No. 23-3281

Appeals from the United States Tax Court

Submitted: May 13, 2025 Filed: September 3, 2025Page 2 Before COLLOTON, Chief Judge, SMITH and SHEPHERD, Circuit Judges.

COLLOTON, Chief Judge.

For nearly fifteen years, the Commissioner of Internal Revenue and Medtronic, Inc. have been embroiled in litigation over the amount of income Medtronic’s 2005 and 2006 consolidated tax returns attributed to a subsidiary, Medtronic Puerto Rico. In Medtronic, Inc. & Consolidated Subsidiaries v. Commissioner , 900 F.3d 610 (8th Cir. 2018), this court vacated the tax court’s first decision in the case and remanded with instructions to make additional fact findings. The findings were necessary to facilitate review of whether the tax court applied the best method for calculating an arm’s length price for intangible property that was transferred between two Medtronic entities. Id. at 615.

The Commissioner now appeals the tax court’s decision on remand. He argues that the tax court erred by rejecting his proposed transfer pricing method and by adopting a different method that the Commissioner maintains is prohibited under the applicable regulations. Medtronic cross-appeals and asserts that the tax court clearly erred in rejecting the taxpayer’s proposed transfer pricing method. Alternatively, Medtronic argues that this court should uphold the tax court's selected transfer pricing method, but direct certain adjustments. We vacate the tax court’s order and remand for further proceedings consistent with this opinion.

Medtronic is a medical device company that produces and markets class III devices, which include implantable cardiac rhythm stimulation and neurostimulation devices as well as the electrodes or “leads” that connect those devices to the human body. Medtronic’s parent company, Medtronic US, and its distributor, Medtronic Page 3USA, Inc. (Med USA), are located in Minnesota. The company’s class III device manufacturer, Medtronic Puerto Rico Operations Co. (Medtronic Puerto Rico), is located in Puerto Rico.

Medtronic allocates the profit earned from its devices and leads between Medtronic US, Med USA, and Medtronic Puerto Rico through intercompany licensing agreements. This appeal concerns agreements under which Medtronic US granted Medtronic Puerto Rico the exclusive right to use intangible property to manufacture and sell devices and leads in exchange for Medtronic Puerto Rico’s agreement to pay a royalty based on net sales to Medtronic US. We refer to these agreements as the Technology Licenses.

The Internal Revenue Code empowers the Commissioner to allocate gross income between or among commonly controlled parties 'if he determines that such . . . allocation is necessary in order to prevent evasion of taxes or clearly to reflect the income of any such” entities. 26 U.S.C. § 482. * The taxable income attributed to a controlled taxpayer participating in a controlled transaction is determined as if the parties were “dealing at arm’s length.” 26 C.F.R. § 1.4821(b)(1). “A controlled transaction meets the arm’s length standard if the results of the transaction are consistent with the results that would have been realized if uncontrolled taxpayers had engaged in the same transaction under the same circumstances . . . .' Id.

The regulations specify four methods to evaluate whether the amount charged in a controlled transfer of intangible property meets the arm’s length standard. See id. § 1.482-4(a). Three are relevant to this case. The comparable uncontrolled transaction method “evaluates whether the amount charged for a controlled transfer Page 4of intangible property was arm’s length by reference to the amount charged in a comparable uncontrolled transaction.' Id. § 1.482-4(c)(1). The comparable profits method “evaluates whether the amount charged in a controlled transaction is arm’s length based on objective measures of profitability (profit level indicators) derived from uncontrolled taxpayers that engage in similar business activities under similar circumstances.' Id. § 1.482-5(a). The regulations also permit the use of 'unspecified methods' if they satisfy applicable requirements. Id. § 1.482-4(d). The transaction must be evaluated under the “best method”—that is, “the method that, under the facts and circumstances, provides the most reliable measure of an arm's length result.' Id. §§ 1.482-4(a), -1(c)(1).

As explained in Medtronic I , this case began with a dispute about Medtronic’s 2002 consolidated tax return. See 900 F.3d at 612. That return used the comparable uncontrolled transaction method to determine the royalty rates paid on its intercompany licensing agreements. After an audit in which the Internal Revenue Service disputed Medtronic’s transfer pricing method and profit allocation, the IRS and Medtronic entered into a memorandum of understanding in which Medtronic Puerto Rico would pay wholesale royalty rates of 44% for devices and 26% for leads on its intercompany sales. These royalty rates resulted in an overall profit split of approximately 55.6% for Medtronic US/Med USA and 44.4% for Medtronic Puerto Rico.

The IRS and Medtronic could not agree on how the memorandum of understanding should apply to Medtronic Puerto Rico's royalty payments to Medtronic US for the 2005 and 2006 tax years. After auditing Medtronic's consolidated returns, the IRS determined that the comparable profits method was the best way to determine an arm’s length price for Medtronic’s intercompany licensing agreements for those two years, and the IRS charged Medtronic with a tax deficiency. Medtronic disputed the adjustment and filed suit in the United States Tax Court, Page 5arguing that the comparable uncontrolled transaction method was the best method for determining an arm’s length royalty rate for the intercompany licensing agreements.

After a trial, the tax court rejected both parties’ royalty rate valuations, and conducted its own valuation analysis. The court ultimately decided that Medtronic’s comparable uncontrolled transaction method was the best way to determine an arm’s length royalty rate for the intercompany licensing agreements, but made a number of adjustments. The tax court then determined arm’s length wholesale royalty rates for intercompany sales of 44% for device licenses and 22% for leads licenses, which resulted in an overall profit split of 54.1% to Medtronic US/Med USA and 45.9% to Medtronic Puerto Rico. The court then issued an order concluding that Medtronic had an income tax deficiency in 2005 and an overpayment in 2006. The Commissioner appealed and sought a reevaluation of the best transfer pricing method and a recalculation of the arm’s length royalty rate.

On appeal, this court held that the tax court’s factual findings were insufficient to allow a determination whether the tax court correctly accepted a patent-licensing agreement between Medtronic US and Siemens Pacesetter as a comparable uncontrolled transaction, and whether the tax court “applied the best transfer pricing method.' Medtronic I , 900 F.3d at 614-15. Accordingly, this court vacated the tax court’s order and remanded the case for further consideration and factual findings “necessary to our determination whether the tax court applied the best transfer pricing method for calculating an arm’s length result or whether it made proper adjustments under its chosen method.' Id. at 615.

On remand, Medtronic again maintained that the Pacesetter Agreement is comparable to the Technology Licenses and that the comparable uncontrolled transaction method is the best method to determine an arm’s length royalty. The Commissioner argued that the comparable profits method is the best method to price the Technology Licenses, and proposed a modified version of that method which Page 6relied on a set of five comparable companies, narrowed from a set of fourteen offered at the first trial. Medtronic also proposed, in the alternative, a three-step unspecified method that incorporated aspects of the parties’ comparable uncontrolled transaction method and comparable profits method and then allocated residual profits between Medtronic US and Medtronic Puerto Rico.

After another trial, the tax court abandoned its previous conclusion that a Pacesetter Agreement-based application of the comparable uncontrolled transaction method was the best method to determine an arm's length royalty rate for the Technology Licenses. The court found that three of the five general comparability factors described in § 1.482-1(d)(1) were not satisfied. First, the court determined that Medtronic Puerto Rico and Pacesetter 'did not perform the same functions.' See id. § 1.482-1(d)(1)(i), (3)(i). Second, the court found that the economic conditions between the agreements were not comparable because the property involved in the two agreements did not have similar profit potential. See id. §§ 1.482-1(d)(1)(iv), 4(c)(2)(iii)(B)(1)(ii). And third, the court concluded that the property licensed under the agreements was not similar: the Pacesetter Agreement licensed only patents, while the Technology Licenses encompassed the “full array of intangible property” needed to manufacture devices and leads for sale, including patents, know-how, regulatory approvals, secret processes, technical information and expertise, and copyrights. See id. § 1.482-1(d)(1)(v), (3)(v).

The tax court next determined that the Commissioner’s modified comparable profits method could not be the best method to price the Technology Licenses. The court reasoned that the Commissioner erroneously relied on companies that made different products and carried different asset bases, functions, and risks than Medtronic Puerto Rico. The tax court also rejected the Commissioner’s proposed adjustment to account for differences in the product liability risk borne by Medtronic Puerto Rico compared to the proposed comparable companies. See id. §§ 1.4825(c)(2)(iv), -1(d)(2).

Page 7Having rejected the preferred methods of both parties, the tax court applied Medtronic’s three-step unspecified method, with some adjustments. Step one applied a modified version of Medtronic's Pacesetter Agreement-based comparable uncontrolled transaction method as a 'starting point' to price Medtronic US's research and development activities. Step two applied a modified version of the Commissioner’s comparable profits method to allocate profit to Medtronic Puerto Rico’s finished-device manufacturing functions. Step three split the remaining profit from the devices and leads between Medtronic US and Medtronic Puerto Rico. The tax court determined a wholesale royalty rate of 48.8% for both leads and devices, which resulted in an overall profit split of 68.7% to Medtronic US/Med USA and 31.3% to Medtronic Puerto Rico. The court then issued an order concluding that Medtronic had income tax deficiencies in 2005 and 2006.

The Commissioner appeals the tax court’s order, asserting several legal errors in the decision to adopt the three-step unspecified method and reject his proposed comparable profits method. Medtronic cross-appeals and argues that the tax court clearly erred in finding that the Pacesetter Agreement is not a comparable uncontrolled transaction. Alternatively, the taxpayer contends that we should uphold the tax court’s unspecified method but remand for reconsideration of its adjustments at step three. We review the tax court's legal conclusions de novo and factual findings for clear error. Medtronic I , 900 F.3d at 613.

Medtronic first challenges the tax court’s rejection of the Pacesetter Agreement as a comparable uncontrolled transaction. The taxpayer argues that an application of the comparable uncontrolled transaction method based on the Pacesetter Agreement is the best method for pricing the Technology Licenses. That method “evaluates whether the amount charged for a controlled transfer of intangible property was arm’s length by reference to the amount charged in a comparable uncontrolled transaction.” 26 C.F.R. § 1.482-4(c)(1).

In the absence of an uncontrolled transaction involving the same intangible property, application of the comparable uncontrolled transaction method “requires that the controlled and uncontrolled transactions involve . . . comparable intangible property,' and that the transactions arise 'under comparable circumstances.' Id. § 1.482-4(c)(2)(iii)(A), (ii). To be “comparable” in the relevant sense, the intangible property involved in an uncontrolled transaction “must . . . [h]ave similar profit potential' to the intangible property involved in the controlled transaction. Id. § 1.482-4(c)(2)(iii)(B)(1)(ii).

The tax court found that the intangible property licensed under the Pacesetter Agreement and the Technology Licenses do not have similar profit potential. The court cited, inter alia , a report by the Commissioner's expert Michael Heimert showing that Pacesetter’s product profit margin from intellectual property licensed under the Pacesetter Agreement averaged 29% from 1993 to 1995, compared to Medtronic’s 54% average product profit margin in 2005 to 2006 from intellectual property licensed under the Technology Licenses.

Page 9Medtronic challenges the tax court’s finding on profit potential. The taxpayer first argues that the tax court failed to consider the effect of a clause in the Pacesetter Agreement that permitted either Pacesetter or Medtronic US to license any of the other party’s relevant patents developed during the term of the agreement for an aggregate rate of no more than 15%, with an exception for certain “key patents.” Medtronic asserts that this maximum rate clause “accounts for any difference in [the] profit potential” of the property licensed under the Pacesetter Agreement compared to the profit potential of the property licensed under the Technology Licenses. But this provision applied exclusively to licenses for patents-the only intangible property transferred under the Pacesetter Agreement. In contrast, Medtronic Puerto Rico obtained the right to exploit “the full array of intangible property” necessary to manufacture and sell devices and leads, including “patents, trade secrets, know-how, copyrights, and all regulatory approvals associated with” the products.

Medtronic's reliance on the 15% maximum rate clause is premised on its contention that the “core value” of the Technology Licenses is derived from the same patent rights that Medtronic US licensed to Siemens Pacesetter. The taxpayer further asserts that the benefits Medtronic Puerto Rico obtained from the non-patent intellectual property licensed under the Technology Licenses—particularly knowhow and regulatory approvals—were small compared to the patents, and should be disregarded in the comparability analysis.

The tax court found, however, that the know-how and regulatory approvals were valuable, and this finding is supported by the record. The Commissioner’s expert Heimert attributed the difference in profit margins to the difference in the intangible property involved in each transaction. He opined that “[t]he broader range of IP subject to the [Medtronic Puerto Rico] License conveys significantly more value than the bare patent license conveys to Siemens [Pacesetter].” “Because IP with higher profit potential should, all else equal, attract higher royalty rates,” Heimert explained, “the IP licensed under the [Medtronic Puerto Rico] License would attract Page 10a much higher royalty rate than the IP licensed under the Pacesetter [Agreement].” See id. § 1.482-4(c)(2)(iii)(B)(1)(ii) (“The profit potential of an intangible is most reliably measured by directly calculating the net present value of the benefits to be realized (based on prospective profits to be realized or costs saved) through the use or subsequent transfer of the intangible . . . .”). Because the non-patent intellectual property licensed under the Technology Licenses conveyed additional profit potential, the 15% maximum royalty rate on patents stipulated in the Pacesetter Agreement does not account for differences in the profit potential of the property licensed under the Pacesetter Agreement and the Technology Licenses.

Medtronic also asserts that differences in the profit potential between property involved in uncontrolled and controlled transactions does not preclude comparability “[i]f the difference can (as here) be accounted for” by making certain adjustments. Medtronic cites the general comparability standard described in § 1.482-1(d). That standard provides that even when there are “material differences” between controlled and uncontrolled transactions, 'adjustments must be made if the effect of such differences on prices or profits can be ascertained with sufficient accuracy to improve the reliability of the results.' Id. § 1.482-1(d)(2). The regulations specific to transfers of intangible property, however, permit adjustments to account for differences in circumstances of controlled and uncontrolled transactions, not for differences in the intangible property transferred, which ' must . . . [h]ave similar profit potential.' Id. §§ 1.482-4(c)(2)(iii)(A), (B)(1)(ii) (emphasis added).

We conclude that the tax court did not clearly err by finding that the Pacesetter Agreement and the Technology Licenses did not transfer comparable intangible property, and that the Pacesetter Agreement cannot be used as a comparable uncontrolled transaction to determine an arm’s length royalty rate for the Technology Licenses. See id. § 1.482-4(c)(2)(iii)(B)(1)(ii). The comparable uncontrolled transaction method is therefore not the best method to determine an arm’s length royalty rate for the Technology Licenses.

Page 11The Commissioner challenges the tax court’s three-step unspecified method, which uses the Pacesetter Agreement as a “starting point” for the analysis. The regulations permit the use of “methods not specified” in § 1.482-4(a)(1)-(3)— i.e. , the comparable uncontrolled transaction method, the comparable profits method, and the profit split method—to evaluate whether the amount charged in a controlled transfer of intangible property is arm's length. Id. § 1.482-4(d)(1). The Commissioner argues that because the Pacesetter Agreement fails the similar-profit-potential requirement of the comparable uncontrolled transaction method, see § 1.482-4(c)(2)(iii)(B)(1)(ii), the tax court may not use the agreement to determine an arm’s length price for a controlled transaction under an unspecified method.

We find merit in the Commissioner’s position. Section 1.482-4(d)(1) instructs that “[c]onsistent with the specified methods, an unspecified method should take into account the general principle that uncontrolled taxpayers evaluate the terms of a transaction by considering the realistic alternatives to the transaction, and only enter into a particular transaction if none of the alternatives is preferable to it.' Id. '[A]n unspecified method should provide information on the prices or profits that the controlled taxpayer could have realized by choosing a realistic alternative to the controlled transaction.” Id. “[T]he reliability of a method will be affected by the reliability of the data and assumptions used to apply the method.' Id. The purpose of the regulations is thus to identify a realistic alternative to the controlled transaction at issue, and to determine a price for the controlled transaction based on what the controlled taxpayer could have realized if it had undertaken the realistic alternative.

Section 1.482-4(a) lists three examples of methods used to find a realistic alternative to a controlled transfer of intangible property. The provisions that follow explain when data weighed in the application of each of those methods are reliable. The comparable uncontrolled transaction method “compares a controlled transaction Page 12to similar uncontrolled transactions to provide a direct estimate of the price the parties would have agreed to had they resorted directly to a market alternative to the controlled transaction.' Id. § 1.482-4(d)(1). Section 1.482-4(c)(2)(iii)(B)(1)(ii) specifies that an uncontrolled transaction constitutes reliable data with which to evaluate a controlled transactioni.e. , it is a realistic alternative to which the controlled taxpayer could have resorted directly in the free market-only if the property involved in the uncontrolled transaction has “similar profit potential” to the property involved in the controlled transaction.

As applied to this case, the tax court found that the property transferred under the Pacesetter Agreement and the property transferred under the Technology Licenses do not have similar profit potential. The Pacesetter Agreement therefore does not provide reliable data to use in evaluating whether the amount charged in a controlled transaction is consistent with the amount to which the parties would have agreed in a transaction taking place at arm’s length. See id. § 1.482-4(d)(1). We therefore reject the tax court’s use of the Pacesetter Agreement under an unspecified method to determine an arm’s length price for the Technology Licenses.

The Commissioner next challenges the tax court’s rejection of his proposed application of the comparable profits method. This method evaluates whether the amount charged in a controlled transaction is arm’s length by comparing the amount of operating profit that one party to the transaction (the “tested party,” here Medtronic Puerto Rico) would have earned “if its profit level indicator were equal to that of an uncontrolled comparable (comparable operating profit).' Id. § 1.482-5(b)(1). 'Profit level indicators are ratios that measure relationships between profits and costs incurred or resources employed.' Id. § 1.482-5(b)(4). An 'uncontrolled comparable' is an unrelated taxpayer that engages in similar business activities under similar circumstances. Id. § 1.482-5(a).

Page 13Under the comparable profits method, a profit level indicator is selected and applied to the financial information of comparable parties to determine their profitability. Id. § 1.482-5(b)(1), (4). After application of the profit level indicator, '[t]he tested party's reported operating profit is compared to the comparable operating profits derived from the profit level indicators of uncontrolled comparables to determine whether the reported operating profit represents an arm’s length result.” Id. § 1.482-5(b)(1). The comparable profits method 'relies on the general principle that similarly situated taxpayers will tend to earn similar returns.” Intercompany Transfer Pricing Reguls. Under Section 482, 59 Fed. Reg. 34971, 34985 (July 8, 1994).

The Commissioner first contends that the tax court erred in rejecting the comparable profits method on the ground that his proposed comparable companies “did not make solely class III medical devices,” but also made class I and class II medical devices. The tax court also noted that Medtronic Puerto Rico manufactured cardiological and neurological devices, while the proposed comparable companies made orthopedic, vascular, and urology products.

We conclude that the tax court applied an incorrect standard in rejecting the comparable profits method. Under that method, '[t]he degree of comparability between an uncontrolled taxpayer and the tested party is determined by applying the provisions of § 1.482-1(d)(2).' 26 C.F.R. § 1.482-5(c)(2)(i). The general comparability standard of § 1.482-1(d)(2) does not require that proposed comparable companies produce the same products as the tested party, but requires only that the comparable companies are “sufficiently similar” to the tested party to provide “a reliable measure of an arm’s length result.” Id. § 1.482-1(d)(2). The regulations emphasize that the general comparability provisions described in § 1.482-1(d) should be tailored to the specific transfer pricing method under consideration. Id. § 1.482Page 141(d)(1). “[T]he comparable profits method is not as dependent on product similarity” as are other methods, “[b]ecause operating profit usually is less sensitive than gross profit to product differences.” Id. § 1.482-5(c)(2)(iii); see also id. § 1.482-8, Ex. 6 (comparable profits method is best method despite “significant differences” between the products manufactured).

In rejecting the Commissioner’s proposed comparable companies because they “did not make solely class III medical devices,” the tax court overemphasized the importance of product similarity under the comparable profits method. On remand, the tax court should apply the correct legal standard, and consider whether the proposed comparable companies were “sufficiently similar” to Medtronic Puerto Rico and, if not, whether adjustments can be made to account reliably for any material differences. See id. §§ 1.482-1(d)(2), -5(c)(2)(iv).

The Commissioner next disputes the tax court’s conclusion that his proposed comparable companies 'have fundamentally different asset bases and involve different functions and risks' than Medtronic Puerto Rico. See id. § 1.482-5(c)(2)(ii) (identifying resources employed, risks assumed, and functions performed as particularly important considerations in determining the degree of comparability between the tested party and an uncontrolled taxpayer). The tax court did not make sufficient factual findings to support this conclusion.

i.

The Commissioner proposed the “return-on-assets” profit level indicator to determine the profitability of the comparable parties. See id. § 1.482-5(b)(1), (4)(i). This indicator measures a company’s profit level by comparing its operating profit to its operating assets. Id. § 1.482-5(b)(4)(i). The Commissioner cited two reasons Page 15for his proposed indicator: (1) this profit level indicator is most reliable when a company’s operating assets play a critical role in its ability to generate operating profits, and (2) the level of functional comparability required between the controlled and uncontrolled transactions is reduced when using the return-on-assets profit level indicator as compared to ratios that measure relationships between profit and costs or revenue. See id. § 1.482-5(b)(4)(i), (ii).

In rejecting the Commissioner’s proposed comparable companies on the basis of their 'fundamentally different asset bases,' the tax court made no findings regarding what those differences were, what effect the differences had on the amount of profit allocated to Medtronic Puerto Rico, or whether adjustments could be made to account reliably for any material differences that negatively impacted Medtronic Puerto Rico's profit allocation. See id. § 1.482-5(c)(2)(iv). Medtronic contests this point on the view that the tax court rejected altogether the use of the return-on-assets profit level indicator. That is not our understanding of the tax court’s decision. Although the tax court in 2017 cast doubt on use of the return-on-assets profit level indicator, the court did not do so on remand in 2022. In fact, the tax court employed the return-on-assets profit level indicator as part of its three-step unspecified method.

Medtronic also contends that the tax court found that Medtronic Puerto Rico “had specialized, valuable intangible assets related to its expertise in manufacturing class III medical devices, while the proposed comparables did not.” That assertion is not supported by the record. Although the tax court found that Medtronic Puerto Rico had a 'highly skilled workforce,' the court made no finding that the Commissioner’s five proposed comparable companies did not. The tax court found that Medtronic Puerto Rico and the Commissioner’s proposed comparable companies had “fundamentally different asset bases,” with no further explanation.

On remand, the tax court should make findings as to what were the purported differences in asset bases, what effect any differences had on the amount of profit Page 16allocated to Medtronic Puerto Rico, and, if necessary, whether adjustments can be made to account reliably for any material differences that negatively impacted Medtronic Puerto Rico’s profit allocation.

The Commissioner next challenges the tax court’s rejection of the comparable profits method on the basis that Medtronic Puerto Rico and the proposed comparable companies perform “different functions.” The Commissioner acknowledges that Medtronic Puerto Rico performed only one function—finished-product manufacturing—while the proposed comparable companies all performed additional functions, including research and development and distribution. But, he argues, the comparable companies’ performance of additional functions does not render the comparable profits method unreliable.

We conclude that performance of different functions by proffered comparable companies is insufficient reason to reject the comparable profits method. The comparable profits method accounts for functional differences between the tested party and comparable companies because “differences in functions performed often are reflected in operating expenses,' so that companies 'performing different functions may have very different gross profit margins but earn similar levels of operating profits.” Id. § 1.482-5(c)(2)(ii). Indeed, the comparable profits method may be used to price intangible property even where significant functional differences “are likely to materially affect gross profit margins, but it is not possible to identify the specific differences and reliably adjust for their effect on gross profit.' Id. § 1.482-8, Ex. 6. That the Commissioner’s proposed comparable companies perform functions other than finished-product manufacturing does not render the comparable profits method inapplicable.

Page 17Accordingly, on remand the tax court should reconsider the comparability of the functions performed by Medtronic Puerto Rico and the Commissioner’s five proposed comparable companies, and apply the approach to functional differences envisioned by the regulations. See id. §§ 1.482-5(c)(2)(ii), -8, Ex. 6.

We also conclude that the court did not make sufficiently specific findings regarding the amount of product liability risk borne by Medtronic Puerto Rico during tax years 2005 and 2006. See id. §§ 1.482-5(c)(2)(ii), -1(d)(3)(iii)(A)(5). The court noted the parties’ contrasting evidence about the amount of risk borne by Medtronic Puerto Rico during those years: the Commissioner's expert quantified the subsidiary’s risk at around $25 million, and Medtronic’s expert placed the amount at $220 million to $235 million. The tax court also observed that Medtronic Puerto Rico had incurred product liability costs ranging from $117 million to $324 million in four previous recalls of devices and leads. The tax court, however, did not resolve the factual dispute about quantity of risk. The tax court instead rejected the Commissioner’s $25 million valuation and corresponding adjustment as “not in line with the costs associated with prior recalls,” and found simply that “all the product liability risk was allocated to, and borne by” Medtronic Puerto Rico. Without a finding on the amount of product liability risk borne by Medtronic Puerto Rico in tax years 2005 and 2006, we cannot properly evaluate the tax court’s rejection of the Commissioner's proposed comparable profits method. See Medtronic I , 900 F.3d at 615.

Nor did the tax court make sufficient findings to support its implicit conclusion that any difference in risks borne by the proposed comparable companies and those borne by Medtronic Puerto Rico was material. Although the tax court credited testimony from Medtronic’s expert Glenn Hubbard that “quality is more important for a manufacturer solely of class III devices than for the companies [the Commissioner’s expert] Heimert selected as comparables,' the findings do not explain why the manufacturer of class III devices necessarily bears more financial risk than a company Page 18that manufactures class I, II, and III devices. On remand, the tax court should quantify the amount of product liability risk borne by Medtronic Puerto Rico and the Commissioner’s proposed comparable companies, evaluate whether any difference between them is material with respect to the comparability of profits earned, and, if necessary, consider whether an adjustment can be made to account reliably for any material difference. See 26 C.F.R. § 1.482-5(c)(2)(iv).

After making these findings, the tax court should reconsider its conclusion that the Commissioner’s allocation of 12-14% of the profit from the devices and leads to Medtronic Puerto Rico was “unreasonable.” The tax court reached that conclusion 'for the same reasons' given in its 2017 opinion after observing that 'the five comparables are not identified as solely class III products.” The court should reassess its conclusion after reevaluating the comparable companies under the comparable profits method as discussed above.

In addition to the comparable profits method, the Commissioner offered evidence that Medtronic could replace Medtronic Puerto Rico in its function as a finished-device manufacturer by relying on a different Medtronic facility or building a new manufacturing facility. See id. §§ 1.482-4(d)(2) (royalty paid by foreign subsidiary to U.S. parent to license a proprietary widget production process is not arm’s length if the U.S. parent could have earned more profit by producing the widget itself), -1(d)(3)(iv)(H) (comparability of economic conditions analysis requires considering the “alternatives realistically available to the buyer and seller”). The tax court acknowledged this point, saying that “there is a possibility that [Medtronic Puerto Rico] could have been replicated,” albeit “not without substantial time and cost.” The tax court, however, made no findings about how much time and cost Page 19Medtronic would have to incur to replicate Medtronic Puerto Rico’s role in the conglomerate’s structure. On remand, the tax court should make those findings and determine whether manufacturing the devices and leads in a different Medtronic facility or building a new manufacturing facility was a realistic alternative to the Technology Licenses. This determination is necessary to facilitate review of the tax court’s ultimate profit allocation to Medtronic Puerto Rico.

For these reasons, we vacate the tax court’s order and remand the case for further proceedings consistent with this opinion.