NINTH CIRCUIT ISSUES NEW OPINION AFFIRMING TAX COURT IN XILINX TRANSFER PRICING CASE.

Ronald B. Schrotenboer, Kenneth B. Clark (argued) and Tyler A. Baker, Fenwick & West LLP, Mountain View, California, for the petitioner-appellee.

Alice E. Loughran, Steptoe & Johnson LLP, Washington, D.C., for amici curiae Cisco Systems, Inc., and Altera Corporation.

A. Duane Webber, Baker & McKenzie LLP, Washington, D.C., for amici curiae Software Finance and Tax Executives Council and AeA.

On this appeal from the tax court, we must decide whether, under the tax regulations in effect during tax years 1997, 1998 and 1999, related companies engaged in a joint venture to develop intangible property must include the value of certain stock option compensation one participant gives to its employees in the pool of costs to be shared under a cost sharing agreement, even when companies operating at arm's length would not do so. The tax court found related companies are not required to share such costs and ruled that the Commissioner of Internal Revenue's attempt to allocate such costs was arbitrary and capricious. We affirm.

In 1995, Xilinx and XI entered into a Cost and Risk Sharing Agreement ("the Agreement"), which provided that all right, title and interest in new technology developed by either Xilinx or XI would be jointly owned. Under the Agreement, each party was required to pay a percentage of the total R&D costs in proportion to the anticipated benefits to each from the new technology that was expected to be created. Specifically, the Agreement required the parties to share: (1) direct costs, defined as costs directly related to the R&D of new technology, including, but not limited to, salaries, bonuses and other payroll costs and benefits; (2) indirect costs, defined as costs incurred by departments not involved in R&D that generally benefit R&D, including, but not limited to, administrative, legal, accounting and insurance costs; and (3) costs incurred to acquire products or intellectual property rights necessary to conduct R&D. The Agreement did not specifically address whether employee stock options (ESOs) were a cost to be shared.

Xilinx offered ESOs to its employees under two plans. Under one plan, employees were granted options as part of the employee hiring and retention program. The options were of two varieties: incentive stock options (ISOs) and nonstatutory stock options (NSOs). Employees could exercise these options two ways: (1) by purchasing the stock at the market price on the day the option was issued ("exercise price") regardless of its then-current market price or (2) by simultaneously exercising the option at the exercise price and selling it at its then-current price, pocketing the difference. Under the other plan, employees could acquire employee stock purchase plan shares (ESPPs) by contributing to an account through payroll deductions and purchasing stock at 85 percent of either its exercise price or its market price on the purchase date. Employees must always pay taxes on NSOs, see 26 U.S.C. § 83, but have to pay taxes on ISOs and ESPPs only if they sell acquired stock shares before a specified waiting period has expired ("a disqualifying disposition"), see 26 U.S.C. § 421(b). In determining the R&D costs to be shared under the Agreement for tax years 1997, 1998 and 1999, Xilinx did not include any amount related to ESOs.

In tax years 1997, 1998 and 1999, Xilinx deducted as business expenses under 26 U.S.C. §§ 83 and 162 approximately $41,000,000, $40,000,000 and $96,000,000, respectively, based on its employees' exercises of NSOs or disqualifying dispositions of ISOs and ESPPs.¹ It also claimed an R&D credit under 26 U.S.C. § 41 for wages related to R&D activity, of which approximately $34,000,000, $23,000,000 and $27,000,000 in the respective tax years were attributable to exercised NSOs or disqualifying dispositions of ISOs and ESPPs.² Furthermore, in 1996 Xilinx and XI entered into two agreements that allowed XI employees to acquire options for Xilinx stock. Both agreements provided XI would pay Xilinx for the "cost" of the XI employees' exercise of the stock options, which was to equal the stock's market price on the exercise date minus the exercise price. In the 1997, 1998 and 1999 tax years, XI paid Xilinx $402,978, $243,094 and $808,059, respectively, under these agreements.

The Commissioner of Internal Revenue ("Commissioner") issued notices of deficiency against Xilinx for tax years 1997, 1998 and 1999, contending ESOs issued to its employees involved in or supporting R&D activities were costs that should have been shared between Xilinx and XI under the Agreement. Specifically, the Commissioner concluded the amount Xilinx deducted under 26 U.S.C. § 83(h) for its employees' exercises of NSOs or disqualifying dispositions of ISOs and ESPPs should have been shared. By sharing those costs with XI, Xilinx's deduction would be reduced, thereby increasing its taxable income. The Commissioner's determination resulted in substantial tax deficiencies and accuracy-related penalties under 26 U.S.C. § 6662(a).

Xilinx timely filed suit in the tax court. The tax court denied cross motions for summary judgment. After a bench trial, the tax court found that two unrelated parties in a cost sharing agreement would not share any costs related to ESOs. After assuming ESOs were costs for purposes of 26 C.F.R. § 1.482-7(d)(1), the tax court then found 26 C.F.R. § 1.4821(b)(1) -- which requires cost sharing agreements between related parties to reflect how two unrelated parties operating at arm's length would behave -- dispositive and concluded the Commissioner's allocation was arbitrary and capricious because it included the ESOs in the pool of costs to be shared under the Agreement, even though two unrelated companies dealing with each other at arm's length would not share those costs.

The Commissioner timely appealed. On appeal, the parties focused primarily on whether the requirement in 26 C.F.R. § 1.482-7(d)(1) that "all costs" be shared between related parties in a cost sharing agreement or whether the controlling requirement was 26 C.F.R. § 1.482-1(b)(1) that all transactions between related parties reflect what two parties operating at arm's length would do. After oral argument, we requested supplemental briefing on whether ESOs were "costs" and whether they were "related to" the intangible product development for purposes of 26 C.F.R. § 1.4827(d)(1), and whether a literal application of 26 C.F.R. § 1.482-7(d)(1) would conflict with a tax treaty between the United States and Ireland that was in effect during the 1998 and 1999 tax years.

[1] Ambiguity. Congress has authorized the Secretary of the Treasury to allocate income and deductions among related business entities to prevent tax avoidance.

Another section, however, specifically governing cost sharing agreements between controlled parties to develop intangible property, authorizes the Internal Revenue Service "to make each controlled participant's share of the costs (as determined under paragraph (d) of this section) of intangible development under the qualified cost sharing arrangement equal to its share of reasonably anticipated benefits attributable to such development. . . ." 26 C.F.R. § 1.482-7(a)(2). Controlled participants, under paragraph (d) of § 1.482-7, must include "all" costs in the pool of costs to be shared proportionally (the "all costs requirement"):

[2] Section 1.482-1(b)(1) specifies that the true taxable income of controlled parties is calculated based on how parties operating at arm's length would behave. The language is unequivocal: this arm's length standard is to be applied "in every case." In the context of cost sharing agreements, this rule would require controlled parties to share only those costs uncontrolled parties would share. By implication, costs that uncontrolled parties would not share need not be shared. In contrast, § 1.482-7(d)(1) specifies that controlled parties in a cost sharing agreement must share all "costs . . . related to the intangible development area," and that phrase is explicitly defined to include virtually all expenses not included in the cost of goods. The plain language does not permit any exceptions, even for costs that unrelated parties would not share. Each provision's plain language mandates a different result. Accordingly, we conclude that when related to each other, the two provisions establish an ambiguous standard for determining which costs must be shared between controlled parties in cost sharing agreements specifically related to intangible product development.

Given the resultant ambiguity, our choice is to:

[3] Often the specific controls the general. This rule has been used by the Supreme Court. E.g., Long Island Care At Home, Ltd. v. Coke, 127 S. Ct. 2339, 2348 (2007). Apply this simple rule here, and section 1.482-7(d)(1) controls. The conflict dissolves. The Commissioner is vindicated.

[4] This simple solution is all too pat. It gives controlling importance to a single canon of construction. But, as every judge knows, the canons of construction are many and their interaction complex. The canons "are not mandatory rules." Chickasaw Nation v. United States, 534 U.S. 84, 94 (2001). They are guides "designed to help judges determine the Legislature's intent." Id. They can be "overcome" by "other circumstances" manifesting that intent. Id. The canons are "tools designed to help courts better determine what Congress intended, not to lead courts to interpret the law contrary to that intent." Scheidler v. National Org. of Women, Inc., 547 U.S. 9, 23 (2006). In the light of these principles, two considerations show the Commissioner's position to be untenable.

[5] Purpose. Purpose is paramount. The purpose of the regulations is parity between taxpayers in uncontrolled transactions and taxpayers in controlled transactions. The regulations are not to be construed to stultify that purpose. If the standard of arm's length is trumped by 7(d)(1), the purpose of the statute is frustrated. If Xilinx cannot deduct all its stock option costs, Xilinx does not have tax parity with an independent taxpayer.

[6] Treaties. The "arm's length" standard used in the United States Ireland Tax Treaty RIA Int. Tax Treaty 3057, aids in understanding the mind and practice of the Treasury. A tax treaty is negotiated by the United States with the active participation of the Treasury. The Treasury's reading of the treaty is "entitled to great weight." United States v. Stuart, 489 U.S. 353, 369 (1989) (quoting Sumitomo Shoji America, Inc. v. Aragliano, 457 U.S. 176, 184-185 (1982)). Simultaneous with the signing of the treaty into law, the Treasury issued its "Technical Explanation." As to Article 9, the Explanation reads:

We do not, however, need to decide in this case whether the treaty obligations "constitute binding federal law enforceable in United States courts." Medellin v. Texas, 128 S. Ct. 1346, 1356 (2008). It is enough that our foreign treaty partners and responsible negotiators in the Treasury thought that arm's length should function as the readily understandable international measure.

The judgment of the tax court is AFFIRMED.

² Under 26 U.S.C. § 41(b)(2)(A), companies can claim a tax credit for "wages paid or incurred to an employee for qualified [research] services performed by such employee."

³ Controlled taxpayer is defined as "any one of two or more taxpayers owned or controlled directly or indirectly by the same interests, and includes the taxpayer that owns or controls the other taxpayers." 26 C.F.R. § 1.482-1(i)(5).

I concur, but write to explain my particular reasons for rejecting the Commissioner's position in this case.

The parties provide dueling interpretations of the "arm's length standard" as applied to the ESO costs that Xilinx and XI did not share. Xilinx contends that the undisputed fact that there are no comparable transactions in which unrelated parties share ESO costs is dispositive because, under the arm's length standard, controlled parties need share only those costs uncontrolled parties share. By implication, Xilinx argues, costs that uncontrolled parties would not share need not be shared.

On the other hand, the Commissioner argues that the comparable transactions analysis is not always dispositive. The Commissioner reads the arm's length standard as focused on what unrelated parties would do under the same circumstances, and contends that analyzing comparable transactions is unhelpful in situations where related and unrelated parties always occupy materially different circumstances. As applied to sharing ESO costs, the Commissioner argues (consistent with the tax court's findings) that the reason unrelated parties do not, and would not, share ESO costs is that they are unwilling to expose themselves to an obligation that will vary with an unrelated company's stock price. Related companies are less prone to this concern precisely because they are related -- i.e., because XI is wholly owned by Xilinx, it is already exposed to variations in Xilinx's overall stock price, at least in some respects. In situations like these, the Commissioner reasons, the arm's length result must be determined by some method other than analyzing what unrelated companies do in their joint development transactions.

Under Xilinx's interpretation, § 1.482-1(b)(1) and § 1.4827(d)(1) are irreconcilable. The latter specifies that controlled parties in a cost sharing agreement must share all "costs . . . related to the intangible development area," and that phrase is explicitly defined to include virtually all expenses not included in the cost of goods. The plain language does not permit any exceptions, even for costs that unrelated parties would not share, so each provision mandates a different result.

Under the Commissioner's interpretation, § 1.482-7(d)(1)'s "all costs" requirement is consistent with § 1.482-1(b)(1)'s arm's length standard and controls. In particular, the Commissioner argues that, because there are material differences in the economic circumstances of related and unrelated companies in relation to cost-sharing agreements like the one in this case, it was proper for the IRS to require that in this narrow context the arm's length result should be defined by the "all costs" requirement.

Having thoroughly considered not only the plain language of the regulations but also the various interpretive tools the parties and amici have brought before us, including the legislative history of § 482, the drafting history of the regulations, persuasive authority from international tax treaties¹ and what appears to have been the understanding of corporate taxpayers in similar circumstances and of others,² I conclude that Xilinx's understanding of the regulations is the more reasonable even if the Commissioner's current interpretation may be theoretically plausible. Traditional tools of statutory construction do not resolve the apparent conflict in these regulations as applied to Xilinx, and the Commissioner's attempts to square the "all costs" regulation with the arm's length standard have only succeeded in demonstrating that the regulations are at best ambiguous.³

Although I would not go so far as Xilinx in characterizing the Commissioner's interpretation as merely a "convenient litigating position," Bowen v. Georgetown Univ. Hosp., 488 U.S. 204, 213 (1988), we need not defer to it because he has not clearly articulated his rationale until now. See United States v. Thompson/Ctr. Arms Co., 504 U.S. 505, 518-19 & n.9 (1992) (declining to defer to an agency interpretation of a tax statute where no prior guidance went directly "to the narrow question presented"). Indeed, I am troubled by the complex, theoretical nature of many of the Commissioner's arguments trying to reconcile the two regulations. Not only does this make it difficult for the court to navigate the regulatory framework, it shows that taxpayers have not been given clear, fair notice of how the regulations will affect them.⁴

Accordingly, I join Judge Noonan in affirming the tax court. These regulations are hopelessly ambiguous and the ambiguity should be resolved in favor of what appears to have been the commonly held understanding of the meaning and purpose of the arm's length standard prior to this litigation.

² Apparently Xilinx's understanding was widely shared in the business community and tax profession. See Brief of PricewaterhouseCoopers LLP, Deloitte Tax LLP and KMPG LLM as Amici Curiae on the Petition for Rehearing at 5-6 (describing a "global consensus"); Brief of Cisco Systems, Inc. and Thirty-Two Other Affected Companies as Amici Curiae on the Petition for Rehearing at 3 (describing "settled business expectations"); cf. Brief of Former U.S. Treasury and Internal Revenue Service Officials at 3-5.

³ The dissent invokes the prior, withdrawn majority opinion. Dissent at 4613-14; see Xilinx, Inc. v. CIR, 567 F.3d 482 (9th Cir. 2009), withdrawn January 13, 2010. In writing that opinion, I was persuaded that the arm's length standard and the all costs regulation were in conflict, and that the more specific of the two should control. See 567 F.3d at 486. I no longer share Judge Reinhardt's confidence in that resolution because the Commissioner's response to Xilinx's petition for rehearing declined to fully endorse its reasoning. Instead, the thrust of the Commissioner's response was that our result was correct, even though our reasoning was not.

⁴ It is an open question whether these flaws have been addressed in the new regulations Treasury issued after the tax years at issue in this case. See 26 C.F.R. § 1.482-7T(a) & (d)(1)(iii) (2009) (stating explicitly that ESOs are costs that must be shared and that the all costs requirement is an arm's length result).

I have considerable doubt as to whether Xilinx, Inc. ("Xilinx") and Xilinx Ireland allocated the costs associated with employee stock options in a manner that can be characterized as an arm's length result. I will assume, however, that the tax court correctly resolved that issue. If so, there is clearly a conflict between the arm's length regulation codified at 26 C.F.R. § 1.482-1(b)(1), which applies to all transactions between controlled parties, and the "all costs" regulation codified at § 1.482-7(d)(1), which applies only to cost-sharing arrangements between controlled parties.¹ I continue to believe that, as a matter of law, the "all costs" regulation, as the specific of the two provisions, the one designed to deal specifically with the type of question before us, controls. I would therefore reverse the tax court's ruling that the Commissioner's proposed allocation was arbitrary and capricious for the reasons explained in our opinion, Xilinx Inc. v. CIR, 567 F.3d 482 (9th Cir. 2009), withdrawn on January 13, 2010 in anticipation of the issuance of Judge Noonan's and Judge Fisher's new opinions, supra.

I agree with the majority that the canons of construction "are not mandatory rules," and that their interpretive force can be overcome by other circumstances evidencing legislative intent. Maj. op. at 4608-09 (quoting Chickasaw Nation v. United States, 534 U.S. 84, 94 (2001)). Such circumstances, however, are not present here. Contrary to the majority's assertions, the conflict between the arm's length provision and the "all costs" requirement cannot be resolved by looking to the purpose of the regulations or to Treasury's Technical Explanation of the 1997 United States-Ireland Tax Treaty. Judge Fisher also looks to the understanding of the multinational corporations and their business and tax advisors, a dubious practice for which he cites no legal authority.

The stated purpose of the regulatory scheme is "to ensure that taxpayers clearly reflect income attributable to controlled transactions and to prevent the avoidance of taxes with respect to such transactions." 26 C.F.R. § 1.482-1(1)(a). In the context of this case, neither regulation more clearly implements that purpose than the other. Controlled and uncontrolled parties always operate under materially different circumstances with regard to employee stock option costs. Accordingly, the "all costs" regulation may simply reflect the conclusion that, whatever uncontrolled parties might do, requiring controlled parties to share such costs "ensure[s] that taxpayers clearly reflect [the] income attributable to [the] controlled transaction[ ]" as a whole. Nor is it clear that excluding those costs would better achieve tax parity. It is not the identity of treatment with respect to a single item that controls with respect to this general goal, but the overall manner in which the transaction is treated. The Commissioner has determined that including "all costs" is the best manner of achieving this general objective, and his decision does not appear to be unreasonable.

Similarly, Treasury's Technical Explanation of the 1997 United States-Ireland Tax Treaty does not justify disregarding the "all costs" requirement when determining deductible costs. A Technical Explanation is not subject to the APA's notice and comment requirement and does not carry the force of law. See 5 U.S.C. § 553(a)(1) (exempting "foreign affairs function[s] of the United States" from the APA); see, e.g., Explanation of Convention with Ireland (1997) ("The Technical Explanation is an official guide to the Convention and Protocol." (emphasis added)). Certainly, it cannot trump the plain language of the duly enacted "all costs" regulation, which does have the force of law and is entitled to this court's deference. United States v. Mead Corp., 533 U.S. 218, 226-27 (2001); cf. Cent. Laborers' Pension Fund v. Heinz, 541 U.S. 739, 748 (2004) ("[N]either an unreasoned statement in the manual nor allegedly longstanding agency practice can trump a formal regulation with . . . the force of law.").

I recognize that Xilinx and amici have raised serious doubts as to whether the result that I believe to be legally required is, from both a practical and an equitable standpoint, the proper one. I am particularly troubled by the international tax consequences that such a result would apparently create. Tax law, however, involves the resolution by Congress of complex political and economic issues that sometimes may affect business or individual interests in unforeseen ways, and sometimes in ways that benefit one political or economic interest at the expense of another. These resolutions are not always arrived at in an open, objective, or non-political manner. To put it plainly, fairness is not always Congress's ultimate objective in enacting tax legislation. Accordingly, some provisions of tax law may appear to some businesses, individuals, or even judges to be in conflict with reasonable or sensible tax or national policy. Yet they may reflect the intent of Congress when it enacted the statute. Tax regulations are frequently even more complex than the legislation they implement, and it is often difficult for judges to clearly resolve their meaning. Still, it is the job of the courts to make the necessary determinations and in doing so to apply established legal rules and principles. By contrast, it is the business of Congress and the Treasury, not the courts, to correct any errors in those statutes or regulations, especially as they can be readily corrected once they are called to their attention.

For the reasons I have explained, it is particularly inappropriate for courts to resolve tax cases on a practical or equitable basis or to interpret tax statutes and regulations other than strictly in accordance with settled legal principles. The canon of construction under which the specific controls the general is one such settled legal principle, and one that is especially pertinent here. Indeed, it is controlling. I adhere to the previous majority opinion of this court.²

For these reasons, I respectfully dissent.

² I, like Judge Fisher, am less than enthusiastic about the Commissioner's explanation of how he believes we should resolve this case. His preference is that we find somehow that the arm's length standard is met by way of the all costs requirement. I must confess that I have difficulty following his reasoning and, like Judge Fisher, am not persuaded by that argument. However, the Commissioner then says that if we still believe that the two provisions are in conflict, we must apply the rule on which Judge Fisher originally relied and on which I continue to rely. I guess I am just not as sensitive as Judge Fisher. Simply because the Commissioner advanced an argument that we reject, but then argued that if we reject it, we should apply the rule that we held applicable in our opinion is hardly a reason for abandoning the rule that we believed to be correct. We can't expect anyone, let alone the Commissioner of Internal Revenue, to agree completely with everything we say. Rejecting the Commissioner's first argument leaves us exactly where we were before he advanced it: The two regulations are in conflict, and (as Judge Fisher and I once agreed) that conflict must be resolved by applying the specific regulation rather than the general one.