A Counterfactual Analysis of Amazon’s Acquisitions Under the 2023 Merger Guidelines

Abstract

The fact that Amazon was allowed to acquire hundreds of companies as it rose to become the fourth most valuable U.S. company in terms of market capitalization and a leader in three lines of business has been viewed by some as damning evidence of underenforcement by the United States antitrust authorities. In this article we ask the obvious question: If the 2023 Guidelines had been in place instead of prior guidelines, what effects would they have had on Amazon’s development? To provide an answer, we identify relevant changes in the guidelines and then select for review a subset of Amazon’s 280 acquisitions over the period 1998 to 2022. In our counterfactual, we analyze five horizontal acquisitions, four vertical acquisitions, and two sets of serial acquisitions. We find that the 2023 Guidelines would have broadened the bases for potential challenges and thereby would have increased the likelihood that Amazon would have faced greater resistance from antitrust authorities. The lack of safe harbors, the plasticity of individual Guidelines, and the optionality to challenge mergers under alternative theories would have exposed most of Amazon’s acquisitions to challenge. The lack of meaningful guidance about which individual transactions would have been challenged suggests that going forward enforcer discretion will play a yet larger role. Regarding Amazon’s serial acquisitions of nearly one hundred technology firms, we find that the 2023 Guidelines would have provided multiple rationales for intervention. Therein lies a weakness in the Guidelines and in antitrust policy: the lack of a framework for assessing both the anticompetitive and procompetitive effects of such acquisitions in high-tech industries.

1 Introduction

In the nearly three decades since it began selling books online, Amazon developed into the fourth most valuable United States company and a leader in three lines of business: eCommerce, cloud computing, and interactive technologies for individuals and homes (Snyder, Canaday, and Hughes, 2022).¹ The fact that Amazon was allowed to acquire hundreds of companies has been viewed by some as damning evidence of underenforcement by the United States antitrust authorities.²

Here, we ask: If the 2023 Merger Guidelines³ (the “Guidelines”) were in place when Amazon launched, how would they have affected its advance? Our counterfactual analysis leverages data on 280 Amazon acquisitions over the period 1998 through 2022.⁴ Included are: (a) acquisitions that allowed Amazon to rapidly broaden the scope of its product and service offerings; (b) acquisitions of actual competitors; (c) high-profile acquisitions that expanded Amazon vertically; and (d) “serial acquisitions” of relatively small companies with technological capabilities.

Given the breadth and volume of Amazon’s acquisitions, the counterfactual is an informative means of assessing how the Guidelines would be applied to cases that would involve actual and potential competition, nascent competition, vertical integration, monopolization, and innovation.

In Sect. 2 we discuss four features of the Guidelines that potentially could have been used by enforcers to challenge Amazon mergers in our counterfactual: (a) tougher standards for mergers between actual competitors (whether in concentrated markets or not); (b) capacious theories for challenging acquisitions of potential or nascent competitors; (c) renewed emphasis on vertical foreclosure concerns, including issues stemming from platforms acquiring participants; and (d) a new focus on the cumulative effects of serial acquisitions. Reasonable minds can differ but in our view the Guidelines’ focus on nascent competition–and potential competition–generate the most noteworthy new source of scrutiny.⁵

Section 3 provides context about Amazon’s three lines of business and then selects a subset of acquisitions for our counterfactual analysis. We note that many of Amazon’s eCommerce-related acquisitions expanded its business horizontally: most commonly, by adding or strengthening product lines on the platform. By contrast, most acquisitions that were related to its two other lines of business—cloud computing, and interactive technologies for homes and individuals—added to technological capabilities.

The acquisitions that we selected for our counterfactual analysis include all of Amazon’s major acquisitions: Zappos (2009); Whole Foods (2016); Ring.com (2018); and MGM Studios (2021). We also offer comments on other acquisitions that pose novel fact patterns with respect to actual and potential competition. Together the selected acquisitions create a lens through which we can assess whether and, if so, to what extent the Guidelines evince a far-reaching expression of enforcement.

In Sect. 4 we proceed with our counterfactual analysis and assess the implications for antitrust policy and for Amazon’s development. We find that the 2023 Guidelines would have provided potential pathways to challenging a majority of Amazon’s acquisitions that we selected for our counterfactual analysis. Its horizontal acquisitions could have received serious scrutiny based on actual competition, potential and nascent competition, entrenchment, and platform competition concerns. Its vertical acquisitions would have been analyzed not only based on traditional foreclosure concerns, but also through the prism of platform economics (and its attendant concerns about self-preferencing and depriving rivals of network efficiencies) and the Guidelines’ overarching focus on combating entrenchment of firms that are deemed to be dominant. In addition, patterns of small acquisitions would have received additional scrutiny given the Guidelines’ wholistic approach to evaluating serial acquisitions.

We do not claim that most of Amazon’s acquisitions would have or should have been challenged. The options provided to the Agencies in the new Guidelines result in a dramatic imbalance between the number of transactions that could be challenged under the Guidelines and the number that antitrust enforcers could realistically pursue—making prosecutorial discretion a more important factor in determining the shape of merger enforcement policy. And the new Guidelines create substantial uncertainties about which of Amazon’s many acquisitions would have received the greatest scrutiny. With little in the way of quantitative guidance or bright-line rules, no safe harbors, and no precise criteria for analyzing key issues such as the prospects of a nascent competitor’s developing into a viable rival, the Guidelines leave many questions unanswered. In certain respects, a cynic might say that these are Guidelines without guidance—or at least without sufficient guidance to allow businesses to predict reliably which of the many mergers that potentially could attract scrutiny would actually be challenged.

The other theme that emerges from our analysis is that the Guidelines may over-deter mergers due to insufficient consideration of procompetitive effects—especially in the context of dynamic industries. Our analysis shows that the Guidelines could provide a narrative for a challenge in connection with virtually all of Amazon’s acquisitions of individual firms with technological capabilities and its serial acquisitions of start-ups. Under the Guidelines, the Agencies could seek to stop the transfers of emerging technologies that are developed by start-ups to firms with commercial traction by claiming that either: (a) the start-ups could have developed into a powerful rival to the acquiring firm; or (b) the acquiring firm could use its control of the technology to foreclose actual or potential competition.

The Guidelines miss the other side of the ledger: the principle that start-ups are motivated by exit strategies—including acquisition—and not by the typically doubtful prospect of start-ups’ developing into a real enterprise. Given that entry anticipates all states of the world, deterring acquisitions of startups will lead to fewer startups. An overarching question for antitrust policy is how to account for efficiencies from new technologies and the market for corporate control: the transferring of assets and capabilities to firms that can maximize their value.

Of course, there is more than one level of unreality to the journey here that we readily acknowledge: When in our counterfactual would Amazon brandish the scarlet letter of a “platform”? Amazon technically became a platform when it launched its third-party marketplace in November 2000; but it is unclear when its importance as a platform operator superseded its importance as a first-party seller, which would refocus the lens through which antitrust enforcers would evaluate Amazon’s conduct and acquisitions. The precise timing is relevant because it could match when the Agencies would increase their scrutiny of individual acquisitions or deem an ongoing series of acquisition as anticompetitive. We do not attempt to demarcate the precise timing of this transition. Suspending disbelief on our journey nevertheless has benefits in showing how elastic the Guidelines are and how they may (and likely will) be brandished against firms that are deemed to be dominant platforms.

With regard to Amazon itself, our analysis leads to the conclusion that the Guidelines could have substantially stunted its development. The net effects on Amazon in the counterfactual would turn on the interaction between the judiciary and the Guidelines: whether the latter would get traction with the former, and, importantly, whether Amazon could adjust its strategy and its ability to substitute internal investments for acquisitions. With regard to the latter, the evidence is sufficient to establish a mix of cases: The effect of some blocked acquisitions would be minimal because of Amazon’s demonstrated ability to develop similar capabilities by internal investments. In other cases, the exact opposite conclusion could be drawn, and, as a result, the scope of the Amazon enterprise would be more limited and market structures in its major lines of business could be altered.

The larger issue is, of course, whether a less ubiquitous Amazon would be good for consumers and society.

2 Provisions in the 2023 Merger Guidelines that are Most Relevant to Evaluating Amazon’s Acquisitions

The Guidelines exhibit a broad sweep for potential Agency challenges compared to prior iterations. In this Section we focus on several concepts that likely would have been particularly relevant to evaluations of Amazon’s acquisitions. We discuss: the Guidelines’ tougher stance against horizontal mergers between actual competitors; expansive theories for preserving competition from potential or nascent rivals; a greater focus on vertical mergers (including a new focus on vertical acquisitions by platforms); and a new lens for evaluating serial acquisitions.

2.1 Horizontal Mergers: New Structural Presumptions and No Safe Harbor

The structural presumption—that mergers that significantly increase concentration in highly concentrated markets are likely to harm competition—is an enduring feature of merger enforcement. The latest Guidelines continue the reliance continue the reliance on the structural presumption,⁶ and contain two changes from the 2010 Guidelines that herald tougher enforcement against horizontal mergers⁷: (i) They endorse a presumption that almost any horizontal acquisition by a firm with a market share of 30% or higher is anticompetitive⁸; and (ii) they eliminate the safe harbor for horizontal mergers in unconcentrated markets, and make clear that any merger may be subject to challenge “[i]f evidence demonstrates substantial competition between the merging parties prior to the merger,” regardless of market shares or concentration.⁹ These changes are especially relevant when a large company such as Amazon acquires even a small competitor.¹⁰

2.2 Acquisitions of Potential or Nascent Competitors: Supplementing Traditional Theories with Tools to Challenge Acquisitions by “Dominant” Firms

The Guidelines contain significant changes for acquisitions of potential or nascent competitors. The prior version of the Horizontal Guidelines offered little guidance on this subject¹¹; the new Guidelines specify the two potential competition theories that have long been part of merger enforcement, and add an aggressive theory for challenging acquisitions of nascent competitors by dominant firms.

2.2.1 Actual Potential Competition and Perceived Potential Competition Doctrines

Under the actual potential competition theory, the Agencies may challenge a merger based on the loss of the procompetitive effects that could have been realized upon entry by the firm that is identified as a potential competitor. Under the perceived potential competition theory, the Agencies may challenge a merger based on the loss of a perceived competitive threat that has already spurred greater competition—regardless of whether the potential entrant actually would have entered the market.¹²

The Agencies have not had much recent success blocking mergers under these theories: The FTC lost its 2022 challenge to the Meta-Within merger¹³ and its 2015 challenge to the Steris-Synergy Health merger,¹⁴ which prompted some commentators to opine that current doctrine makes it difficult for the Agencies successfully to challenge acquisitions of potential competitors.¹⁵

2.2.2 Acquisitions of Nascent Competitors by a Dominant Firm

The Guidelines open an additional avenue for challenging acquisitions of nascent competitors by a dominant firm. The draft Guidelines set a 30% market share threshold for identifying dominant firms¹⁶; the final Guidelines offer less specific guidance; they indicate that the Agencies “assess whether one of the merging firms has a dominant position based on direct evidence of market shares showing durable market power.”¹⁷ If a merging firm is found to be dominant, the Agencies can challenge the merger either under Clayton Act Sect. 7 (the merger statute) or Sherman Act Sect. 2 (the monopolization statute).

A Sect. 2 challenge may allow the Agencies to bypass showing that the nascent competitor would have been a successful entrant in the near term: According to the Guidelines, “a firm that may challenge a monopolist may be characterized as a ‘nascent threat’ even if the impending threat is uncertain and may take several years to materialize.”¹⁸

The nascent competitor theory that is featured in Guideline 6 could be used to challenge several types of mergers, including:

• Acquisitions of niche rivals that offer only a subset of the products that the “dominant” acquiring firm offers, but that have the potential to expand product offerings and thereby compete with the acquiring firm more broadly. The Guidelines explain that “the nascent threat may be a firm that provides a product or service similar to the acquiring firm that does not substantially constrain the acquiring firm at the time of the merger but has the potential to grow into a more significant rival in the future.”¹⁹ A firm can start out as a niche competitor, but “[o]nce established in its niche, a nascent threat may be able to add features or server additional customer segments, growing into greater overlap of customer segments or features over time, thereby intensifying competition with the dominant firm.”²⁰

• Acquisitions of firms that may challenge a platform’s status as a “one-stop” shop and encourage users to multi-home. As the Guidelines state, “[a] nascent threat may… facilitate customers aggregating additional products and services from multiple providers that serve as a partial alternative to the incumbent’s offering,” which “may facilitate competition or encourage entry by other, potentially complementary providers that may provide a partial competitive constraint.”²¹

• Acquisition of firms that could have facilitated the growth of the dominant firm’s rivals: e.g., by providing some key technology or other input that spurred rivals’ development. This is a variant of a vertical foreclosure theory, which will be discussed further below. The Guidelines indicate that a nascent competitive threat may be “a firm that could… facilitate other rivals’ growth,”²² and expound on a similar theory (not necessarily limited to nascent firms) in the discussion on mergers that involve platforms, indicating the Agencies would scrutinize platform acquisitions “of firms that provide services that facilitate participation on multiple platforms… [or] provide other important inputs to platform services.”²³

2.3 Vertical Mergers: Emphasis of Traditional Theories of Foreclosure, a New Theory of Harm, and Greater Attention to Platforms

The Guidelines place great emphasis on vertical issues, including the incentives and likelihood of upstream or downstream foreclosure. Two aspects are particularly relevant going forward²⁴:

First, the Guidelines provide a detailed framework for evaluating the merged firm’s ability and incentive to foreclose competitors, including a new share-based presumption. Guideline 5 explains that Agencies will assess the merged firm’s ability and incentive to foreclose competitors’ access to products, services, or routes to market based on: (i) the availability of substitutes; (ii) the competitive significance of the product, service, or route to market that is controlled by the merged firm; (iii) the effect of potential foreclosure on competition in the relevant market; and (iv) the closeness of competition between the merged firm and the firms that rely on the merged firm for products, services, or routes to market.²⁵

Consistent with the Guidelines’ emphasis on market structure, Guideline 5 introduces a form of structural presumption for vertical mergers: The Agencies “will generally infer, in the absence of countervailing evidence” that if the merged firm has at least 50% share of the market for a competitively significant product, service, or route to market that its rivals use to compete, then “the merged firm has the ability to weaken or exclude” rivals.²⁶

Second, the Guidelines explain the vertical issues to which Agencies will pay particular attention when analyzing platform competition; the focus on platform-specific issues is a change from prior guidelines, which did not mention platforms. According to Guideline 9, the Agencies will scrutinize mergers between a platform operator and users so as to assess whether the merger creates a conflict of interest that harms competition. According to the Guidelines, “[a] platform operator that is also a platform participant may have a conflict of interest whereby it has an incentive to give its own products and services an advantage over other participants competing on the platform,” which “can harm competition in the product market for the advantaged product.”²⁷ The Agencies may be concerned that an ecommerce platform such as Amazon may self-preference its proprietary products or otherwise induce other parties to use Amazon’s services: e.g., its fulfillment services or cloud computing services.

The Agencies also will scrutinize a platform’s acquisition of a participant where the merger may “entrench the operator’s position by depriving rivals of participants and, in turn, depriving them of network effects.”²⁸ As is well known, attracting a critical mass of users is vital to a platform’s success and can result in a virtuous cycle of making it easier for the platform to attract users on multiple sides of the platform.²⁹ While the obverse can also occur—a death spiral, where losing participants on one side causes loss of participants on the other side, and so on—the Guidelines encourage the Agencies to enhance actual and potential competition among networks.

2.4 Serial Acquisitions

Guideline 8 makes clear that the Agencies will focus on relevant trends, including (i) industry trend toward concentration; (ii) industry trend toward vertical integration; and (iii) patterns of multiple acquisitions in the same or related business lines.³⁰ Even where a single acquisitions may not warrant scrutiny, the Agencies may analyze a group of transactions as a unified business strategy of “growth through acquisition,”³¹ and assess a broad range of potential anticompetitive effects, including: the elimination of actual and potential competition; reduced access to products and services; and entrenchment of the acquirer. It is clear that the Guidelines were written against the backdrop of competition policy concerns with platforms. Hence, serial acquisitions by platforms would likely be a particular focus for the Agencies in implementing the new Guidelines.

3 Amazon’s Three Lines of Business and the Selection of Acquisitions for the Counterfactual Analysis

In this section we review Amazon’s three major lines of business, and, for each, and provide market-share information that is relevant as to whether structural presumptions would apply to Amazon. We then identify a subset of acquisitions for the counterfactual analysis.

3.1 Amazon’s Three Lines of Business

3.1.1 eCommerce

Amazon’s original line of business—eCommerce—was developed through internal investments in its online retail interface, warehouses and fulfillment centers, and technologies to support customer search and transactions. Early on Amazon developed features such as: robust, hyper-text enable search; customer recommendations; retention of address and payment information; and its Buy-Now algorithm.³²

While other online retailers focused on single-product categories, Amazon moved quickly to make “Amazon.com the shopping destination to find anything.”³³ Its rapid expansion is analogous to “adding departments to a department store but without the physical constraint of traditional retailers.”³⁴ Amazon also moved toward a multi-sided platform: The company introduced Amazon Marketplace for the sale by third-parties of used merchandise and, more significant, established the Associate Program for independent retailers to sell on Amazon.com.³⁵ With these asset-light changes that avoided inventory burdens, Amazon realized network efficiencies and added sources of revenue (Rochet & Tirole, 2006).

But acquisitions are also an important part of Amazon’s eCommerce history. In 1999, Amazon paid $250 M to acquire Alexa Internet, which generated insights about Amazon’s rivals and its customers by tracking user traffic on millions of websites. Subsequently, Amazon acquired 14 relatively small technology companies. Amazon’s largest eCommerce acquisition was the $1.2 B Zappos acquisition in 2009, 15 years after Amazon’s launch. Amazon acquired in 2010 another established online retailer, Quisdi, for $545 M. Its main product line was diapers. Amazon’s third significant horizontal or product-extension acquisition was of Twitch Interactive in 2014 for $842 M. Twitch expanded Amazon’s gaming products by adding an interactive platform for users.

Three other of Amazon’s major acquisitions were vertical in nature: the $775 M acquisition of Kiva Systems in 2012; the 13.7B acquisition of Whole Foods in 2017, and the $10.B acquisition of MGM in 2021. Kiva systems provided integrated hardware and software for warehouse operations. While a brick and mortar business, Whole Foods expanded Amazon’s eCommerce business through synergies with fulfillment and its Prime Membership program. We view the MGM acquisition as backward vertical integration: MGM provided a stock of video content for Amazon’s Prime Video services.

What is Amazon’s eCommerce market share? Former CEO Jeff Bezos testified in 2020 before Congress that Amazon accounted for less than 1% of the $25 trillion global retail sales and less than 4% of U.S. retail sales.³⁶ Relevant for our analysis, however, multiple sources indicate that Amazon’s share of eCommerce revenues in the U.S. has exceeded the 30% threshold and, even in recent years, exceeds the next 10 rivals combined (Chevalier, 2023).³⁷

3.1.2 Cloud Computing

Amazon’s second line of business—Cloud Computing—developed out of its eCommerce business. Amazon engineers “confronted the problem of how to organize, store, and analyze the volumes of data that are generated by its sales” (Snyder et al., 2022). Once these capabilities were developed, Amazon recognized that the services could be marketed to: independent retailers in its Associates Program; sellers on Amazon Marketplace; and a broad base of potential customers. The dedicated subsidiary—Amazon Web Services (AWS)—followed with massive investments in data centers in the United States and globally. An ongoing focus of AWS has been the development of robust service that is related to: computing; data base storage and security; support for web and mobile apps; machine learning and Artificial Intelligence; and internet-of-things buildouts.

Objectively, internal investments have been more consequential than acquisitions in building AWS. The absence of any acquisitions of significant cloud computing rivals is also noteworthy. Amazon, however, has been active in acquiring technical capabilities. The most significant is the 2016 $350 M acquisition of Annapurna Labs: a fabless supplier of leading-edge chips for cloud computing. Over its history, Amazon acquired 29 companies with technological capabilities. Amazon was especially acquisitive in the seven-year period beginning in 2015, when Amazon accelerated its efforts to improve services that were related to: data base migration; security; analytics; machine learning; and blockchain systems.

AWS has become host to some 50 million websites and the leader in cloud-based computing services.³⁸ Most recent data put its global share of cloud computing revenue at 32% and indicate a higher United States share.³⁹ In contrast to its eCommerce business, Amazon faces ongoing competition from rivals such as Alphabet, IBM, and Microsoft who operate at scale.

3.1.3 Interactive Technologies for Individuals and Homes

Apple’s 2003 introduction of iTunes may have motivated Amazon’s development of its third line of business: One year later, Amazon began planning for the launch of the Kindle in 2007. While considered to be an eReader, Amazon recognized the value of the Kindle as a consumer interface (Brandt, 2011). Amazon’s subsequent success with interactive technologies is attributable in part to internal investments and marketing of devices—e.g., smart speakers—to its established user base. Over time, Amazon has expanded services to include virtual assistants, home security systems, streaming devices, and Wi-Fi systems (Snyder et al., 2022). These smart devices have “skills” that allow individuals to use other services: e.g., listen to Amazon Prime Music; run iPhone applications; conduct internet searches; and task management.

Amazon substantially advanced this business through acquisition. We view the $993 M acquisition of Ring.com (2018) as horizontal in nature given that Amazon had begun development of security systems—including cameras.⁴⁰ However, most of Amazon’s acquisitions in this line of business focused on technology and, more specifically, on startups that would “create momentum for voice and smart devices and show the art of what’s possible.”⁴¹ Amazon allocated $200 M to the Alexa Fund, an internal private equity fund, with which it completed 70 acquisitions of start-ups in just seven years. A majority were partial acquisitions; this is consistent with Amazon’s interest in gaining information about new technologies and positioning itself to invest more if the technology progressed.

Market shares of smart speakers are a rough inferential proxy for market shares in interactive devices for homes and individuals. Since 2016 Amazon’s over 90% share of global sales has eroded, but it continues to lead with a 28% share (Laricchia, 2023) Its major rival is Alphabet with a 17% global share. Sources indicate that Amazon has a two-thirds share of voice-activated smart speakers in the U.S.,⁴² and that most owners of Amazon’s speakers have integrated them with other Amazon products and services (Richter, 2023b).⁴³

3.2 Selection of Acquisitions for the Counterfactual Analysis

The available data on Amazon’s 280 acquisitions over the period 1998 to 2022 include the target, the date of the acquisition, the up-front acquisition cost, and whether the acquisition was full or partial.⁴⁴ This total is greater than the totals that are in the US House Report and the Stigler Center Report.⁴⁵ From Table 1, which breaks down the acquisitions across Amazon’s three lines of business, we observe that: (a) nearly half of the 280 acquisitions are related to its original line of business, eCommerce; and (b) the vast majority of acquisitions that were related to interactive technologies for individuals and homes were partial.

Table 1 Summary data on Amazon’s acquisitions by line of business: 1998–2022

The available data also allow categorization as (i) horizontal expansion; (ii) vertical expansion; or (iii) adding technological capabilities. While acquisitions of technological capabilities could be viewed as a form of vertical integration, the rationale for categorizing these separately is that adding technological capabilities typically did not expand the scope of the Amazon enterprise. In addition, their frequency matches to the rubric of serial acquisitions.

Table 2 summarizes our selection of individual transactions for our for our counterfactual analysis by line of business, type of acquisition, year, and acquisition cost in millions of dollars. We include eight of Amazon’s eleven acquisitions that cost $500 M or more.⁴⁶ As indicated, our set includes five horizontal acquisitions across the three lines of business; these provide opportunities to review acquisitions that raise concerns about actual and potential competition. We also include four vertical acquisitions that allow for exploration of concerns about foreclosure and entrenchment.

Table 2 Selection of individual acquisitions for counterfactual analysis

Recognizing that the Agencies might view Amazon’s serial acquisitions of technology-related companies as anticompetitive under Guideline 8, we also analyze two sets of acquisitions that are listed in Appendix A. As summarized in Table 3, Amazon acquired 29 technology companies that match to its Cloud Computing business over the period April 2015 to November 2021: an average of 1.35 acquisitions per quarter. The second set of 70 acquisitions matches to Interactive Technologies for homes and businesses. This series involves yet greater frequency: an average of 2.5 acquisitions per quarter over a seven-year period.

Table 3 Two sets of serial acquisitions for counterfactual analysis

Many of the targets were start-ups; but the first set includes several companies with customers and revenues. Cost information is spotty; but the cumulative cost of the first series of acquisitions would not be expected to exceed $2B.⁴⁷ The cumulative cost of the second series likely matches to the $200 M that Amazon allocated to the Alexa Fund, whose stated purpose was to invest in start-up technologies.

4 Counterfactual Analysis

Having selected a total of 107 acquisitions for our counterfactual analysis, our main questions are: First, would these actual acquisitions have been challenged under the new Guidelines?; and, second, what would have been the bases for the challenges?

4.1 Horizontal Acquisitions

4.1.1 Zappos

Amazon’s July 2009 $807 M acquisition of Zappos.com, which was reviewed by the FTC at the time, likely would have received greater scrutiny under the new Guidelines. Zappos—an online retailer that focused on a single product line—had surpassed $1 billion in shoe sales.⁴⁸ Zappos had created an iconic brand image arising from its strong customer service and returns policy, which was an especially important feature for online sales of footwear. Public statements by the parties emphasized potential learnings that would enhance customer experience for both customer bases.⁴⁹

But under the counterfactual, the Agencies would have focused on both actual competition between the two parties in shoe sales and potential competition in other product lines such as hiking gear and exercise apparel. In December 2007, 19 months before the Zappos acquisition, Amazon touted shoes sold as one of their best product lines.⁵⁰ Zappos was the stronger brand, had more sales, and could be viewed as a maverick that had established a superior reputation for customer service. As of June 2007, Zappos had 4.5 M visitors to its site compared to Amazon’s 777,000 visitors to Endless.com—Amazon’s site for shoe sales and related products.⁵¹

The potentially deeper concerns with the Zappos acquisition would have focused on potential competition: Amazon famously started off in 1995 selling only books and then quickly expanded into other product lines to become the virtual ‘one stop’ shopping experience it is now. At the time of the Zappos acquisition, the Agencies likely would have asked, who could challenge Amazon in becoming an online “department store”? The list would include Walmart.com with its breadth of products, but also Zappos if it chose to expand its product lines through internal investments.

As the 2023 Complaint in Federal Trade Commission v. Amazon makes plain, the FTC believes Amazon is a monopolist in online shopping and that absent enforcement actions (that would seek behavioral and perhaps structural outcomes) the best path forward is for a maverick firm to gain traction and expand outwards from a core experience into other lines of business. Could Zappos have expanded from shoes into other product lines and become a “second” Amazon? The quirkiness of Zappos and its former CEO would argue against Zappos’ mounting a broad-based effort to broaden its product lines. But the scenario was certainly not implausible,⁵² and under the counterfactual the scenario would have been front and center in the merger review.

One advantage of retrospection is that we know that Amazon chose to maintain Zappos.com as a separate, consumer-facing site. This preserved Zappos’ brand-name capital while allowing operational integration. Post-acquisition, Zappos revenues grew 50% in the first year post-acquisition, and online sales of shoes grew relative to the larger volume of shoe sales through traditional retailers.⁵³ These facts might counter the view that the merger did not reduce actual competition in shoe sales. Yet even if these outcomes would have been anticipated at the time of the merger review, the Agencies might have considered an intervention based on a reduction in potential competition in a broader set of product lines.

4.1.2 Quidsi

Like the Zappos acquisition, the Quidsi acquisition in November 2010 raises issues with respect to actual and potential competition, as well as other issues with respect to pre-acquisition predation by Amazon. Quidsi—the parent company of online baby care retailer Diapers.com⁵⁴—was founded in 2005 and grew into a $300 M revenue business.⁵⁵ Amazon, having expanded into selling diapers in 2009 or 2010, was a direct competitor in the online sales of diapers. Earlier, the consensus view at Amazon was that diapers were too bulky and low-margin to yield profits.⁵⁶ With the benefit of hindsight, the profitability of online diaper sales did indeed prove to be elusive.⁵⁷

The FTC issued a second request and investigated the proposed acquisition for several months the Agency ultimately cleared the deal after concluding that a sufficiently large number of other companies sold diapers online.⁵⁸ Would the new Guidelines have made a difference? The review of actual competition might have emphasized: (i) Quidsi’s surprising and early success; and (ii) Quidsi’s dedicated customer base. Guideline 2 states that “[i]f evidence demonstrates substantial competition between the merging parties prior to the merger, the Agencies can determine that the merger may substantially lessen competition”—regardless of market shares or market concentration. According to press reports, Amazon was tracking Quidsi prices and adjusting its prices in response.⁵⁹

As we explained in Sect. 2, while the concern over elimination of competition between close competitors is not new (and was reflected in the unilateral effects theory in the 2010 Guidelines), Guideline 2 in the 2023 edition is notable for its breadth and lack of a limiting principle: It applies to cases where there is “substantial competition between the merging parties” and does not contain a presumptive safe harbor for mergers resulting in unconcentrated markets or increases in HHI under 100. ⁶⁰ Hence, the new Guidelines allow for challenges of horizontal mergers where there is no evidence of market concentration and plenty of remaining competitors (as was the case in the diaper market)

An agency under the 2023 Guidelines could have focused not only on the existing product overlap, but on Quidsi’s potential to expand into baby care products and beyond. If the Agency had found Amazon to be dominant, it likely would have investigated under Guideline 6 to assess whether the acquisition threatened to entrench Amazon’s position by removing a nascent threat. And the Guidelines specifically contemplate potential challenges to acquisitions of “[f]irms with niche or only partially overlapping products or customers [that] can grow into longer-term threats to a dominant firm.”⁶¹ The outcome of an investigation and the decision as to whether to challenge the deal likely would have, therefore, depended on the Agency’s assessment of Amazon’s dominance and Quidsi’s potential to transform itself from a niche baby-care specialist into a broader e-commerce rival.⁶²

The allegations of predatory conduct by Amazon pre-acquisition also would have been grist for a challenge. When it was engaged in negotiations to acquire Quidsi, contemporaneous reports indicated that Amazon engaged in a price war. If documented, this at the very least would have contributed to the view that Quidsi and Amazon were close competitors, and would have strengthened the case for a challenge based on traditional concerns about elimination of actual competition.

4.1.3 Twitch Interactive

Amazon’s $842 M acquisition of Twitch Interactive in 2014 features a different dimension of competition: to add features to platforms that add users and increase their engagement. Amazon described the target as follows: “Twitch is the leading live video platform and community for gamers where more than 55 million have gathered to broadcast, watch and talk about video games. Twitch’s video platform is the backbone of both live and on-demand distribution for the entire video game ecosystem.”⁶³ Of note, in 2011, before Twitch Interactive was acquired, Amazon began offering its Prime Video Services; and in 2014 Amazon raised annual member fees from $79 to $99.⁶⁴

After acquiring Twitch Interactive, Amazon integrated gaming benefits into its platform:

Amazon first launched gaming benefits in 2016 as Twitch Prime. Today, Prime Gaming brings more new content for more games than ever before, plus more free games, and a monthly Twitch channel subscription. Now the more than 150 million paid Prime members around the world can get even more value out of their Prime membership with the best of gaming.⁶⁵

Amazon was not, however, considered to be a leader in games at the time—in part because its entry into smart phones failed. Hence, the prospective loss of actual competition from the Twitch Interactive acquisition would not have been a likely reason for a challenge.

The Guidelines provide a separate basis for a potential challenge. If Amazon is viewed as a leading platform, an acquisition that incorporates millions of users onto the platform could be characterized as entrenchment.⁶⁶ Of the horizontal acquisitions that are included in our analysis, this case appears as an unlikely challenge: While the Twitch acquisition brought a significant number of users to Amazon’s platform, it would have been difficult to argue that these users were locked into Amazon. The users remained free to use other platforms, and it would have been a stretch to argue that the acquisition made it more difficult for other platforms to compete by adding users.

Yet a fallback theory enunciated in the Guidelines could have motivated a challenge: The Agencies could have claimed that competition would be enhanced by forcing Amazon to develop improved video-encoding software; and then (the argument would go) Twitch Interactive and Amazon would compete in video-encoding software. A challenge based on this theory would require a combination of indifference to the realization of efficiencies that would benefit customers and confidence that Twitch Interactive—operating independently—would do more to enhance competition than being integrated into Amazon’s platform. While the potential efficiencies that are associated with transferring technologies internally and supporting them on an ongoing basis are often substantial, the major point that would have favored a challenge is that, prior to the acquisition, Twitch Interactive was a viable commercial enterprise with actual customers and revenues.

4.1.4 Elemental Technologies

Amazon made one major horizontal acquisition related to cloud computing: its $500 M purchase of Elemental Technologies in 2015. The target had developed software for compressing and formatting massive video files for different devices. Elemental Technologies, which had received substantial venture funding from sources other than Amazon, had established itself as a technology leader: with annual revenues were $50 M at the time of acquisition and a reported customer base of 700 that included ABC, the BBC, CNN, Comcast, Ericsson, and ESPN.⁶⁷ Prior to the acquisition, Amazon and Elemental Technologies—which was featured on AWS’s partner page—had been “working together on shared sports and entertainment industry accounts for 4 years.”⁶⁸ Thus, the acquisition brought an external technology partner inside the Amazon enterprise.

The 2023 Guidelines provide two bases for a challenge to this acquisition: First, according to Guideline 2, the authorities could have viewed Elemental Technologies as both an actual and potential competitor in cloud computing. The difficulty with this basis is that the target’s role in providing services was narrow in contrast to the breadth of services that were offered by major cloud computing firms. Given the massive investments that are required to build data centers and to develop a comprehensive suite of services, it is hard to conceive how Elemental Technologies could shift from a specialized actual competitor into a broad-based competitor in cloud computing.

The second and more viable basis for a potential challenge would be that Amazon would limit access of the acquired technologies to its major rivals: If the record indicated that Elemental Technologies was licensing its software to rival cloud computing providers or working in partnerships with them to serve downstream customers such as those identified above, and if these rivals had few or no alternatives to turn to, then the Agencies could have challenged the acquisition based on Guideline 5.

We hasten to state, however, that the merits of a challenge based on reduced access to technologies would depend in part on the efficiencies from shifting from: (a) an external partnership relationship with the attendant transactions costs; to (b) having Elemental Technologies inside the firm, which would allow for closer coordination with AWS and foster specific investments in a context of ongoing innovation.

4.1.5 Ring.com

Amazon acquired video doorbell maker Ring on April 12, 2018 (Amazon, 2018). It was reported that Ring had 97% share of video doorbell sales the year before the merger; but others, including Amazon itself, were entering or preparing to enter (Adams, 2018). If those shares are accurate and if the Agencies found video doorbells to be a relevant market (and not just part of a broader market for security cameras), the acquisition certainly would have attracted close scrutiny under the Guidelines. If Amazon had even a small presence in video doorbells, its acquisition of Ring likely would have triggered the stringent structural presumptions under Guideline 1 (including the presumption against mergers that result in a firm with a market share above 30% and increase HHI by more than 100).⁶⁹

Alternatively, Amazon’s plans to enter or expand would likely have triggered scrutiny under the actual potential competition theory: Antitrust enforcers would have evaluated whether Amazon would have continued with its independent entry absent the merger and whether the market would have been more competitive with Amazona and Ring as standalone competitors.

Beyond the concerns over actual and potential competition, the merger could have attracted attention for its potential impact on competition in smart home ecosystems. At the time, Amazon was clearly in the process of developing such an ecosystem (Lunden, 2018), and, as indicated, Ring was the leading maker of smart doorbells. Enforcers operating under the 2023 Guidelines would have focused on how the integration of Ring’s products into Amazon’s ecosystem would have influenced both competition for smart doorbells and competition in smart home ecosystems more broadly. The Agencies likely would have asked whether Amazon could hamper rival ecosystem providers by making Ring products exclusive to the Amazon ecosystem.

Guideline 9 warns that a platform’s acquisition of an important platform participant may “entrench the operator’s position by depriving rivals of participants and, in turn, depriving them of network effects.”⁷⁰ If the Agencies found that (i) Amazon would have the ability and incentive to make Ring products exclusive to the Amazon ecosystem, and (ii) Ring products were sufficiently popular that the exclusivity would make it difficult for rival smart home ecosystems to compete, the enforcers would have strongly considered a challenge.

With regard to competition in smart doorbells, the Agencies also would have asked whether Amazon would have “an incentive to give its own products and services an advantage other participants competing on the platform”—in this case, making it more difficult for rival smart doorbells to compete by making it more difficult for users to integrate these rival products with other Amazon smart home devices.⁷¹

In sum, it is clear that the Guidelines would have provided the Agencies with additional tools to scrutinize the competitive effects of this acquisition. Whether blocking the transaction would have necessarily benefited competition is, again, a different and difficult question. With the benefit of retrospection, we know that Amazon reduced the price of the Ring video doorbell upon acquisition (Amazon, 2018); and it is likely that the acquisition allowed consumers to integrate Ring products more easily with other Amazon devices. Such gains, however, might have been difficult to quantify and credit. Moreover, a full analysis would require comparing actual outcomes to a but-for world absent the acquisition.

4.1.6 Other Horizontal Acquisitions

If the Agencies had reached the judgment that Amazon, early in its history, was becoming a dominant platform in eCommerce, then many other acquisitions that built out that platform and increased the size of its user base could have been challenged. For example, a prescient Agency could have even challenged Amazon’s 1999 acquisition of Tool Crib of the North: an online seller of tools to professionals. While the target only had a small user base relative to Amazon’s, the Agencies could have argued under Guideline 6 that this and subsequent acquisitions—e.g., the acquisition of Shelfari in 2013—would allow Amazon to realize powerful network effects, would entrench Amazon, and would make it more difficult for rival platforms to develop.

4.2 Vertical Acquisitions

4.2.1 Kiwa Systems

This 2012 acquisition of Kiva Systems for $775 M likely could have been scrutinized under Guideline 5 that is concerned with acquisitions that would limit “access to products or services that its rivals use to compete.”⁷² At the time of the acquisition, this leader in robotic fulfillment systems earned an estimated $100 M of annual revenues by licensing its technology to Amazon and to eCommerce rivals, including Footlocker, Home Depot, The Gap, Saks, Staples, Quidsi, and Zappos (Contreras, 2023; Tobe, 2016). With the acquisition, Amazon brought Kiva inhouse and reportedly stopped “supporting their existing client base … [to] focus entirely on Amazon.” (Tobe, 2016). Having renamed Kiva Systems as Amazon Robotics in April 2015, Amazon “encouraged prospective users of Kiva technology to let Amazon Robotics and Amazon Services provide fulfillment within Amazon warehouses using Amazon robots” (Tobe, 2016).

In implementing Guideline 5, the Agencies would have analyzed whether Amazon’s rivals had alternative suppliers of robotic fulfillment systems to turn to; and if not, whether other fulfillment systems were an adequate substitute for the robotic systems. The key questions would have been: (i) whether Amazon had the ability and incentive to cut off rivals from using Kiva; and (ii) whether being deprived of Kiva’s services would have significantly hampered rivals’ ability to compete.

If farsighted, the Agencies could have considered a challenge based on: (i) the acquisition’s restriction of access to services that other eCommerce retailers were using; and (ii) the entrenchment of Amazon’s fulfillment system (“Fulfillment By Amazon”).⁷³ With regard to the former, Amazon’s acquisition of Kiva Systems came in the same time frame as its acquisitions of Quidsi and Zappos. Hence, acquiring these other businesses meant that they would continue to be able to rely on Kiva Systems technologies. With regard to the potential entrenchment effect, Amazon has 110 fulfillment centers in the U.S. and Walmart has 40.⁷⁴ In addition, Amazon has become the largest delivery service in the U.S.; it has overtaken Fedex, UPS, and USPS (Mattioli & Fung, 2023; Morris, 2023). Moreover, nearly two-thirds of third-party sellers who operate on Amazon.com use FBA, and, conditional on doing so, these third-parties secure advantages such as two-day Prime shipping. This nexus to Prime features prominently in the FTC’s current monopolization case against Amazon.

Again, our confidence that the 2023 Guidelines would have provided pathways to challenge to the Kiva Systems acquisition does not necessarily translate into support for a challenge. By integrating the not well-capitalized target,⁷⁵ Amazon had a stronger incentive to make greater specific investments in improved services and customized services. It also could realize savings in the transactions costs that were associated with being a licensee. While the post-acquisition conduct identified above would have raised issues under Guidelines 5 and 6, available evidence on deployments of Kiva Systems post-acquisition points to pro-competitive benefits from a dramatic increase in output. This assessment is supported by the target’s co-founders.⁷⁶

Thus, analysis of the Kiva Systems acquisition underscores the importance of taking into account the efficiencies from transferring resources to parties who might be higher-valued owners. The Guidelines do not offer a framework for comparing potential anticompetitive gains for the acquirer with gains from better management of the technological resource. The increase in output is an important indicator of a pro-competitive result—especially given the actual competition between Amazon Robotics and rival robot services providers.⁷⁷

4.2.2 Whole Foods

The Whole Foods acquisition in 2017 broadened Amazon’s ecosystem and allowed it to offer premium groceries on its eCommerce platform with local delivery options. Here the abundance of substitutes (other grocers) with which rival platforms could potentially partner make competitively significant foreclosure unlikely and a “dominance entrenchment” case difficult to establish.

It is possible, however, that a particularly aggressive enforcer would attempt to use the dominance entrenchment prong of Guideline 6 to argue that adding more and more features to Amazon’s platform (video, groceries) in combination with the increasing loyalty of a larger user base makes it all the more difficult for rivals to replicate Amazon’s value proposition and cements Amazon’s leadership position In our view, it would have been an uphill struggle to prevail on this theory for as long as antitrust doctrine maintains the distinction between procompetitive improvement of one’s own products and anticompetitive impairment of rivals’ opportunities.

4.2.3 MGM

Amazon announced in 2021 its planned acquisition of MGM: the owner of iconic video content such as James Bond, The Handmaid’s Tale, and thousands other films and TV shows.⁷⁸ The FTC investigated the deal, but ultimately chose not to bring a challenge. It was reported that FTC Chair Lina Khan did not call for a Commission vote to issue a complaint because she lacked a Democratic majority at the time, and anticipated that the two Republican Commissioner’s would have opposed a challenge (Nylen, 2022).

In our view, neither traditional doctrine or the new Guidelines would justify a challenge to this acquisition. The concern is whether Amazon would have ability and incentive to use its control of the MGM content to deprive Amazon Prime Video’s rivals of a critical input, and thereby weaken the rivals and harm competition in the video streaming market. The main problem with this theory is that MGM does not control a high percentage of video content. In 2021, MGM’s United Artists studio accounted for only 7.2% of the domestic box office revenue.⁷⁹ While box office revenues is not the only relevant measure, there is little doubt that there is a vast amount of video content that is not controlled by MGM. Even if Amazon were to refuse to license MGM content to rival streamers, they would have plenty of content available to them to attract viewers. The availability of substitutes and the fact that MGM content would likely not be critical to rival streamers’ ability to compete weighs against challenging the merger under Guideline 5.⁸⁰

The video streaming market appears to have remained highly competitive in the aftermath of the acquisition: There are seven different providers with at least 7% share, and the largest provider has only a 21% market share (Stole, 2023). And in 2023, Amazon announced the launch of the Amazon MGM Studios Distribution unit, which will be responsible for licensing MGM’s content to third parties (Harding, 2023). Foreclosure concerns appeared unlikely at the time and appear just as unlikely in retrospect.

Beyond a traditional vertical theory of harm, antitrust enforcers could have evaluated the MGM acquisition under the “dominance entrenchment” framework that is specified in Guideline 6. Prime Video is not a standalone streaming service; it is a component of Prime membership service that allows users to receive free and expedited shipping from Amazon’s ecommerce platform. As Amazon founder Jeff Bezos made clear, Prime Video helps attract more people to Prime, which in turn keeps people shopping on Amazon: “We get to monetize [our subscription video] in a very unusual way… When we win a Golden Globe, it helps us sell more shoes. And it does that in a very direct way. Because if you look at Prime members, they buy more on Amazon than non-Prime members” (McAlone, 2016). Antitrust enforcers could have investigated whether by making Prime membership even more attractive, the MGM acquisition could have helped lock customers into the Amazon ecosystem, which would make it more difficult for ecommerce rivals to emerge.

In our view, even under the 2023 Guidelines, this theory would have been a step too far. A challenge would require establishing a very tenuous causation chain: (i) MGM acquisition makes Prime Video significantly more compelling; (ii) more customers subscribe to Prime; (iii) the Prime subscribers use Amazon to the exclusion of other ecommerce channels; and (iv) potential ecommerce rivals would be impaired. While the Agencies generally need not establish the merger’s anticompetitive effects with perfect certainty, courts are reluctant to block mergers based on theoretical suppositions that are not supported by concrete evidence.

Beyond the factual challenges, the “dominance entrenchment” theory would have been on shaky legal footing. It would have been difficult to claim that making the Prime subscription more attractive—without doing anything to block competitors from making their own offerings more attractive⁸¹—is anticompetitive or constitutes “exclusionary” conduct. Rival ecommerce providers could provide their own enticements to attract shoppers—Amazon’s MGM acquisition does not take away their ability to compete. While Guidelines discuss dominance entrenchment through the lens of “create[ing] or enhanc[ing] barriers to entry or expansion by rivals,”⁸² it would take a significant expansion of antitrust doctrine to claim that improving one’s own product—without impairing the opportunities of rivals to improve theirs—constitutes an actionable barrier to entry.

4.2.4 Annapurna Labs

Annapurna Labs—a fabless supplier of fast, low-energy chips for servers—was acquired by Amazon in January 2015 for $350 M; this was 3 years after the Israeli start-up began designing chips. Prior to the acquisition, the Annapurna Labs operated in a nearly stealth mode, and there is no reliable information about its customers or revenues (Whitmam, 2015). Amazon had become disillusioned with the supply of chips from AMD to the point where it was considering developing its own chips for AWS. Presumably, Amazon made efforts to identify potential suppliers, found Annapurna Labs, and then proceeded with the acquisition: This preempted internal investments and positioned Annapurna Labs as its internal supplier.

Amazon viewed the target’s chips as complements to Amazon’s Elastic Compute Cloud (EC2) capabilities for computation, storage, and networking. AWS claimed that the acquisition led directly to innovation in many of its services and that the speed of Annapurna’s chips provided, in effect, an accelerator that increased responsiveness to the variable workflows that are central to Infrastructure as a Service (IaaS).⁸³ The greater efficiency in core functions increased AWS’s competitiveness versus its primary rivals, Microsoft and Alphabet (MSV, 2019). In addition, post-acquisition AWS was positioned to challenge AMD and Intel in the sale of niche chips for cloud computing (MSV, 2019). The acquisition thereby contributed to Amazon’s leading position at a time when, but its share of cloud computing revenues was falling in the U.S.

Would the acquisition have been challenged under the 2023 Guidelines? In this case, we lack information about the target’s customer base prior to acquisition and the complementarities of its products with Amazon’s services. Evidence on these could support a foreclosure claim. But the facts that AMD and other established chip manufacturers represented viable alternative supply sources for AWS rivals would argue against such a claim.

4.3 Serial Acquisitions of Technologies

When Agencies observe a “pattern or strategy of multiple acquisitions, Guideline 8 states that Agencies may “examine the whole series” to assess whether the mergers run afoul of Guidelines 1–6.”⁸⁴ The questions that are posed therein are whether serial acquisitions: (i) significantly increase concentration; (ii) eliminate substantial actual competition; (iii) increase the risk of coordination; (iv) eliminate potential competition; (v) limit access to services; or (vi) entrench a dominant position.

The two sets of Amazon’s acquisitions that we selected for our counterfactual analysis stand out.⁸⁵ As summarized in Table 3, over the period 2011–2021Amazon acquired 29 technology companies that match to its Cloud Computing business; and over the period 2015–2021 Amazon acquired 70 technology companies that match to its Interactive Technologies for Individual and Homes business. (See Tables 4 and 5 in the Appendix.) Many of the acquisitions were partial, and some involved multiple rounds of investments. The targets were typical start-ups: with little or no revenues. An example is Thalmic Labs: a developer of “a connected band that can pick up electrical impulses from a wearer’s arm and translate them into input for a computing platform” and whose more commercially viable products at the time of Amazon’s investment were “in the pipeline” (Etherington, 2016). Others, however, such as SQRLL—a provider of innovative software to improve data security (Novet, 2018)—had a base of customers (Miller, 2018).

The frequencies of acquisitions—an average of 1.35 per quarter for the first series and 2.5 per quarter for the second series—likely would have prompted Agency review. The question in our counterfactual then becomes: Which of the potential anticompetitive concerns enumerated above would be most salient? Given the profiles of the two groups of targets, neither set would have a discernable effect on concentration in the line of business. AWS revenues in 2015 reached nearly $8 B and were growing rapidly (Vailshery, 2023); and in 2022 Amazon was the leader in the $113 B “smart home market” (Tarasov, 2022).

In a similar vein, while an occasional start-up rises to such heights, a theory that the acquisitions in question would eliminate potential competitors would confront the huge gap between a typical start-up and the profile of a firm that could actually challenge Amazon or its major rivals in these two lines of business. As the Guidelines state, a challenge that is based on the elimination of potential competition depends on the target’s “overall capability, which can include the ability to expand or add to its capabilities on its own or in collaboration with someone other than the acquisition target.”⁸⁶ As was explained in Sect. 3 above, competitors in cloud computing provide large numbers of individual services, and competitors in interactive technologies focus on the development of increasing numbers of “skills” that are integrated into smart systems. Targets with demonstrated capabilities in one service or skill lack the range of capabilities to compete for customers. For example, a target that developed promising technologies to improve voice recognition by smart devices would not have the complementary capabilities to compete against Amazon or its rivals.

For these reasons, it is likely that the Agencies would put aside concerns about industry structure and reductions in actual competition (Guidelines 1 and 2), and instead favor theories that the serial acquisitions would: limit access to technologies; snuff out nascent competitors; or entrench Amazon’s dominant positions. With regard to access, post-acquisition Amazon might not have an incentive to license a useful technology to rivals. This theoretical beachhead, however, exhibits vulnerabilities. Amazon’s ongoing competition against rivals with large market shares suggests that actual foreclosure was not consequential. Nevertheless, one can conceive that the Agencies would intervene based on a preference that the target firms would develop into licensors with robust customer bases.

The concern about nascent competitors could also motivate a challenge to Amazon’s serial acquisitions.⁸⁷ One can conceive of Agency interventions premised on the idea that some nascent competitors would develop into substantial business and increase rivalry in Amazon’s lines of business. However, the appeal of an Agency challenge would be countered by, again, the gap between the profiles of Amazon’s targets and the profiles of firms that are effective competitors in industries such as cloud computing. The likelihood of successful entry or expansion by the targets into the provision of a broad range of products and services is vanishingly small.

Challenges to the serial acquisitions based on “entrenchment of a dominant position” would appear to have the greatest appeal in large part because of the vulnerabilities that were identified above: An entrenchment claim would not require a coherent explanation by the Agencies of how the target firms would actually compete against Amazon. Instead, the threshold issue would appear to be whether serial acquisitions entrenched Amazon’s position “through exclusionary conduct, weakening competitive constraints, or otherwise harming the competitive process.”.⁸⁸ The patterns of acquisitions by themselves in the two series that we have identified reflect Amazon’s belief that the acquisitions strengthened Amazon’s position, but do not necessarily indicate whether the acquisitions constituted “growth or development as a consequence of increased competitive capabilities,”⁸⁹ which the Guidelines take no issue with, or anticompetitive entrenchment that would be likely to draw a challenge.

As an addition to the likelihood of a challenge: Under the counterfactual the Agencies might recognize that many of the acquisitions are partial and result in joint ownership of the targets by Amazon and other major firms.⁹⁰ In light of such outcomes, the Agencies might question whether the cross-firm investments in start-ups would increase the risks of coordination among them: for example, to allocate technologies and the services that they support, or to decide jointly when to cease further investment in their jointly-owned start-ups.

The main insight from this analysis is that Agencies operating under the Guidelines would have had many bases for challenging Amazon’s serial acquisitions. We (of course) cannot know when they would intervene, whether they would have been successful, and what relief the Agencies might have obtained. With regard to the latter, the “unscrambling the eggs” problem with prior acquisitions of knowledge-based firms is severe.

Related uncertainties concern how potential acquisitions of individual firms would be evaluated by the Agencies and the courts. Our earlier point that the Guidelines lack safe harbors means that a successful intervention could yield different results. Amazon might have been prevented from acquiring any small tech firms. Alternatively, Amazon might have been prevented from acquiring firms that were licensing technologies to Amazon and others.

We should make clear that our confidence that Agencies that operated under the 2023 Guidelines would have considered challenges to Amazon’s serial acquisitions does not extend to a belief that successful Agency interventions would have enhanced competition. Indeed, given the fundamental principle that entry anticipates all states of the world, if technology start-ups could not exit through an acquisition by major firms such as Amazon, then interventions under Guideline 8 would reduce subsequent entry by technology start-ups.

The potential for adverse effects is indicated by the fact that the United States has the most developed tech ecosystems in the world. Data from PitchBook indicate the following: The top two global ecosystems for start-ups, deal flow, and exits are San Francisco and New York; seven of the top 20 ecosystems are in the United States; and seven of the 20 emerging ecosystems are in the United States.⁹¹

The clear deficiency in the Guidelines—and we might add in current antitrust policy generally—is the lack of an adequate framework for accounting both for the anticompetitive effects of acquisitions and for their procompetitive effects. In the case of acquisitions of tech start-ups that are not real businesses, challenges by the Agencies that are based on the loss of potential competition are weak relative to the gains from sustaining more vibrant tech ecosystems. In classic antitrust terminology, the gains in static competition—even taking into account that the gains may be realized over time—are likely to be dwarfed by gains from dynamic competition: greater innovation in products and services.

From our analysis of the two sets of serial acquisitions, a case can be made that the Guidelines put excessive emphasis on adding new competitors compared to the importance of strengthening competition. The market for corporate control oftentimes improves overall efficiency by transferring capabilities that are embedded in one firm to firms that can integrate them into existing services and products. Even if interventions to prevent such transfers resulted in the development of some new bona fide businesses, the question would remain: Why is that outcome better than having incumbents acquire technologies and quickly bring them to market? This deficiency in the Guidelines will be of ongoing relevance.

5 Effects of the Guidelines

Our counterfactual analysis indicates that Amazon would have encountered substantially more resistance to its acquisition strategies. Stricter merger enforcement—enabled by expansive individual guidelines and the options to pursue challengers under claims related to dominance—necessarily would have constrained Amazon’s development.

The extent of the constraint depends on four factors: (i) how enforcers would have exercised their considerable discretion in challenging Amazon’s many acquisitions; (ii) how Courts would have resolved divergences between the Guidelines and established doctrine; (iii) Amazon’s ability to substitute internal investments for acquisitions; and (iv) Amazon’s ability to contract with firms that remained independent.

An assessment of these factors would require in-depth analysis that we have not conducted. We note, however, that the evidence on the third factor is mixed: As Amazon made acquisitions to expand its eCommerce product lines, it was also making internal investments for similar purposes. This suggests an ability to substitute internal investments for acquisitions—but does not rule out the possibility that acquisitions allowed Amazon to acquire superior technology, capabilities, or know-how compared to the internal options. On the other hand, other acquisitions concern specialized resources such as MGM’s library of content and the leading-edge video compressing technologies that were developed by Elemental Technologies. In these cases, internal investments would not have been viable substitutes for the acquisitions.

However the four factors that we identified above would have played out, it is safe to say that in the counterfactual Amazon would be smaller and less ubiquitous. Of course, that conclusion raises the question of whether such an outcome would be good for consumers and society.

Our counterfactual raises larger questions about the role of guidelines in antitrust policy and enforcement. While reasonable minds may differ as to the wisdom or prosecutorial prudence of the Guidelines, we accept the enforcers’ professed animating purposes: that antitrust enforcement—particularly in the era of high technology—had effectively abandoned its mission by being fixated with false positives, and had thereby turned a blind eye to the risk of false negatives; that it is important to heed the lesson that the Obama-era enforcers missed that markets where products appear “free” can engender market power and competitive concerns because “free” products are simply a link in a chain of monetization; and that seeking to return some fealty to controlling Supreme Court precedent in the merger doctrine should not be controversial as a legal matter: After all, district court and court of appeals opinions cannot overrule Supreme Court precedent; and the High Court has not heard a merger case in 50 years.

Accordingly, the Guidelines evince a paradox: Their vanguard wear the uniform of novel and up-to-date thinking on platform economics and the like, while their rearguard wear very traditional Beefeater outfits: old Supreme Court cases permeate the Guidelines. Our analysis indicates that the current Guidelines would have provided potential pathways for considering challenges to most of Amazon’s acquisitions given the actual competition, nascent and potential competition, entrenchment, foreclosure, and platform competition concerns that are outlined in the Guidelines. And our analysis suggests strongly that enforcer discretion has become paramount. What also is clear is that this discretion lacks well-articulated, antecedent, measurable, and principles-based metes and bounds.

Francis Bacon’s famous statement comes to mind when we reflect on the Guidelines; and if it was not in the minds of the authors, it should well have been: “He that will not apply new remedies must expect new evils, for time is the greatest innovator.”

Appendix A: Serial Acquisitions by Amazon

Here we provide additional details on Amazon’s serial acquisitions of technology companies that were related to two lines of business: Cloud Computing; and Interactive Technologies for Individuals and Homes.

We identified from Snyder and Hughes (2023): (i) 29 acquisitions of typically small technology companies that match to Cloud Computing over the period 2011–2021; and (ii) 70 acquisitions of technology companies that match to Interactive Technologies for Individuals and Homes over the period 2015–2021. In contrast to these high volumes, that Amazon made only 14 technology acquisitions that related to eCommerce from 1998 to 2021. Given the lower number of acquisitions and the longer period of time, we presume that the Agencies would not consider these to constitute a series of acquisitions. Under the new Guidelines, the Agencies would more likely focus on Amazon’s horizontal acquisitions.

Table 4 lists the serial acquisitions of typically small technology companies that match to Amazon’s cloud computing business. Table 4 does not, however, the January 2019 acquisition of Cloud Endure for $250 M. Nevertheless, the remaining acquisitions constitute a substantial series that could have been subject to Agency review under Guideline 8. If one focuses on the time period April 2015 to November 2021—a period of 20 quarters—AWS averaged 1.35 acquisitions per quarter.

Table 5 presents similar information for Amazon’s acquisitions of technology companies that match to Amazon’s third line of business: Interactive Technologies for Individuals and Homes. All 70 acquisitions were acquired by the $200 M that was to allocated Alexa Fund. As is indicated, cost information is not available for these acquisitions; some of them involved multiple funding rounds. This series involves greater frequency: 70 acquisitions over 7 years: an average of 2.5 acquisitions per quarter. If one focuses on the time period April 2015 to November 2021—a period of 20 quarters—AWS averaged 1.35 acquisitions per quarter.

Table 5 does not include three significant acquisitions that were not funded by the Alexa Fund: the $90 M acquisition of Blink in December 2017; the $993 M acquisition of Ring.com in February 2018; and the $97 M acquisition of Eero in February 2019. Nor do we include the $1.7 B proposed acquisition of iRobot, which was announced in August 2022 and was abandoned in January 2024. Despite these exclusions, the remaining acquisitions constitute a substantial series that could have been the subject of Agency review under Guideline 8.

Table 4 Serial acquisitions related to cloud computing: 2015–2021

Table 5 Serial acquisitions related to interactive technologies for individuals and homes: 2015–2021