This section draws in large part on Scholz et al. (2020).
The Decision of 18 July 2019 on the Qualcomm (predation) caseFootnote 26 is the European Commission’s first predatory pricing case since the Wanadoo Decision of 2003.Footnote 27 It concerns the market for “baseband” chipsets of the third generation—Universal Mobile Telecommunications System (UMTS) chipsets—which are used in mobile devices (such as mobile phones, tablets, and “dongles”) to enable calls and data exchange via the cellular network.Footnote 28
The Commission concluded that between 1 July 2009 and 30 June 2011, Qualcomm abused its dominant position in the UMTS chipset market by selling certain volumes of three of its UMTS baseband chipsets to two of its key customers—Huawei and ZTE—below cost, with the intention of foreclosing Icera: its most important competitor in the leading-end segment at the time. For this infringement of Article 102 of the Treaty on the Functioning of the European Union (‘TFEU’), the Commission imposed a fine of EUR 242 million on Qualcomm. Qualcomm lodged an application for annulment of the Decision in November 2019, which is currently pending before the General Court.
Facts of the Case
During the investigated period, Qualcomm was the dominant developer and manufacturer of mobile communications technologies and baseband chipsets in a variety of standards—in particular, baseband chipsets that were based on UMTS. Qualcomm’s customers are device makers—original equipment manufacturers (OEMS)—that buy baseband chipsets and integrate them into their mobile devices before selling them to mobile network operators (MNOs) or wholesalers of electronic devices.
Icera—a semiconductor start-up that was founded in 2002 and was headquartered in Bristol (United Kingdom)——specialised in the development of UMTS baseband chipsets. In May 2011—towards the end of the period for which the Commission found predation—Icera was acquired by the US technology company Nvidia. In 2015, Nvidia decided to wind down its baseband chipset business, thus withdrawing the particular chipset technology that had been developed by Icera (so-called soft modems) from the market. This market exit led to a reduction in product variety and a possible slowdown in technological progress and therefore had a negative impact on consumers.
The Qualcomm case differs from previous European predatory pricing cases, such as the Wanadoo decision of 2003, mainly because almost all of the investigated prices were above average variable costs (AVC) but below long-run average incremental costs (LRAIC). By contrast, in the previous predation cases only part of the sales had been made at a price between AVC and LRAIC, whereas the price was below AVC for a significant volume of sales. In the Qualcomm predation case, the finding that Qualcomm’s conduct was intended to foreclose Icera from the market was therefore an essential element of the Decision that complemented the results of the price-cost test. This finding of intent is based on a thorough analysis of Qualcomm’s contemporaneous internal documents, which were supplemented by third-party information. The analysis of the documentary evidence gave a clear picture of the market environment around the period of the infringement and Qualcomm’s desire to defend its dominant position by means of an abusive predatory pricing strategy against Icera.
The Rationale of Qualcomm’s Conduct: Nipping Competitors in the Bud
While Qualcomm’s conduct during the investigated period aimed at protecting Qualcomm’s dominance in the entire UMTS chipset market—and in particular its strong position in the high-volume segment of baseband chipsets for use in mobile phones—its pricing strategy was selective: in terms of the market segment and also the customers that were affected by predatory prices.
First, Qualcomm’s strategy focused on UMTS chipsets that offered advanced data rate performance: the “leading-edge segment”. In this segment, Icera had started to gain traction in 2008/2009 due to: the software upgradability of its chipsets to leading-edge data rates; its smaller die size; and its competitive pricing.Footnote 29
Second, Qualcomm's strategy focused on Huawei and ZTE: the two strategically most important customers for leading-edge UMTS chipsets during the relevant period. These two customers were the main OEMs of “mobile broadband” (MBB) devices at the time.Footnote 30 While the market for MBB devices was (and still is) relatively small compared to the mobile phone market, they were particularly important for the leading-edge chipsets that were supplied by Icera and Qualcomm.
By containing Icera’s growth at the two key customers in the leading-edge segment, which consisted at the time almost exclusively of chipsets that were used in MBB devices, Qualcomm intended to prevent Icera—a small and financially constrained start-up—from gaining the reputation and scale necessary to challenge Qualcomm’s dominance in the wider UMTS chipset market. Qualcomm was particularly concerned by Icera’s threat because of the expected growth potential of the leading-edge segment due to the global take-up of smart mobile devices.
Thus, while only a small part of the UMTS chipset market was affected by below-cost pricing, this selective and targeted predatory strategy had an adverse impact on the entirety of the market that the strategy intended to foreclose. Moreover, Qualcomm’s conduct also affected the contestability of future markets, beyond the UMTS chipset market on which the predatory attacks occurred. With revenues from MBB sales falling below Icera’s targets, Icera was forced to scale back its R&D in voice functionality/LTE, which was crucial for Icera to be able to enter the LTE smartphone segment as scheduled by the end of 2011. Instead, Icera’s entry into the LTE smartphone segment was delayed to February 2013, by which time Icera had already lost the commercial opportunity of being the first entrant in this segment; and its acquirer, Nvidia, eventually wound down Icera’s modem operations in May 2015.
Economic Analysis of Predatory Pricing: Principles
Predatory pricing cases are notoriously difficult to pursue, which may explain why they are “something between a white tiger and a unicorn”.Footnote 31 At a conceptual level, economists have long struggled to come up with a convincing rationalisation of below-cost pricing that lives up to the standards of modern game theory.Footnote 32 The idea that predation may involve a dynamic strategy where market entrants can at first only compete on a sub-segment of the market, which may become the theatre of a predatory attack to protect future markets, has only recently been explored.Footnote 33
At an operational level, the difficulty of establishing predatory pricing lies in the fact that “low” prices can arise both under predation and under a pro-competitive reaction to entry, with prices “naturally” declining as the competitive pressure experienced by the incumbent increases. In technical terms, this inherent ambiguity may lead to two types of errors: A Type I error consists in failing to identify a predatory attack when in fact it has occurred. If such a Type I error is likely to occur, this weakens antitrust enforcement and encourages unlawful behaviour. A Type II error consists in finding predation when in fact there was none. The problem with Type II errors is that they tend to have a chilling effect on competition, as incumbents may hesitate to react pro-competitively to entry for fear of triggering an antitrust investigation.
Different jurisdictions have tried to solve this conundrum in different ways. In the US, the legal standard for finding predatory pricing is based on the “Areeda-Turner” test, which requires that prices are below short-run AVC, and that recoupment is at least possible.Footnote 34
The legal framework applicable in the EEA builds on two different tests, known as the AKZO I and AKZO II tests. With respect to the relevant cost benchmark, the case law states that “first, […] prices below average variable costs must be considered prima facie abusive inasmuch as, in applying such prices, an undertaking in a dominant position is presumed to pursue no other economic objective save that of eliminating its competitors. Secondly, prices below average total costs but above average variable costs are to be considered abusive only where they are fixed in the context of a plan having the purpose of eliminating a competitor.”Footnote 35
In other words, there is a presumption that prices below AVC are abusive (AKZO I). However, even prices above AVC (but below average total costs) may be found abusive, provided that there was exclusionary intent (AKZO II). The Qualcomm predation case discussed in this article falls squarely within this second category. While prices above AVC do not generally give rise to antitrust concern (as they are compatible with many forms of pro-competitive behaviour), they are not a safe haven in the EEA, as the existence of evidence on intent may give rise to a finding of abusive pricing nonetheless.
Economic Analysis of Predatory Pricing: The Qualcomm Case
As explained above, Qualcomm’s predatory pricing strategy was targeted at certain products and customers, which were indispensable for Icera’s development prospects on the UMTS market and beyond, but represented only a small part of Qualcomm’s product portfolio or turnover. The Commission therefore focused its price-cost test on those three chipsets and the two customers on which Qualcomm’s predatory pricing strategy was based. The test was carried out in three steps, which are explained in more detail below: (1) the reconstruction of the effectively paid prices; (2) the comparison of these prices with Qualcomm’s AVC; and (3) the calculation of the non-variable production costs and the comparison of the prices with Qualcomm’s LRAIC.
For reasons of legal certainty, the relevant cost benchmark for both the AVC and non-variable costs included in the calculation of LRAIC were the costs of the dominant undertaking (i.e., Qualcomm), not the costs of the foreclosed undertaking (i.e., Icera), as the latter would not normally be known with the necessary degree of precision to the dominant undertaking. Therefore, the Commission’s price-cost test is equivalent to an “as-efficient-competitor” test: It indicates whether—under the assumption that Icera was as efficient as Qualcomm in producing the products under investigation—Icera could have survived in the market in the long run in the face of the prices that were charged by Qualcomm.Footnote 36
From Accounting Prices to Effectively Paid Prices
In the Statement of Objections of December 2015, the Commission had used Qualcomm’s average prices from its formal accounting in its price-cost test. However, these data only partially reflected the unit prices that were effectively paid by Huawei and ZTE for the products that were under investigation. In addition to the realised revenues from the quantities that were sold at any given time, the accounting prices also included the release of reserves that had been created to cover possible (eventually non-realised) rebate claims for units that were shipped in previous quarters, which were therefore unrelated to the units that were shipped in the quarter in which the release was recorded. Following an in-depth analysis of the resulting distortions, the Commission considered that it was necessary to correct the accounting price data for these releases by allocating them to the sales to which they effectively related.
Furthermore, the Commission’s investigation revealed two one-off payments that were made by Qualcomm to Huawei and ZTE, respectively. In Qualcomm’s accounts, these one-off payments had been attributed to a specific chipset model. However, it was apparent from internal Qualcomm documents (sometimes supported by third-party documents) that these one-off payments were intended as a discount for certain quantities of other chipsets for which Qualcomm did not wish to reduce the price directly at that time. The Commission therefore allocated the one-off payments in question to the specific sales that resulted from those documents, which significantly reduced the unit price effectively paid for the sales in question.
The effectively paid prices thus established were then cross-checked against relevant internal documents that reflected Qualcomm’s plan to foreclose Icera from the UMTS chipset market. The price reconstruction carried out by the Commission made it possible to link each of the price points discussed in those documents to the actual transactions that had occurred at these prices. In doing so, the Commission provided the necessary proof that the predatory pricing strategy that was envisaged in the Qualcomm documents had actually been implemented.
Comparison of Prices and Variable Costs
Since Qualcomm does not manufacture its chips itself, but outsources this to third-party chip manufacturers (“foundries”), the AVC was calculated on the basis of the unit prices that were invoiced by Qualcomm’s foundries for the shipments of the respective chipsets to Qualcomm. In order to take account of the fact that not all chipsets were immediately resold but sometimes kept in stock for some time, the Commission reconstructed the respective stocks so as to allow for a meaningful comparison of the acquisition cost of any given chipset to the price that was charged for this chipset by Qualcomm in the quarter in which it was eventually sold.
As mentioned above, the unit prices effectively paid by Huawei and ZTE were always above the AVC calculated by the Commission, with very few exceptions. The legal standard to be applied is therefore the AKZO II standard: There may also be abusive pricing at levels that are between AVC and average total cost (ATC) if these prices were intended to foreclose a competitor from the market.
Comparison of Prices and LRAIC
As in previous cases (see, for example, Wanadoo), the Commission considered LRAIC—the average of all (variable and fixed) costs that were incurred by Qualcomm in the production of the products under investigation—to be the most appropriate cost measure for the purposes of the price-cost test. LRAIC is below ATC because Qualcomm is a company that produces a large number of different products, of which only three specific chipset models were relevant for the implementation of its predatory pricing strategy vis-à-vis Icera. In the case of a multi-product company, in addition to the costs that can be attributed to the production of certain products, there are also genuine “common costs” (such as commercialisation or administrative costs) that could not have been avoided if a single product had not been produced. Such common costs are therefore not taken into account in the calculation of LRAIC, but would be fully included in the calculation of ATC. LRAIC is therefore a conservative cost benchmark as compared to ATC.
In order to calculate LRAIC for the investigated chipsets, the Commission relied, in addition to the AVC, on product-specific development costs that were obtained from internal Qualcomm documents. The latter costs do not vary with the short-run volume of sales (and are therefore not part of variable costs), but are unavoidable in the long run to achieve the entirety of the chipset sales under investigation. As most of these development costs are incurred before the sale of the chipsets to which they relate, the Commission had to allocate these development costs to the investigated chipset sales, with the use of a revenue-based allocation method. This approach takes into account the fact that leading-edge chipsets tend to generate the highest margins at the beginning of their product life cycle, since the innovative added value of these products is highest when they are first launched on the market.
In this context, if instead a quantity-based allocation were to be applied—as is common in business cost accounting—this would significantly increase the LRAIC for the products under investigation towards the end of their life cycle, when the sales volumes are still high, but margins tend to be much lower. As a result, costs would mechanically exceed prices in later periods, even though no predatory pricing may have been intended.
For innovative products whose margins are subject to such strong variations over the life cycle, a revenue-based allocation takes account of the impact of these life cycle dynamics on the calculated minimum margin in a realistic manner. This approach also allowed the Commission to take into account spill-over effects between two of the three products under investigation.
The Commission then compared the effectively paid prices with the respective LRAIC. This comparison was carried out on a quarterly basis for each of the three investigated chipset models and for each of the two key customers. The test showed that almost all sales to Huawei and most of the sales to ZTE were made at prices that were below the respective LRAIC. Qualcomm’s total revenue from these predatory sales was well above Icera’s total turnover in 2011 (the last year in the investigation period).
At the beginning of the period under investigation, Icera had already established a nascent sales relationship with ZTE, and so Qualcomm’s strategy focussed mainly on Huawei: By selling at below-cost prices to Huawei, the latter was to gain a competitive advantage at the next level of the supply chain where chipset suppliers such as Qualcomm and Icera compete indirectly for orders of MNOs through the MBB devices of OEMs that incorporate their respective chipsets. Only later did Qualcomm extend its predatory strategy also to ZTE.
The established case law does not require the Commission to prove recoupment after a predatory episode. In the Qualcomm case, however, it was apparent from the dominant undertaking’s internal documents that such recoupment was never envisaged, and would not have been possible, over the remaining life cycle of the investigated products. This is because (as was described above) the margins of chipsets in the leading-edge segment tend to fall rapidly after a short period of time, even absent any predatory pricing strategy. Instead, the “payoff” from Qualcomm’s strategy was that Icera was prevented from entering the much larger and more profitable segment of smartphone chipsets, which would have resulted in a much more significant loss of profits for Qualcomm than that caused by the predatory pricing episodes on the smaller segment of MBB devices.
Yet, in addition to the price-cost test that was described above, the Commission also examined whether the two high-volume chipsets under investigation had eventually recovered their aggregated LRAIC when their total life-cycle revenues are considered. The main differences with the above-mentioned price-cost test were as follows: (1) The Commission took into account the whole life cycle revenues of the products, including sales in the period following the infringement period (when one of the two high-volume products was still on sale); (2) the Commission also took into account all of the revenues that were obtained from other customers of these two products (not only Huawei and ZTE); and (3) the Commission carried out the test not only for each chipset, but also at the level of the chipset family: The Commission pooled the revenues and the development costs of the two products. Thus, although this overall life cycle analysis took into account far higher revenues than the above price-cost test, the investigation showed that—even at the product-family level—the two products did not fully recover their aggregate product-specific production costs (including development costs). This result does not depend on the specific assumptions that were made about the allocation of development costs and thus corroborates the result of the price-cost test in this case.