Group Suggests Expanding Scope of Cloud Transaction Regs

November 11, 2019

Hon. Charles P. Rettig
Commissioner
Internal Revenue Service
1111 Constitution Avenue, NW
Washington, DC 20224

CC:PA:LPD:PR (REG-130700-14)
Room 5203
Internal Revenue Service
P.O. Box 7604
Ben Franklin Station
Washington, D.C. 20044

Re:Comments on:
Prop. Reg. Sec. 1.861-7(c)
Prop. Reg. Sec. 1.861-18
Prop. Reg. Sec. 1.861-19
Classification of Cloud Transactions and Transactions Involving Digital Content

Dear Commissioner Rettig:

The Software Finance and Tax Executives Council (“SoFTEC”) appreciates the opportunity to provide comments with respect to REG-130700-14, Classification of Cloud Transactions and Transactions Involving Digital Content. SoFTEC's members support the development of these proposed regulations and look forward to having guidance with respect to these rapidly evolving business models. Our comments include suggestion for expanding the scope of the proposed regulations and clarifying certain provisions.

SoFTEC is a trade association providing software industry-focused public policy advocacy in the areas of tax, finance and accounting. SoFTEC is the voice of the software industry on matters of tax policy. Because all SoFTEC members distribute their products worldwide, they have an interest in the application of the U.S. tax laws to the profits they earn from foreign distribution and thus have an interest in commenting on these proposed regulations.

A. Comments with respect to proposed changes to 1.861-18:

1. Expansion to other digital content:

Currently, Reg. Sec. 1.861-18 (hereinafter referred to as the “-18 regs) is limited to classifying transfers of computer programs as either the transfer of a copyright right or the transfer of a copyrighted article. The proposal is to expand the coverage of the -18 regs to include transfers of other digital content, such as movies music and books, that either are or were subject to copyright protection.

The preamble to the proposed regulations specifically asks for comments on whether “the definition of digital content should be defined more broadly than content protected by copyright law and content that is no longer protected by copyright law solely due to the passage of time.” We believe it should. We know of no reason why the classification of digital content rules should turn on whether the content is copyrightable.

There is a wide variety of digital content that is not eligible for copyright protection. See, e.g., Feist Pubs., Inc. v. Rural Tel. Svc. Co., Inc., 499 U.S. 340 (1991). Developers of such content may rely on a variety of other forms of intellectual property protection. Uncopyrightable digital content can be transferred under contractual restrictions as to its use. See ProCD, Inc. v. Zeidenberg, 86 F.3d 1447 (7th Cir. 1996). Use of non-disclosure agreements under trade secret law to protect non-copyrightable elements of digital content also is available. See MAI Systems Corp. v. Peak Computer, Inc., 991 F.2d 511 (9th Cir. 1993).

Under the current -18 regs, classification of computer programs turns not on whether the computer program is eligible for copyright protection, it turns on what the recipient of the copy is allowed to do with it. If the recipient is allowed to make copies of the program and resell the copies, then the transaction is a license generating royalty income. If the recipient does not obtain the right to make copies and resell them, then the transaction is either a sale or exchange or a lease. See Reg. Sec. 1-861-18(a)(2).

We believe such rules are easily adaptable to transactions involving categories of digital content not eligible for copyright protection. If the recipient of the digital content is allowed to further exploit it in the marketplace, then the income should be classified as royalty income. If the recipient is not allowed to further exploit the digital content but will be making use of the content, then the income should be classified as either from a sale or exchange or lease income.

In addition, if the final regulations expand the scope beyond copyrighted digital content, then we suggest the phrase “copyright right” be replaced with the phrase “rights with respect to digital content” throughout.

2. Treatment of public performance and public display rights:

Because computer programs themselves generally are not susceptible to public performance or public display, there was no reason for the current -18 regs to provide guidance on transfers of public performance and public display rights. However, the proposal is to expand of the scope of the -18 regs to include other digital content, such a movies, music and books, which raises all kinds of opportunities for transfers of such rights. We believe the proposed regulations rightly characterize transfers of such rights for the limited purpose of advertising as not transfers of copyright rights.

3. Expansion of the application of 1.861-18 to other areas of the Internal Revenue Code:

Reg Sec. 1.861-18 (a)(1) provides that the -18 regs only apply to certain, enumerated, parts of the Internal Revenue Code. We believe the classification rules of the -18 regs should apply any time the facts and circumstances make them relevant to the proper determination of a taxpayer's liability under any provision of the Code. Because both taxpayers and the IRS will utilize the guidance in the -18 regs by analogy despite the limitation in Reg. Sec. 1.861-18(a)(1), it only makes sense to make clear the provisions apply whenever they might be relevant to a tax determination and not make one party have to go through the rigors of trying to convince the other of such.

B. Sourcing Sales of Digital Content:

As part of the expansion of the scope of the -18 regulations, the proposal includes changes adding specific rules for sourcing sales of digital content. The proposal is to add a new Sec. 1.861-18(f)(2), along with a cross-reference in Reg. Sec. 1.861-7(c). Under new -18(f)(2)(ii), sales of digital content are to be sourced to the download location and in cases where the seller is without information regarding the download location, the taxpayer is permitted to use a location based on the recorded sales data. We do not believe sourcing sales of digital content to the location of the purchaser at the time of download provides a useful sourcing touchstone as it would require the capture and retention of information that may not provide reliable information as to user-location. We believe that in all cases, recorded sales information, if available, is the most reliable proxy for the location of the end user of a cloud computing service as it is possible to capture, retain and audit and is less susceptible to manipulation through the use of virtual private networks and other technology that masks the user's location. The final regulations also should clarify that “recorded sales data” with respect to a transaction includes the customer's billing address or the mailing address in the seller's books and records associated with the credit card used by the customer.

We therefore suggest that the final regulations eliminate references to the location of download or installation and instead provide that the taxpayer may source a sale of digital content to the country of the billing address of the first unrelated purchasing entity. While many taxpayers currently may not have the capability to look through the purchaser's billing address to the countries where the purchaser will be deploying the digital content, such capability may become available in the future. Also, a taxpayer may have the ability to look through the billing address to the country of actual use for some customers and not others. We suggest that taxpayers be allowed to elect to source sales of digital content using such a look through method. Under this election, we request that the taxpayer must only show that the supporting information, if accurate, would allow the taxpayer to make reasonable conclusions on the locations of the actual use. We also suggest that the final regulations make clear that the title transfer rule of proposed -18(f)(2)(ii) only applies for purposes of Reg. Sec. 1.861-7(c).

The following proposed examples illustrate application of these two proposals:

In cases where the seller is without recorded sales data with respect to a category of transactions, as discussed below, other proxies for user-location may be appropriate. We envision fact patterns where a taxpayer may be selling digital content and allow its customers to make payment through a third-party payment facilitator under circumstances where no location information or billing address for the customer is disclosed to the taxpayer. In such cases, the taxpayer will have no “recorded sales data” for such sales. We suggest the final regulations include guidance on where a taxpayer should source such sales. We propose that taxpayers be given a choice of one of three countries to which such sales may be sourced: (1) the county of the headquarters of the seller's business, (2) the country where the seller has the greatest number of employees or (3) a country from which the seller makes digital content available for electronic delivery. The country chosen by the taxpayer must be used for sourcing all sales of digital content that fall into this category.¹

C. Classification of Cloud Transactions:

While the proposal to add a new Section to the guidance under Section 861 providing rules for the classification of so-called “cloud transactions” is welcome, we do not know of any cloud transactions that could also be characterized as leases. For this reason, as described below, we recommend the final regulations clarify that transactions classified as leases are not cloud transactions and that all transactions characterized as cloud transactions are classified as the provision of services. In the alternative, the final regulations should include a rebuttable presumption that all cloud transactions are classified as the provision of services.

1. Predominant Character Test:

The proposed regulations in Sec. 1.861-19 (a)-(c) provide a confusing structure for analyzing so-called “cloud transactions” as either the provision of services or leases of property. We believe the analysis can be simplified first by removing the terms “arrangement” and “de minimis.” Second, we believe that the term “transaction” should be defined based on the facts and circumstances. Such facts and circumstances may include the availability of separate pricing, the use of separate SKUs, the taxpayer's definition for non-tax purposes, etc. Third, the character of a transaction should be determined by the predominant character of the transaction. The predominant character should be determined according to the facts and circumstances. These may include the overall commercial purpose, the taxpayer's treatment for non-tax purposes, the relative cost of each component, a comparison of unit prices for components sold separately, etc. The facts and circumstances must provide at least a reasonable basis for determining the predominant character of the transaction.

Following is an example that illustrates the application of the proposal described above:

Guidance on when access to digital content constitutes a lease or is a cloud transaction also would be helpful. For instance, whenever the customer's rights are limited to the download of a discrete item of digital content, such as a specific movie, and view it for a set period of time for a set price, the transaction should be characterized as a lease. However, transactions where the customer, for a monthly fee, can access and download as many movies as desired from a catalog of thousands of movies, and once the subscription ends access to the downloaded movies ends, should be characterized as a cloud transaction and classified as a service. We believe whether the customer can download the content should not be the axel around which characterization of the arrangement as the supply of a service or a lease turns. A taxpayer's characterization of such transactions as cloud transactions is presumptively reasonable.

2. Rebuttable Presumption Proposal:

The preamble asks that commenters provide realistic examples of cloud transactions that could be treated as leases of property; we have been unable to conjure any such examples. If Treasury is unable to incorporate the predominant character test into the final regulations, as outlined above, we suggest that the final regulations include a rebuttable presumption that all cloud transactions are the supply of a service. The presumption might be rebutted with evidence that the taxpayer transferred possession of digital content or tangible personal property and any service supplied in connection with the cloud transaction is de minimis. The so-called “streaming” of digital content would be considered the supply of a service that is more than de minimis. Use of such a presumption would simplify the analysis and require application of a facts and circumstances analysis for only the few outlier transactions. Also, the rules for determining when a transaction is comprised of more than one transaction, some of which are cloud transactions and some of which are not, further narrows the possibility for what otherwise would be the supply of a service to be characterized as the lease of property.

3. Application to Related Party Transactions:

Cloud transactions, as defined in the Proposed Regulations, can be interpreted as including purchases of data center services that allow purchasers to operate dedicated servers in a remote location without the ownership of a physical data center (“DC Operations”). For purposes of the Proposed Regulations, DC Operations are distinct from end-user cloud transactions where a consumer purchases software, platform, and/or infrastructure as a service.

The distinction between DC Operations and an end user cloud transaction is particularly relevant where DC Operations are provided to a related party. Global cloud service providers serve their customers through distributed networks of data centers organized into regions. The largest cloud service providers (“CSP”) currently have 20+ regions. There may be multiple regions in one country (e.g., US East 1, 2 and 3). Separate legal entities in each country generally own and operate the local data center(s). Such legal entities are remunerated under an intercompany agreement (“DC Agreement”) with the CSP principal entity. Taxpayers currently determine the characterization of payments made under the DC Agreements under general US federal income tax principles, including §7701(e) and its judicial interpretations.²

Prop. Treas. Reg. §1.861-19(c)(1) states that a cloud transaction is classified as solely a lease or provision of services by taking into account all relevant factors including the factors set forth in paragraph (c)(2). The relevance of any factor varies depending on the factual situation, and one or more of the factors set forth in paragraph (c)(2) may not be relevant in a given instance. Further, the preamble to the proposed regulations explains that “the interpretation of factors and their application to cloud transactions require an analysis that is sensitive to the inherent differences between transactions involving physical access to property and transactions involving on-demand network access.”

The factors for distinguishing a lease from a service, as set out in Prop. Reg. 1.861-19(c)(2), are appropriate only for third-party cloud transactions. In these instances, factors such as whether the customer controls the property, whether the customer bears risk of substantially diminished receipts or is receiving 'concurrent services' from an unrelated person (as examples) will be unambiguous in determining “services” characterization. However, applying factors appropriate for third-party transaction characterization is not appropriate for determining the character of related-party DC Agreement payments. This is because a more expansive factual inquiry is required to address the factors relevant under existing law and their application to the legal and commercial arrangements between a CSP, related DCs and other related and unrelated suppliers of components making up the final cloud services offered by the CSP to its third-party customers. Specifically, the four modifications to the factors listed in §7701(e), as provided in Prop. Reg. 1.861-19(c)(2), will increase the complexity in charactering related-party DC Agreements Further, these modifications will result in, with respect to related-party DC Agreements: i) characterization rules that are difficult to administer; ii) unneeded additional law without a clear benefit; and iii) a perverse incentive to move functions to local country markets. As discussed below, these modifications to the specifically enumerated characterization factors in Sec. 7701(e) should not be applied to related-party DC Agreements.

First, §7701(e)(1)(B) requires the determination of whether “the service provider controls the property”. Under current law, the interpretation of §7701(e)(1)(B) and related authorities provide a reasonable basis supporting the conclusion the CSP customer did not have control of the datacenter assets. The proposed regulations also add the following clause, “beyond the customer's network access and use of the property” to the factor under §7701(e)(1)(B). It appears the proposed regulation attempts to define what “control” means (or doesn't mean) by this clause. For unrelated third-party cloud services transactions, this is appropriate. However, this clause causes confusion in the context of a DC Agreement. CSPs often utilize centralized cloud operations teams that are responsible for global delivery of cloud services, which includes repairing any breakages or curing outages. To achieve this, the cloud operations teams remotely monitor performance of datacenter assets, upload software to data center assets, and can turn off servers for remote maintenance checks as needed. It is unclear if these activities could constitute “control” under the proposed regulations. Application of this rule will require an analysis of how much “remote control” and “oversight” is enough to transform a DC Agreement from what appears for all other purposes to be a services transaction into a lease. This is not administrable and will create areas of controversy that is not in the interest of either the government or the taxpayer. To reduce this uncertainty, taxpayers could transfer more functions to data center entities (that are in market jurisdictions outside the US) for support of a services characterization. This would be a perverse result of the proposed regulation.

Secondly, the proposed regulations add the following factor that is not in §7701(e), “the provider has the right to determine the specific property used in the cloud transaction and replace such property with comparable property.” This factor is appropriate in unrelated third-party cloud transactions but creates uncertainty in related-party DC Agreements. In infrastructure-as-a-service transactions with unrelated third-party customers, it is common for CSPs to provide a menu of hardware components (e.g., Intel chip with 4 cores, 16 gigabytes of memory, and solid state disk drives) that allow customers to configure their virtual machine (VM) in a manner appropriate for their specific use.³ To provision these VM requests, CSPs place such components throughout their network of data centers. This allows CSPs to replace components at any time (e.g., in case of failure) and supports high levels of availability. This feature is referred to as “resource pooling” and is a fundamental element of cloud services.⁴ CSPs facilitate this pooling by centrally designing their data centers and specifying all the components to be included in each data center. When applied to a related-party DC Agreement, this factor does not contemplate the common situation described above where the CSP can specify the type of hardware used in the DC Agreement as a result of the CSP's centralized DC design function. Taxpayers would need to argue that the CSP does not have the “right to determine specific property used in the cloud transaction” notwithstanding the centralized design process. Similar to the prior factor, this could provide an incentive for taxpayers to transfer more functions to datacenter entities.

Third, the proposed regulations add the following factor that is not in §7701(e), “the property is a component of an integrated operation in which the provider has other responsibilities, including ensuring the property is maintained and updated.” Related-party data center entities typically have a single purpose and operation (i.e., the ownership and operation of a data center). The proposed regulation does not define an “integrated operation” and is not clear if this factor is determined by the functions of the provider (i.e., the data center entity) or related parties. Assuming this test applies solely at the provider level, DC Agreements likely would result in lease income because the “provider” generally does not have functions beyond maintaining and updating property even though the CSP has no right of entry to the data center's premises or possession of the individual hardware assets located at the data center. These functions appear to fall short of an “integrated operation” if the result of the operation is the delivery of a cloud service.

Lastly, the proposed regulations add the following factors that is not in §7701(e), “the provider's fee is primarily based on a measure of work performed or the level of the customer's use rather than the mere passage of time.” Cross-border DC Agreements are typically structured to satisfy the transfer pricing requirements in the country of the CSP and data center. A common compensation model is for payments under the DC Agreement to be equal to cost plus a certain percentage. This is neither a measure of a CSP's use nor the passage of time. Application of this factor has no relevance to these related-party DC Agreements.

As a result of these concerns, the characterization rule of Prop. Reg. 1.861-19(c)(2) should be applied only to transaction with unrelated third-party customers t and not to related-party cloud transactions. Related-party DC Agreements should be characterized under existing law. We suggest the scope of the application of the factors set forth in Prop. Reg. Sec. 1-861-19(c)(2) be limited to transactions with unrelated parties and that the final regulations clarify that existing law, such as Sec. 7701(e), and the judicial interpretations thereunder, apply to transactions between unrelated parties. The following example, which we suggest being included in the final regulations, illustrates application of this limitation:

D. Sourcing Income from Cloud Transactions:

In the Proposed Regulations, Treasury specifically asked for comments regarding “administrable rules for sourcing income from cloud transactions in a manner consistent with section 861 through 865”. We are not aware of any commercially significant cloud transactions that result in lease income. Therefore, our comments focus on the appropriate method for sourcing cloud transactions characterized as the provision of services under the Proposed Regulations.

CSPs generally source services income based on the location of assets and functions of the entity that recognizes such income. Such sourcing takes into account the impact of each function and asset on the delivery of the cloud service. This sourcing is based on existing law. In general, services income is sourced to where the services are performed under §§861(a)(3) and 862(a)(3). Under Treas. Reg. §1.861-4(b)(1), income for services performed partly within and partly without the US by persons other than individuals is allocated on the basis that most correctly reflects the proper source of the income under the particular facts and circumstances. Further, such Regulations provide that in many cases the facts and circumstances will be such that an apportionment on the 'time basis” will be acceptable. Time basis is defined as the ratio of the taxpayer's compensation expense paid to its employees performing services within and without the United States, attributing such compensation to the location where such employees perform their functions (on a daily basis taking into account days worked within and without the United States).

The location where the service was performed also is relevant for determining foreign base company services income (“FBCSI”) under §954(e). Under Treas. Reg. §1.954-4, the determination of whether a CFC performs services outside of its country of incorporation is based on all facts and circumstances. As a general rule, the FBCSI regulations consider services as performed “where the persons performing services for the CFC which derives income in connection with the performance of technical, managerial, architectural, engineering, scientific, skilled, industrial, commercial, or like services are physically located when they perform their duties in the execution of the service activity resulting in such income.” Consistent with the limited guidance under §§861 and 862, the location of the people functions is determinative for the sourcing of FBCSI to a specific country.

Both sources of law also consider the value of the functions performed. Where an allocation is required for sourcing purposes, the time basis allocation under Treas. Reg. §1.861-4(b)(1) requires the consideration of the compensation received by each individual. For the determination of the place of performance for FBCSI purposes, the Regulations require that “. . . relative weight must also be given to the value of the various functions performed by persons . . . For example, clerical work will ordinarily be assigned little value, while services performed by technical, highly skilled, and managerial personnel will be assigned greater values in relation to the type of function performed by each individual.”

The above authorities are generally silent with respect to role of equipment or capital in the sourcing of services. Our view is that significant equipment is also required to deliver cloud services, which is consistent with analogous case law.⁵ In the event an entity has functions and assets directly related to the delivery of cloud services that are located in different jurisdictions (e.g., through a branch or disregarded entity), services income should be sourced to functions and assets consistent with the arms-length principle.

1. Discussion of Engineering Functions that Support Cloud Service Delivery

As noted above, people functions are critical to sourcing cloud services. The below provides an overview of the various cloud engineering disciplines. This overview supports why service and site reliability engineers are particularly important in cloud service delivery, and therefore should be heavily weighted to determine the source of cloud services income.

Global CSPs require both significant people functions and assets to meet service level agreements that are defined as percentage of uptime (e.g., a common agreement is set at 99.99% that allows for a maximum of 52.56 minutes of downtime per year). Failure to maintain a specified service level typically results in the automatic refund of fees by CSPs. A clear benefit of cloud computing is the useful automation of manual tasks previously performed by humans. However, automation alone has not and cannot provide a sufficiently reliable system for providing cloud services on a global scale. As a result, the people-functions with the highest level of impact in the delivery of a cloud service should be given the most weight for purposes of sourcing cloud services.

Cloud service outages are a common occurrence that affect service level availability. In the past two years, major outages have occurred at every major CSP and are often a result of inadvertent factors. Here are some examples: In September 2018, a major CSP's data center in San Antonio, Texas experienced an electrical surge as a result of thunderstorms. The spike in voltage disrupted cooling to the data center, forcing the shutdown of certain servers to prevent overheating. As a result of inadequate fail-over capacity, multiple cloud services were taken offline for up to 24 hours while engineering teams transferred workloads to functioning servers. A separate outage occurred in February 2017 in a Northern Virginia data center of another major CSP as a result of an incorrect command to remove certain servers for routine maintenance. This resulted in a larger set of servers being removed from service than was intended, triggering an automatic restart of related systems and the failure to process all storage requests. During this period, all storage services dependent on that data center were taken offline for a period of at least five hours. Engineering teams again played a critical role during the outage to restore offline servers, and such teams implemented changes to prevent a similar failure, including limiting how many servers can be removed at a particular time.

In each of these examples, automation (i.e., hardware and software) alone was unable to prevent failures, and the intervention of significant people functions were required to manually fix, debug, and mitigate the effect of the outages caused by such failures. These outages demonstrate that cloud computing is far from an automated service that depends solely on software running correctly on hardware. Instead, engineering functions are critical and support the successful delivery of cloud services.

Cloud engineering functions are generally grouped into three disciplines:

Service engineers and SREs are the engineering functions that have the most impact on cloud service delivery. These engineers have primary responsibility for delivery of the service and respond to incidents (such as the outages above) that could affect service level availability. Often such responsibility is distributed globally to regional teams under a “follow-the-sun” service model. This distribution allows for each engineering team to work during their normal business hours and support the global delivery of cloud services. Further, some CSPs may utilize an engagement model where SREs work with both product and service engineering teams to identify and fix issues in software or hardware that reduce reliability. SREs will utilize their knowledge of the “full stack” to improve how software runs on a distributed network of hardware and vice versa. SREs likely were involved in the root cause analysis and postmortem analysis of the outages described above.

Another result of the SRE function is the reduction of manual work required for the delivery of cloud services. Manual work that is repeatable and increases as the services grow is referred to as “toil.” The reduction of toil and increase in reliability of cloud services (both outcomes of SRE functions) have contributed to a higher rate of cloud computing adoption. Historically, enterprises owned and operated their own hardware that supported their business processes. The availability and reliability of hardware resources in the cloud has increased to a point where enterprises are transferring the more significant workloads from their own infrastructure to CSPs. Gartner estimated in 2019 that the market for cloud services is expected to grow by +$30 billion annually.⁶ In order for CSPs to meet this demand, engineers from all three disciplines will need to develop more automated processes, but it is not expected such advancements will reduce the importance of people functions. Rather, they will instead obviate the simpler problems and give more time to tackle the more complex ones.

While the proposed regulations do not include guidance on sourcing income from cloud transactions, the preamble asks for comments regarding administrable rules that are consistent with Sections 861-865. The software industry believes existing guidance on sourcing cloud transaction is sufficient and further guidance is not needed. However, if Treasury plans on developing proposed regulations in this area, we suggest they make clear that cloud services income is sourced on an entity-by-entity basis and in a manner that takes into account all facts and circumstances of the people functions and assets that are directly related to generating service income from cloud transactions. The following example illustrates application of these principles:

E. Conclusion:

SoFTEC thanks you for the opportunity to provide these comments and we look forward to working with you as the proposed regulation moves towards finality. I can be reached at (202) 486-3725 or mnebergall@softwarefinance.org with any questions or requests for further information.

Respectfully submitted,

Mark E. Nebergall
President
Software Finance and Tax Executives Council