Permanent Establishment Aspects of Transaction Processing on U.S. Servers

The Tax Management Transfer Pricing Report ™ provides news and analysis on U.S. and international governments’ tax policies regarding intercompany transfer pricing.

By Gary D. Sprague, Esq.
Baker & McKenzie LLP, Palo Alto, CA

Transaction processing undertaken for the benefit of a nonresident enterprise, executed by software running on servers located in the United States, where the servers are not owned or leased by the nonresident enterprise, normally should not give rise to a U.S. permanent establishment (PE) of the nonresident enterprise.  This question arises frequently in practice. U.S. multinationals commonly concentrate their data processing capacity for global transactions in the United States. Many foreign multinationals utilize the technological infrastructure of the United States to support their global businesses.  The absence of guidance on the issue means that tax uncertainty unnecessarily exists with respect to an important element of the commercial infrastructure of many multinationals.

E-commerce business models have presented challenging issues in many areas of international tax law. In the PE area, the issue in many cases involves the ability of an enterprise to sell goods or services into a market from a remote location, while maintaining a smaller business presence in the market jurisdiction than has been the case historically for many industries. The natural tendency of tax administrations is to question whether the PE standards should be modified in an attempt to capture more of the profit from those transactions.

At the same time, and probably driven partly by the adoption of e-commerce business models, tax administrations and policy makers recently have given increased attention to the interpretation of the PE standard generally. This greater focus on PEs has been manifested in the field in a large increase in the number of controversies over PE issues. On the policy level, issues arising under the PE rules of Article 5 of the OECD Model Tax Convention have been something of a constant focus for several years at the OECD. As noted at the recent International Fiscal Association Congress in Vancouver, Working Party 1 will be taking another look at certain difficult and contentious PE issues over the next couple of years, with the goal of clarifying the commentary on the Model with respect to those issues.

The issue of how to apply the PE rules to transaction processing executed through software residing on a server established at a particular location, therefore, sits at the center of this evolving area. All companies that enter into contracts through online transaction processing for sales booked by an entity taxable in a jurisdiction other than the one in which the server is located must address this issue. The impact is not limited to pure e-commerce companies; an enterprise that sells widgets on-line must address the same questions.

The business model is as follows. A nonresident enterprise is engaged in the business of selling goods or services.  In some cases, prospective purchasers are visiting the website of a particular merchant and have decided to purchase the goods or services.  In other cases, the solicitations may have taken place through other channels, and the prospective purchaser has been referred to the website to complete the transaction. In either case, purchasers are required to enter into contracts for the purchase of such goods or services through a website hosted on servers that could be located outside the jurisdiction of either the buyer or the seller. As part of the online contracting process, the customer may be required to submit credit card and other identifying information. Credit verification happens automatically, through established routines in which the potential customer's creditworthiness is established. Normally the customer is presented with standard terms; price and other conditions of sale are not negotiated.

At the end of the process, a contract has been formed. Offer and acceptance have been communicated. The customer has agreed to pay a certain amount in exchange for the delivery of a certain good or service. The enterprise on whose behalf this contract has been formed has agreed to supply such good or service. Does any of this create a U.S. PE if the server through which this contract conclusion has been automated is in the United States, but the enterprise selling the goods or services is not?

Under the OECD Model Tax Convention, there are two principal ways in which a PE can be created. First, under the "basic rule" PE of Article 5(1) of the OECD Model, a PE exists if the nonresident enterprise maintains a fixed place of business through which the business of the enterprise is wholly or partly carried on.  The commentary elaborates on this basic rule, providing that the basic rule requires the following conditions to exist:

— the existence of a "place of business," which can be constituted, in certain instances, by machinery or equipment;

— this place of business must be "fixed," i.e., it must be established at a distinct place with a certain degree of permanence; and

— the carrying on of the business of the enterprise through this fixed place of business, which can be done through persons who, in one way or another, are dependent on the enterprise.

Alternatively, under Article 5(5) of the Model, a PE can be deemed to exist through the actions of certain dependent agents. Even if a basic rule PE does not exist, the nonresident enterprise can be deemed to have a PE if a person, other than an agent of an independent status as defined in Article 5(6), 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 (unless the activities of such person are limited to preparatory or auxiliary activities as mentioned in Art. 5(4)).

So how should these rules be applied in the case of automated transaction processing?

First, the basic rule. If the nonresident enterprise establishes a U.S. data center and owns or leases the equipment directly, it could well have a PE in the United States. The commentary is clear that the ownership of automated equipment can give rise to a fixed place of business. The more recent commentary relating to computer servers extends that principle to computer equipment. If the business activities conducted through that equipment constitute only preparatory or auxiliary activities, then no PE would be created. The commentary in paragraph 42.7 provides examples of activities that might be performed through computer equipment and would generally be regarded as preparatory or auxiliary activities, as follows:

— providing a communications link — much like a telephone line — between suppliers and customers;

— advertising of goods or services;

— relaying information through a mirror server for security and efficiency purposes;

— gathering market data for the enterprise;

— supplying information.

The acceptance of orders for the sale of goods or services, however, would seem to be a core activity of a business seeking profits from those sales, and thus likely would not be excluded from PE treatment on the basis of the Art. 5(4) preparatory or auxiliary exclusions, in cases where the nonresident has a fixed place of business through the ownership or leasing of computer equipment.

As a result, it is common for nonresident enterprises to not establish U.S. data centers comprised of servers that are owned or leased by the nonresident enterprise itself. Instead, the servers normally would be owned or leased by a U.S. affiliate, which then provides hosting services to the nonresident enterprise. Alternatively, if the group prefers to outsource the function entirely, the nonresident enterprise may enter into an outsourcing contract with a third-party hosting services provider. The U.S. affiliate or third-party service provider normally would employ those personnel responsible for the management and maintenance of the data center premises and the hardware contained within it. The software applications residing on the U.S. servers could be accessed by the nonresident enterprise, and personnel of the nonresident enterprise would be able to manipulate remotely the data hosted on the U.S.-located server.

In this case, there should be no basic rule PE on the basis that the nonresident enterprise does not have any premises or equipment at its disposal at the location of the data center. The Article 5 commentary is clear that a fixed place of business does not exist unless the enterprise has at its disposal the physical place that is asserted to constitute the PE. It is clear that in most cross-border hosting arrangements the nonresident purchaser of the hosting services has the ability to manipulate data hosted on the servers from a remote location. The customers too can access the server remotely, and obviously do so every time a transaction is entered into through a server located in a jurisdiction different from that of the purchaser. The ability to access data or manipulate the software application resident on the server, however, should not be relevant for the "basic rule" analysis. The basic rule looks to a physical presence, and the only enterprise that maintains a physical presence in the United States in this fact scenario is the provider of the hosting services which owns and operates the assets comprising the data center.

This conclusion is confirmed by paragraph 42.3 of the commentary in the case of an internet service provider (ISP) by distinguishing the physical assets of the server from the website (and presumably other software applications) hosted by the server:Although the fees paid to the ISP under such arrangements may be based on the amount of disk space used to store the software and data required by the web site, these contracts typically do not result in the server and its location being at the disposal of the enterprise…, even if the enterprise has been able to determine that its web site should be hosted on a particular server at a particular location. In such a case, the enterprise does not even have a physical presence at that location since the web site is not tangible. In these cases, the enterprise cannot be considered to have acquired a place of business by virtue of that hosting arrangement.

While the authors of this paragraph may have had unrelated ISPs in mind, there is no reason to distinguish between unrelated- and related-party services in this context.

The commentary dealing with electronic commerce also holds some guidance regarding whether Article 5(5) could apply to transaction processing carried out through a source state server to cause a deemed PE to exist under the dependent agent rules. Referring again to ISPs, paragraph 42.10 provides:The issue may then arise as to whether paragraph 5 may apply to deem such ISPs to constitute permanent establishments of the enterprises that carry on electronic commerce through web sites operated through the servers owned and operated by these ISPs. Whilst this could be the case in very unusual circumstances, paragraph 5 will generally not be applicable because the ISPs will not constitute an agent of the enterprises to which the web sites belong, because they will not have authority to conclude contracts in the name of these enterprises and will not regularly conclude such contracts or because they will constitute independent agents acting in the ordinary course of their business, as evidenced by the fact that they host the web sites of many different enterprises. It is also clear that since the web site through which an enterprise carries on its business is not itself a "person" as defined in Article 3, paragraph 5 cannot apply to deem a permanent establishment to exist by virtue of the web site being an agent of the enterprise for purposes of that paragraph.

This paragraph takes something of a "cover the waterfront" approach to the analysis, in that it refers to several possible theories by which the ISP should not be regarded as creating a deemed PE under Article 5(5). One of the theories mentioned, however, is the critical one, as it is the correct application of an analysis that should apply to many questions arising from e-commerce transactions. The theory is that the ISP itself does not have or exercise authority to conclude contracts in the name of the nonresident enterprise. This is the correct conclusion because even though contracts indeed are formed between buyer and seller through user interactions with the website, the authority to conclude those contracts cannot be said to be exercised by an economic actor whose only role is to maintain computer equipment. Similarly, the software application itself cannot be regarded as exercising an authority to conclude contracts; a software program exercises no more authority than the servers do.  This analysis applies regardless of whether the hosting services provider is a related or an unrelated entity.

Despite the central position that the dependent agent concept holds in the PE firmament, the Article 5 commentary is relatively sparse regarding the contours of what it means to have and habitually exercise an authority to conclude contracts. What guidance as does exist reflects the push and pull in the recent past over certain current developments, including commissionaire structures, national court decisions, and the like.

Of relevance to the issue of automated transaction processing is language that seems to express the economic purpose of the deemed PE rule. Paragraph 33 provides:A person who is authorized to negotiate all elements and details of a contract in a way binding on the enterprise can be said to exercise this authority "in that State," even if the contract is signed by another person in the State in which the enterprise is situated … .

Furthermore, paragraph 32 of the commentary notes that the application of the dependent agent rule is intended "to be limited to persons who in view of the scope of their authority or the nature of their activity involve the enterprise to a particular extent in business activities in the State concerned."

Both passages seem to reflect the judgment that the dependent agent rule is to be applied to cases where the agent is exercising business discretion, in a way that constitutes an extension of the business of the nonresident enterprise into the source state.  The execution of a software application cannot be considered the exercise of business judgment. Even in the case of a program that incorporates artificial intelligence, the business intelligence is that of the persons who set the terms of trade, not the application files that execute instructions provided by others.

Accordingly, there is no policy reason that would justify transaction processing, standing alone, as justifying the constitution of a PE under dependent agent theory. A confirmation of this point, as well as the analysis above under the basic rule, would provide a useful incentive for multinational groups to take advantage of the technological resources of the United States without fear of unwarranted tax consequences.

This commentary also will appear in the January 2010 issue of the Tax Management International Journal.  For more information, in the Tax Management Portfolios, see Isenbergh, 900 T.M., Foundations of U.S. International Taxation,  and in Tax Practice Series, see ¶7140, U.S. Income Tax Treaties.


Request Transfer Pricing Report