customers are based in country T. of its profits should be sourced to and taxed by this jurisdiction. This view seems to be supported, for example, with respect to Indian law, in the case of a non-resident cloud service provider who manages and controls servers in India rather than merely accessing the data. ⁷⁸Rohan Shah, Pranay Bhatia and Nishant Shah, Taxing the Cloud: India, 70 Tax Notes Int’l 159, 160. The point is far from clear, however. For a contrary view under Article 5 OECD Model generally, see Kristina Dautrich, Server as a PE: Have Your Server’s Activities Created a Taxable Presence? Transfer Pricing Int’l J. (Sept. 14, 2012). For an argument that the existence of a server which is effectively dedicated by the cloud service provider to a customer offers at least “strong evidence” for the existence of the latter’s PE, see Oliver Heinsen and Oliver Voss, Cloud Computing Under Double Tax Treaties: A German Perspective, 40 Intertax, at 584, 592.

It should become immediately clear that such a disparate approach to the determination of whether a PE exists, based on the underlying contractual arrangement on which allocation of control will depend, is an open invitation to engage in tax avoidance or is otherwise likely to facilitate Base Erosion and Profit Shifting by manipulating the contractual provisions governing the use and access to the server. This is particularly likely given that there is a widespread consensus that using another entity’s server will rarely, if ever, lead to the creation of a PE. ⁷⁹Shakow, above n. 46, at 348. In effect, even if the most “permanent” IaaS cloud model is used, it may nonetheless be relatively easy for the customer to avoid tax liability in the source country by ensuring that it does not have the requisite degree of control and thus that the fixedness threshold is not reached. The likelihood of tax liability will be even lower when SaaS or PaaS platforms are used because then it is even easier to separate the ownership of software from that of the server.

D. Agent PEs in the Cloud

Articles 5(5) and 5(6) of the OECD Model deal with the so-called “unassociated permanent establishments,” i.e., not facilities but persons different than the taxpayer herself, ⁸⁰Reimer et al., above n. 62, at 34. that are constituted by the dependent and independent agents, respectively. As a rule, dependent agents will be considered PEs of the principal whereas independent ones, subject to certain qualifications, will usually not constitute a PE of the taxpayer. ⁸¹Id. at 95. It is therefore the former category of dependent agents regulated by Article 5(5) ⁸²Article 5(5) reads as follows:

Notwithstanding the provisions of paragraphs 1 and 2, where a person — other than an agent of an independent status to whom paragraph 6 applies — is acting on behalf of an enterprise and has, and habitually exercises, in a Contracting State an authority to conclude contracts in the name of the enterprise, that enterprise shall be deemed to have a permanent establishment in that State in respect of any activities which that person undertakes for the enterprise, unless the activities of such person are limited to those mentioned in paragraph 4 which, if exercised through a fixed place of business, would not make this fixed place of business a permanent establishment under the provisions of that paragraph.

on which I focus in this subsection.

Before engaging with the analysis of the existing provisions, it is instructive to briefly consider how the concept of agents as permanent establishments evolved over time. The inclusion of “agency” as an instance of a PE in the 1928 Business Profits Article highlights the flexibility of that provision and therein the source jurisdiction’s potential to capture income. Assuming that the 1928 Business Profits Article (or a substantively equivalent provision) would be in force today, an argument could be made that a website or similar form of cloud provider interface, i.e., the SaaS model, could be classified as an “agency” thereby constituting a PE for purposes of such provision. Consider the following hypothetical. Enterprise E, domiciled in residence jurisdiction J, uses software S, which is based on the SaaS platform and is provided by cloud provider P. S is a software agent able to answer questions about E’s products and services, provide pricing information and negotiate terms of payment within pre-specified limits. E uses S to serve its customers, all of whom are based in the (source) jurisdiction T in which E otherwise has no physical presence. The servers on which S is hosted are owned by P and located in a tax haven. Assume further that a tax treaty, based on the OECD Model, is in force between countries J and T. Under this hypothetical, and based on the current wording of Article 5, it is virtually impossible for source jurisdiction T to tax any of the profits derived by E. There is no server in T which could serve as a permanent establishment. There is also no other fixed place of business in T falling under Article 5(1) or Article 5(2). The dependent agent provision (Article 5(5)) is equally problematic. This provision refers to a “person,” which is in turn defined in Article 3(1)(a): “the term ‘person’ includes an individual, a company and any other body of persons.” In effect, the human element is a necessary condition for a taxpayer’s agent to constitute a PE under Article 5(5). ⁸³The same conclusion is reached in Cockfield et al., above n. 41, at 123. Going back to the discussion in Part III.A above, an argument based on the agency concept similar to the one contained in 1928 Business Profits Article could be made, that the software agent is effectively E’s agency in source country J. Such broad and flexible approach to agency, which would be beneficial to T’s taxing rights and would also fairly accurately reflect the economic reality of the arrangement, is, however, ultimately precluded by the existing OECD Model provisions and the insurmountable reference to “persons.” Analysis of the policy implications of this conclusion is deferred to Part V.

In sum, it can be observed that the existing provisions of Article 5 in most cases cannot be counted on to ensure that cloud-based income is captured by the source jurisdiction. The potential reforms of Article 5 will be considered in Part V. Having addressed the difficulties generated by the cloud for the PE model embodied in Article 5, I shall now move on to evaluate the implications of this phenomenon for Article 7.

IV. PEs UNDER ARTICLE 7 — ATTRIBUTION OF BUSINESS PROFITS

This Part analyses the crucial provision of OECD Model, Article 7, which deals with business profits attribution. ⁸⁴Article 7 reads as follows:

1. Profits of an enterprise of a Contracting State shall be taxable only in that State unless the enterprise carries on business in the other Contracting State through a permanent establishment situated therein. If the enterprise carries on business as aforesaid, the profits that are attributable to the permanent establishment in accordance with the provisions of paragraph 2 may be taxed in that other State.

2. For the purposes of this Article and Article [23 A] [23B], the profits that are attributable in each Contracting State to the permanent establishment referred to in paragraph 1 are the profits it might be expected to make, in particular in its dealings with other parts of the enterprise, if it were a separate and independent enterprise engaged in the same or similar activities under the same or similar conditions, taking into account the functions performed, assets used and risks assumed by the enterprise through the permanent establishment and through the other parts of the enterprise.

3. Where, in accordance with paragraph 2, a Contracting State adjusts the profits that are attributable to a permanent establishment of an enterprise of one of the Contracting States and taxes accordingly profits of the enterprise that have been charged to tax in the other State, the other State shall, to the extent necessary to eliminate double taxation on these profits, make an appropriate adjustment to the amount of the tax charged on those profits. In determining such adjustment, the competent authorities of the Contracting States shall if necessary consult each other.

4. Where profits include items of income which are dealt with separately in other Articles of this Convention, then the provisions of those Articles shall not be affected by the provisions of this Article.”

I begin by introducing the arm’s-length principle, on which Article 7 is based, and I indicate the most important shortcomings and weaknesses of this principle, which have been exposed by the shift of MNEs activities into the cloud. I do so within the context of both U.S. domestic tax law and international tax law. I then explain the basic concepts behind Article 7 and its importance for the attribution of income. Last, I offer a critical appraisal of Article 7 as applied to international tax problems generated by cloud computing.

A. The Arm’s-Length Principle

At the heart of Article 7 lies the arm’s-length principle (hereinafter ALP) which informs the attribution of profits mechanism under this provision. This Part is aimed at introducing the ALP and explaining its conceptual underpinnings. I will then deploy the concepts introduced here in Part IV.B to analyze whether the ALP is a suitable mechanism for dealing with profit allocation in the cloud.

In general, the ALP serves to price transactions between related parties, such as companies within one group, as if they were separate and independent entities. ⁸⁵Reuven S. Avi-Yonah and Ilan Benshalom, Formulary Apportionment — Myths and Prospects, Public Law and Legal Theory Research Paper Series, University of Michigan Law School, Empirical Legal Studies Center, Working Paper No. 10-029. Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1693105. The arm’s-length principle features prominently in OECD’s Transfer Pricing Guidelines and serves as a lynchpin of a large part of international transfer pricing regime. According to a 2010 OECD report, ⁸⁶OECD, Report on Attribution of Profits to Permanent Establishments (2010) (hereinafter “2010 OECD Attribution Report”) . the 2010 OECD Guidelines are also applicable to the determinations made for the purposes of attribution of profits under Article 7. ⁸⁷Id., at 13, ¶9:

Accordingly, the profits to be attributed to a PE are the profits that the PE would have earned at arm’s length, in particular in its dealings with other parts of the enterprise, if it were a separate and independent enterprise engaged in the same or similar activities under the same or similar conditions, taking into account the functions performed, assets used and risks assumed by the enterprise through the permanent establishment and through the other parts of the enterprise, determined by applying the [Transfer Pricing] Guidelines [for Multinational Enterprises and Tax Administrations] by analogy [emphasis added].

Traditionally, the ALP has informed a process by which the transfer price between affiliated taxpayers was arrived at by means of comparing transactions involving the same product sold/bought by one of the affiliated parties to/from an unrelated party, or transactions concerning the same product sold between parties unrelated both to the affiliated parties and to each other. ⁸⁸Reuven S. Avi-Yonah, The Rise and Fall of Arm’s Length: A Study in The Evolution of U.S. International Taxation, 15 Va. Tax Rev. 89, 91–92 (1995–1996). This method of comparison is often referred to as the “comparable uncontrolled price,” or CUP method. The traditional ALP can also underlie two other methods that rely on comparable transactions that do not necessarily involve the same product. ⁸⁹Reg. §1.482-2(e)(2), §1.482-3(b). Under the “cost plus” method, the transfer price is determined by comparing the manufacturer to a similar entity (under a more flexible standard than the one employed under the CUP) that deals with unrelated parties, and allocating to the manufacturer the costs borne by the unrelated comparable party, increased by the latter’s profit margin. ⁹⁰Reg. §1.482-2(e)(4), §1.482-3(d). Lastly, the “resale price” method is identical to cost plus except that it is applied to resellers rather than manufacturers. ⁹¹Reg. §1.482-2(e)(3), §1.482-3(c).

Contrasting the above formulation of the ALP are the so-called “unitary” or “formulary apportionment” methods (hereinafter the “FA Formula”), such as those deployed by some U.S. states. ⁹²For some of the most recent writings on the ALP and the FA Formula, see, for example: Reuven S. Avi-Yonah, Kimberly A. Clausing and Michael C. Durst, Allocating Business Profits for Tax Purposes: A Proposal to Adopt a Formulary Profit Split, Law & Economics Working Papers Archive: 2003–2009 (2008); Kerrie Sadiq, Unitary Taxation — The Case for Global Formulary Apportionment, 55 Bulletin for Int’l Tax’n, 275. Under the FA Formula, the entire profit of an affiliated group is allocated among its constituent members by means of a formula based on such inputs as each group member’s assets, payroll or sales (or a combination of any of those). The major difference between the two approaches is immediately apparent. The ALP begins with treating each entity in an affiliated group as a separate taxpayer, while the FA Formula treats the entire affiliated group as a single, unitary enterprise.

Importantly it might be more helpful to think of the ALP and the formulary method as being on two opposing ends of a spectrum rather than belonging to two separate categories. ⁹³Avi-Yonah, above n. 88, at 93. This approach can be justified as follows. The cost plus and resale price methods, which are included in the traditional ALP-based approach, already represent a move away from treating each entity in the group as independent, because they involve considering the totality of the group’s profits, from which a profit margin allocable to the manufacturer or the reseller on the basis of the comparables is then subtracted, and the residual profit is allocated to the other party. ⁹⁴Id.

The ALP has been widely used in international tax law to determine the prices between affiliated enterprises for more than 50 years. Tax treaties that are based on the OECD Model are premised on the arm’s-length principle. Both Article 7 (attribution of profits) and Article 9 (associated enterprises) treat the head office and permanent establishments, and associated enterprises, respectively, as separate entities operating at arm’s length. Yet despite its widespread and long-standing use, the formula has proved very problematic and has generated a lot of criticism. ⁹⁵See Willard Taylor, Testimony before the President’s Advisory Panel on Federal Tax Reform (Mar. 2005) and in Tax Notes 2005-6654 (Apr. 4, 2005), where the ALP was described as “a cumbersome creation of stupefying complexity.” In what follows I offer a brief summary of the most important shortcomings of the ALP formula: administrative burdens associated with it and the fact that it does not reflect economic and business reality. Other undesirable consequences of using the ALP, such as the facilitation of profit shifting, are discussed in more detail in Part V. Whilst the examples used below are taken from the area of transfer pricing between different entities of one enterprise, the discussion in Parts IV.C and IV.D below, will make clear that both attribution of profits between such units and the attribution of profits between the head office and its permanent establishment(s) follow the same arm’s-length concept. ⁹⁶2010 OECD Commentaries at 134, ¶16. See also Reimer et al., n. 62, at 158.

The first major problem with using the ALP is that it creates uncertainty for MNEs because of the difficulties with establishing appropriate transfer prices and the tax authorities’ discretion to adjust them. This has resulted in expensive litigation because the courts are required to rule on the validity of the adjustments. ⁹⁷Kobetsky, above n. 48, at 78. Already in 1992 the IRS was spending about $15 million on expert witnesses in §482 cases. ⁹⁸Avi-Yonah, above n. 85, at 151. In 2013, 24% of companies surveyed by Ernst & Young reported having been subject to penalties when they had a transfer pricing adjustment. ⁹⁹Ernst & Young, 2013 Global Transfer Pricing Survey, at 7. Retrieved 25 April 2014 from: http://bit.ly/1ky5dLP

The second, and indeed main, difficulty is that the ALP formula can, and often does, lead to results that are completely unrealistic in economic terms. Cases ¹⁰⁰These examples are quoted in: Avi-Yonah, above n. 85, at 119. such as R.T. French Co. v. Commissioner, ¹⁰¹60 T.C. 836U.S. Steel Corp. v. Commissioner ¹⁰²617 F.2d 942, 2d Cir. or Bausch & Lomb
¹⁰³92 T.C. 525 illustrate this undesirable outcome where inexact comparables are used because the market had changed (R.T. French), or because the relationship between the parties makes for a different nature of transaction (U.S. Steel Corp. and Bausch & Lomb). The reason why the courts nonetheless chose to rely on ALP in those cases was because this method turned out to be a lesser evil. In the absence of a viable alternative, the only available solution would have been to determine the prices without reference to comparables, or indeed, to any underlying formula — a result which would be even more unpredictable and arbitrary than the one arrived at via the ALP route. ¹⁰⁴Avi-Yonah, above n. 85, at 119. The debate preceding the amendment of §482 resulted largely from the recognition of absence in many cases of comparable arm’s-length transactions between unrelated parties. This was recognized in the House Report on House Bill 3838 where it was emphasized that fundamental problems with the ALP stem from the fact that “the relationship between related parties is different from that of unrelated parties” and that “multinational companies operate as an economic unit, and not ‘as if’ they were unrelated to their foreign subsidiaries.” ¹⁰⁵H.R. Rep. No. 426, 99th Cong., 1st Sess. 423–24 (1985) quoted in Avi-Yonah, above n. 85, at 131. The very existence of integrated multinationals thus constitutes evidence that the ALP does a very poor job in reflecting economic reality because, according to the internalization theory, multinationals exist mostly as a result of market and non-market advantages that are derived from their structure. ¹⁰⁶Avi-Yonah, above n. 85, at 149. In effect, a market rate of return applied separately to each of the components of the multinational will generate a result which will be less than the actual return of the organization as a whole. ¹⁰⁷Id. I shall now consider how those fault lines inherent in the ALP are exposed and deepened by the cloud economy.

B. The Arm in the Cloud

Both of the problems identified above become arguably more acute when the ALP is applied to transactions that are concluded in the cloud. Some intangibles, such as rights relating to important patents, software, or proprietary expertise, are often a crucial component of the competitive advantage of a multinational corporation, and will rarely if ever be transferred between unrelated parties, meaning that no comparable uncontrolled transactions might be available. ¹⁰⁸Michael McDonald, Income Shifting from Transfer Pricing: Further Evidence from Tax Return Data, OTA Technical Working Paper 2 (2008), 5. At least two different scenarios can be distinguished. First, the cloud itself can be the object of “transfer,” e.g., when the IaaS model is deployed and the underlying infrastructure involves some form of intangible property. However, this scenario is unlikely to give rise to qualitatively different considerations than are usually pertinent in case of transfer of intangibles, for example when there are no internal comparables. ¹⁰⁹2010 OECD Guidelines, above n. 36, at 77, ¶2.58 and at 115–116, ¶¶3.27 and 3.28. Second, and perhaps more interestingly, the cloud might provide specific background to a transaction when certain intangibles are transferred between enterprises which are connected by the cloud. Consider the following hypothetical. Subsidiaries S₁ and S₂ both share the same data warehouse provided by their parent via SaaS; a transfer of intangible T from S₁ to S₂ takes place within that (virtual) warehouse. Such arrangement not only will significantly diminish the already limited number of available comparables due to the fact that independent enterprises will be very unlikely to transfer T in a shared cloud, but also will generate difficulties for so-called “package deals.” ¹¹⁰The Guidelines refer to a package deal as a single transaction which “[establishes] a single price for a number of benefits such as licences for patents, know-how, and trademarks, the provision of technical and administrative services, and the lease of production facilities.” Id. at 111. This will be particularly true if it becomes necessary to disentangle a transaction which is intertwined with the shared data warehouse and then evaluate it separately from the virtual storage space in which it occurred. The nature of the problem which arises in the cloud is slightly different from what is normally the case with package deals. This is so because in the cloud it might be relatively easy to conceal a number of intangibles which would normally have to be transferred independently, especially if they are disguised as part of the overarching cloud infrastructure rather than as individual components transferred within it. The picture will become even more blurred if we consider the deployment of an iPaaS model which has been designed precisely for the purposes of building and hosting packaged integration solutions. ¹¹¹See Part II.A, above. If more than just one intangible is transferred between S₁ and S₂ while they are under the iPaaS “umbrella” it will be virtually impossible to evaluate each transfer separately because all of them will be integrated by the iPaaS and are also likely to derive most, if not all, of their functionality from being integrated by it. This difficulty will then trickle down to the next level affecting the choice of differing tax treatments of the respective individual sub-transactions depending on how the latter are classified.

To illustrate this difficulty consider the following example ¹¹²Welsh et al., above n. 8, at 149. under the U.S. Software Regulations. Subsidiary S is allowed, pursuant to a license granted by its parent P, to access and use software in the cloud via SaaS, i.e., without obtaining a copy of the software. Depending on whether the use of the software is considered a “transfer” under Reg. §1.861-18, the transaction will be classified either as a transfer of intangibles (with the cloud serving merely as a new medium via which the transfer takes place) or as a service (with the cloud being a constitutive part of the transaction). ¹¹³Id. at 150. If the latter classification prevails then it will be transfer pricing rules covering controlled service transactions ¹¹⁴Treatment of Services Under §482. Available at: http://www.irs.gov/irb/2009-33_IRB/ar07.html#d0e228. that will apply. ¹¹⁵Welsh et al., above n. 8, at 150. The tax treatment will then vary accordingly depending on which specific rules are applied.

As the foregoing analysis demonstrates, the cloud affects the ALP analysis on every level, beginning with the identification of individual transactions that might be subject to taxation, through their classification and the choice of applicable transfer pricing regime. The cloud removes the ALP-based transfer pricing regime even further from economic reality by making MNEs more closely integrated than ever before and by laying bare the radically counterfactual nature of the “as if independent” approach; it also drastically reduces the number of available comparables. On a more easily quantifiable level, given how costly transfer pricing litigation already is, ¹¹⁶Reuven S. Avi-Yonah, Between Formulary Apportionment and the OECD Guidelines: A Proposal for Reconciliation, Public Law and Legal Theory Working Paper Series, Working Paper No. 152, 10. Available at: http://bit.ly/1BMPDlM. the increase in number of intangibles transferred via the cloud is bound to increase the costs even further, mainly because of the even higher complexity of cloud-based transactions as opposed to “traditional” e-commerce. Not only will very expensive valuations have to be conducted, but also more uncertainty will be created as to appropriate identification and classification of particular transactions. The discussion of potential responses to these undesirable developments is deferred to Part V.

Having addressed the broad implications that cloud computing generally presents to the ALP, I now move on to discuss the issue in a narrower context of Article 7, which is also premised on the arm’s-length standard.

C. Article 7 as Framework for Attribution of Business Profits

When an MNE earns income from sources in two or more states, each of the jurisdictions may assert its taxing rights with respect to different parts of such income. Under the bilateral tax treaty framework, it falls on Article 7 of the OECD Model to deal with the issue of determining how such income should be apportioned between two competing states. Under Article 7, a contracting state cannot tax the profits of an enterprise domiciled in another contracting state unless it carries on its business through a permanent establishment situated there. ¹¹⁷2010 OECD Commentaries, above n. 47, at 92.

At the preliminary stage of determining whether a PE exists, it is necessary to do so pursuant to the provisions of Article 5. As discussed in Part III, if an enterprise E domiciled in country X derives profits from both that country and another country Y (between which a tax treaty is in force), but when no permanent establishment exists in country Y, country X will, as a rule, have exclusive taxation over E’s profits. If, however, E has a permanent establishment in country Y, then, in principle, the host country will have unlimited right to tax the profits attributable to the PE. But Article 5 itself does not allocate taxing rights. When an enterprise of a contracting state carries on business in a different contracting state through a PE situated there, it is necessary to determine what, if any, are the profits that the other state may tax. ¹¹⁸2010 OECD Commentaries, at 130. The worldwide profits will be apportioned between the two countries, and this will be done according to the rules laid down in Article 7. Subject to other provisions of the OECD Model, country Y will be prevented under Article 7 from taxing E’s business profits which are not attributable to its PE in that country. In sum, under Article 7 PE plays the role of a profit center which serves the purpose of assigning assets, liabilities, opportunities and risks or, in other words, the purpose of quantifying the tax base assigned to each of the two contracting states. ¹¹⁹Reimer et al., above n. 62, at 154. The burden of eliminating double taxation will then rest on the home state of the enterprise which must, in accordance with either Article 23 A (exemption) or Article 23 B (tax credit), eliminate such taxation on the profits properly attributable to the PE. ¹²⁰2010 OECD Commentaries, at 134.

For more than 30 years since its initial formulation in 1977, Article 7 remained unchanged, but it was eventually amended in 2010. ¹²¹A detailed discussion of changes introduced in 2010 is provided in Kobetsky, above n. 48, pp. 351–392. The approach developed in the 2008 Report (the so-called authorized OECD approach) was not constrained by either the original intent or by the historical practice and interpretation of Article 7; instead, the focus was on formulating the best approach to attributing profits to a permanent establishment under Article 7 given the challenges posed by the business operations of MNEs. ¹²²2010 OECD Commentaries, at 131. The development of this approach was undertaken to examine how far the approach of treating a PE as a hypothetical separate and independent enterprise could be taken. ¹²³Id. In particular, the examination focused on the extent to which any modifications would be necessary to account for the differences between a PE and a legally distinct and separate enterprise.

Despite long and intense discussions preceding the changes, ¹²⁴OECD’s Committee for Fiscal Affairs acknowledged “the need to provide more certainty to taxpayers: in its report Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations, it indicated that further work would address the application of the arm’s-length principle to permanent establishments. That work resulted, in 2008, in a report entitled Attribution of Profits to Permanent Establishments. […] The focus of the 2008 Report was on formulating the most preferable approach to attributing profits to a permanent establishment under Article 7 given modern-day multinational operations and trade. When it approved the 2008 Report, the Committee considered that the guidance included therein represented a better approach to attributing profits to permanent establishments than had previously been available.” Id. the fundamental reliance on the arm’s-length concept has not been abandoned. The OECD resisted the challenge to adopt a different approach, such as the general force of attraction principle. ¹²⁵Id. at 133. Under this principle, if an enterprise E has a PE in source country Y then its business profits, dividends, interest and royalties derived in source country Y may be treated as having the PE as their source, even if they are not attributable to that PE. But this opportunity to address at least some of the difficulties inherent in the ALP was rejected, and the new Article 7 still treats PEs as entities that are functionally separate and independent from the enterprise located in the residence country. ¹²⁶The so-called Functionally Separate Entity Approach (FSEA), 2010 OECD Attribution Report, above n. 86, at 13. It is true that the commitment to the functionally separate entity approach will allow the source country in which a PE is located to tax more than just a small surplus on the costs incurred by the PE. ¹²⁷Reimer, above n. 76, at 4. However, it does little to address the fundamental problem with the counter-factual assumption which disregards the idiosyncratic or synergistic (as the case may be) nature of dealings between the head office and permanent establishments of an MNE, and the lack of comparables stemming from such idiosyncrasies and/or synergies. The analysis of this phenomenon and conclusions drawn from it in the context of transfer pricing are equally relevant with respect to the attribution of profits to PEs because the 2010 OECD Guidelines are also applicable to the determinations made for the purposes of attribution of profits under Article 7. Overall, the difficulties in using the ALP in the cloud discussed in Part IV.B, above, remain largely unaffected by the changes to Article 7 introduced in 2010. In effect, those changes are unlikely to have any significant impact on how effectively this provision deals with MNEs engaging in cloud-based transactions.

It still remains to be seen how many countries choose to adopt the new Article 7. ¹²⁸For example, Chile, Greece, Mexico and Turkey reserved the right to use the previous version of Article 7 and they did not endorse the change. Portugal reserved its right to continue to adopt in its conventions the text of Article 7 in its pre-2010 wording until necessary changes are made in its domestic law. OECD, The 2010 Update to the Model Tax Convention (July 22, 2010), 41. This uncertainty notwithstanding, the focus here is on the new Article 7 rather than the pre-2010 version, because it is the former that seems to reflect OECD’s forward-looking approach to profit attribution. The new version will also serve as a stepping stone to the normative analysis presented in Part V. Finally, the 2010 version is preferable for the purposes of the normative argument because any potential changes to Article 7 will necessarily have to be based on the current 2010 wording of this provision. As indicated above, however, the reliance on the arm’s-length standard seems bound to continue given the OECD’s commitment to the FSEA. ¹²⁹Reimer, above n. 76, at 4.

D. Attribution of Business Profits in the Cloud

The discussion presented in this section is based on the assumption that, as a preliminary issue, a PE has been found to exist under Article 5 of the OECD Model in the source state. Given the narrow scope of this provision, in particular with respect to cloud-based activities, it is further assumed that the existing PE will have the form of a server, as this is the most probable route via which the cloud-based e-commerce activities of an MNE will be captured (if at all) by the source state. Based on these assumptions, I will now consider the exact mechanics of Article 7: how it operates to effectuate the profit attribution and how these mechanics may be affected by the cloud.

As indicated above, the basic purpose of Article 7 is to provide the framework within which attribution of profits between the PE and other units within the enterprise can take place. This consideration then directly affects the assignment of profits to the PE. Under Article 7(2) the conditions of the transaction can then be adjusted, if necessary, to reflect the conditions of a similar transaction between independent enterprises. ¹³⁰2010 OECD Commentaries, above n. 47, at 136. This analysis is typically presented as a two-step approach. ¹³¹2010 OECD Attribution Report, above n. 86, at 13, ¶10. Under the first step, a functional and factual analysis is undertaken while the second step involves the quantification of the profits connected to any given transaction. The first step can be broken down into following stages, ¹³²The analysis is based on: Reimer et al., above n. 62, at 168–184. which will now be discussed in turn.

Step 1 — Functional and Factual Analysis

• Determining the qualifying units of the enterprise

At this stage it is necessary to determine whether the activity carried through a sub-unit of the enterprise meets the PE threshold as provided for in Article 5. ¹³³This issue was explored in Part III.B, above.

• Identifying the relevant transactions and dealings

In conceptual terms, “transactions and dealings” includes both objective elements, such as different categories of assets, and other items, even if the latter do not constitute assets under traditional accounting rules — e.g., the relocation of know-how when an actual position requiring sophisticated work is transferred from one contracting state to another. ¹³⁴Reimer et al., above n. 62, at 169. In strictly legal terms, assets owned by the enterprise will, as a rule, belong to the enterprise as a whole, of which the PE is only a part. To address this difficulty, OECD has introduced the notion of “economic ownership” in order to attribute economic ownership of assets to a PE. ¹³⁵2010 OECD Attribution Report, above n. 86, at 28, ¶72. Under this approach, in determining the characteristics of a PE for taxation purposes it will be the economic (rather than legal) conditions that are going to be relevant because they are likely to have a greater effect on the economic relationships between the various parts of the single legal entity. ¹³⁶Reimer et al., above n. 62, at 169. The category of relevant transactions and dealings also includes activities generally as broad as the scope of the relevant domestic tax provisions as well as a broad range of services of different types. ¹³⁷Id. at 170. The difficulties with the application of the economic ownership concept to cloud-based activities are discussed in the next subsection.

• Determining for each of these transactions whether they shall be assigned to the relevant PE

This stage can be said to go to the heart of the functionally separate entity approach because all activities and transactions must be notionally assigned to one or more PEs and whatever residue there is left will then be assigned to the state of residence. The assignment of a transaction to a given PE, pursuant to the FSEA, will then depend on whether or not the PE would have been a party to this transaction. ¹³⁸Id. at 171. In order to conduct this determination, Article 7(2) of the OECD Model mandates taking into account “the functions performed, assets used and risks assumed by the enterprise through the permanent establishment and through the other parts of the enterprise.” As a rule, the assignment of transactions will follow the assignment of underlying assets or liabilities. In many cases, especially when operating assets are concerned, it will be pertinent to consider how closely, from an economic viewpoint, the relevant asset is connected to the functions of the respective unit within the enterprise. ¹³⁹Id. at 172. But since in many cases, and especially in the cloud environment, there will be no underlying assets, functions can also exist and be assigned independently, in which case the determination will take into account human activities which shape and define the function. ¹⁴⁰Id. at 175. A similar approach is taken with respect to allocation of risks which are assigned to individuals and thus to the unit which can control or reduce the risks. ¹⁴¹2010 OECD Attribution Report, above n. 86, at 16.

On its face, the economic ownership approach should be quite efficient at allocating the ownership of any given asset or function, in particular when compared to the alternative of attempting to attribute the ownership along legal boundaries that are likely to be blurred and very difficult to determine within a unitary enterprise. ¹⁴²Indeed, Professor Reimer goes so far as to say that many such dealings are “fully invisible from the private law perspective,” Reimer, above n. 76, at 11. But whatever merit there might be in this approach, it is still premised on the flawed fiction of the FSEA and all of the attendant shortcomings become exposed when the economic ownership approach is analyzed in the context of intangibles and against the background of an MNE operating in the cloud. Under the authorized OECD approach, the key question in determining economic ownership of an intangible focuses on risk, i.e., where within the enterprise the significant people functions related to decisions concerning management of risks are undertaken. ¹⁴³2010 OECD Attribution Report, above n. 86, at 32. The 2010 OECD Attribution Report states that these functions might include “the evaluation of the performance of any required follow-on development activity, and the evaluation of and management of risks associated with deploying the intangible asset.” ¹⁴⁴Id. Let us now assume that enterprise E licenses an intangible asset from a provider P. The asset is in the form of a PaaS cloud software which was specifically designed and programmed to fit E’s particular needs and corporate structure. E deals in providing water-purity monitoring services to developing countries and while it is headquartered in country A, its operations are decentralized and take place in multiple jurisdictions other than A, from which it also derives most of its profits. The PaaS software is used as a central platform which aggregates and processes data from all water-purity monitoring stations. Assume further that the PaaS software is, for security reasons, run simultaneously from servers located in each respective country in which E operates. If we now attempt to answer the question as to where within the enterprise “significant people functions related to decisions concerning management of risks” related to the PaaS are undertaken — a question which must be asked under the authorized approach — arguably we will not be able to draw such a boundary on any principled grounds. If we consider the arrangement as a whole, the risk is most likely to be present in country B where the server is, but this is not very helpful for the purposes of assigning risk within E. Given the limited degree of control that any respective unit of the enterprise usually exercises over this type of software, the reference to “follow-on development activities” will also be of limited usefulness. Such activities are likely to be performed by the provider at any given server, depending on when the need arises. Furthermore the management of risks associated with the software is also concentrated in the hands of the provider with only tangential matters being under the control of the respective units of the enterprise. In other words, it can be said that only a small residue of risk-management activities is undertaken at the enterprise level and it is questionable whether such residue allows for any meaningful allocation of risk within E along the lines mandated by the economic ownership approach which is effectively hollowed out when applied in the cloud. ¹⁴⁵It might also be worth noting that some jurisdictions might potentially consider the cloud provider P as a permanent establishment of E, but the analysis of such scenario lies beyond the scope of this article. See Franck Llinas, Abishek Goenka, Pierre-Regis Dukmedjian and Kent Wisener, Cloud Activities: Check Your Tax Forecasts, Transfer Pricing Int’l J. (Dec. 2013), 1.

Step 2 — Quantification

After the first stage has been completed, the second step involves the quantification of the profits connected to any given transaction. The process of quantification will be normally carried out in accordance with the OECD Transfer Pricing Guidelines. ¹⁴⁶Reimer et al., above n. 62, at 136. To that end, the remuneration of any dealings between the hypothesized enterprises is determined by applying by analogy transfer pricing tools used under Article 9 of the OECD Model, by reference to the functions performed, assets used and risks assumed by the hypothesized enterprises. ¹⁴⁷2010 OECD Report on Attribution of Profits to Permanent Establishments, at 13, ¶10. The overall result of these two steps should be to allow the calculation of the profits (losses) of the PE from all its activities. ¹⁴⁸Id. at 13. The fundamental problem which has to be dealt with at this stage is whether the whole transaction can be attributed to a PE or whether a fractional apportionment of profits arising from the transaction is necessary. At this stage, the various methods of profit allocation discussed earlier come into play. ¹⁴⁹See above Part IV.B. The choice of method will usually be up to the taxpayer.

 

The difficulty encountered at the last stage of the first step (i.e., attribution of transactions between the respective units of the enterprise) unfortunately carries over to the second step of the analysis, which involves quantification. Under the authorized approach, where the economic owner makes a licensed intangible available for use by another part of the enterprise, the functional and factual analysis is normally expected to determine the character of that dealing, e.g., as an outright transfer or a licensing of those rights to use, for purposes of attributing profit from that use. ¹⁵⁰2010 OECD Attribution Report, above n. 86, at 55, ¶210. Unfortunately this guidance does not get us very far because of the indeterminacy surrounding the preliminary question of who is the economic owner described above. Yet again it can be seen how the fundamental flaw at the heart of the ALP — i.e., the fiction of treating unitary enterprise as if it were a set of independent entities — makes any solution, which would be more than just arbitrary allocation, virtually impossible. It is therefore perhaps hardly surprising that in the new draft of its Transfer Pricing Guidelines the OECD advocates abandoning the concept of economic ownership. ¹⁵¹See OECD, Discussion Draft on Transfer Pricing Documentation and CbC Reporting (Jan. 30, 2014). But even if we assume that the economic ownership can be determined on non-arbitrary grounds we will still face difficulties when it comes to attributing profits from the use of the intangible. This is so because the cloud which connects all the entities within an enterprise is rather unlikely to exist between unaffiliated enterprises. Assume for example, under a slightly different scenario, that the head office of the enterprise E introduced earlier, acquires a license from the provider P to use the PaaS software and then sublicenses it to its units operating in developing countries. The very reason for such sublicensing consists in the availability of cloud-based synergies inherent in the software which would simply not exist if the head office were to contract with independent entities. ¹⁵²Another example would include the specialized enterprise procurement solutions such as those offered by Oracle under their Oracle Procurement Cloud (“OPC”). Realizing More Savings with Oracle Procurement Cloud
http://www.voiceson-resources.com/rt.asp?I=54CB0XB6F4X8&L=1253253. OPC handles both the transactional aspect of procurement which focuses on making resources such as catalogs and self-service portals available to those in the organization who need to buy supplies, parts or office equipment, as well as the strategic aspect of procurement, which includes sourcing, contract management, and compliance. It is extremely unlikely that any form of such tight integration would exist between unaffiliated enterprises, Id. Under the counterfactual, unaffiliated scenario, the head office would almost certainly sublicense outside of the cloud (mainly due to privacy and security concerns), and even if it did share the cloud with an unaffiliated enterprise the nature of the arrangement would be very different, especially given data protection considerations and general security concerns, from the one deployed within a unitary enterprise. In sum, at both the functional and factual analysis stage as well as at the quantification stage the deficiencies of the ALP are exposed by the cloud whenever one tries to follow the OECD approach to profit allocation

Having identified the challenge posed by the cloud to the ALP as expressed in the profit allocation mechanism under Article 7, I shall now examine how (in)efficient the respective transfer pricing methods are when it comes to dealing with cloud-based or cloud-related transfers of intangibles within an enterprise. This analysis is based on the fundamental insight that the FSEA does not answer the question on how any added value from the fact that the units form part of the same enterprise should be allocated between those units and, by extension, between the respective jurisdictions in which those units are located. ¹⁵³Reimer, above n. 76, at 4. As for the CUP method, the 2010 OECD Guidelines suggest that it can be deployed where the same owner has transferred or licensed comparable intangible property under comparable circumstances to independent enterprises. ¹⁵⁴2010 OECD Guidelines, above n. 36, at 199, ¶6.23. As explained above, however, it would be very rare for any such transfer to occur given that mostly affiliated enterprises (units of an enterprise) are likely to be connected by the cloud. The CUP is thus not very helpful for determining the appropriate price for the transfer. For the same reasons, similar considerations apply to all traditional transaction methods and to the transactional net margin method. ¹⁵⁵Id. at 200, ¶6.26.

Recognizing that those methods might not be adequate, tax authorities in many countries have recently begun to rely on the so-called profit-split method (hereinafter the PSM) to price affiliated transactions where there are no good comparables. ¹⁵⁶Avi-Yonah & Benshalom, above n. 85, at 14. Instead of trying to proceed based on a counter-factual assumption as to how unaffiliated parties might price the transaction, the PSM aggregates the income generated from the transactions under consideration and divides it according to the contribution of the respective units within the enterprise; this is essentially a quasi-formulary approach. ¹⁵⁷Id. The normative implications of this tacit yet telling admission of ineptness of the ALP are considered in detail in Part V.C, below. It should be observed, however, that the allocation then takes place based upon an economically valid approximation of the division of profits that would have been anticipated and reflected if an agreement was entered into at arm’s length. ¹⁵⁸2010 OECD Guidelines, above n. 36, at 28. Indeed, the OECD in its Transfer Pricing Guidelines goes so far as to suggest that the “overriding objective” under the PSM should be to approximate as closely as possible the split of profits that would have been effectuated had the parties been transacting at arm’s length. ¹⁵⁹Id. at 95. Thus the same objections that were levelled against the ALP in the context of the profit attribution under Article 7 of the OECD Model apply here as well, and perhaps with even greater force. Instead of trying to make the profit split fit the straightjacket of the ALP it would be preferable to harness the PSM’s relative flexibility and use a different proxy for profit allocation than hypothesizing from unaffiliated enterprises. As was pointed out in the context of cost sharing regulations, affiliated transactions related to intangible assets are difficult to appraise in accordance with an arm’s-length standard not only due to lack of comparables, but also because of the high-risk nature of intangibles, which means that their value and correlative risks cannot be accurately inferred from the costs incurred in their creation, and their enormous value variation which depends on the economic environment in which they are deployed. ¹⁶⁰Ilan Benshalom, Sourcing the ‘Unsourceable': The Cost Sharing Regulations and the Sourcing of Affiliated-Intangible Related Transactions, 26 Va. Tax Rev. 631, 649 (2007). Accordingly, the value of many intangible assets, especially those that are inchoate and have not yet been tested or are still in the process of formation, are extremely difficult to appreciate even for MNE insiders. ¹⁶¹Id. at 649.

Yet when one looks at the 2010 OECD Guidelines section which focuses specifically on arm’s-length pricing when valuation is highly uncertain at the time of the transaction, there is no recommendation that PSM ought to be followed. ¹⁶²2010 OECD Guidelines, above n. 36, at 201, ¶¶6.28–6.35. Rather, the uncertainty is to be resolved by reference to what independent enterprises would have done in comparable circumstances to take account of the valuation uncertainty in the pricing of the transaction. ¹⁶³Id. at 201. The specific provisions of this set of guidelines suggest that independent enterprises could use anticipated benefits as a means for establishing the pricing, adopt shorter-term agreements or include price adjustment clauses. ¹⁶⁴Id., ¶¶6.29 and 6.30, respectively. They might also simply decide to bear the risk of unpredictable subsequent changes with the understanding that major unforeseen developments would lead to the renegotiation of the pricing arrangements. ¹⁶⁵Id., ¶6.31. However, the benefits derived from the cloud by the head office of enterprise E introduced earlier, will be contingent upon the participation of other units of the enterprise because the synergies from sharing and processing water-purity data in the cloud can be realized only with the participation of those units that operate in developing countries — i.e., outside of E’s country of residence. In other words, the benefits accruing to the head office will be realized only as long as the cloud continues to be available to the entire enterprise. This contingent element would have to be discounted when the initial price is set and also in any type of price-adjustment clause or when attempting to quantify the anticipated benefits. The price of transferring the PaaS device would thus have to be reduced in order to reflect the fact that such transfer is a necessary condition to any profits accruing to the head office and, indeed, being generated by the enterprise as a whole. This presents at least two problems with respect to the application of the arm’s-length standard. First, as discussed in the preceding paragraph, there is the problem of finding a comparable based on unaffiliated enterprises deciding to share data in the cloud, because such data sharing is either unlikely to occur between unrelated entities or, if it indeed does occur, its nature will be very different from an internal data sharing arrangement due mainly to security and data protection constraints. As a result, the lack of comparable circumstances will make it difficult to quantify the necessary price adjustment. Second, and more generally, because the cloud makes the enterprise much more unified, almost any form of hypothesizing from unaffiliated enterprises dealing at arm’s length will be very far removed from economic and business reality. Even assuming that unaffiliated enterprises would be sharing some data between them in the cloud, the nature of such arrangements is likely to be radically different from those between affiliated units, and much more limited, and the cloud infrastructure much less integrated, not the least because of security considerations and concerns about access to sensitive data.

Before concluding this Part, I shall analyze an alternative, income-based approach to transfer pricing. The income-based approach relies principally on three methods to arrive at the appropriate price: the “relief from royalty method,” the “excess earnings method” and the “with or without method.” ¹⁶⁶Steven van Wijk, Pooja Kalra and Joel Wilpitz, Towards a New OECD Valuation Standard for Intangibles? Transfer Pricing Int’l J. (Sept. 2012), 3. They shall be now considered in turn. The relief from royalty method, which looks at the amount of income that a company would be “deprived” of if it did not own the intangible, but was required to license it from a third party. To this extent this method is therefore of little help when, as in our scenario, the intangible is already being licensed from a third-party provider. The excess earnings method relies on the present value (PV) of earnings attributable to the intangible assets after providing for the proportion of earnings that should be attributed to returns from contributory assets. ¹⁶⁷Id. In the case of cloud-based software, it might be difficult to disentangle the contributory assets from the “main” asset the value of which we are attempting to determine. Finally, the “with or without method” estimates the fair value of an asset by comparing the value of the business inclusive of that asset to the value of such business without the relevant asset. ¹⁶⁸Id. This is perhaps the most promising method which might be most efficient when applied to cloud intangibles, although it is considerably weakened by the fact that it necessitates a number of assumptions such as business forecasts, competition and market. ¹⁶⁹Id. Once the appropriate share of profits (losses) that are attributable to a permanent establishment has been quantified in accordance with Article 7(2), it will then fall on the domestic law of the relevant jurisdictions to determine whether and how such profits should be taxed. ¹⁷⁰2010 OECD Commentaries, above n. 47, at 139, ¶30.

As the foregoing discussion hopefully makes clear, the cloud renders the ALP increasingly obsolete and exposes its ineptness to efficiently deal with taxation of highly integrated MNEs, especially when such integration is achieved through the use of cloud software. Given the need for a shift farther away from the ALP and toward some type of a formulary approach, I will now analyze various proposals to reform the profit allocation mechanism.

V. IMPACT OF CLOUD COMPUTING ON PE REFORM PROPOSALS

In the preceding Parts, I have examined how the rules governing permanent establishments under the OECD Model are constructed, how they function and whether they deals adequately with apportioning income within unitary MNEs operating in the cloud. The analysis has revealed that the design of Article 5 prevents or at least makes it very difficult to apply the PE rules to many instances of cloud-based activities, even if such activities are conducted through a server located in the source jurisdiction. It should also have been made clear that the arm’s-length principle, as embodied in Articles 7 and 9, is a serious obstacle to dealing with situations that require apportionment of income within an MNE in cases in which the income is generated by, or in some cases just related to, cloud computing. In this Part, I attempt to evaluate various reform proposals that have been put forward, with the aim of addressing some of the difficulties identified in Parts III and IV. First, I explain what are the main policy reasons which inform the desirability of reform in general (Part V.A). Second, I analyze the changes proposed in relation to Articles 5 and 7 (Parts V.B and V.C, respectively). It is against the background of the aforementioned policy goals that the relevant reform proposals are evaluated. In the case of Article 7, I place particular emphasis on reform suggestions based on introducing some form of a formulary apportionment mechanism.

A. The Reform Agenda — An Overview

It is possible to identify at least three major reasons why the PE concept embodied in the OECD Model should be remodeled. ¹⁷¹For a comprehensive, albeit now somewhat dated, overview of reform proposals, see OECD Technical Advisory Group on Monitoring the Application of Existing Treaty Norms for Taxing Business Profits, Are the Current Treaty Rules for Taxing Business Profits Appropriate for E-Commerce? (June 2004) (hereinafter 2004 OECD Report). Available at: http://www.oecd.org/tax/treaties/arethecurrenttreatyrulesfortaxingbusinessprofitsappropriatefore-commerce.htm. First, a significant amount of electronic commerce, including cloud-based e-commerce, falls below the existing PE threshold. This means that even though the cloud-based activity of an MNE which generates significant amounts of income economically sourced to country X, will nonetheless be sourced and to taxed in residence country Y (hereinafter the “High Threshold Problem”). The income might also be subject to no taxation at all if the residence country is a tax haven or is subject to a disproportionately low tax rate if the MNE’s residence is in a country with a specific tax regime for holding companies. ¹⁷²Avi-Yonah, above n. 49, at 527–528. One of the most common cross-border tax optimizing structures involving such residence countries, the so-called “Double Irish with a Dutch Sandwich,” has been widely used to take advantage of Ireland’s relatively low corporate income tax rate of 12,5%; the Irish-based royalties are then channeled to a Dutch affiliate free of withholding taxes because the transfer takes place between two related, European Union–based companies. ¹⁷³Cockfield et al., above n. 41, at 188–189. The attempt to deal with this type of problem by ensuring that the income is taxed at source is also in line with the approach expressed by the OECD in 1998 in the so-called Ottawa Taxation Framework. ¹⁷⁴OECD Committee on Fiscal Affairs, Electronic Commerce: Taxation Framework Conditions (1998) (hereinafter Ottawa Taxation Framework). In this document, the OECD stressed the goals of “fair sharing of the tax base” and the need to avoid “unintentional non taxation.” ¹⁷⁵Id. at 3. Moreover, and regardless of the potential under- or non-taxation, in such situations there exists an undesirable disjointedness between the economic and legal source of income which is being taxed. Under the Benefits Principle, the division of income between jurisdictions laying claim to it should ensure that the source jurisdiction has the primary right to tax active (business) income. ¹⁷⁶Avi-Yonah, above n. 49, at 520. Traditionally, in situations involving taxation of MNEs, this allocative outcome was justified by the fact that the host country furnished certain benefits to an MNE, such as the provision of infrastructure or education, as well as more specific government policies such as keeping the exchange rate stable or interest rates low. ¹⁷⁷Id., at 521. Source-based corporate taxation was thus thought to be justified because the host country’s government bore some of the costs of providing the benefits that were necessary for earning the income. ¹⁷⁸Id.

This justification seems problematic, however, when applied in the context of MNEs operating in the cloud. Not only may the servers on which the cloud is run be located outside of the source jurisdiction, but even if they are within such jurisdiction, it is unclear whether the provision of electricity and broadband capacity for those servers can be considered sufficient to carry the burden of justifying the allocation of income which would be mandated by the Benefits Principle. Other potential justifications might also be offered, such as developed property rights protection system, or, even more generally, rule of law. Whilst these considerations are admittedly important, their relevance is arguably less significant in the relatively narrow context of provision of IT infrastructure, even if broadly conceived, that is necessary to maintain a server. It might therefore be necessary to look elsewhere for the appropriate justification for the Benefits Principle when it is applied to cloud-based income. Perhaps a better view would be to take a more flexible approach and decouple the justification for the right to tax from provision of any goods or services by the source state and instead focus on such factors as the size of the customer base located in the source country which could be quantified, for example, by reference to the amount of data processed in the cloud and traced to the source jurisdiction. The fact that customers are located in the source jurisdiction would then furnish, via the proxy of data transfer originating in that jurisdiction, the requisite justification for the assertion of taxing rights by the source country. This view is a variation on what has been referred to as the “supply-demand” view under which the interaction of supply and demand is what creates business profits of an enterprise and the use of the source country’s legal and economic infrastructure should allow it to claim source taxing rights on a share of the enterprise’s profits. ¹⁷⁹2004 OECD Report, above n. 171, at 14. Indeed, a not too dissimilar approach has been proposed in the U.S. context with respect to determining whether a threshold has been reached for a permanent establishment to exist based on the quantum of sales into a given host jurisdiction. ¹⁸⁰Reuven S. Avi-Yonah, Virtual PE: International Taxation and the Fairness Act, Public Law and Legal Theory Research Paper Series, Paper No. 328 (2013), 2. http://ssrn.com/abstract=2259127. Under this approach, a taxpayer whose U.S. sales (defined by the destination principle, as under value-added tax and sales taxes) exceed $500,000 in any given calendar year would be deemed to have a PE in the U.S. and would be taxed on profits attributable to those sales. ¹⁸¹Id. Similarly, the right to tax could be potentially justified based on the amount of Internet traffic and attendant data transfer generated in the source jurisdiction.

The “supply-demand” approach is not, however, uncontroversial. Some OECD members do not regard an enterprise which may have access to a country’s market, even though it is not physically present in that country, as “using” that country’s infrastructure and, even if that were the case, they consider that such use would be too incidental to the business profit-making process to allow for attribution of significant portion of income to that country. ¹⁸²2004 OECD Report, above n. 171, at 14. But when these objections are considered in the context of the cloud, they lose a lot of persuasiveness. Consider the following hypothetical scenario: A cloud provider furnishes the full spectrum of cloud services, including the most comprehensive BPaaS-based BPO ¹⁸³See Part II.A, above. and IaaS models, and maintains a server in source country C. All of the provider’s clients are based in country C. Given how much economic activity can be carried on in such a comprehensive cloud infrastructure it would be rather misleading, if not outright counterfactual, to claim that the use of country C’s infrastructure is “incidental” to the provider’s business profit-making process, even though it might just be a single server. Even if some of the clients were based in jurisdictions other than C, all of the provider’s business profits derived in country C would still come through the cloud and should not be classified as “incidental.” It is therefore suggested that developments in cloud computing lend support to this variation on the “supply-demand” view which can then, in turn, be used to justify the allocation of taxing rights to the source country.

The second reason for amending the existing rules governing permanent establishments is that the existing PE concept is becoming increasingly detached from economic reality and does not reflect the way in which integrated enterprises operate ¹⁸⁴See the discussion in Part IV.D, above. (hereinafter the “Detachment Problem”). There are many soft advantages, disadvantages and spill-over effects occurring within MNEs that fall outside the range of the FSEA. ¹⁸⁵Reimer, above n. 76, at 5. As a consequence, it is very difficult under the arm’s-length approach to arrive at an accurate picture of the transactions taking place between related entities operating as part of a highly unified enterprise. Wherever size matters, both the FSEA and the arm’s-length tests for transfer pricing purposes under Article 9 are likely to be inoperable. ¹⁸⁶Id. As already explained in Part IV.D, this chasm between legal and economic reality is particularly wide in the case of MNEs that conduct their businesses in the cloud. Many commentators have suggested adopting some form of a formulary apportionment model which would either replace or complement the ALP thus narrowing the gap between the two spheres. ¹⁸⁷See Avi-Yonah, above n. 116 and Kobetsky, above n. 48.

Third, and more generally, the shortcomings of the PE concept, both as a threshold under Article 5 and as a profit center under Article 7, facilitate Base Erosion and Profit-Shifting (hereinafter the “BEPS Problem”). In its BEPS Action Plan, ¹⁸⁸BEPS Action Plan, above n. 37. the OECD has emphasized the challenges posed to international taxation by the digital economy. ¹⁸⁹Id. at 10. The OECD calls for a careful examination of how MNEs that derive their profits from the digital economy add value, in order to determine whether, and to what extent, it may be necessary to adapt the current rules to prevent BEPS. ¹⁹⁰Id. In light of the high likelihood of under- or non-taxation at the source country level, as well as having regard to the potential for transfer pricing manipulation, it is not surprising that cloud computing is expected to facilitate tax avoidance. ¹⁹¹Primavera De Filippi, Taxing the Cloud: Introducing a New Taxation System on Data Collection? Available at: http://policyreview.info/articles/analysis/taxing-cloud-introducing-new-taxation-system-data-collection. The BEPS Report itself highlights the need to amend the PE rules in order to prevent artificial avoidance of PE status and identifies transfer pricing and enforcement of the ALP as “major issues” in this respect. ¹⁹²BEPS Action Plan, above n. 37, at 19. Valuing intangibles receives particular attention with a call for development of rules that would prevent BEPS generated by moving intangibles among group members, including developing transfer pricing rules or special measures for transfers of “hard-to-value” intangibles. ¹⁹³Id. at 20. BEPS thus constitutes a problem which stems in part from the flaws in the design of Article 5 and from the shortcomings of the ALP which underlies the apportionment mechanism deployed in Article 7. Because the rise of the cloud economy facilitates BEPS-related activities, it also exacerbates this problem and thereby it further strengthens the need for remodeling the existing PE rules.

Having considered various arguments in favor of reform, I move on to assess the particular reform proposals. The policy reasons identified above will serve as criteria for evaluation, in the context of Articles 5 and 7, of the reform proposals and their respective merits and demerits. In order to determine which of the reform proposals ought to be acted upon, it will be necessary to establish, as a pre-condition, how well-positioned they are to deal with the three problems I have just identified.

B. Amending Article 5 OECD Model

It should be made clear from the outset that from the point of view of internal consistency within the OECD Model and with a view to fostering an efficient international tax regime the reforms ideally ought to be applied in both instances, i.e., when the PE serves its function as a threshold (Article 5) and when it serves the role of a profit center (Article 7). Such sweeping reform might, however, be difficult to implement in one stage and so the analysis presented here takes into account the fact that Articles 5 and 7 might have to be amended independently and perhaps also at different times. In other words, while the proposals for reform would benefit from a synergy effect if implemented together, they are not mutually dependent and can achieve their respective goals if implemented separately.

The desirability of lowering the PE threshold has been already recognized for quite some time, and certain recent developments aimed at achieving this goal are undoubtedly a step in the right direction to address the Threshold Problem and the BEPS Problem. Some of the major changes which have taken place over the last few years include acknowledgment of service PEs and unmanned facilities as PEs, a weaker interpretation of the 12-month period under Article 5(3) and finally the extension of the concept of agency PEs to cover cases where the agent does not have the legal authority to bind her principal under private law. ¹⁹⁴Reimer, above n. 76, at 2–3. But they are not sufficient to effectuate the requisite readjustment of taxing rights in favor of source states generally and also they do not go far enough to ensure that cloud-based activities reach that lower threshold.

Some of the de lege ferenda proposals have recommended, mirroring the wording adopted in Article 5(3)(b) of the United Nations Model Tax Convention, lowering the threshold by including continuous and intensive services as a type of activity that might also constitute a PE. ¹⁹⁵Reimer et al., above n. 62, at 196. This change would also eliminate a lot of inconsistencies, such as the one whereby conclusion of contracts may constitute a PE if exercised by a dependent agent, while the same activity will not qualify if exercised by the enterprise itself. ¹⁹⁶Id. Another change which might be relevant in the context of cloud computing would be to relax or abandon the requirement that a PE be “fixed” as well as allowing the business activity to be conducted “in” or “at,” rather than just “through,” the place of business. ¹⁹⁷Id. at 196–197.

But more far-reaching changes are necessary if the PE threshold is to become an appropriate tool for assigning taxing rights to cloud-based income. For example, as indicated earlier, ¹⁹⁸See Part II.D, above. the concept of the agency PE could be extended by remodeling Article 5(5) of the OECD Model in a way that would include agents that are not persons as defined in Article 3(1)(a) of the OECD Model. For example, the references to “person” in Article 5(5) could be replaced by references to “agent” which could in turn be defined in the new Article 3(1)(i) as “a person, an entity or an electronic interface.” The new Article 5(5) would then read as follows:

Notwithstanding the provisions of paragraphs 1 and 2, where an agent — other than an agent of an independent status to whom paragraph 6 applies — is acting on behalf of an enterprise or is made available by the enterprise in a similar capacity and has, and habitually exercises, in a Contracting State an authority to conclude contracts in the name of the enterprise, or such authority is otherwise asserted in a Contracting State by the enterprise with certain regularity, that enterprise shall be deemed to have a permanent establishment in that State in respect of any activities which that agent undertakes for the enterprise, unless the activities of such agent are limited to those mentioned in paragraph 4 which, if exercised through a fixed place of business, would not make this fixed place of business a permanent establishment under the provisions of that paragraph.

 

The new Article 3(1)(i) would read:

(i) the term “agent” includes persons, entities and electronic interfaces.

 

Moreover, to maintain consistency, the reference to “person” in Article 5(6) would also have to be changed to “agent.” The new definition could then apply to websites as well as other electronic interfaces such as those deployed within the SaaS or PaaS frameworks through which “authority to conclude contracts in the name of the enterprise” is exercised. This amendment would have the advantage of having limited disruptive effect on other provisions of the OECD Model while being sufficiently broad in its scope of application to potentially capture significant amounts of cloud-based economic activity, or e-commerce more generally, performed in the source country. At the same time, the additional qualifications contained in Article 5(5), such as the requirement that the authority be exercised/asserted habitually or with a certain regularity ¹⁹⁹The “regularity” requirement would mirror the requirement currently embodied in Article 5(5) that an agent must “habitually” exercise an authority to conclude contracts which reflects the underlying rationale that “the presence which an enterprise maintains in a Contracting State should be more than merely transitory if the enterprise is to be regarded as maintaining a permanent establishment, and thus a taxable presence.” OECD Model, above n. 40, at 106. would ensure that the reference to electronic interfaces does not capture incidental or otherwise auxiliary activities performed by the enterprise in the source country. In sum, the proposed change would address the Threshold Problem by lowering the threshold with respect to non-trivial amounts of electronic commerce, including cloud-based commerce, thereby facilitating appropriate allocation of income between residence and source countries. It could also deal, at least partially, with the problem of artificial avoidance of PE status by expanding its scope, thus mitigating the BEPS Problem.

C. Article 7 and Formulary Apportionment: Evolution or Revolution?

The ALP has been traditionally contrasted with the formulary apportionment method and indeed the debate between the proponents of those two approaches has spanned more than 30 years. ²⁰⁰Avi-Yonah, above n. 116, at 1. The shortcomings of the ALP standard have already been analyzed in Parts IV.B and IV.D and it is therefore now pertinent to focus on the advantages and disadvantages of the FA method and to assess how, if at all, it can address the Detachment Problem and the BEPS Problem.

Two types of formulary apportionment can be distinguished. One refers to using a formula to attribute the profits of a company to different jurisdictions while the other, the so-called unitary formulary apportionment (“UFA”) concerns the treatment of a functionally integrated group of companies, such as an MNE, as one entity for apportionment purposes. ²⁰¹Kobetsky, above n. 48, at 403. An MNE would be treated as one entity for tax purposes under the UFA, but under the ALP it would be considered as a group of independent enterprises dealing at arm’s length. ²⁰²Id. It is essential, however, to emphasize an important distinction at this stage in order to avoid unnecessary confusion as to which concepts are being actually subject to analysis. Although a unitary system such as the UFA requires an allocation formula, the two terms are not equivalent and must be analytically distinguished. ²⁰³Avi-Yonah & Benshalom, above n. 85, at 11. Formulary allocation in itself is concerned exclusively with allocating income through an appropriate formula — it does not attempt to determine the market price of the relevant underlying transactions. ²⁰⁴Id. Thus while formulary sourcing requires distinguishing certain sources of income from other sources in order to identify those to which the formula might be applied, it does not depend, in contrast to the unitary regime, on the ability to consolidate the income of the whole enterprise. ²⁰⁵Id. The focus in this Part is on the formulary allocation which does not have to operate within a consolidated unitary setting. The reason for such preference is largely pragmatic. Given the long-standing resistance of the OECD against the adoption of the UFA, ²⁰⁶See, e.g., BEPS Action Plan, above n. 37, at 14. it is unlikely that such change would be considered let alone implemented. Indeed, even the proponents of the unitary method themselves concede that a worldwide implementation of the UFA on the basis of a global multilateral tax treaty is “idealistic, and support for this approach is unlikely to be readily achieved.” ²⁰⁷Kobetsky, above n. 48, at 403. In these circumstances it is hard to deny the persuasive force of the argument that practical difficulties associated with agreeing to and implementing the details of a new system consistently across all countries are very significant. ²⁰⁸Id. at 20. A less radical and wide-ranging reform might still achieve considerable success in dealing with both the Detachment Problem and the BEPS Problem without the need for a very costly and time-consuming multilateral action. The analysis presented here will therefore follow the narrower approach which was outlined by Professors Avi-Yonah and Benshalom in their 2010 paper Formulary Apportionment — Myths and Prospects, ²⁰⁹Avi-Yonah & Benshalom, above n. 85. as complemented by Professor Avi-Yonah’s more focused analysis devoted specifically to the 2010 OECD guidelines. ²¹⁰See n. 116, above. In the following paragraphs, I assess to what extent these proposals are capable of dealing with the Detachment Problem and the BEPS Problem in the context of cloud computing.

From a theoretical perspective, as far as the degree of arbitrariness is concerned, the formulary alternatives are probably as arbitrary as the ALP, but from a revenue-generating perspective, formulary arrangements are probably less arbitrary — because they are less susceptible to manipulation by intra-MNE contractual arrangements. ²¹¹Avi-Yonah & Benshalom, above n. 85, at 13. Moreover, as already signaled above, ²¹²See Part IV.D, above. the increasing reliance by tax authorities in transfer pricing cases on the Profit-Split Method further strengthens the case in favor of the FA approach. Formulary Apportionment would then be narrowly deployed in order to apportion the unallocated residual which arises precisely because MNEs exist to generate returns that would be unobtainable if their respective units operated at arm’s length. The assets and activities which form such residual are generally thought to generate most of current transfer pricing compliance and administrative costs, as well as tax avoidance opportunities. ²¹³Avi-Yonah & Benshalom, above n. 85, at 2. The proposal put forward by Professor Avi-Yonah is thus particularly relevant because it offers a solution to how the residual should be allocated. ²¹⁴Avi-Yonah, above n. 116, at 12.

Yet this solution generates its own problems, quite apart from the problem of how to distinguish between the residual and the non-residual, ²¹⁵Avi-Yonah & Benshalom, above n. 85, at 16. especially when considered in the context of transfer of intangibles within MNEs, an issue which is particularly relevant for MNEs operating in cloud environment.

The first problem is more general and relates to the potential inaccuracy of formulary methods, in particular the distorting effect of the labor cost factors. ²¹⁶James R. Hines Jr., Income Misattribution Under Formula Apportionment, 54 Euro. Econ. Rev. 108–120, 118 (2010). Yet as argued above, the scope of application of the FA would be relatively narrow and would only deal with the unallocated residual thereby limiting the distortive effect. Moreover, given the nature of e-commerce generally and of cloud computing in particular, the labor cost factor would provide very poor guidance on income allocation and so the future FA formula would in any case have to be tweaked in order to address this issue, with more reliance placed on other factors. I defer the discussion of what these factors could be to the paragraph in which I deal with the inclusion of intangibles in the formula.

The second problem relates more specifically to Professor Avi-Yonah’s proposal under which an apportionment formula based on payroll, tangible assets and sales would be adopted to determine how the residual is to be allocated between the respective units of the enterprise, but it would exclude intangibles. ²¹⁷Id. at 12. This aspect of the proposal is very unsatisfactory not only because, as already established, the ALP is inadequate to deal with the valuation of intangibles, but also, more generally, because the market value of intangible assets is likely to be very high ²¹⁸C.E. McLure, U.S. Federal Use of Formula Apportionment to Tax Income from Intangibles, 14 Tax Notes Int’l, 859, 865 (1997). and so their exclusion from the formula is likely to be far removed from economic reality. This is particularly likely for MNEs that transfer cloud-related intangibles because of the ease with which such assets can be transferred or deployed within the cloud. Thus the proposed solution, insofar as it completely excludes intangibles from the apportionment formula, cannot be counted on to deal effectively with the Detachment Problem. Similarly, when one looks at how the FA formula would address the BEPS Problem, it is far from clear that the outcome would be satisfactory. As such, we are faced here with the following paradox. On one hand, the inclusion of intangibles in the formula should be supported on the grounds that it would mitigate the Detachment Problem. On the other hand, such inclusion might facilitate the abuse of any given formulary apportionment formula by allowing MNEs to manipulate the location of intangibles, especially when they operate in a cloud environment — thereby exacerbating the BEPS Problem. There are two potential responses to this conundrum.

First, accept those deficiencies of the FA model, which excludes intangibles but is still arguably superior to the ALP formula. It narrows the gap inherent in the Detachment Problem, and its impact on the BEPS Problem is unlikely to be more negative than that of the ALP-based formulas. The FA also seems to be preferable because of its potential to decrease uncertainty surrounding the outcomes of transfer pricing disputes. The degree of uncertainty for taxpayers can be said to be inversely correlated with the likelihood of enterprises to anticipate how tax authorities or judicial authorities will apply the ALP. When we consider the capacity of the MNEs to anticipate tax consequences of their cross-border transactions, the effectiveness of formulary methods is likely to be higher than that of the methods informed by the ALP. ²¹⁹Andrea Musselli and Alberto Musselli, Saving Arm’s Length Pricing: From Economists’ Myths of Tax Avoidance by Taxpayers, to the Reality of Uncertain Application of Rules, Int’l Transfer Pricing J., at 402 (Nov./Dec. 2012).

Second, and in the alternative, the FA formula could be modified so as to include intangibles while at the same time attempting to mitigate the risk of manipulation of apportionment outcomes. This solution, however, generates its own difficulties and at least two objections could potentially be raised against it: (1) that because the value of intangibles results from physical and human capital and from the market, and those elements are already included in the formula, ²²⁰Avi-Yonah, above n. 116, at 12. the marginal increase in accuracy of income apportionment will be very small; and (2) that any such increase, especially one on the margin, will be outweighed by the increased likelihood of abuse or manipulation. As for the first objection, it seems questionable in the context of cloud computing that the value of PaaS or SaaS is to any significant extent derived from physical or human capital, especially after the initial stage of development. The market, understood as a customer base, is perhaps a better proxy for the value of an intangible but in many cases the software might generate value by means of endogenous synergies, as might be the case under the BPaaS or iPaaS models, without any exogenous market factors coming into play. In the context of cloud-related intangibles it is therefore unlikely that their value can be accounted for by reference to capital, and the residue for which the market can serve as proxy is likely to be insufficient. These are therefore compelling reasons to expand the FA formula so that it can take into account the cloud-based intangibles.

The second objection is more serious, and it might help to introduce a distinction between acquired and developed intangibles in order to better elucidate the inherent difficulty. Intangibles acquired from third-party providers could be valued by reference to the acquisition cost. While definitely imperfect, the acquisition price would still be a reasonably accurate proxy for the value of an intangible and the scope for manipulation would be quite limited. As for the internally developed cloud software a similar solution could be adopted to the one employed by the EU when valuing intangibles within the framework of the Common Consolidated Corporate Tax Base. ²²¹Joann Martens Weiner, CCCTB and Formulary Apportionment: The European Commission Finds the Right Formula in Dennis Weber, CCCTB: Selected Issues (2012), at 259. The factors would include costs incurred for research and development and potentially also costs associated with marketing and advertisement within a pre-specified time period. ²²²Id. Of course, this solution has its own difficulties, as the value of intangible might be much higher than the cost incurred for R&D which led to its creation, and the probability of abuse is likely higher than in case of externally acquired intangibles. This difficulty could be dealt with by some of the anti-abuse rules contained in the OECD Model. ²²³According to the 2010 OECD Commentaries, one of the guiding principles underlying the Model Convention prescribes that “the benefits of a double taxation convention should not be available where a main purpose for entering into certain transactions or arrangements was to secure a more favourable tax position and obtaining that more favourable treatment in these circumstances would be contrary to the object and purpose of the relevant provisions.” 2010 OECD Commentaries, above n. 47, at 61. For a recent discussion of the nature of anti-avoidance rules in tax treaties, see Shee Boon Law, Anti-Avoidance Rules in Recent Tax Treaties, Bulletin for Int’l Tax’n (June 2012), 319–322. Whether the dispute resolution costs associated with the application of anti-abuse provisions would be higher than those currently incurred due to transfer pricing litigation is unclear. This would to a large extent depend on whether the existing, relatively broad, anti-abuse measures would be deployed, or if a more targeted rule, tailored specifically to deal with the abuse in the context of formulary apportionment would be adopted. Despite its salience, however, the question of applicability of anti-avoidance rules and their capacity to deal with the potential abuse of intangibles valuation lies beyond the scope of this article. The use of anti-avoidance rules and their capacity address the weaknesses of the FA formula undoubtedly merits much closer scrutiny.

If the foregoing analysis is correct and the attendant difficulties which I have identified can be satisfactorily resolved, then it might be possible to amend the profit attribution mechanisms under the OECD Model in such a way as to introduce some elements of the formulary apportionment approach along the lines advocated by Professor Avi-Yonah, which would, however, be modified to account for intangible assets. Such a modified FA formula would then have the capacity to successfully deal with the Detachment Problem while simultaneously addressing the BEPS Problem. If implemented in conjunction with Article 5 amendments that the OECD proposed (Part V.B above), the overall reform might potentially solve the Threshold Problem as well.

VI. CONCLUSION

The analysis presented in this article has identified certain key challenges raised by cloud computing with respect to permanent establishments under the OECD Model. While the nature of these challenges does not seem to be significantly different from those raised by e-commerce in general, the cloud nonetheless exhibits many idiosyncrasies which exacerbate the problems currently encountered in the areas of threshold determination, allocation of business profits and the deployment of arm’s-length methods for the purposes of transfer pricing in general.

With respect to the normative debate and the need to address increasing disjointedness between the provisions of OECD Model and economic reality as well as in relation to Base Erosion and Profit Shifting, the rise of cloud computing lends significant support to the following arguments. First, making the threshold under Article 5 of the OECD Model more flexible so that more income can be allocated to the source jurisdictions, in line with the volume of economic activity taking place in those countries. Second, readjusting the profit allocation mechanisms under Article 7 by greater reliance on profit allocation methods based on formulary apportionment at the expense of the increasingly inadequate arm’s-length standard.