Policy Research Working Paper 11106 Massive Modular Ecosystems A Framework for Understanding Complex Industries in the Digital Age Eric Thun Daria Taglioni Timothy J. Sturgeon Mark P. Dallas Development Economics Development Research Group April 2025 Policy Research Working Paper 11106 Abstract The rapid evolution of the global economy, driven by study of the mobile phone industry, the paper depicts a digitization and modularization, has outpaced existing Massive Modular Ecosystem characterized by a dynamic theoretical tools. This paper proposes a unified analytical mix of internalization and outsourcing, with a few domi- framework for understanding the complexity of digital nant players in each sub-ecosystem but no dominant actor industries through the lens of Massive Modular Ecosystems. overall. The study explores the broader implications of Mas- To operationalize this concept, the paper develops a system- sive Modular Ecosystems, including their impact on market atic methodology for mapping prevailing and alternative structure, technological innovation, and global-scale eco- governance structures across interconnected, layered, and nomic geography. Increasing geographic specialization and nested ecosystems. Synthesizing insights from the global clustering of sub-ecosystems create strong interdependen- value chain governance, modularity and standards, and cies across countries, making aggressive policy interventions business ecosystem literatures, the paper demonstrates how risky and potentially leading to unintended consequences. modular governance enables distributed innovation, while Efforts to build self-sufficient domestic industries may be non-modular governance structures—including captive impractical and costly, while attempts to decouple from supply relationships, relational partnerships, and hierar- Massive Modular Ecosystems can lead to exclusion from chical control—support greater value capture and strategic critical innovation networks and standard-setting processes. control. The defining element of a Massive Modular Eco- By providing an integrative analytical tool, this paper offers system is not that every relationship is purely modular, but policymakers, business leaders, and scholars a systematic that non-modular relationships operate in a “sea of mod- approach to navigating the complexities of modern digital ularity,” a context where modularity prevails. Using a case industries. This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at dtaglioni@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Massive Modular Ecosystems: A Framework for Understanding Complex Industries in the Digital Age1 Eric Thun,† Daria Taglioni,‡ Timothy J. Sturgeon,§ Mark P. Dallas¶ † University of Oxford Saïd Business School. E-mail: eric.thun@sbs.ox.ac.uk. ‡ World Bank Development Research Group. E-mail: dtaglioni@worldbank.org. § MIT Industrial Performance Center. E-mail: sturgeon@mit.edu. ¶ Union College. E-mail: dallasm@union.edu. JEL codes: L14 Transactional Relationships • Contracts and Reputation • Networks L22 Firm Organization and Market Structure L23 Organization of Production L25 Firm Performance: Size, Diversification, and Scope O14 Industrialization • Manufacturing and Service Industries • Choice of Technology O32 Management of Technological Innovation and R&D O33 Technological Change: Choices and Consequences • Diffusion Processes Keywords: Industrial Organization, Innovation, Modularity, Platforms, Standards, Global Value Chains, Business Ecosystems, Socioeconomic Megastructures 1 Acknowledgments: This work has been supported by a World Bank trust fund with the Republic of Korea, acting through the Korean Development Institute School of Public Policy and Management (KDIS) on the KDI School Partnership for Knowledge Creation and Sharing (TF0B0356) as well as the World Bank’s Multi-Donor Trade Umbrella Trust Fund “Umbrella Facility for Trade”. We would like to thank Jing-ming Shiu at National Cheng Kung University for his insights and for sharing his original data on mobile phones, telecommunication standards and Android Open-Source Project code commits. In addition, we wish to thank Irene Iodice, Ritika Khandelwal, Jose Marzluf and Isak Falk for their excellent research assistance. For very insightful comments on a draft manuscript, we thank Ari Van Assche, Carliss Baldwin, Judith Biewener, Florian Butollo, John Humphrey, Devika Narayan and Lindsay Whitfield. Furthermore, for their useful feedback, we thank participants in the 2022 Comparative Political Economy of Global Business Research Conference at the Saïd Business School, University of Oxford, and the 2022 and 2024 Society of Socio-economic Annual Conference Network O Sessions. MASSIVE MODULAR ECOSYSTEMS TABLE OF CONTENTS 1. INTRODUCTION ................................................................................................................................................................................................ 1 2. LITERATURE REVIEW ........................................................................................................................................................................................ 2 2.1 MODULARITY AND STANDARDS ................................................................................................................................................................................ 2 2.2 GLOBAL VALUE CHAINS .......................................................................................................................................................................................... 3 2.3 BUSINESS ECOSYSTEMS AND PLATFORMS................................................................................................................................................................. 4 2.4 SHORTCOMINGS OF EXISTING LITERATURES .............................................................................................................................................................. 5 3. TOWARD AN INTEGRATED GOVERNANCE FRAMEWORK ............................................................................................................................. 6 3.1. DEFINING MASSIVE MODULAR ECOSYSTEMS.............................................................................................................................................................. 6 3.2. MAPPING TECHNICAL SUBSYSTEMS TO ORGANIZATIONAL SUB-ECOSYSTEMS ................................................................................................................. 7 3.3. IDENTIFYING ROLES IN SUB-ECOSYSTEMS ................................................................................................................................................................ 8 3.4. IDENTIFYING GOVERNANCE STRUCTURES IN COMPLEX INDUSTRIES: AN INTEGRATIVE FRAMEWORK .................................................................................. 9 4. THE MOBILE PHONE INDUSTRY AS A MASSIVE MODULAR ECOSYSTEM .................................................................................................... 11 4.1 T HE DESIGN AND PRODUCTION OF MOBILE PHONE SYSTEMS ................................................................................................................................................................................................12 4.2 THE DESIGN AND PRODUCTION OF KEY MOBILE PHONE SUBSYSTEMS ...........................................................................................................................14 4.3 THE DISTRIBUTION AND CONSUMPTION OF MOBILE PHONES........................................................................................................................................17 5. TECHNOLOGY AND INNOVATION, MARKET STRUCTURE, AND GEOGRAPHIC OUTCOMES IN THE MOBILE PHONE INDUSTRY ....... 18 5.1 TECHNOLOGY AND INNOVATION............................................................................................................................................................................18 5.2 MARKET STRUCTURE .............................................................................................................................................................................................20 5.3 Economic GEOGRAPHY .......................................................................................................................................................................................25 6. DISCUSSION ................................................................................................................................................................................................... 27 6.1. IMPLICATIONS FOR THEORY..................................................................................................................................................................................27 6.2. IMPLICATIONS FOR STRATEGIC AND INNOVATION MANAGEMENT ...............................................................................................................................29 6.3. IMPLICATIONS FOR POLICY ...................................................................................................................................................................................30 7. CONCLUSION ................................................................................................................................................................................................. 31 REFERENCES ....................................................................................................................................................................................................... 32 APPENDIX A. GLOSSARY OF KEY TERMS AS USED IN THE PAPER ...................................................................................................... 38 APPENDIX B. FIGURE: A DECISION TREE FOR IDENTIFYING GOVERNANCE STRUCTURES .......................................................................... 41 APPENDIX C. DATA SOURCES ............................................................................................................................................................................ 41 APPENDIX D. HOW MODULARITY EMERGED IN THE MOBILE PHONE INDUSTRY.......................................................................................... 42 APPENDIX E. HOW THE RADIO FREQUENCY SUBSYSTEM WORKS ................................................................................................................ 44 TABLES TABLE 1. GOVERNANCE VARIABLES WITH RESEARCH QUESTIONS .............................................................................................................................. 9 TABLE 2. MOBILE PHONE DESIGN, SUB-SYSTEM SOURCING, AND ASSEMBLY: GOVERNANCE STRUCTURES, ACTORS AND ROLES, KEY INTEROPERABILITY STANDARDS AND MODULAR INPUTS ....................................................................................................................................................... 13 TABLE 3. MOBILE PHONE SUBSYSTEM DESIGN AND PRODUCTION: GOVERNANCE STRUCTURES, ACTORS AND ROLES, KEY INTEROPERABILITY STANDARDS AND MODULAR INPUTS ......................................................................................................................................................................... 15 TABLE 4. MOBILE PHONE DISTRIBUTION AND CONSUMPTION: MOBILE PHONE RETAIL SALES AND MOBILE APP STORE FUNCTIONS, GOVERNANCE STRUCTURES, ACTORS AND ROLES, AND KEY STANDARDS ........................................................................................................................... 17 TABLE 5. MOBILE PHONE PERFORMANCE AND FUNCTIONAL IMPROVEMENTS BY SUBSYSTEM (2009 AND 2020) .......................................................... 19 TABLE 6: MOBILE PHONE SUB-ECOSYSTEM MARKET AND GEOGRAPHIC OUTCOMES CORRELATED WITH RISING MODULARITY IN MOBILE PHONE TECHNICAL ARCHITECTURE..................................................................................................................................................................... 21 Table 7. Analysis of IHS mobile phone teardown database ............................................................................................................................ 26 FIGURES FIGURE 1. FIVE GOVERNANCE STRUCTURES POSITED BY GLOBAL VALUE CHAIN THEORY WITH KEY VARIABLES AND ROLES .............................. 4 FIGURE 2. CONCEPTUAL FRAMEWORK: SIX GOVERNANCE STRUCTURES IN MASSIVE MODULE ECOSYSTEMS ................................................. 9 FIGURE 3. LAYERED MODULAR ECOSYSTEMS IN THE MOBILE PHONE MME AND LINKS TO ADJACENT INDUSTRIES ........................................ 12 FIGURE 4. MARKET CONCENTRATION LEVELS CORRELATED WITH GOVERNANCE STRUCTURES IN THE MOBILE PHONE MME . .................................................................................................................................................................. 22 ii 1. Introduction The global economy is undergoing profound transformations driven by digitization, modularization, and the rise of platform-based business models. In this new landscape, industries that were once organized along linear, easily understood value chains and discernible industry boundaries are now interconnected in ways that span multiple sectors and geographies, and simultaneously incorporate a variety of governance structures. Growing complexity creates both opportunities and tensions, raising critical challenges for businesses and policymakers alike. Drawing on key concepts from the global value chains (GVCs), modularity, and business ecosystems literatures, this paper develops a framework for understanding complex industries in terms of Massive Modular Ecosystems (MMEs). The paper argues that existing concepts, while insightful, struggle to capture the layered, decentralized nature of complex digitally-mediated industries, where multiple forms of governance operate simultaneously in a system dominated by modularity, enabling distributed innovation and rapid scaling across multiple interlinked sub-ecosystems. Our aim is to overcome these limitations by combining key insights from several literatures. Specifically, GVC governance theory accounts for multiple governance forms and excels at mapping vertical linkages but underemphasizes indirect ties between actors. It therefore offers a limited view of how modular governance structures evolve and is ill- equipped to explain the dynamics of modular platforms. The modularity and standards literature usefully highlights the role of standardized technical interfaces in product design and inter-firm coordination but has little to say about how modularity intersects with various forms of innovation or co-evolves with non-modular forms of governance. Meanwhile, the business ecosystems literature emphasizes the key concepts of complementarity and co-specialization, especially in digitally-enabled platforms, but often disregards geographic and institutional frictions that can profoundly affect the scope of firm agency, and fails to adequately engage the reality of linked, layered, and nested ecosystems. The concept of MMEs addresses these gaps by linking technical modularity (and its absence) to organizational roles, and by highlighting the contributions of both modular and non- modular governance structures in organizationally and spatially fragmented industries. Our goal is to provide a comprehensive yet parsimonious framework for analyzing complex digital industries as they face supply chain disruptions, the rise of increasingly powerful digital platforms, and more aggressive state interventions. We illustrate our framework through a detailed case study of the mobile phone industry. Mobile phones provide an ideal case for examining the complexities of linked modular sub-ecosystems because the industry is characterized at the technical level by multiple highly complex, rapidly evolving subsystems, linked by standardized interfaces. Distributed innovation at the subsystem level enables recombinant innovation at the systems level. Our analysis of the industry proceeds in three steps: first, it identifies key mobile phone subsystems according to their function and maps each to a corresponding sub-ecosystem; second, it assigns roles to the most important firms within each sub-ecosystem; and third, it identifies the prevailing and alternate governance structures between and within in each sub-ecosystem. We perform this exercise across three technical levels of the mobile phone industry: final system, key subsystems, and post- production distribution and consumption. (A glossary of terms used in the paper is provided in Appendix A.) The paper is structured as follows: section 2 discusses the literature; section 3 develops our conceptual framework; section 4 applies the framework to the case of the mobile phone industry; section 5 discusses industry outcomes in terms of technological innovation, market structure, and geography; and section 6 explores the implications for policymakers, business leaders, and researchers confronting the 1 MASSIVE MODULAR ECOSYSTEMS challenges of engaging complex, digital, and globally integrated industries. Section 7 provides a concluding statement. 2. Literature review This section selectively reviews the literature on modularity and standards, GVCs, and business ecosystems and platforms to contextualize the conceptual framework developed in section 3. 2.1 Modularity and standards A complex system, according to Simon (1962), almost always has a decomposable, hierarchical structure. Complex systems will have successively fewer complex subsystems nested within them, until, at the lowest layer, we can observe simple, single-function inputs (parts and components) or discrete tasks. Unless a firm coordinates activities in all layers internally, open interface standards will typically be used to link up parts, components, and subsystems into a final system. When standard interfaces are the prevailing form of coordination, we can say that the complex system has a modular character. Modularity in Design. Interface standards often play a role in modularizing both the design of complex systems and the organization of complex industries. The literature on modular design focuses on how modular systems can be intentionally created as a means of conserving system design effort accelerating recombinant innovation (Schumpeter, 1934, 1939; Ulrich, 1995). Sanchez and Mahoney (1996; 65) point to design modularity as “a special form of design which intentionally creates a high degree of independence or ‘loose coupling’ between component designs by standardizing component interface specifications.” When sourcing modules externally, design engineers can incorporate advanced features and functions without needing in-depth knowledge of every part of the system. Designers of complex systems need only to understand what each module can do, select the right modules, and use them to construct downstream systems that perform to requirements. Similarly, the designers of complementary subsystems must be experts in their specific area and also know the relevant interface standards, but need not know the inner workings of other modules or the downstream system in which their module will be used. In this way, modularity spurs recombinant innovation at the system level because modules can be used, re-used, re-combined, and quickly adapted to alter the functionality of the overall system. At the same time, module designers can work (and innovate) on improving subsystems with a large degree of independence from system designers (Meyer and Lehnerd, 1997; Thomke and Reinertsen, 1998). Modularity in Value Chains. The “mirroring hypothesis” (Colfer and Baldwin, 2016) suggests that the design of technical systems has implications for how industries are organized. For example, when complex systems are produced in an industry where modularity prevails, participating firms can focus on a narrower set of “core competencies,” leading to contemporaneous innovation, greater efficiency, and improved economies of scale (Prahalad and Hamel, 1999; Baldwin and Clark, 2000). This logic can repeat at different technical levels, adding depth to the hierarchical structure of an industry. In short, the mirroring of technical and organizational systems can help to explain how modularity shapes industry dynamics by encouraging specialization and coordination through standardized interfaces. Interface Standards as Modular Interfaces. While the technical standards governing modularity can be closed – developed by a firm entirely for its own use – standards that are open for public use simplify the process of outsourcing (and offshoring) by offering “a pre-established way to resolve potential conflicts between interacting parts of a design” (Baldwin and Clark, 2000; 73). At “thick crossing points” within modules, interactions between agents will tend to be complex, uncertain, and iterative, 2 MASSIVE MODULAR ECOSYSTEMS often involving the creation and exchange of tacit knowledge via relational ties. By contrast, “thin crossing points” governed by standardized interfaces simplify information exchange, lower transaction costs, and facilitate externalization (Baldwin and Clark, 2000; Sturgeon, 2002; Gereffi et al., 2005; Baldwin, 2007). At the same time, highly modular industries risk a ‘modularity trap,’ where reliance on common subsystems can reduce differentiation, intensify competition, and create risks of commodification (Chesbrough and Kusunoki, 2001). Proprietary and Non-Proprietary Interface Standards. Open interface standards have two basic types – proprietary or non-proprietary – and can emerge in a variety of ways. Proprietary interface standards create revenue for their sponsors, and can become established in an industry by dominant firms that have prevailed in a standards war (Cusumano et al., 1992; Shapiro and Varian, 1999), by aggressive intellectual property litigation (i.e., standards as IP), and by firms that have launched successful platforms (Gawer and Cusumano, 2014). Open, non-proprietary interface standards, even though they take thousands of expensive engineering hours to create and maintain, are non-rivalrous and non-remunerative. They may be sponsored by governments (Thun and Sturgeon, 2019), formal standard setting organizations (Chiao et al., 2008), open-source communities (West, 2003), or alliances and consortia (Wiegmann et al., 2017). Closed, firms-specific standards sometimes evolve into open industry standards over time, and eventually shift from proprietary to non-proprietary (Sturgeon, 2019). This dynamic can also flow in reverse, with elaborated versions of non-proprietary standards created in open-source projects ring-fenced by firms and rendered proprietary (Dallas et al, 2019). 2.2 Global value chains The prevailing nature of technical interface standards also influences their geographic scope. Specifically, loose and tight coupling of industry actors have been associated with geographic dispersion and clustering, respectively. The global value chain (GVC) governance literature maps the various modes of coordination (governance structures) across global industries, with a particular focus on how less-developed economies experience different patterns of growth and technological learning depending on the roles they play in various GVC governance structures (Humphrey and Schmitz, 2002; Gereffi and Sturgeon, 2013). To specify this, Gereffi et al. (2005) developed a conceptual model to characterize different governance forms, ranging from more tightly coordinated and asymmetrical forms to more loosely coordinated forms. The ‘GVC governance framework’ hinges on three independent variables: transaction complexity, codifiability, and supplier competency. Hierarchical governance is anticipated with highly complex, co-specialized transactions, while market-based relations are simpler and price-driven. In between the traditional market-hierarchy dichotomy are modular, relational and captive governance. Modular governance is expected when complex yet codifiable interactions are mediated by open interface standards, and when supplier capabilities are high enough to meet requirements and fully manage activities on their side of the transaction (on a ‘turnkey’ basis). Relational governance is expected when information requirements are complex and uncodified, requiring partnerships based on trust and the expectation of repeat interactions (Granovetter, 1973, 1985; Powell, 1990). Captive governance is expected when powerful lead firms exert detailed control over less capable suppliers (Humphrey and Schmitz, 2002). The predicted (binary) values for the three independent variables associated with each of the five governance forms are shown in Figure 1, with the forms arranged left to right according to the need for explicit coordination. To help convey the structure of governance, stylized diagrams are provided above each column. The blue block arrows depict the direction and extent of explicit coordination within each diagram. These governance forms are further discussed in Section 3, where we develop our expanded framework. 3 MASSIVE MODULAR ECOSYSTEMS Figure 1. Five governance structures posited by global value chain theory with key variables and roles Source: Gereffi et al, 2005, with diagrams from Inomata (2017). 2.3 Business ecosystems and platforms The concept of business ecosystems expands the linear view of buyer-supplier linkages offered by the GVC governance framework to include a wider set of actors and relationships (Normann and Ramirez, 1993; Rosenbloom and Christensen, 1994). Although there is disagreement over boundary conditions (what is included and excluded), the most general definition of a business ecosystem is “an interdependent network of self-interested actors jointly creating value” (Bogers et al., 2019, p. 2), with the common qualifications that ecosystem actors must be “autonomous” and provide “complementary” inputs (Baldwin et al., 2024). As Jacobides et al. (2018) note, these actors maintain autonomy but coordinate, sometimes indirectly, to achieve shared goals. Adner (2017) emphasizes that a business ecosystem is inherently multilateral, as opposed to a series of bilateral interactions. With many actors influencing outcomes, ecosystems are generally perceived to be highly dynamic. Stability in ecosystems, according to Baldwin (2020), comes at a fragile balance point between interdependence, which binds the members of an ecosystem together, and autonomy, which allows them to act independently. If interdependence is too powerful, the system collapses into a hierarchy; if independence is too dominant, there is no incentive to interact at all, leading to market governance (Baldwin, 2020). Complementarity and Platforms in Business Ecosystems. The concept of complementarity is central to business ecosystems, where joint efforts produce more value than would be the case with independent work (Jacobides et al., 2018; Baldwin et al., 2024). This is what Baldwin et al. (2024; p. 1) refer to as a “complementarity surplus.” Two key dimensions of complementarity in business ecosystems are strength and symmetry. When complementarity is strong between functions in a technical system or the actors who provide them, system elements will be useless and actors unproductive in the absence of their co-specialized inputs and partners. Strong complementarity – associated with high asset specificity in transaction cost theory or co-specialized assets in the dynamic capabilities literature – will incentivize internalization, or hierarchical governance. With weak complementarity governance can shift toward the market end of the governance spectrum because there will be many substitutes for each system element and actors who can provide them. Symmetrical complementarity arises in situations when all system elements/actors are similarly strong or weak complements. With asymmetrical complementarity, some system elements/actors are essential (strong complementarity) while others are optional (weak complementarity). Asymmetrical 4 MASSIVE MODULAR ECOSYSTEMS complementarity forms the basis of a particular form of ecosystem: the platform. A platform ecosystem is defined by Baldwin and Clark (2000) as “a technical system comprising a core set of essential functional elements (the platform) plus a set of optional complements”. In terms of actors, platforms have a ‘sponsor’ that specifies the requirements for accessing the platform (the interface standards), ‘complementors’ that provide goods and services that can be accessed via the platform, and the platform’s ‘users,’ who access goods and services over the platform and, in some cases, enrich the platform with user-supplied content, usage data, personal information, and/or peer-to-peer networking, all of which increase the value of the platform for other users and for complementors. Because complements are optional, and not co-designed with the platform, and interoperability is ensured by means of a standard interface, a platform ecosystem always depends on modular governance. To reinforce this point, this form of governance is referred to as a ‘modular platform’ in the framework we develop below. Unlike linkages in a ‘modular value chain,’ where suppliers provide a specific complementary function sequentially and at a preordained time, complements to platforms can provide almost any function at any time, opening vast space for distributed innovation and rapid scaling. Even so, unlike arm’s length market governance, the linkages between complements and platforms still have significant interdependencies because they must adhere to (open-yet-proprietary) interface standard set by the platform’s sponsor (Adner and Kapoor, 2010; Jacobides et al., 2018). A remarkable feature of modular platforms, and a reason they are a major focus of the business ecosystem literature, is their ability to leverage network effects to scale up rapidly. This is explained by the concept of supermodularity, the idea that more of one good will make another good more valuable in a positive feedback loop which results in a growing pool of complementors (Jacobides et al., 2018; Baldwin, 2020). As Jacobides et al. (2018) note, supermodular network effects can occur both in production and consumption, but there is greater potential for strong network effects and rapid scaling on the consumption side because the vast number of end use consumers relative to industrial users (Van Alstyne et al., 2016; Cusumano et al., 2019). In addition, innovation is accelerated by expanding access to novel ideas and resources supplied by autonomous complementors. 2.4 Shortcomings of existing literatures Despite the valuable insights offered by the GVC governance and business ecosystem literatures – and the modularity and standards literature, upon which they both partially rely – they each fall short of capturing the full complexity of global industries in the digital age. The GVC governance framework effectively highlights the variety of governance structures linking activities across organizational and geographic space. However, it has several shortcomings. First, the framework lacks the tools needed to differentiate between a modular linkage, where standard-based ties between buyers and suppliers are symmetrically weak, and a modular platform, where asymmetric ties to a platform are strong for complementors, but ties to complementors are weak for platform sponsors. Second, GVC governance theory has focused mainly on vertical relationships, for example by tracing the flow of activity needed to bring a product from conception to production to end use, while failing to adequately conceptualize indirect ties between actors or the horizontal and diagonal connections to adjacent products, industries, and standards. The business ecosystem literature highlights the asymmetric patterns of complementarity that underpin platforms, but also has shortcomings. First, it underemphasizes the varied forms of governance that characterize complex industries, especially how firms and other organizations rely on non-modular forms of governance to work at the global technology frontier, or how these non-modular forms can carry on as pockets of more co-specialized activity along a modular backbone. Second, the business ecosystem literature neglects the importance of geography. The dominant conception of modular 5 MASSIVE MODULAR ECOSYSTEMS ecosystems in this literature is that they consist mainly of information flows, and with this comes the implicit assumption that business decisions are unimpeded by external factors, such as international trade frictions, logistics disruptions, or other points of geoeconomic tension. While this may have been a reasonable analytical choice in an era when barriers to flows of trade, investment, and data were on the decline, the trend in the current geoeconomic climate is shifting in the opposite direction. Finally, while both the GVC and business ecosystem literatures have strengths, their units of analysis are too narrow. For example, industry research using a GVC approach typically begins by tracing economic geography of the individual actors – lead firms and suppliers – responsible for producing a particular product or commodity (Fernandez-Stark and Gereffi, 2019). The business ecosystem literature takes a broader view by explicitly including non-firm actors and indirect, non-contractual relationships. Still, there are limitations. First, the nested character of business ecosystems might be acknowledged, but the analysis and empirical illustrations tend to be limited to one ecosystem at a time (e.g. Teece, 2018; Jacobides et al., 2018). Kenney and Zysman (2016) note the vertically layering of digital ecosystems, each building on the platforms below, and quote Stuart Feldman’s intriguing comment that the structure of a platform ecosystem is “platforms all the way down.” Nevertheless, the literature neglects the essential role played by sub-ecosystem layers governed by means other than hierarchies or modular platforms. What is lacking are systematic tools for understanding how ecosystems with various governance structures are layered and interlinked. Second, although ecosystems and platforms have sometimes been presented as an alternative to value chains (e.g., Van Alstyne et al., 2016; Adner, 2017), the two views are not mutually exclusive, since linked, layered, and nested ecosystems have a strong vertical dimension. However, vertical hierarchies may operate from a technical perspective, since there is no clear organizational hierarchy present when ecosystems are layered, nested, and extend into and across adjacent industries. In the next section, we offer an approach for identifying prevailing and alternate governance structures in complex global industries, including linkages across layers of sub-ecosystems and to adjacent industries. 3. Toward an integrated governance framework In this section, we advance beyond insights from the three literatures discussed in section 2 to propose a unified analytical framework that seeks to capture the complexity of digital industries in a systematic, comprehensive, yet parsimonious manner. Our main arguments are that (1) complex, digitally-mediated industries are comprised of layered and nested sub-ecosystems, and (2) non-modular governance structures not only persist in industries where modular governance prevails, but are essential for maintaining their innovative dynamism. The framework provides a method for identifying the key functional sub-ecosystems, actor roles, and prevailing and alternative governance forms in an industry. For clarity, the italicized terms in the remainder of this section are included in the glossary in Appendix A. 3.1. Defining massive modular ecosystems We define a massive modular ecosystem (MME) as a collection of autonomous yet interdependent and complementary sub-ecosystems coordinated primarily through open interface standards to produce a group of related products or services that fulfill a similar purpose. When sufficiently widespread, open interface standards allow sub-ecosystems to be joined up, creating ‘massive’ modularity. Sub- ecosystems in an MME are complementary, meaning their combined value-added exceeds the sum of 6 MASSIVE MODULAR ECOSYSTEMS their individual contributions.2 An MME is a higher-order structure than a single business ecosystem, as it integrates multiple linked sub-ecosystems to create a productive macro organizational structure. Just as Baldwin (2020: 6) defines an ecosystem as a “meta-organization” that provides a level of structure above that of Herbert Simon’s firms and transactions, an MME provides a view of industrial organization at a level higher than an individual business ecosystem. We refer to these higher-order structures in the global economy as socio-economic megastructures, of which an MME is an example. Crucially, our definition of a business ecosystem goes beyond the traditional focus on linkages and relationships among specific firms or technical functions: it encompasses all products and services that fulfill a similar purpose. Thus, the boundaries of an MME and its sub-ecosystems will align with the colloquial concept of industry: a group of related products and services that share common actors, technologies, standards, and end uses.3 Common technical standards, technologies, and suppliers create specific routines, capabilities, interdependencies and relationships, even if some are indirect, including among firms that may be competitors (e.g., two mobile phone firms). For example, aside from direct business relationships, participating firms may influence and be influenced by the interface standards that structure their interactions with shared sub-ecosystems. At the same time, firms in other sub- ecosystems may operate under different design rules and therefore work with a high degree of independence as they contribute inputs to downstream sub-ecosystems. We include generic inputs accessed through market governance in our definition of business ecosystems, diverging from the definitions offered by Jacobides et al. (2018) and Baldwin et al. (2024). Indeed, these inputs may normally be generic from a technical perspective, with low switching costs due to limited co-specialization, interdependencies, or requirements for explicit coordination. However, operational or geographic constraints, such as changes in industrial and trade policies and force majeure supply-chain disruptions, can all render them unique and create critical dependencies, sometimes in short order. We also include vertically integrated firms (i.e. hierarchical governance). A single firm in isolation is not an ecosystem, given that it is only a single actor, but as part of an MME it will have interdependencies with other actors in the MME, again, due to shared interface standards, technologies, and suppliers. In this way the concept of MMEs is able to include organizational forms across the governance spectrum, from hierarchy to markets, within a unified conceptual framework. The rest of this section illustrates three steps in our analytical framework: first, identifying key technical subsystems and mapping them to their organizational sub-ecosystems (subsection 3.2); second, assigning roles to participating firms (subsection 3.3); and third, mapping prevailing and alternate governance structures between and within each sub-ecosystem (subsection 3.4). 3.2. Mapping technical subsystems to organizational sub-ecosystems As a first step in our methodology, we set out to identify the functions in a generic technical system produced in a specific industry (e.g. car, airplane, AI chatbot, complete software system, or mobile 2 Our definition of an MME and its sub-ecosystems follow closely on the Bogers et al.’s (2019) definition of an ecosystem referred to in Section 2.3. 3 Our definition of ‘industry’ includes all inputs and activities needed to bring products with a specific end-use to market. This is unlike the concept in most economic statistics, which focuses more narrowly on final products and their specific inputs (e.g., automobiles and automobile engines, but not steel specifically used to manufacture automobiles). This broad boundary definition helps to accommodate the fact of an MME’s inevitable overlap with adjacent industries and sub-ecosystems given that many if not most sub-ecosystems contribute to multiple industries. 7 MASSIVE MODULAR ECOSYSTEMS phone). As Holgersson et al. (2022) explain, value within a complex system is created by specific functions. These technical functions, which typically lack a standalone practical use (such as a jet engine, for example), can be conceptualized as subsystems that can be linked to others to form a complete technical system with practical uses and applications (such as an airplane). Mapping technical functions to the actors producing them reveals that subsystems are typically produced, not by single a firm, but by groups of actors operating in organizational sub-ecosystems. Each sub-ecosystem can usually be defined as an industry in its own right. It will have its own prevailing and alternative governance structures, and include corresponding interface standards and standard setting routines. We define governance structures as the rules and norms coordinating actors’ efforts and resources toward common goals.4 Our methodology approaches complex industries as an amalgam of linked sub- ecosystems, enabling analysis of the linkages between and within ecosystems rather than only between individual firms or functions. When modular governance prevails at the final product design and component sourcing stages, then the benefits of loose coupling, such as risk reduction, and supply chain resiliency, and recombining the fruits of distributed innovation can be realized across an entire industry. The distinction between technical systems and sub-ecosystems is crucial to our analytical method. The former comprises the material subsystems defined by function and the interoperability standards that link them. Understanding them requires at least a layperson’s understanding of the technical details of products, processes, and the interdependencies and linkages between final systems and their intermediate inputs (parts, components, modules and subsystems). The latter are populated by actors playing one or more roles. Identifying them requires an examination of the roles that the most important firm and non-firm actors play in producing specific systems and subsystems. While complementarity exists at both levels, the technical and organizational features of an industry do not mirror each other perfectly (Baldwin, 2023). For example, it is much easier to replicate (and therefore substitute) a complex subsystem (e.g., by reverse engineering), than it is to replace entrenched actors due to path dependence, network effects, intellectual property ownership, or geographic advantages. In addition, actors have agency to select roles and governance structures aligned with their strategic aims with at least a partial degree of independence from technical requirements. Thus, autonomous actors in business ecosystems play different roles, and do not align their activities uniformly. In sum, analysis of both technical and organizational systems is essential for mapping governance structures in complex industries. 3.3. Identifying roles in sub-ecosystems The second step in our methodology is to identify roles played by actors within sub-ecosystems. We begin with the distinction between parent/headquarters and subsidiary/group/team functions within an integrated, hierarchically-managed firm. Next, buyers and suppliers can be readily identified in external relationships, as well as relational partners in collaborative efforts typically formed with the specific aim of developing new or synergistic technologies, products, processes, routines, or business models. In platform ecosystems sponsors, users, and complementors can be identified. We describe these roles and link them to specific governance structures in more detail below. Determining “who does what” within a business ecosystem is inherently provisional due to evolving technological, managerial, and regulatory conditions. In addition, firms and other actors can choose and shift roles, and typically play multiple roles at the same time. However, general patterns can be discerned and “snapshots” taken when examining an industry. Identifying roles is crucial because in many cases “interdependencies tend to be standardized within each role” (Jacobides et al., 2018). In an app store, for example, all apps have the role of complements to their corresponding platform – and it 4 In the ecosystem literature, governance structures are sometimes referred to as “alignment structures” (Adner, 2017). 8 MASSIVE MODULAR ECOSYSTEMS is this role that defines their interdependencies within the ecosystem rather than their individual traits. 3.4. Identifying governance structures in complex industries: An integrative framework The third step in our methodology is to identify the presence of prevailing and alternative governance structures in each of the major functional sub-ecosystems that comprise a specific industry. The framework we present below differentiates governance structures using five independent variables, three of which are derived from the GVC literature (complexity, codifiability, actor capabilities) and two from the ecosystem literature (complementarity strength and symmetry). Although this paper does not provide specific proxies for these variables, Table 1 highlights key questions for evaluating them within any technical system. Table 1. Governance variables with research questions Governance Research question variable Complexity How complex is the information and knowledge required to sustain a particular function, particularly with respect to product and process specifications? Codifiability To what extent can a function’s requirements be codified and transmitted efficiently without transaction-specific investment? Actor capabilities Are the capabilities of actual and potential actors in an ecosystem adequate in relation to the requirements of transactions? Complementarity Strong complementarity: Are inputs to a final system “unproductive unless used together” (Baldwin, Strength 2020: 13)? Are they specific and unique inputs, inseparable from the system they are part of? Weak complementarity: Are substitutes available? Can one component be withdrawn without destroying the value of the others? Complementarity Symmetric complementarity: Are all complementary functions equally strong or equally weak? Symmetry Asymmetric complementarity: Are both strong and weak complements present in a technical system? Do non-unique components depend on the unique ones for their value, but not the reverse? Source: Authors; strength and symmetry questions are inspired by Baldwin, 2020, Section 3.5 Obtaining answers to the questions in Table 1 makes it possible to map governance structures in complex industries according to the integrated framework shown in Figure 2. Figure 2. Conceptual framework: six governance structures in massive module ecosystems Source: Authors’ elaboration Building from Figure 1, the framework identifies six governance structures, each characterized by 9 MASSIVE MODULAR ECOSYSTEMS specific combinations of governance variables: complexity, codifiability, and actor capabilities (from the GVC literature) and complementarity strength and symmetry (from the business ecosystem literature).5 While these variables are continuous in practice, they are presented in Figure 2 mainly in binary terms (e.g., high/low, strong/weak) to simplify interpretation. Arrows below the table depict the continuous nature of explicit coordination and the number of actors typically present in each governance arrangement. A decision tree for identifying governance structures is provided in Appendix B. This unified framework adds the platform governance structure to the five governance structures derived from GVC governance theory (i.e. corporate hierarchies, captive buyers and suppliers, relational partners, modular value chains, and arms-length market linkages) as follows: • Hierarchical governance occurs within firms, where managers oversee complex interactions via bureaucratic control. This enables firms to manage high complexity while maintaining strong complementarity and symmetrical dependencies among co-specialized inputs. Even when complexity is high, hierarchical governance remains viable because firms can develop proprietary interface standards to codify the exchange of complex information for internal use or simply tolerate reliance on non-codified information exchange in one-off or highly profitable projects. Through bureaucratic control, co-specialized inputs within the firm can reach maximum complementarity, becoming symmetrically strong in support of in-house system design and integration. • Captive governance emerges when powerful buyers exert tight control over less capable suppliers through detailed instructions, required equipment, approved vendor and component lists, and proscribed financing or procurement arrangements. Buyers manage complexity by strictly controlling key supplier activities, including industrial procedures, machinery, and software. A supplier’s relationship becomes captive when the buyer represents a significant share of its revenue, making the supplier’s survival contingent on maintaining the relationship with its main buyer. Complementarity is strong but asymmetric in that suppliers rely heavily on buyers for survival, while buyers face low switching costs due to relatively low supplier capabilities. • Relational governance involves partnerships between actors with high, symmetric capabilities, fostering joint innovation through mutual expertise and complementary co-specialization. When codification is not (yet) possible, relational partners manage high complexity by jointly creating idiosyncratic routines and, possibly, emergent standards. Co-specialization leads to strong complementarity, where actors become interdependent, motivating them to protect shared intellectual property as they collaborate on new products, technologies, and processes. Risk stemming from incomplete contracts and deep interdependence is mitigated the shared expectation of continued collaboration and high returns from innovation (Granovetter, 1985; Adler, 2001). • Modular governance includes two governance categories: modular value chains and modular platforms, both characterized by high informational complexity, high capability and knowledge requirements, and well-defined open interface standards that serve as coordination mechanisms. o Modular value chains have low switching costs and scalable buyer-supplier relationships due to high codifiability and robust interface standards. Complementarity is relatively weak and symmetrical, as both buyers and suppliers have high, functionally-specific (“turnkey”) capabilities that allow them to support multiple business relationships with relatively low levels of explicit coordination. o Modular platforms actors include platform sponsors, users, and complementors, each playing distinct roles. Platform sponsors define interface standards, users access platform- enabled goods and services, and complementors contribute products, data, and technical tools. Complementarity is asymmetric: platform sponsors are essential to the platform’s 5 These variables can interact with one another: for example, complexity heightens interdependencies (and hence the strength of complementarity) while codification relaxes interdependencies. 10 MASSIVE MODULAR ECOSYSTEMS operation, while any single user or complementor is not. When platforms are successful, network effects can take hold since users and complementors can be added to (or subtracted from) from the ecosystem without the need for explicit coordination as long as linkages adhere to the standards set by platform sponsors.6 • Market governance is characterized by fully codified, arms-length transactions with minimal complexity and weak, symmetrical complementarity. Interactions are price-driven, with limited interdependencies between actors. The framework presented in Figure 2 and elaborated here incorporates insights from management and innovation studies about how platform sponsors set design rules, manage complementor interactions, and influence market structure. The role of interface standards is fundamental in this context, especially when they codify complex information exchange, as emphasized in the modularity and standards and GVC governance literatures. By positioning standards as central to inter-firm coordination, Figure 2 facilitates analysis that transcends specific industry boundaries, since interface standards generally apply to functions and processes that are deployed across multiple industries. The framework distinguishes between strong and weak complementarity, acknowledging that business ecosystems must balance modularity with interdependence. Crucially, the governance categories in Figure 2 are not mutually exclusive. Firms can and do adopt multiple roles simultaneously within and across different sub-ecosystems in an industry, and different governance structures can and do co-exist in the same sub-ecosystem. This comprehensive approach illustrates, for example, how modular platforms can coexist and create synergies with other forms of governance to create identifiable megastructures.7 The case study on the smartphone industry, which we turn to next, illustrates this approach. 4. The mobile phone industry as a massive modular ecosystem Because they are extremely complex, produced in high volumes, and contain many complex and technologically dynamic subsystems, mobile phones provide an ideal case study for examining how multilayered ecosystems emerge, impact competition, and shape innovation, organization, and geography in complex industries where modular governance prevails.8 There is also rich, high-quality and firm-specific longitudinal data available on the industry. The analysis is supported by a comprehensive dataset drawn from multiple sources, including: component lists for 456 mobile phones introduced between 2008 and 2019; data on 15,544 mobile phone components and performance specifications from 2009-2020; data on nearly 10 million company contributions to various releases of Google’s Android operating system from 2008-2020; and data on about 16,000 company contributions to each generation 6 Platform sponsors can be a firm or consortium of firms with a proprietary (yet open) interoperability standard, a government agency or private-sector-led standard, or a non-proprietary interoperability standard promulgated by an open-source community. A platform’s users can be consumers or industrial users. Complementors can contribute products to be accessed over the platform, according to the rules set by the platform sponsor, but also data, tools, intellectual property and other technical inputs. The stylized modular platform shown in Figure 2 also includes a modular arrow representing user-created content, as when users post their own content on social media or ecommerce sites according to the rules of the platform. 7 We can speculate that industries will have different corresponding megastructures due to the architecture of their technical systems, the strategic approach to interface standards, their articulation with the state, and their economic geography, among other influences. For example, the automotive industry has historically exhibited more co-specialization between final systems and sub-systems, which has helped to shape the industry’s global value chain (Sturgeon et al, 2008). It is an open question if the more modular technical subsystems in electric vehicles will be mirrored in the industry’s organization (Sturgeon, 2022). 8 An account of when and how modularity emerged in the mobile phone industry can be found in Appendix D and in Thun et al (2022). 11 MASSIVE MODULAR ECOSYSTEMS of mobile telecom standards from 2001-2019 (see Appendix C for more detail). From the analysis of these data, secondary literature, interviews with engineers and industry analysts, an extensive review of the trade press, company reports, and the on-line technical interface requirements available for complementor design engineers, we are able to deploy the analytic framework developed in section 3 to depict the mobile phone industry as a set of layered and nested sub-ecosystems, each associated with key subsystem function in a generic mobile phone, as suggested by Figure 3.9 In this figure the vertical and horizontal dimensions of the MME are apparent, as well as the links to adjacent industries that share actor groupings and adhere to common standards; in short, the sub-ecosystems within the mobile phone MME span multiple industries. Figure 3. Layered modular ecosystems in the mobile phone MME and links to adjacent industries Note: There are linkages from “other wireless” subsystems to both the mobile phone system and the application processor subsystem because these functions have gradually become “encapsulated” within application processors (see section 5.2 under the heading “Other factors as influencing market structure”). Source: Authors’ elaboration In the remainder of this section we organize the discussion into three subsections: the design and production of mobile phone systems (4.1), the design and production of key mobile phone subsystems (4.2), and the distribution and consumption of mobile phones (4.3). 4.1 The design and production of mobile phone systems Table 2 shows governance at the mobile phone system level, detailing how mobile phones are designed and developed, key subsystems are sourced – a set of activities commonly referred to as ‘system integration’ (Prencipe et al, 2003) – and how final products are assembled. It identifies the primary 9 Due to space limitations, Figure 3 can only hint at the complexity underlying the mobile phone MME. Missing, for example, are the interface standards that set the rules for modular linkages, and any assignation of alterative governance structures. In addition, Figure 3 mainly depicts the mobile phone design and sourcing stages of the MME, elaborated in more detail below in Table 2, leaving out the design and production of subsystems depicted in Table 3. Finally, the app stores elaborated in Table 4 are shown at the top of Figure 3, but not the retail distribution of mobile phones. 12 MASSIVE MODULAR ECOSYSTEMS governance structure linking each subsystem to mobile phone designs,10 the main actors and their roles, some of the key interoperability standards and modular inputs, and notes alternative structures when applicable. Table 2. Mobile phone design, sub-system sourcing, and assembly: governance structures, actors and roles, key interoperability standards and modular inputs Source: Authors’ elaboration The first row of Table 2 depicts governance of the mobile phone design function. Due to their immense complexity, all mobile phones are designed on modular platforms. Mobile phone designers (the platform’s users) utilize computer aided design (CAD) software for system-level industrial and mechanical design and electronic design automation (EDA) software for the design of electronic circuitry. Outputs are generally in industry-standard formats so they can be easily received and acted upon by suppliers, regulators, and manufacturers. Mobile phone designers can access a variety of optional platform-specific complements, which have evolved over time to comprise vast libraries of complements that allow designers to easily combine functions, features, and specific modules and components in their mobile phone designs. These are available to platform users either when the software subscriptions are first purchased or as optional ‘plug-ins’ after the fact. Linkages between design platform sponsors (the design software vendors) and their complementors is typically asymmetric in that the platforms are essential for compatible complements, but individual complements are optional for users and non-essential for the platform’s functioning. 10 The answer to the question of which governance structure in a sub-ecosystem can be considered “prevailing” can depend on how dominance is measured, according to revenues, profits, or number of users and complementors. For example, Apple iPhones have higher profits than most Android-based phones but many more Android phones are sold. Here we use market share within the sub-ecosystem’s industry as an indicator. 13 MASSIVE MODULAR ECOSYSTEMS Subsystems providing core functionality – including application processors, memory modules, display modules, and radio frequency modules – are primarily sourced through modular value chains, as illustrated in the second row of Table 2.11 Interfaces are codified by a myriad of proprietary and non- proprietary standards, with a few key examples shown.12 While these interoperability standards facilitate scalability and reduce transaction costs, high complexity often requires collaboration between buyer and supplier engineers, especially for higher-end mobile phones, either as a routine part of the design process or later on if problems arise that require joint attention. Notably, some mobile phone companies opt for non-modular governance structures despite clear standards enabling modular linkages, usually to avoid ‘modularity traps’ (Chesbrough and Kusunoki, 2001). Prominent examples of alternative governance structures are reflected in the fifth and six columns of Table 2, showing that multiple governance structures coexist in an MME. The third row of Table 2 focuses on the assembly of the mobile phone. The prevailing governance structure for this function is also a modular value chain, enabled by industry-standard formats for design outputs and the global supply base of contract manufacturers that emerged in the 1990s and 2000s in the information and communications technology (ICT) industries to provide manufacturing as a service (Sturgeon, 2002; Raj-Reichert, 2018). However, just as with sourcing key subsystems, some firms make different choices. For example, to maintain the utmost control over quality, cost, and component purchasing, Apple has developed a captive governance relationship with its largest contract manufacturer, Foxconn of Taiwan, China, which has long assembled nearly all iPhones (Grimes and Sun, 2016), while Samsung has a long-standing policy to assemble nearly all of its mobile phones in- house for similar reasons (Zylberberg, 2017). 4.2 The design and production of key mobile phone subsystems Having examined how mobile phone systems are designed, key sub-systems sourced and final products assembled, we now turn to the governance structures, actors, roles, and interface standards within key sub-ecosystems in the mobile phone MME. The method of analysis is the same as at the mobile phone system level. Each subsystem corresponds to a specific function, involving actors that adhere to design rules and standards that help to structure their activities and interactions. 11 Application processors are the “brains” of mobile phones that run applications and perform processing and control functions; core memory modules that provide long-term information storage and retrieval; display modules that provide the touchscreen user interface; and radio frequency modules that manage connections to mobile phone cellular infrastructure for voice and data communications. Mobile phones typically include between 500 and 1,500 individual components, modules, and sub-assemblies. There are various ways to customize complex subsystems and modules, for example through relational partnerships between top firms or formalized licensing agreements that provide buyers with degrees of functional flexibility. Simple components can be purchased from component manufacturers or distributors through simple market linkages, since the standards and specifications are well-known and features and functionality do not vary. 12 For memory, a set of standards issued by the Joint Electron Device Engineering Council (JEDEC) help to “eliminate misunderstandings between manufacturers and purchasers, facilitating interchangeability and improvement of products, and assisting the purchaser in selecting and obtaining with minimum delay the proper product for use.” For example, JEDEC’s The Common Flash Interface (CFI) specification outlines “a device and host system software interrogation handshake that allows specific software algorithms to be used for entire families of [flash memory] devices…It allows flash vendors to standardize their existing interfaces for long-term compatibility” (JEDEC undated). For displays, key standards include the Mobile Display Digital Interface (MDDI) and Mobile Industry Processor Interface Alliance (MIPI) standards, which enable low-power signal transfer between the display module and the mobile phone processor. A discussion of standards for RF modules in mobile phones is included in Appendix E. 14 MASSIVE MODULAR ECOSYSTEMS For brevity’s sake Table 3 summarizes governance in just three subsystems: the phone's application processor (which contains three nested sub-ecosystems: application processor design, application processor architecture, and application processor manufacturing), operating systems, and core memory. We chose these for their importance for mobile phone functionality, and to demonstrate the diversity of governance structures that coexist within the mobile phone MME. Table 3. Mobile phone subsystem design and production: governance structures, actors and roles, key interoperability standards and modular inputs Source: Authors’ elaboration Application processors are designed using EDA software platforms in a manner similar to what was described in the previous discussion of mobile phone design (with the same software vendors 15 MASSIVE MODULAR ECOSYSTEMS dominating). Several “fabless” design firms dominate (the United States’ Qualcomm and Taiwan, China’s MediaTEK), but top mobile phone manufacturers such as Apple, Samsung, and Huawei have sufficient resources and volume requirements to design application processors internally – using the same design platforms and complements as fabless design firms – as a way to differentiate and optimize mobile phone performance. Again, EDA software platforms have vast IP and tool libraries containing features and functions supplied both internally and from many 3rd party complementors. Within the application processor design sub-ecosystem, one complementor stands out: UK-based ARM Holdings. ARM provides a proprietary ‘reduced instruction set architecture’ (ISA) that enhances power efficiency across the mobile phone, a feature essential for extending mobile phone battery life. To do this, ARM’s algorithms, which reside within the application processor, define data flows within the processor, how data are formatted, and how the processor interacts with other subsystems.13 ARM’s technology licenses range from fully functional systems-on-a-chip core design code to á la carte licenses that allow more customization by application processor designers. A wealth of optional third- party tools and add-ons accessed over the ARM’s platform amplify network effects and have strengthened ARM’s position in the mobile phone MME to a virtual monopoly. In the context of the mobile phone MME, ARM’s ISA can be seen as a modular platform nested within the application processor design platforms. As the third row of Table 3 shows, application processor manufacturing primarily follows a modular value chain structure, facilitated by micro-circuitry layout and hardware description standards that help to streamline design-to-production work flows. These standards, initially proprietary, are now overseen by standard setting organizations such as the IEEE. Given their complexity – with 20 billion micro- components or more – most application processors are produced by a handful of top semiconductor foundries (essentially semiconductor contract manufacturers), heavily concentrated in Taiwan, China. However, leading ‘fabless’ application processor designers, such as Apple and Qualcomm, have established close, relational partnerships with top foundries to produce the most advanced processors. At the leading-edge, both EDA software firms and semiconductor manufacturing equipment firms also collaborate relationally, highlighting the coexistence of modular and relational governance in this sub- ecosystem. Finally, Samsung relies on their internal manufacturing divisions to produce a subset of their mobile phone application processors, adhering to the same standard interfaces between design and manufacturing as companies using the fabless-foundry business model. Operating systems provide the interface between end users and all mobile phone subsystems. Android and iOS are both based in part on open-source software ‘kernels,’ and co-evolve with ARM’s instruction set architecture, but Google and Apple take different approaches to operating system development. Apple develops successive versions of iOS internally, as shown in the fifth column of the fourth row of Table 3, while Google sponsors a modular platform that invites 3rd party complementors to contribute code through the Android Open Source Platform (AOSP). AOSP is a multi-tiered modular platform that crowd-sources code intended for the next version of Android while allowing Google to maintain “architectural control” over final versions to ensure the integrity and backward compatibility of the operating system (Shultz et al, 2011). Google also requires that mobile phones using official versions of the operating system be approved through Google’s Android Compatibility Program. However, user’s linkages to Google’s vary: the company maintains relational ties with a small set of advanced mobile phone companies to support deeper collaborations on next- generation versions. These partners gain early access to these more sophisticated and customizable versions of Android, whereas firms that are less sophisticated, demanding, and powerful can only 13 Because of this thoroughgoing influence, both iOS and Android operating systems are optimized for ARM, depicted by the extension of the ARM sub-ecosystem across the application processor and operating system sub- ecosystems in Figure 3. 16 MASSIVE MODULAR ECOSYSTEMS access the fully developed and standardized versions of the operating system (Kenney and Pon, 2011). In contrast to the sub-ecosystems highlighted in this section so far, the design and manufacturing of core memory are always governed internally, within corporate hierarchies (fifth row of Table 3). One reason is because the design content of memory is low, relative to application processors, mainly consisting of three-dimensional grids of cells that can record data by changing their value to 1 or 0. On the one hand, the lack of complexity and co-specialization within mobile phone design commodifies these subsystems. On the other hand, there is a premium placed on process excellence – to drive extreme miniaturization and device performance in the context of high-volume production – and this drives up capability and capital investment requirements, motivating hierarchical governance. Governance structures in the display sub-ecosystem are very similar to memory, for similar reasons. However, large and sophisticated buyers such as Apple are reported to have relational ties with key memory and display suppliers, including Samsung, its top rival in mobile phones (Adamu, 2024). The governance of the radio frequency module sub-ecosystem is a more complex mix of hierarchical and relational, in both design and manufacturing, due to some extreme design and manufacturing challenges arising from the ambiguities inherent in analog design and the need to embed circuitry on exotic substrates (see Appendix E for more detail). 4.3 The distribution and consumption of mobile phones The governance of mobile phone retailing is market-based, involving straightforward, arms-length transactions with retailers and resellers, as depicted in the first row of Table 4. Consumers evaluate mobile phone quality through standardized specifications such as processor speed, display resolution, battery life, and memory capacity, contributing to a highly fragmented retail sector. Table 4. Mobile phone distribution and consumption: mobile phone retail sales and mobile app store functions, governance structures, actors and roles, and key standards Source: Authors’ elaboration In our framework, consumers play the role of platform ‘users’ as well mobile phone owners. The second row of Table 4 depicts how users access mobile applications on modular platforms (app stores), sponsored by either Apple or Google, and supplied by complementors (3rd party app developers). Developers create applications according to an extensive set of proprietary and non-proprietary design rules – called application programming interfaces (APIs) – that are specified by platform sponsors (Apple and Google). Modular linkages also commonly exist between users and complementors, for example for the collection of usage data or user-provided content. In-app payments also flow between 17 MASSIVE MODULAR ECOSYSTEMS users and complementors over the platform, with the sponsor taking a share of the proceedings.14 Application store platforms foster network effects, since their value increases as the number of available apps and user participation grow. The complementarity between the platform and app developers is asymmetric in both cases: while apps and users depend on the platform, the platform does not rely on any single app or user. To summarize, this section illustrates four important points. First, as a class of products, mobile phones are made possible by linking functionally-specific sub-ecosystems in a layered and nested socioeconomic megastructure that crosses multiple industries (i.e., the operating systems, design software, logic semiconductors, memory, displays, and many other subsystems used in mobile phones are produced in sub-ecosystems that serve multiple other industries, as shown in Figure 3). Second, governance structures may vary within the same function. Third, firms have some agency in choosing governance structures, even when modular governance prevails. Fourth, MMEs are supported by a plethora of proprietary (remunerative) and non-proprietary (non-remunerative) interoperability standards that are open for use. 5. Technology and innovation, market structure, and geographic outcomes in the mobile phone industry This section examines how the evolution of technology and governance in the mobile phone MME – and in particular how increasing modularity – correlates with technological, market, and geographic outcomes. We pay particular attention to changes between 2009 and 2020, a transformative phase of the industry during which increasingly complex and modular touchscreen ‘smartphones’ gradually supplanted the simpler and less modular ‘feature phones’ that had dominated the market until the late 2000s.15 5.1 Technology and innovation Technological progress across mobile handset subsystems has been remarkable, as shown in Table 5. For example, between 2009 and 2020, processing speed tripled, supported by denser semiconductor circuitry16 and additional application processor ‘cores’ for multitasking.17 Over this period the average number of processing cores increased from one to eight, and the capacity of cache memory circuits (DRAM) encapsulated within the application processor18 increased at an average annual rate of 37% per year. Integration of one or more graphics processing (GPU) cores, originally aimed at enhancing photo and video processing, and more recently used for artificial intelligence processing, increased from about a quarter of phones in 2009 to become ubiquitous by 2020. These leaps were enabled by improved application processor designs, design software, and manufacturing, and also by advances in the sub-ecosystems nested within application processor designs and design platforms.19 Over the same period, core memory (SRAM) capacity grew at an annual rate of 39%, 14 The division of revenue from these purchases has proven to be extremely controversial, triggering lawsuits between the platform sponsors and the most popular app developers (Scarcella, 2025). 15 Again, a brief narrative of how this transition took place is provided in Appendix D. 16 For example, the 2023 iPhone 15 Pro's A17 Bionic application processor had 19 billion transistors, up from 8.5 billion in the application processor powering the 2019 iPhone 11 series mobile phone (Friedman, 2024). 17 A processing core is essentially a separate embedded processor. Multiple cores speed data processing by performing calculations in parallel rather than linear fashion. 18 Encapsulated DRAM provides the volatile memory used to cache working data and speed processing. 19 For example, Apple’s application processors have utilized ARM processing cores and GPU cores from UK- based PowerVR. 18 MASSIVE MODULAR ECOSYSTEMS while display resolution improved by 26% per year on average, and video resolution saw a 42% per year alongside the integration of optical zoom and the deployment of multiple cameras. Advances in connectivity – such as Wi-Fi, GPS, nearfield communications, and wireless charging – necessitated sophisticated antenna systems to manage multiple radio signals and power transmission with minimal interference. Astonishingly, these dramatic performance and functional improvements were achieved with little change in the physical size of handsets, and with improved battery life through battery and power management innovations. The technical complexity of software (operating systems, apps, and design tools) has paralleled growth in mobile phone performance and functionality. For example, iOS and Android have surpassed 12-15 million lines of code each, placing them among the world’s largest software systems (Fernández, 2023), and app developer documentation (APIs etc.) has become correspondingly voluminous and complex. Growing technical performance and complexity in mobile phones has come with increasing scale in production and consumption. After 2009, when platform operating systems became increasingly dominant, sales of mobile phones increased at an average annual rate of 33.7%, peaking at 1.46 billion units shipped in 2016 (Thun et al, 2022). Table 5. Mobile phone performance and functional improvements by subsystem (2009 and 2020) Note: Bluetooth wireless connectivity is not included in the “Other wireless functions category because it was already ubiquitous prior to 2009. Source: Authors' calculations using data from PhoneDB (https://phonedb.net/), a mobile phone specifications database containing specifications for 15,544 mobile phone models. In the mobile phone MME, recombinant and distributed innovation make up two sides of the same coin. On one side, innovation distributed within each sub-ecosystems helps to improve subsystem performance and generate new features. This in turn provides more options for mobile phone designers to engage in recombinant innovation using EDA software.20 Crucially, EDA software, and the 20 Examples of recombinant innovation include Apple’s in-house A-series application processors, which have utilized processing cores from ARM and GPU cores from UK-based PowerVR. Apple also outsourced manufacturing of its application processors, first to Samsung and later to TSMC. In low-end mobile phones, 19 MASSIVE MODULAR ECOSYSTEMS encapsulation of features in ‘reference’ circuit board designs from application processor vendors, provides mobile phone design engineers with predetermined solutions that implement the evolving interface standards needed to integrate various features and functions. App distribution platforms offer a different and extremely powerful vector for distributed innovation as users can access millions of compatible apps serving a seemingly endless variety of purposes. While the leading-edge of innovation in sub-ecosystems often takes place under hierarchical, captive, and relational governance, the prevalence of modular governance at the mobile phone system level means that actors in the MME are not able to strike out entirely on their own; they must adhere to lion’s share of prevailing design rules even as they try to gain temporary advantage over rivals by breaking others. 5.2 Market structure Table 6 illustrates the evolving market and geographic outcomes across mobile phone sub-ecosystems, contrasting the ‘less modular’ feature-phone era (2002–2009), dominated by more tightly coupled governance structures (hierarchies, captive linkages, and relational partnerships), with the ‘more modular’ smartphone era (2017–2023), characterized primarily by modular platforms and value chains.21 What is revealed are patterns of both high and low market concentration, with each pattern intensified when the geography of firm ownership is taken into account. The outcomes shown in Table 6 were likely shaped not only by the type of governance structure that prevails overall and in each mobile phone sub-ecosystem, but also by other factors creating barriers to entry such as high capability and capital expenditure requirements. While a causal exploration of how various factors shape market structure in the mobile phone industry is beyond the scope of this paper, the remainder of this section highlights several correlations and offers a few likely explanations – both related and unrelated to governance structures – that might be tested in future research. We sequence the discussion according to the type of observed governance structure, running from hierarchical governance to simple markets, as laid out in Figure 4. Hierarchical governance correlates with concentrated markets. As shown in the second column of Figure 4, sub-ecosystems with hierarchical governance, assigned when dominant firms internalize both design and manufacturing, tend to operate in highly concentrated markets. For example, in our IHS teardown dataset Samsung accounted for 73% of both core memory and mobile phone displays in the post-smartphone era (2017-2022 average), up from about a third from the pre-smartphone era.22 We posit that increasingly high technical and capital expenditure requirements created rising barriers to entry that gradually thinned the field. In RF modules, capital requirements are somewhat lower, but as already alluded to in section 4.2, specialized and proprietary material science and manufacturing capabilities across a handful of dominant firms are associated with a complex mix of hierarchical and China’s Shenzhen Transsion Holding, which started selling feature phones in Africa in 2006, gained a 50% market share on the continent in touchscreen smartphones by 2024. Drawing on the resources of the MME, the company is able to offer a range of mobile phone brands and models priced from $20-$200, a branded earbud, and its own streaming service similar to Spotify. As incomes rise in Africa, however, it is unclear if the company will be able to compete long term with global leaders (Carr, 2024). 21 The specific dates of the comparison periods in Table 6 reflect both the years covered in the secondary sources used and the timing of the documented shift toward modularity in the mobile phone industry. As discussed in detail in Appendix D, the introduction of Apple’s iOS and Google’s Android in 2007-09 marked a pivotal shift by decoupling operating systems from applications. This change empowered third-party app developers to innovate independently of hardware constraints, driving technological convergence, introducing the platform governance model, and replacing vertically integrated, analog systems with more modular digital sub-ecosystems. 22 However, since this data set is biased toward higher-end phones, lower end phones are more likely to use subsystems produced by less dominant suppliers. 20 MASSIVE MODULAR ECOSYSTEMS relational governance (between hierarchies) that creates barriers to new entrants (see Appendix E for more detail). In these subsystems, it is possible that deep interdependencies and co-specializations between design and manufacturing, as well as control over critical manufacturing assets, create barriers to entry and drive internalization. Table 6: Mobile phone sub-ecosystem market and geographic outcomes correlated with rising modularity in mobile phone technical architecture * Mentor Graphics was acquired by Siemens of Germany in 2016, but the subsidiary is still mainly headquartered in Wilsonville, Oregon, USA. Sources: CAD-CAM software: Svitlo (2009: citing Jon Peddie Research); 6Sense (2024). Phone System: Gartner (2004-07) and IDC (2017-20) via Statista (2024). Contract Manufacturers: Sturgeon and Lee (2005); Weiss (2019). application processor, core memory, display, and RF module subsystems: IHS Market Teardowns (see Appendix C). ISA: Fitzpatrick (2011); ARM Holdings (2023), Yarrow (2010). Operating systems: Statcounter 21 MASSIVE MODULAR ECOSYSTEMS (2024). EDA Design Software: StatCounter (2024); IC Manufacturing: LaPedus (2005), TrendForce (2024). App distribution platforms: Bigabid (undated). Figure 4. Market concentration levels correlated with governance structures in the mobile phone MME Source: Authors’ elaboration based on Table 6 Captive governance correlates with market concentration when suppliers are tied to dominant buyers. Contract manufacturing tends to be governed by modular value chains, as discussed below, but more tightly coupled captive ties can allow suppliers to ‘ride the coattails’ of dominant buyers, consolidating their market position. The most prominent example in the mobile phone MME is the captive governance linkage Apple has – as an alternative to modular value chains – with its primary contractor, Taiwan, China’s Foxconn, as the exclusive maker of iPhones (Grimes and Sun, 2016). Apple’s sustained success with the iPhone has produced a long-term relationship between the two firms able to tolerate the strong, if asymmetric complementarity that comes with relationship-specific assets and routines. As a result of these barriers to entry, Foxconn’s market share, which was negligible in the pre-smartphone era, rose to 46% by 2018. The third column of Figure 4 depicts this alternative governance option. Relational partnerships among leading players correlate with concentrated markets. Relational governance between leading firms appears to have played a significant role in shaping market structures and innovation trajectories of the mobile phone MME. In the higher-end of the market, mobile phone brands firms like Apple, Samsung, Oppo, and Huawei dominate by combining modular with non- modular governance to help differentiate their products and command premium pricing. This includes relational linkages to key suppliers to customize key features and gain priority access co-developed inputs. Thus, the higher-end of the market remained relatively concentrated across the two periods, transitioning from Nokia’s 47% market share during 2007–2009 (pre-smartphone) to a slightly less consolidated structure by 2017–2019, with Samsung (25%), OPPO (19%), and Huawei (17%) and Apple (17%) leading the market in terms of unit shipments.23 Relational linkages also may have contributed to high market concentration levels in the EDA software, application processor, and semiconductor manufacturing sub-ecosystems as leading firms regularly collaborate to develop and 23 According to Jones (2018), Apple iPhones captured 19% of global smartphone shipments in the fourth quarter of 2017, yet took in 87% of the profits. 22 MASSIVE MODULAR ECOSYSTEMS (eventually) codify the most advanced processes. Such ‘innovation clubs’ ensure that a small group of firms remain at the cutting edge and wield disproportionate influence over the broader MME. By leveraging relational governance within ecosystems where modularity prevails, these firms can consolidate leadership, sustain competitive advantages, and drive the innovative trajectory of their sub- ecosystems forward. The fourth column of Figure 4 depicts this alternative governance option. Modular platform sponsorships correlate with highly concentrated, ‘winner-take-most’ markets. Successful modular platforms in the mobile MME phone display extremely high levels of market concentration. EDA software has long been dominated by three firms emerging from Silicon Valley: Cadence, Synopsys, and Mentor Graphics, which together held a 92% market share in 2023, while Autodesk has held more that 50% of the CAD market for many decades. ARM holds a near-monopoly in the mobile device instruction set architecture sub-ecosystem, as already mentioned. Google’s Android dominated with 74% of the operating system market (2017–2020), while Apple’s iOS accounted for the remaining 24%. Mobile app distribution outside China is controlled by just two firms, with Apple capturing 67% of revenues in 2023 and Google capturing the remaining 23%. Qualcomm expanded its share in application processors from 27% (2004–2007) to 67% (2017–2022).24 The association of “winner-take-all” and “winner-take-most” market outcomes for successful modular platform markets have been widely noted and analyzed (Sun and Tse, 2007; Cennamo and Santalo, 2013), attributed to network effects arising from asymmetric complementarity. In the mobile phone MME, network effects create lock-in, not only for end-users to specific operating systems and app stores, but also for software engineers adhering to the app store specifications, and phone designers to EDA and CAD platforms since network effects are intensified when platform users invest in highly specific skills or face path dependence due to requirements for backward compatibility to prior products and standards.25 As already noted, mobile phone designers also rely on reference design elements issued by application processor companies, which in turn rely ARM instruction set licenses, revealing the nested structure of the MME. The fifth column of Figure 4 depicts how concentrated markets are associated with successful platform sponsorship. Modular platform complements correlate with fragmented markets. Because complementarity in modular platforms is asymmetric, markets for complements are as fragmented as successful platform sponsors are concentrated. According to company documents, Qualcomm has a network of 453 partners (including 112 firms offering optimized products) while ARM has more than 1,000 firms offering complements. The number of mobile apps peaked at 1.3 million apps in 2017, with significant contributions from populous developing countries such as China (16%), India (5%), and Brazil (3%) (Thun et al, 2022). The sixth column of Figure 4 depicts fragmentation in markets for platform complements. Modular value chains correlate with fragmented markets. In contrast to high-end mobile phones, the medium and low-cost portions of the market are extremely fragmented, as reflected by the share of “other brands”, which accounted for 40% of the total market over the 2017–2019 period in comparison to 29% in 2007-2009 period. Globally, according to DeviceAtlas,26 more than 5,000 firms compete in this segment with more than 70,000 phone models. They rely on commoditized subsystems and complements included in EDA libraries to design weakly differentiated, price-driven products, or 24 As explained in Appendix C, the dataset these figures are drawn from is biased toward higher-end phones. When lower end phones are analyzed separately, the market share of other application processor vendors, especially Taiwan, China’s MediaTEK, is much higher. 25 Network effects are accentuated when a platform’s user must invest in highly specific skills in order to use a complement, and hence is reluctant to switch, and when a need for backward compatibility to prior standards creates path dependence. 26 See https://deviceatlas.com/ accessed August 2024. 23 MASSIVE MODULAR ECOSYSTEMS contract designs to 3rd parties.27 In contract manufacturing, the market has always been extremely fragmented (outside of the exceptional case of Apple-Foxconn discussed above) with Pegatron, the second place firm in 2018, accounting for just 12% of revenues. Because of weak, symmetric complementarity, contract manufacturers are highly substitutable. The seventh column of Figure 4 depicts market fragmentation in mobile phone sub-ecosystems governed by modular value chain linkages. Market governance correlates with fragmented markets. As depicted in eighth column of Figure 4, mobile phone sales are the most fragmented part of the MME, with innumerable retail outlets governed primarily by simple markets. Exceptions and other factors influencing market structure. Of course, the mobile phone MME has exceptions to the patterns just discussed. In addition, because firms operate under multiple governance arrangements at the same time, market structure outcomes are not likely to be due to a single factor.28 The semiconductor foundry sub-ecosystem is instructive. Companies in this sub- ecosystem have no overriding need to combine design and manufacturing because interface standards are very well established and complementarity is weak and symmetric, features that place governance in the modular value chain category. However, TSMC, the leading company, held 58% of the IC contract manufacturing market in 2016 and 2022, on average, a 10 percentage-point increase over the 2011- 2016 average. This high market share may result from TSMC’s participation in the ‘innovation clubs’ formed through the relational linkages with leading EDA, application processor, and top mobile phone companies. It is also true that TSMC operates in a highly concentrated sub-ecosystem along with only a handful of other firms, including Samsung, Global Foundries, UMC, and Chartered Semiconductor. Since top innovation clubs have very restricted membership, concentration in high volume logic semiconductor manufacturing services may also arise from the extreme capability and capital expenditure requirements needed to stay at the leading edge of manufacturing logic semiconductors, investments that require amortization across a large customer base. Another example can be found in the application processor sub-ecosystem. Extremely high capabilities and tight control over IP, combined with aggressive legal protections, a combination that is descriptive of Qualcomm, produces barriers to entry that are only loosely related to governance. Qualcomm is aggressive about inserting its technology in mobile telecom standards making their technology ‘essential’ (i.e. a monopoly), through which it has been able to extract licensing fees for several decades since new standards tend to encapsulate older technologies to maintain backward compatibility (Thun et al, 2022). Furthermore, the dynamic of ‘encapsulation’ can also drive market concentration by integrating functions from one or more subsystems into another. An example are the sub-ecosystems denoted as “other wireless subsystems” in Figure 3 (WIFI, GPS, Bluetooth, and nearfield connectivity). These were excluded from the deeper analysis provided in Section 4 because these technical functions 27 According to Brandt and Thun (2010, 2011, 2016) cross-fertilization between these high- and low-end segments as they vie for mid-range market segments can be crucial for capability building, as firms producing higher-end firms take advantage of the low-cost inputs serving of the low-end segment, while firms producing low-end products gradually access more sophisticated inputs as a way of building the capabilities needed to break out of their modularity traps. 28 As we have seen in this section, Apple designs subsystems that it deems to be sources of core competitive advantage in-house (e.g. application processors), has relational ties with key subsystem providers (e.g. chip fabricators such as TSMC), and exerts tight control over its most important (captive) contract manufacturer (Foxconn/Hon Hai). Samsung internalizes the design and manufacturing of both memory and display, allowing it to speed time to market, retain control over strategic assets, and ensure product differentiation, as well as to earn revenue by selling them to other mobile phone firms. 24 MASSIVE MODULAR ECOSYSTEMS have gradually become encapsulated in application processors, as depicted in Figure 3.29 While each of these "other wireless" functions is associated with a distinct sub-ecosystem, with evolving standards, standard setting routines, and standard setting organizations, they have become increasingly commodified along with the encapsulation trend, with many firms offering stand-alone wireless components exiting the mobile phone MME altogether. In summary, barriers to entry in the mobile handset MME include a non-exclusive set of governance and non-governance factors. They include high capital expenditure requirements (e.g., high-end phones, foundries, memory, and displays), tight IP control (e.g., Qualcomm application processors), specialized capabilities (e.g., RF modules), lock-in from network effects (e.g., operating systems, app stores, application processors, and ARM instruction set architecture), and captive linkages to dominant buyers (Foxconn). In general, Table 6 shows a clear trend toward increasing market concentration in most sub-ecosystems in the mobile phone MME. It just as clearly shows that no single actor dominates the entire MME. 5.3 Economic geography In addition to showing market structure outcomes correlated with rising modularity in the mobile phone industry, Table 6 illustrate geographic outcomes at the mobile phone and subsystem levels using company headquarters as a proxy for location (see the right hand columns in both time periods).30 Given high market concentration in many key sub-ecosystems, geographic concentration is partly a function of the global dominance of leading firms. However, due to agglomeration forces, it is often the case that several – or all – top firms in a specific sub-ecosystem are headquartered in the same country, amplifying concentration and driving both supply chain risk and geoeconomic tensions.31 Table 6 also shows a remarkable degree of stability over time in terms of country dominance, and a marked increase in geographic concentration (from an already high base), especially in sub-ecosystems governed by modular platforms and hierarchies. In these sub-ecosystems, top firms may jockey for position, but country dominance appears to shift rarely, suggesting the powerful roles played of geo- graphically embedded capabilities and support institutions. However, changes over time in the geographic center of gravity for sub-ecosystems can occur, as illustrated by final mobile phone systems (from Europe to the Republic of Korea, the United States, and China), core memory (from the United States and Europe to Korea), and contract manufacturing (from the United States to Taiwan, China). In general, the geography of ownership in the mobile phone MME has drifted toward East Asia over time as laggard countries and regions faded in importance (e.g., Japan) or dropped out entirely (many European countries). US-based companies, however, maintain a very strong position in several key sub-ecosystems (electronic design software, operating systems, and application processors). Meanwhile, app ecosystems demonstrate a more 29 In Figure 3, this encapsulation process is suggested by the two modular linkages extending from the “other wireless” sub-ecosystems to both the mobile phone system, where they might be incorporated as stand-alone devices or modules, and to the application processor, where they might be incorporated as ‘IP blocks’ or ‘chiplets’ within the processor. 30 While operations span many countries, we define ownership in terms of the location of strategic and operational headquarters, since this shapes political and regulatory vulnerabilities experienced by firms. 31 For example, as table 6 shows, Taiwan, China, accounted for 90% of contract manufacturing revenues in 2018 (with most production taking place in China) and 69% of worldwide foundry revenues on average across 2016 and 2022. At the same time, US-headquartered companies accounted for 61% of CAD software in 2020, 98.4% of mobile operating systems and 67% of application processors on average across 2017-2020, and 92% of EDA software in 2021, 62% of RF modules on average across the 2017-2020 period, and 100% of app store platforms in 2023. Korea is also a major node for key mobile phone sub-ecosystems, with an 87% of core memory and 81% share of display on average across 2017-2020. As already mentioned, UK’s ARM has a monopoly position in instruction set architecture IP. 25 MASSIVE MODULAR ECOSYSTEMS geographically distributed pattern, with significant contributions from populous countries like China and India, which are emerging as critical markets and sources of innovation in software complements (Thun et al, 2022). Looking solely at the IHS mobile phone teardown database, which includes 456 models averaging 38 phone models per year (as described in Appendix C), significant transformations in the mobile phone MME over the period under examination can be detected, including shifts in sourcing patterns, the rise of new locations, and the geographic clustering of key sub-ecosystems. Table 7a provides numeric evidence of a more concentrated network over time, with fewer, larger players, even as both the average number of components per phone and the number of supplier countries increased substantially, as shown in the figures in Table 7b. Analysis of firm entrants and exits Table 7c shows a growing number buyers from more countries, but that they linked to fewer supplier countries, suggesting geographic clustering. The concentration ratios shown in Table 7d also suggest more geographic concentration of supply and more diversification of demand. Table 7. Analysis of IHS mobile phone teardown database Table 7a. Mobile phone network analysis, 2009-2020 2009 - 2012 2013 - 2016 2017 - 2020 # of different countries (nodes) 21 16 15 # of different Buyer nodes 18 12 11 # of different Supplier nodes 13 10 9 Number of supply links (edges) 224 154 130 SD of # links by Supplier (indegree) 4.68 4.16 3.81 Table 7b. Average number of components and supplier headquarter countries per mobile phone Average number of components Average number of sourcing countries Table 7c. Buyer and country entrants and exiters, 2009-2020 Entrants Exiters Number of buyers 14 12 Number of countries of origin in buyer nodes 4 6 Avg. number of supplier countries per device 4.53 3.59 26 MASSIVE MODULAR ECOSYSTEMS Table 7d. Geographic clustering in key mobile phone sub-ecosystems < 2009 2009 - 2012 2013 - 2016 2017 - 2020 Apps processor 0.72 0.82 0.79 0.83 Display 0.28 0.47 0.40 0.77 Core memory 0.43 0.53 0.52 0.65 RF module 0.41 0.44 0.49 0.67 Source: IHS Market (see Appendix C for more detail) Thus, the evidence from the IHS dataset reveals two intertwined dynamics. On one hand, geographic specialization and clustering consolidate innovation within the boundaries of fewer countries, particularly at the subsystem level. On the other hand, the production and distribution of mobile phones has become increasingly global and dispersed, involving a broader set of firms and regions. Judging from the increasing market share of Chinese and ‘other brands” from 29% to 40% over past several decades, as shown in the fourth row of Table 6, market diversification has likely been driven by the rise Chinese mobile phone firms, which entered the market in force, while several incumbents from Japan, the EU, Korea, and the United States either exited or lost ground. So the rise of Chinese firms in the mobile phone MME has primarily occurred on the buyer side and has not been accompanied by a similar expansion on the supply side. These parallel trends – consolidation of input sub-ecosystems and global dispersion of mobile phone systems firms – reflect the rising complexity of the MME, creating both sourcing risks and geopolitical tensions. As we discuss in section 6.3 below, trade disputes, military tensions, and force majeure disruptions can and do disrupt global supply chains, highlighting the vulnerabilities that have come with a geographically distributed industry structure. Modular value chain governance mitigates these risks to some degree by easing supplier switching and redundancy in supply chains, but such solutions remain challenging to implement rapidly and at scale, especially in sub-ecosystems where geographic concentration is highest, including semiconductor foundries, application processors, EDA software, contract manufacturing, core memory, and instruction set architecture. In segments of sub-ecosystems characterized by hierarchical, captive, relational, and modular platform (in regard to sponsors) governance, any notion of radically reconfiguring patterns of spatial ownership, even over the medium term, is especially daunting, if not quixotic. 6. Discussion In this section we briefly discuss the significance of our approach for theory, strategic and innovation management, and policy. Extended argumentation and comparative case studies are planned for future papers. 6.1. Implications for Theory Complexity and scale are characteristics often in tension, particularly in industries with short product cycles. However, increasing modularity appears to have allowed these tensions to be mitigated in the mobile phone MME, which has combined rapid performance improvements and functional efflorescence with rapid scaling and geographic expansion. By leaving subsystem innovation largely to sub-ecosystems of specialized suppliers, modular value chain governance at the level of final systems design has likely facilitated the distributed, functionally-focused innovation and rapid improvements 27 MASSIVE MODULAR ECOSYSTEMS documented in Table 5. Our approach has allowed us to highlight the variety of governance forms present in an industry where modular governance prevails. Modular forms of governance ease tensions between complexity and scale, but non-modular forms of governance (hierarchical, relational, and captive) produce the complex interdependencies and tacit interactions that drive much of the industry’s cutting-edge innovation. Even so, it is unfeasible to build new ecosystems entirely divorced from a modular backbone that has co- evolved over many decades. When innovations comply with existing interoperability standards, they can be adopted by more actors, allowing them to propagate rapidly across the MME. In sum, non-modular forms of governance within the mobile phone MME have helped the industry achieve the delicate balance envisioned by Baldwin (2020, chapter 5): non-codified linkages – and especially internal and relational linkages – are what keeps the innovation wheels spinning, but without the broader context of modularity, these would be isolated cases of innovation rather than a scalable contribution to a broader matrix that includes both leading-edge and more incremental innovation based on recombining and adapting existing subsystems. The defining element of an MME is not that every relationship is purely modular, but that non-modular relationships operate in a ‘sea of modularity,’ a context where modularity prevails. In terms of non-modular governance, Baldwin’s optimal balance point lies between markets and hierarchy, with modular business ecosystems offering a middle ground where coordination is rendered more supple. But since multiple, interlinked and nested ecosystems are at play an MME, there is no way for it to collapse into hierarchy because of durable dependencies between multiple subsystems and the crucial role played by platform complementors. Our case study method rests on the key distinction between technical and organizational perspectives. From a technical perspective, the mobile phone MME is observable as a layered and nested set of specific intermediate functions which are linked in a rough vertical sequence to arrive at a final product. However, from an organizational perspective, any notion of a vertical or sequential hierarchy breaks down. Due to distributed innovation, new or improved features and functions are constantly becoming available from across the MME. And, because of modularity, this happens irregularly and in no particular order, since each sub-ecosystem follows distinct technology standards and roadmaps. So, the technical hierarchy in an MME is not mirrored by an organizational hierarchy. There is no centralized control or ‘lead firm’ orchestrating the MME, and products are not developed or produced in a sequential manner. Sometimes a firm – even a small one like ARM, along with its complementors – can provide a subsystem so fundamental that it requires actors in other sub-ecosystems to adjust their technology and product roadmaps to remain relevant. Furthermore, mobile phones are never finalized because new mobile applications alter their functionality after they are purchased. This open-endedness repeats across the MME. For example, semiconductor foundries update their interface standards in the form of ‘process development kits’ every month for the leading edge, and these are made available to chip designers as EDA software library updates. As a result, the clear hierarchies of power and sequentialism central to GVC analysis are absent.32 In this way, an MME might be viewed as a self- organizing system that organically tends toward a stable and scalable modular structure with sharply defined edges, similar to what has been observed in biological systems (see McGovern Institute for Brain Research, 2025). Finally, the geographic distribution of sub-ecosystems within the mobile phone MME reflects deep- seated patterns in economic geography, with top firms in specific sub-ecosystems often clustered in specific locations. Geographic specialization is cumulatively reinforced by agglomeration economies, creating remarkably durable spatial advantages based on anchor firms, specialized labor markets and 32 This lack of sequential processes and hierarchical power relations is why we refer to all ecosystems within the MME as ‘sub-ecosystems’ regardless of position in the layered and nested ‘stack’ of ecosystems. 28 MASSIVE MODULAR ECOSYSTEMS industry-specific institutional supports (Stinchcombe, 1965; Myrdal, 1974; Scott, 1988, Storper 1995, Asheim and Gertler, 2006; Saxenian, 1996, Sturgeon, 2003, Thun, 2006). Because sub-ecosystems are often – and in the case of mobile phones increasingly – spatially clustered and concentrated, interdependencies across countries can be much stronger than interdependencies between a final system and any specific subsystem. All mobile phones need core memory, for example, and it would be relatively simple for a phone designer to switch from Samsung to SK Hynix, for example, but because design and production of core memory are heavily concentrated in Korea, it would be exceedingly difficult to shift core memory sourcing to a different country. Even simple components sourced via markets may create powerful supply-chain vulnerabilities if they are only available in one or a few locations, as was seen during the Covid-19 pandemic (Wu and King, 2022; Xiong et al, 2024). 6.2. Implications for strategic and innovation management To effectively navigate in an MME, a firm must effectively manage a competitive landscape that is dominated by the duality of concentration and fragmentation. Since many sub-ecosystems feature dominant actors, but no single firm dominates the system as a whole, firms working in any particular sub-ecosystem must create bridges to other actors in the MME while simultaneously erecting barriers to entry to protect their specific advantages. In short, they must navigate within technical and organizational structures over which they have limited control. The case study of mobile phones offers two important lessons for strategic managers. The first is the importance of balancing modular and non-modular forms of governance, with the former enabling distributed innovation and scaling, and the latter mitigating the risks of commoditization and modularity traps. Non-modular governance is particularly important for developing leading-edge innovations. Apple’s success, for example, stems from adopting multiple governance arrangements at the same time. For example, the company accesses capabilities in multiple sub-ecosystems through modular ties, internalizes of a few key functions (e.g. the operating system and application processor design), forms relational ties with key suppliers, exerts a high degree of control over its captive contract manufacturer, and benefits from powerful network effects from its highly successful App Store platform. At the same time, simultaneous innovations in sub-ecosystems can push the technological frontier forward for the entire industry, since modularity at the product design and sourcing level make it easier to include technological advances in final products. Ironically then, final products can become increasingly commodified even as its complexity and scalability skyrocket. The existence of ‘complex commodities’ is yet another paradox of MMEs and a boon to lower income consumers who can gain access to advanced features at low cost (see Thun et al, 2022 for an extended discussion of the paradoxical nature of MMEs). The second is the strategic role of standards. The benefits of sponsoring a successful modular platform with a set of proprietary, revenue-generating interoperability standards are well-known, but the mobile phone case also points to the crucial roles played by hundreds of “mundane” standards (Langlois, 2006) that provide the basic plumbing of an MME. These are voluntarily established and maintained by a wide variety of companies, consortia, industry associations, formal standard-setting organizations, and open-source communities. Participation in standard-setting can allow powerful firms to maintain their network centrality, speed time to market, and make their proprietary technologies indispensable in key standards (potentially generating perpetual licensing fees and royalty payments). As a result, firms contribute substantial engineering hours to the organizations and open-source engineering projects that create and maintain open industry standards and open-source technologies (Shiu et al., 2023, Shiu, 2024). While leading companies aggressively promote their proprietary technologies to become industry standard when standards undergo generational changes, robust interoperability standards create space for less powerful firms to connect as suppliers and platform complementors offering 29 MASSIVE MODULAR ECOSYSTEMS compatible products, tools and services. Finally, the layers of an MME have no set boundaries. Because modular governance prevails, innovation from functional recombination and the addition of new sub-ecosystem layers is constantly taking place. The result is an industry that continues to grow and evolve in dynamic and unpredictable ways, for example, when the addition of super-apps such as WeChat generate vast sub-ecosystems of their own by running over the top of mobile phone operating systems. 6.3. Implications for policy Because the mobile phone MME is profoundly global in scope, characterized by geographic specialization at the sub-ecosystem level, interdependent innovation, and many national and international standard setting mechanisms, policy interventions are exceedingly challenging to craft and execute. On the one hand, policies that aim to minimize value chain vulnerabilities by building complete domestic industrial systems – such as the “Made in China 2025” policies – are unlikely to succeed because even the most dominant nations within an MME will not have all of the required capabilities within their own borders to produce complete products at scale and at the leading-edge. The costs associated with attempting to achieve self-sufficiency in semiconductors, for example, have been projected to be extraordinary.33 For mobile phones, the increasing Hirschman-Herfindahl Index (HHI) values for subsystems shown in Table 7d illustrate the economic and logistical challenges of self-sufficiency. Efforts to decouple from MMEs risk marginalizing national firms from the intricate dance of rapid innovation that takes place independently in each sub-ecosystem, exclusion from the relational partnerships crucial to cutting edge innovation, and isolation from key standard setting processes and organizations. Furthermore, MMEs pose difficulties for even more limited forms of industrial policy. For example, if a country tries to increase its market share in a particular sub-ecosystem through a combination of subsidies or import restrictions, it will in all likelihood introduce many new technological and supply- chain vulnerabilities due to the layering and nesting of sub-ecosystems. Identifying and controlling ‘strategic nodes’ and ‘chokepoints’ within MMEs has become more challenging as growing geographic specialization has increased reliance on cross-border partnerships. This is a break from the past, when sectoral targeting might more easily be based on clear technological pathways and a full roster of domestic firms. With MMEs, policymakers must determine the key loci within sub-ecosystems where value and power are being produced and concentrated, understand the degree and type of interdependencies in adjacent layers of MMEs, and bear in mind how they connect to other essential sub-ecosystems, many of which will be concentrated in foreign jurisdictions seeking to fulfill their own policy goals. In short, the depth, complexity, and fluidity of an MME make it difficult for policymakers to intervene aggressively without the risk of unintended consequences.34 To navigate the inherent dependencies in MMEs, countries should cultivate cross-border alliances that ensure access to essential capabilities and technologies and participate in standard-setting initiatives in good faith to maintain influence over global technology trajectories. By fostering a national and international environment conducive to distributed innovation, policymakers can mitigate the risks of exclusion or systemic collapse from the implementation of overly restrictive or precipitous policy initiatives. 33 According to the analysis of the Semiconductor Industry Association and the Boston Consulting Group (Varas et al., 2021), the cost of creating a fully localized semiconductor supply chain would be $0.9 trillion to $1.2 trillion in upfront investment and $45 billion to $125 billion in annual costs (resulting from the loss of efficiencies coming from geographic specialization). 34 On this point, see also Gereffi et al. (2021). 30 MASSIVE MODULAR ECOSYSTEMS 7. Conclusion The global economy has changed dramatically over the past several decades, and our theoretical tools have not kept up. The primary purpose of this paper has been to develop a framework for understanding how layers of business ecosystems are structured and interlinked. Our hope is that our framework, and the case study example we have provided, will serve as a foundation for research into other sectors. Mobile phones, like most electronic systems, lie at the far end of the digitization/modularity spectrum, but there are many similar industries, such as cloud computing and AI; others that are not far behind, such as banking and finance; and others that are rapidly advancing, such as the automotive industry.35 As Holgresson et al. (2022: 6) observe, digitization is evident across broad swaths of the global economy, not only transforming products and services, but fostering the adoption of organizational forms and business models that originated in the computer and software industries. Our hope is that the framework developed in this paper will be a useful tool for understanding the distinct organizational structures of complex, global, digitally-mediated industries, and for developing strategies and policies to cope with the challenges that these essential organizational structures create. 35 See the recent April 2024 issue of Research Policy (Volume 53, Issue 3) for examples. Recent research has examined the rise of platform intermediaries even in labor intensive sectors such as metal machining and apparel manufacturing (López et al., 2022, Schneidemesser and Butollo, 2022). 31 MASSIVE MODULAR ECOSYSTEMS References 6Sense (2024). “Autodesk” Technographics / CAD Software." from https://6sense.com/tech/cad- software/autodesk-market-share. Adamu, H. (2024) Did you know: Samsung makes a lot of money from iPhones - Samsung and Apple: Rivals in the spotlight, partners behind the scenes. Android Authority, General Technology. Adler, P.S., 2001. Market, hierarchy, and trust: The knowledge economy and the future of capitalism. Organization science, 12(2), pp.215-234. Adner, R. (2017). "Ecosystem as structure: an actionable construct for strategy." Journal of Management 43(1): 39-58. Adner, R. and R. Kapoor (2010). "Value creation in innovation ecosystems: How the structure of technological interdependence affects firm performance in new technology generation." Strategic management journal 31: 306–333. ARM Holdings (2023). "Form F-1 Registration Statement. U.S. Securities and Exchange Commission." from https://investors.arm.com/sec-filings/sec-filing/f-1/0001193125-23-216983. Asheim, B. T.; and M. S. Gertler (2006). “The geography of innovation: regional innovation systems.” Chapter 11 in: Jan Fagerberg and David Mowery (eds.), The Oxford Handbook of Innovation, Oxford University Press, Oxford, UK. Baldwin, C. Y. (2007). "Where do transactions come from? Modularity, transactions, and the boundaries of firms." Industrial and Corporate Change 17(1): 155-195. Baldwin, C. Y. (2020). Ecosystems and Complementarity. Harvard Business School Working Paper. Boston, Harvard Business School. Baldwin, C., M. Bogers, R. Kapoor and J. West (2024). "Focusing the ecosystem lens on innovation studies." Research Policy 53. Baldwin, C. Y. and K. B. Clark (2000). Design rules: The power of modularity. Boston, MIT Press. Bigabid (undated). "Key 2023 Apple App Store vs Google Play Store Differences for Developers and Marketers." Retrieved November 26, 2024, https://www.bigabid.com/apple-app-store-vs-google- play-store-differences/ Bogers, M., J. Sims and J. West (2019). "What is an ecosystem? Incorporating 25 years of ecosystem research." Academy of Management Proceedings 1. Brandt, L. and E. Thun (2010). "The fight for the middle: Upgrading, competition, and industrial development in China." World Development 38(11): 1555-1574. Brandt, L. and E. Thun (2011). "Going mobile in China: shifting value chains and upgrading in the mobile telecom sector." International Journal of Technological Learning, Innovation, and Development 4(1/2/3): 148-180. Brandt, L. and E. Thun (2016). "Constructing a Ladder for Growth: Policy, Markets, and Industrial Upgrading in China." World Development 70: 78-95. Carnabuci, G. and Operti, E. (2013). "Where do firms' recombinant capabilities come from? Intraorganizational networks, knowledge, and firms' ability to innovate through technological recombination." Strategic Management Journal, 34(13): 1591-1613. Carr, A. (2024). “How an Unknown Chinese Phonemaker Took Over Africa.” Bloomberg BusinessWeek, September 16, https://www.bloomberg.com/news/articles/2024-09-16/africa-s- shift-to-smartphones-is-creating-a-new-chinese-phone-giant Cennamo, C. and J. Santalo (2013). "Platform competition: Strategic trade‐offs in platform markets." Strategic Management Journal, 34(11), pp.1331-1350. 32 MASSIVE MODULAR ECOSYSTEMS Chiao, B., J. Lerner and J. Tirole (2008). "The rules of standard-setting organizations: an empirical analysis." The RAND Journal of Economics 38(4): 905-930. Colfer, L. J. and C. Y. Baldwin (2016). "The mirroring hypothesis: theory, evidence, and exceptions." Industrial and Corporate Change 25(5): 709-738. Cusumano, M. A., A. Gawer and D. B. Yoffie (2019). The Business of Platforms: Strategy in the Age of Digital Competition, Innovation, and Power. New York, HarperBusiness. Cusumano, M. A., Y. Mylonadis and R. Rosenbloom (1992). "Strategic Maneuvering and Mass- Market Dynamics: The Triumph of VHS over Beta." Business History Review 66(1): 51-94. Dallas, M.P., Ponte, S. and Sturgeon, T.J., 2019. "Power in global value chains." Review of International Political Economy, 26(4): 666-694. Fernández, M. (2023) Code by the Numbers: How Many Lines of Code in Popular Programs, Apps, and Video Games? SoftTonic Fernandez-Stark, K. and G. Gereffi (2019). Global value chain analysis: A primer. Handbook on global value chains. S. Ponte, G. Gereffi and G. Raj-Reichert. Cheltenham, UK, Edward Elgar Publishing: 54-76. Fitzpatrick, J. (2011). "An interview with Steve Furber." Communications of the ACM 54(5): 34-39. Friedman, A. (2024). "Apple’s A19 Pro AP, built by TSMC, could sport up to 30 billion transistors." PhoneArena. Retrieved November 19, 2024, from https://www.phonearena.com/news/apple-to- take-advantage-of-tsmc-third-gen-3nm-node-2025_id165009. Gartner and IDC (2024). "Smartphone shipments worldwide from 4th quarter 2009 to 3rd quarter 2024." Statista, Technology & Telecommunications / Telecommunications. from https://www.statista.com/statistics/728644/quarterly-global-smartphone-shipments-by-quarter/. Gawer, A. and M. A. Cusumano (2014). "Industry platforms and ecosystem innovation." Journal of Product Innovation Management 31(3): 417-433. Gereffi, G., J. Humphrey and T. Sturgeon (2005). "The governance of global value chains’." Review of International Political Economy 12(1): 78-104. Gereffi, G., H. C. Lim and J. Lee (2021). "Trade policies, firm strategies, and adaptive reconfigurations of global value chains." Journal of International Business Policy. Gereffi, G. and T. J. Sturgeon (2013). Global Value Chain-Oriented Industrial Policy: The Role of Emerging Economies. Global Value Chains in a Changing World. D. K. Elms and P. Low. Geneva, World Trade Organization, Fung Global Institute and Temasek Foundation Centre for Trade & Negotiations: 329-360. Granovetter, M. (1985). "Economic Action and Social Structure: the Problem of Embeddedness." American Journal of Sociology 91: 481-510. Granovetter, M. S. (1973). "The strength of weak ties." American Journal of Sociology 78(6). Grimes, S. and Y. Sun, Y. (2016). "China’s evolving role in Apple’s global value chain." Area Development and Policy, 1(1): 94-112. Hobday, M. (1998). "Product complexity, innovation and industrial organisation." Research Policy 26: 689-710. Holgersson, M., C. Y. Baldwin, H. Chesbrough and M. L. A. M. Bogers (2022). "The Forces of Ecosystem Evolution." California Management Review 64(3): 5-23. Humphrey, J. and H. Schmitz (2002). "How does insertion in global value chains affect upgrading in industrial clusters?" Regional Studies 36(9): 1017-1027. Imai, K. and J. M. Shiu (2010). Value chain creation and reorganization: The growth path of China's mobile phone handset industry. The dynamics of local learning in global value chains: 33 MASSIVE MODULAR ECOSYSTEMS Experiences from East Asia. M. Kawakami and T. J. Sturgeon. Houndmills, UK, Palgrave Macmillan. Inomata, S. (2017). Analytical Frameworks for Global Value Chains: An Overview. Global Value Chain Development Report 2017. Jacobides, M. G., C. Cennamo and A. Gawer (2018). "Towards a theory of ecosystems." Strategic management journal 39: 2255–2276. JEDEC (undated). "JEDEC: Global Standards for the Microelectronics Industry." Retrieved November 26, 2024, from https://www.jedec.org/document_search/field_committees/25. Jones, C. (2018). "Apple Continues To Dominate The Smartphone Profit Pool." Forbes Markers, Mar 02, https://www.forbes.com/sites/chuckjones/2018/03/02/apple-continues-to-dominate-the- smartphone-profit-pool/ Kenney, M. and B. Pon (2011). “Structuring the smartphone industry: is the mobile internet OS platform the key?” Journal of Industry, Competition and Trade, 11: 239-261. Kenney, M. and J. Zysman (2016). "Rise of the platform economy." Issues in Science and Technology in Society 32(3): 61-69. Lamberg, J.-A., S. Lubinaitė, J. Ojala and H. Tikkanen (2021). "The curse of agility: The Nokia Corporation and the loss of market dominance in mobile phones, 2003–2013." Business History 63(4): 574-605. Langlois, R. N. (2006). "The secret life of mundane transaction costs." Organization studies 27(9): 1389-1410. LaPedus, M. (2005). "China gains in 2004 pure-play foundry rankings." EE Times. López, T., T. Riedler, H. Köhnen and M. Fütterer (2022). Digital Value Chain Restructuring and Labour Process Transformations in the Fast-Fashion Sector: Evidence from the Value Chains of Zara & H&M. Society of Socio-economics Annual Conference. Amsterdam. McGovern Institute for Brain Research (2025). "How nature organizes itself, from brain cells to ecosystems." MIT News, March 10. Accessed at: https://news.mit.edu/2025/how-nature- organizes-itself-from-brain-cells-to-ecosystems-0310 Meyer, M. H. and A. P. Lehnerd (1997). The Power of Product Platforms. New York, Free Press. Myrdal, G. (1974). "What is development?" Journal of Economic Issues, 8(4): 729-736. Normann, R. and R. Ramirez (1993). "From value chain to value constellation: designing interactive strategy." Harvard Business Review 71. Park, Y. and K. Ogawa (2009). "The impact of product architecture on supply chain integration: a case study of Nokia and Texas Instruments." International Journal Services and Operations Management 5(6): 787-798. Ponte, S. and T. Sturgeon (2013). "Explaining Governance in Global Value Chains: Towards a Broader Analytical Framework." Review of International Political Economy 21(1): 195-223. Powell, W. (1990). "Neither Market Nor Hierarchy: Network Forms of Organization Research." Organizational Behavior 12: 295-336. Prahalad, C. K. and G. Hamel (1999). The Core Competence of the Corporation. Knowledge and Strategy. M. H. Zack, Routledge. Prencipe, A., A. Davies and M. Hobday, eds. (2003). The Business of Systems Integration. Oxford, Oxford University Press. Raj-Reichert, G. (2018). The Changing Landscape of Contract Manufacturing in the Electronics Industry Global Value Chain. Upgrading and Innovation in Global Value Chains in Asia. Cambridge, Cambridge University Press. 34 MASSIVE MODULAR ECOSYSTEMS Rosenbloom, R. and C. M. Christensen (1994). "Technological Discontinuties, Organizational Capabilities, and Strategic Commitments." Industrial and Corporate Change 3(3): 655-685. Sanchez, R. and J. T. Mahoney (1996). "Modularity, flexibility, and knowledge management in product and organization design." Strategic management journal 17(S2): 63-76. Saxenian, A. (1996). Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge, MA, Harvard University Press. Scarcella, M. (2025). “Apple and Epic Games head back to court over App Store order.” Reuters, February 19. Schmitt, A. and J. Van Biesebroeck (2017). "In-house production versus specific forms of supplier governance: testing predictions of the global value chains model." International Journal of Automotive Technology and Management 17(1): 26-50. Schneidemesser, L. and F. Butollo (2022). Distribution-Centred Digital Transformation – New Business Models and Spatial Reorganization of Production and Work in the Mechanical Component Industry. Society of Socio-economics Annual Conference. Amsterdam. Schumpeter J. A. (1934, reprinted in 1962), The theory of economic development: An inquiry into profits, capital, credit, interest and the business cycle. Cambridge, MA: Harvard University Press. Schumpeter J. A. (1939). Business cycles. New York: McGraw-Hill. Scott, A. J. (1988). Metropolis: From the Division of Labor to Urban Form. Berkeley, University of California Press. Shapiro, C. and H. R. Varian (1999). "The art of standards wars." California Management Review 41(2): 8-32. Shiu, J.-M., M. P. Dallas and H.-H. Huang (2023). "A friend of a friend? Informal authority, social capital, and networks in telecommunications standard-setting organizations." Technological Forecasting and Social Change 189. Shiu, J. M., Dallas, M. P., & Lin, P. H. (2024). "Collaboration and social capital in meta- organisations: bonding or bridging?" Technology Analysis & Strategic Management: 1-13. Schultz, N., J. Wulf, R. Zarnekow, Q.T. and Nguyen (2011). "The new role of developers in the mobile ecosystem: An Apple and Google case study." In: IEEE 15th International Conference on Intelligence in Next Generation Networks. October: 103-108. Simon, H. A. (1962). "The Architecture of Complexity." Proceedings of the American Philosophical Society 106(6): 467-482. StatCounter (2024). "Market share of mobile operating systems worldwide from 2009 to 2024, by quarter." Statista, Technology & Telecommunications / Software. from https://www.statista.com/statistics/272698/global-market-share-held-by-mobile-operating- systems-since-2009/. Stinchcombe, A. (1965), "Social Structure and Organizations." In J. March ed.: Handbook of Organizations, Chicago: Rand McNally. Storper, M. (1995). "The Resurgence of Regional Economies, Ten Years Later: The Region as a Nexus of Untraded Interdependencies." European Urban and Regional Studies 2(3): 191-221. Sturgeon, T. J. (2002). "Modular production networks: a new American model of industrial organization." Industrial and Corporate Change 11(3): 451-496. Sturgeon, T. J. (2003). "What really goes on in Silicon Valley? Spatial clustering and dispersal in modular production networks." Journal of Economic Geography 3(2): 199-225. Sturgeon, T. J. (2019). "Upgrading strategies for the digital economy." Global Strategy Journal 11(1): 34-57. 35 MASSIVE MODULAR ECOSYSTEMS Sturgeon, T.J. (2022). Reconfiguration of Automotive Global Value Chains, Sectoral Study conducted for the Inter-American Development Bank (IDB) in support of the Strengthening Regional Value Chains in Latin America and the Caribbean Program under the coordination of the Regional Integration Unit of the Integration and Trade Sector of the IDB, September. Sturgeon, T. J. and J.R. Lee (2005). Industry Co-Evolution: Electronics Contract Manufacturing in North American and Taiwan. Global Taiwan: Building Competitive Strengths in a New International Economy. S. Berger and R. L. Lester. Armonk, NY, M.E. Sharpe. Sturgeon, T.J., J. Van Biesebroeck, and G. Gereffi (2008). “Value Chains, Networks, and Clusters: Reframing the Global Automotive Industry.” Journal of Economic Geography, 8: 297-321, 2008 special issue on Global Production Networks, edited by Neil Coe, Peter Dicken and Martin Hess. Sun, M. and E. Tse (2007). "When does the winner take all in two-sided markets?" Review of Network Economics, 6(1). Svitlo, S. (2009). "What is your review of Autodesk: Jon Peddie Research." Quora https://www.quora.com/What-is-your-review-of-Autodesk. Thomke, S. and D. Reinertsen (1998). "Agile Product Development: Managing Development Flexibility in Uncertain Environments." California Management Review 41(1). Thun, E. (2006). Changing lanes in China: Foreign direct investment, local governments, and auto sector development. New York, Cambridge University Press. Thun, E. and T. J. Sturgeon (2019). When global technology meets local standards: Reassessing China’s communications policy in the age of platform innovation. Policy, regulation and innovation in China’s electricity and telecom industries. L. Brandt and T. G. Rawski. Cambridge, Cambridge University Press. Thun, E., D. Taglioni, T. J. Sturgeon and M. P. Dallas (2022). Massive Modularity: Understanding Industry Organization in the Digital Age: The Case of Mobile Phone Handsets. Policy Research Working Paper 10164. Washington D.C., World Bank. Trendforce (2024). "Semiconductor foundries market revenue worldwide 2019-2023, by quarter." from https://www.statista.com/statistics/867210/worldwide-semiconductor-foundries-by- revenue/. Ulrich, K. (1995). "The role of product architecture in the manufacturing firm." Research Policy 24(3): 419-440. Van Alstyne, M. W., G. G. Parker and S. P. Choudary (2016). "Pipelines, Platforms, and the New Rules of Strategy." Varas, A., R. Varadarajan, J. Goodrich and F. Yinug (2021). Strenghtening the Global Semiconductor Supply Chain in an Uncertain Era, Boston Consulting Group and Semiconductor Industry Association. Weiss, D. G. (2019, December 29). "in4ma expanding market research." from https://www.linkedin.com/pulse/in4ma-expanding-market-research-dieter-g-weiss/. West, J. (2003). "How open is open enough?: Melding proprietary and open source platform strategies." Research Policy 32(7): 1259-1285. Wiegmann, P. M., H. J. de Vriesa and K. Blind (2017). "Multi-mode standardisation: A critical review and a research agenda." Research Policy 46: 1370–1386. Wu, D. and I. King (2022). “Shortage of legacy chips keeping Ford CEO up and night.” Bloomberg, November 18, Accessed at: https://www.bloomberg.com/news/articles/2022-11-18/shortage-of- legacy-chips-keeping-ford-ceo-up-at-night Xiong, W., D. D. Wu, and J. H. Y. Yeung (2024). “Semiconductor supply chain resilience and disruption: insights, mitigation, and future directions” International Journal of Production Research, DOI: 10.1080/00207543.2024.2387074. 36 MASSIVE MODULAR ECOSYSTEMS Yarrow, J. (2010). "The Most Important Tech Company You've Never Heard Of: ARM Holdings." Business Insider. Zylberberg, E. (2017). Industrial Policy Refraction: How Corporate Strategy Shapes Development Outcomes in Brazil. DPhil Thesis. Oxford, Saïd Business School, University of Oxford. 37 MASSIVE MODULAR ECOSYSTEMS Appendix A. Glossary of Key Terms as Used in the Paper This glossary provides a detailed overview of the key terms and concepts used in the paper, categorized into two distinct areas: ‘technical systems’ in Appendix Table A1, and ‘organizational systems’ in Appendix Table A2. Each category is further divided into ‘Abstract Concepts’ and ‘Concrete Objects’ or ‘Actor Roles’ to provide clarity on the definitions for both the theoretical ideas and for their practical applications. Appendix A1. Technical Systems, Including Abstract Concepts and Concrete Objects Category and Term Definition Example(s) Technical system An engineered system that combines A mobile phone, airplane, or functional subsystems to create petroleum refinery something of value (or, a value proposition) which may or may not be useful on its own Abstract Concepts Technical function An element within a larger technical Computer memory, jet engines, system that fulfills a specific purpose industrial pumps Interface standard Set criteria for connecting functions in Protocols for accessing computer a technical system memory, bolt patterns for jet engines, and standard pipe threads - Closed interface standard An interface with unpublished criteria IBM mainframe computer architecture where use is controlled by an actor, usually a prior to the 1956 consent decree requiring firm for internal use the company to publish its interface specifications (Morgan, 2014) - Open interface standard An interface with published criteria IBM mainframe computer architecture (also, industry standard interface) after the 1956 consentdecree requiring the company to publish its interface specifications - Proprietary standard An industry standard interface where Specifications for Intel and Qualcomm interface access is controlled by an actor in microprocessors; ARM’s instruction set exchange for remuneration or to achieve a architecturefor mobile computing strategic goal - Non-proprietary interface An industry standard interface with Standards set by 3GPP (for mobile standard published criteria that is free to use telecom), the Linux Foundation (for operating systems), the Bluetooth Special Interest Group of the IEEE Technical architecture A functionally coherent group of interface ARM’s instruction set architecture for standards that allows actors to create mobile devices; TSMC’s design complex technical systems requirements Concrete Objects Parts and components Simple elements in a technical system, A mobile phone case, an airplane tire, a often performing a single function and with length of industrial piping no stand-alone use (also, an intermediate good or service) Subsystem Complex, multi-function elements of a A mobile phone application processor, an technical system, often with no airplane avionics system, an industrial standalone use (also, a module) control system Final system A complete technical system that is useful A mobile phone, an airplane, a on its own (also, a final good or service) petroleum refinery Appendix A2: Glossary of Organizational Systems, Including Abstract Concepts and Concrete Actor Roles Category and Term Definition Example(s) Organizational system A set of bureaucratic and technical rules A private corporation and its groups, and routines by which the combined divisions, and subsidiaries; actors in an efforts of autonomous actors are able to industry adhering to shared standards and produce technical systems protocols 38 MASSIVE MODULAR ECOSYSTEMS Category and Term Definition Example(s) Abstract Concepts Complementarity The idea that autonomous actors or A display producer complements a keyboard components together create more value producer in a personal computer than they would independently (also, a complementarity surplus) Strong complementarity Complementary actors or subsystems When component or subsystem designed that are co-specialized and necessary and produced by a firm for internal use, or one for goals to be met (also, strong ties, produced by a captive supplier to a buyer’s asset specific, customized) specifications, or one aligned with a modular platform sponsor’s standards for complementors Weak complementarity Complementary actors or system When a component or subsystem designed subsystems that are not co-specialized and produced according to open industry and not necessary for goals to be met standard or platform interface, e.g. computer (also, weak ties, generic) disk drive, or mobile app Symmetric complementarity Complementary actors or subsystems When buyers and suppliers that are are equally dependent on the other to mutually dependent (such as a model-specific create a complementarity surplus customized auto part) or mutually independent (off-the-shelf bolts, wiring, or injection molding or other manufacturing services based on industry standards) Asymmetric complementarity One actor or set of actors or When a supplier dependent on a subsystems is essential to the value powerful buyer, such as manufacturer of proposition while another set is customized seat-belts for Tesla or App optional developer distributing through the Apple Store Industry A class of related products and services Mobile phones, automobiles, online video that share common actors, technologies, games standards, and end uses Globalvalue chain The sequence of activities, inter-firm At the product level: a mobile phone; at the linkages, and geographic locations firm level: Toyota and its suppliers; at the present as a product or class of industry level: the motor vehicle GVC, products move from conception to end comprised of the industry’s end users, lead use and recycling firms, and suppliers Business ecosystem An interdependent network of self- Includes actors and the norms, standards, and interested, autonomous actors jointly institutions required to create a product or creating something of value class of products (e.g. the Android mobile ecosystem) Sub-ecosystem A distinct functional ecosystem that Produces functional system elements, such must be linked to other ecosystems to as an automotive braking system or the create something of value mobile phone design, development, and sourcing activities of mobile phone brands Massive modular ecosystem A collection of autonomous yet Mobile phones and possibly other digitally- interdependent and complementary mediated industries such as cloud computing sub-ecosystems coordinated primarily and artificial intelligence chatbots. through open interface standards to produce a group of related products or services that fulfill a similar purpose Socioeconomic megastructure A set of joined-up ecosystems Massive modular ecosystems as a crossing industry boundaries with an type of socioeconomic megastructure identifiable prevailing governance structure Governance structure Rules and norms coordinating actors’ Coordination of economic activities by efforts and resources toward common means of internal corporate hierarchies, goals (also, alignment structure) captive inter-firm relationships, relational linkages, modular value chains, modular platforms, or markets 39 MASSIVE MODULAR ECOSYSTEMS Category and Term Definition Example(s) Prevailing governance structure The most common or important In the mobile phone system design and governance structure in a specific subsystem sourcing, most mobile phone sub-ecosystems as measured either companies have modular linkages to key by market share or number of firms input suppliers Alternative governance structure Governance structures in a specific Apple’s internalization of its iOS mobile sub-ecosystem that falls outside of operating system and A-series application the prevailing form processor design fall outside the norm in the mobile phone industry, as does it’s captive relationship with Foxconn for handset assembly Hierarchical governance Coordination of activities within one Apple proprietary mobile phone hardware (A- organization, controlled by series chips) and software (iOS) management Captive governance Co-specialization where less capable Toyota and a smaller supplier of seats suppliers depend heavily on buyers designed exclusively for specific Toyota car models Relational governance Proprietary, uncodified information Collaboration on novel jet fuel between jet exchange and knowledge generation engine producers and jet fuel refiners through joint activities between two or more firms Modular governance Information exchange across codified Dell laptops and desktops assembled using and standardized interfaces, with standardized components like Intel CPUs, exceptions expected because Samsung RAM, and Western Digital hard information is complex drives, all adhering to industry standards for compatibility. Modular platform governance Information exchange among users, Mobile apps meeting GooglePlay complementors, and platform sponsors requirements according to standards set by sponsors Modular value chain governance Information exchange between buyers Design information from GE Medical for a and suppliers codified according to open ultrasound circuit board transferred to a standards contract manufacturer according to industry standards such as GerberX2. Market governance Fully codified information exchange Generic inputs that can be fully specified based on price and transparent such as fasteners and transistors characteristics Concrete Actor Roles Buyer Firm purchasing inputs to a final system General motors purchases subsystems and or subsystem (also, lead firm) components such air conditioners, braking, electronic subsystems from suppliers Supplier Firm selling inputs or services to a Bosch, a global automotive supplier, sales of buyer automotive parts and subsystems such as sensors, electric motors, braking systems, and software to buyers such as Toyota and Volkswagen. Relational partner Parties to specific collaborative The collaboration between Toyota and partnership formed with the specific Panasonic to develop batteries for electric aim of developing new or synergistic vehicles technologies, products, processes, or business models Platform sponsor An organization or group of organizations The owners or promulgators of platform that set and maintain platform interface specifications can be companies such as Uber standards or Microsoft (Azure), standard setting organizations such as IEEE or 3GPP, or open- source projects such as Linux 40 MASSIVE MODULAR ECOSYSTEMS Category and Term Definition Example(s) Platform user Firm or individual accessing A startup hiring freelance software complements via a modular platform developers via Upwork to build a mobile app, a rider booking a driver through Uber, a data user accessing a country’s trade statistics on UN Comtrade Platform complementor Firm or individual producing A small business selling handmade jewelry complements for a modular platform on Etsy or a restaurant or 3rd party offering delivery services via Uber Eats Appendix B. Figure: A decision tree for identifying governance structures Appendix C. Data sources Estimates for the four key subsystems in Table 6 (application processors, core memory, displays, and 41 MASSIVE MODULAR ECOSYSTEMS RF modules) are derived from a proprietary dataset obtained from IHS Market. Our analysis utilizes 456 mobile phone teardowns, which consist of detailed lists of all electronic, electro-mechanical and mechanical components inside each mobile device from 2009 to 2020, averaging 38 phone models per year. For mobile phones, these data include a bill of materials (BOM) and a pricing model developed by IHS industry experts to calculate the component unit costs as well as the total manufacturing costs (hand assembly, auto insertion, test costs) of each device. Components are categorized by function (e.g. analog, baseband, apps processing, battery and power management, etc.), by which module it appears within (e.g. a printed circuit board, camera module, display module), and by component family and type (accessory, assembly, battery, electro-mechanical component, acoustics, etc.). For each component, IHS reports the dimensions of the package and – whenever possible – the name of component vendor. On average, vendor information is available for 21% of the components in each model. However, for the most important and higher-value components, the company data are nearly complete for all but a few items (e.g., camera sensors). For the list of 1,108 unique component vendors in the dataset, we determined the country of the vendor headquarters and – when relevant – the headquarters country of the parent company. Finally, it should be noted that IHS purposely oversamples flagship brands or product families (e.g., Apple iPhones, Samsung S-series, and Huawei Mate and Ascend brands, etc.) as well as mobile phones produced by new or market-gaining manufacturers and a smaller number mobile phones requested by clients. This means that the data are biased towards higher-end phones and are not representative of the underlying population of the total mobile phone market. We are not aware of a dataset with such high quality data that covers the full population of mobile phones over this period of time. Given the above limitations the authors decided to complement the IHS dataset with data from other main sources such as PhoneDB, and various market research estimates, company documents, and data aggregators. PhoneDB is an online data source that contains information on the technical specifications of over 15,000 phone models. This paper uses phone models between 2009 and 2020 to align with the IHS Market teardown dataset. Technical specifications include those listed in Table 5, in addition to others not reported in this paper. The remainder of the data in Table 6 is drawn from market research reports, company documents, and data aggregators such as Statista. These sources, while indicative of market trends, rely on varied methodologies and inconsistent definitions, and hence should be interpreted cautiously. Appendix D. How modularity emerged in the mobile phone industry Before 1990, mobile telephony was expensive, short-range, and relegated to first responders and other closed commercial networks. Large-scale installation of longer range ‘cellular’ networks that could be used by consumers ramped up in the 1990s.36 Levels of vertical integration in phones was much higher than it is today, and when components and subsystems were outsourced, semiconductor firms and other key suppliers typically had either relational or captive ties to lead firms (Park and Ogawa, 2009: 792). As Table 2 shows, modular governance currently prevails at the level of mobile phone system design and sourcing: modular platforms are used for system design without exception, and for key subsystems and manufacturing services sourced in modular value chains, with exceptions. How did such thoroughgoing modularity come about? One factor is the transition to digital technology. During this 36 These collected analog radio signals from local areas, converted them to digital signals that could be carried long distance between central offices handling wired telephony, and then converted back to analog radio signals that were routed to recipients. 42 MASSIVE MODULAR ECOSYSTEMS period, an underlying shift from analog to digital processing inside mobile phones began, with analog (radio) processing shunted from core processing components to a new “radio frequency” module responsible (along with the mobile phone’s antenna) for communicating with cell towers. To put a very complex set of technical issues in simple terms, analog signals (radio waves) are much more difficult to codify than the 1s and 0s of digital signals. As a result, early cellular phones had integrated product architectures, more or less unique to each device. With digitization, it became easier to codify more elements of the mobile phone’s technical system, and externally sourced components and software that were less co-specialized with mobile phone designs became more common. Because the market was growing, and more componentry could be used across many mobile phone designs and brands, more suppliers entered and grew along with the industry. As modularity increased, specialized semiconductor firms such as Texas Instruments (TI), Analog Devices, Lucent (all USA-headquartered), and Philips (the Netherlands) began bundling the complete stack of interconnect and analog-to-digital conversion protocols into highly functional technology platforms, or “chip sets,” making it easier for a broader range of mobile phone brands to enter the market with basic, low-cost mobile phones for mass markets. MediaTEK, a “fabless” semiconductor design firm based in Taiwan, China, went a step further when it began offering an integrated system-on- a-chip (SOC) for mobile phones in the mid-2000s, along with software and detailed instructions on how to implement them in a mobile phone design. These chip sets and associated “reference designs” covered not only how phones connected to the network, but also circuit board designs that included many of the new functions that were being added at the time, such as address books, audio playback, and image processing for basic cameras, as well as interfaces with higher capacity memory chips needed to accommodate burgeoning information storage needs. These highly integrated modular solutions expanded the market for mobile phones, especially in the price-sensitive Chinese market, but resulted in very similar phones, since they allowed phone designers little leeway for customization. Industry leaders such as Nokia and Motorola responded, on one hand, by deepening relational ties with key suppliers (TI in particular) to develop cutting-edge mobile phones closely aligned to the newest generation of telecom infrastructure, and on the other hand offering their own low-cost mobile phones (Imai and Shiu, 2010, Thun and Sturgeon, 2019). Chinese firms responded by trying to move up- market, setting off a classic “competing for the middle” dynamic as described by Brandt and Thun (2011). In the late 2000s, a less gradual and more dramatic shift toward modularity in mobile phones was led, not by an incumbent from the telecom industry, but by two companies from the world of personal computing and internet search applications: Apple and Google. The industry had struggled to introduce more modularity in software for several years. For example, the market leader, Finland’s Nokia introduced the Symbian operating system software in 1998. It was efficient and reliable, but lacked a standardized modular interface between the operating system and applications – what is referred to as an application programming interface (API) today. This created a barrier for third party app developers, truncating the network effects of the platform and therefore its scalability. Even within Nokia, almost every mobile phone was coupled with a different version of Symbian, and because there were dozens of versions that were not entirely compatible with each other, third-party app developers were frustrated by constant delays and uncertainty (Lamberg et al., 2021). Perhaps not surprisingly, given its origins as a telecom equipment company, Nokia’s focus was on matching software designs with appropriate application processor features in order to optimize network performance (e.g., improved voice clarity and fewer dropped calls) rather than focusing on making it easier for 3rd parties to offer new mobile phone applications. As a result, telecom operators were often relied upon to work with app developers to bundle applications with mobile phones – and in fact, the features available on mobile phones at the time were largely under the control of network operators. The lack of scaling motivated Nokia to develop a consortia including several other mobile phone 43 MASSIVE MODULAR ECOSYSTEMS producers (Samsung, Motorola, and Sony-Ericsson) in the Symbian ecosystem, but the challenge of assuring the compatibility of platforms and complements in an increasingly cost-competitive consumer- oriented market continued to be daunting, especially when spread across several mobile phone brands using different processors and product architectures. Furthermore, there was no centralized marketplace for Symbian applications where users could discover and purchase software. If the app was not bundled with the mobile phone, users had to purchase the software directly from each developer (Kenney and Pon, 2011). With the introduction of Apple’s first iPhone in 2007, applications were delinked from specific versions of the operating system, following a pattern set as it had been with the Macintosh and IBM-compatible PCs. But Apple went a step further by offering applications only over its on-line App Store. This introduced a modular platform to the mobile phone for the first time. As Kenney and Pon (2011, p. 248) write: The first significant new entrant to the emerging mobile Internet space was Apple, which in 2007 introduced the Apple iPhone. It catalyzed a shift in the architecture of the mobile phone industry and set the standard in two ways. First, it provided a fully functional web browser that allowed users to escape previous operator-specific silos of mobile content by linking them directly to the infinitely larger and more diverse Internet value networks. In effect, it collapsed the boundaries between the mobile device and the Internet with its enormous content. Secondly, leveraging the iPod ecosystem, but going far beyond it, the iPhone created a platform sufficiently open and attractive to create an ecosystem of application providers with 250,000 iPhone-specific applications that encouraged billions of downloads. Google’s proprietary distribution of the open-source Android operating system was released the following year along with Google’s own on-line application store, GooglePlay, setting in motion a duopoly that can still be seen today. Because Apple's iOS and Google’s distribution of Android were designed to work on a new class of touch screen mobile phones without fixed keypads, both internal and 3rd party app developers had a flexible palette where users could run a myriad of applications. As millions of apps became available and functionality skyrocketed, sales of touchscreen smartphones boomed, which in turn motivated more developers to create more apps, creating powerful network effects. By 2015 incumbent mobile phone firms and their proprietary OSs essentially disappeared from the market, except for Samsung, which managed to make the transition to Android. Appendix E. How the radio frequency subsystem works The radio frequency (RF) subsystem in mobile phones is composed of dozens of components packaged into one or more discrete modules.37 When the RF subsystem communicates with cell phone towers, the most basic function is to convert analog radio signals, emitted by cell towers and captured by the phone’s antennae, into digital data used by other subsystems, and vice-versa. RF modules contain many different types of components (e.g. tuners, filters, amplifiers) and also link to other components in the mobile phone, such as antennae. Given that the RF spectrum is scarce, national agencies (such as the US Federal Communications Commission) possess ultimate jurisdiction over which entities can use specific RF bands for what purpose (e.g., private sector vs. the military, wireless communication vs television broadcasting, etc.) but since there is a strong motivation for wireless devices to work – and for telecom infrastructure companies to sell equipment – in multiple geographies, many industry standards and spectrum allocations are negotiated through global standard setting organizations (SSOs) with the consent of national agencies. Since the 3G era, which started in 2001, the International Telecommunications Union, a 37 Because they serve as the radio interface with the outside world, RF modules are typically referred to as “front end” modules in the industry. 44 MASSIVE MODULAR ECOSYSTEMS United Nations body, has set the basic parameters and requirements of next generation standards through a collective standard setting process staffed voluntarily by engineers from participating firms. These requirements are then implemented by 3GPP, a voluntary private sector SSO that sets technical standards via a substantially collective process. Finally, these technical standards are converted into national standards and published by national regulatory agencies. RF modules are fabricated on different substrate materials, such as gallium arsenide, ceramics, silicon- on-insulate. Each of these has its own material science challenges and distinct knowledge domains, as does combining them in a single physical package along with silicon-based semiconductors. As mobile phones have become more complex and capable, for example gaining the ability to transmit more data across multiple telecom networks and network standards, the competency requirements have increased dramatically, especially with 4G and even more so with 5G. The evolution of the RF subsystem is different than fully digital subsystems, which have a relatively smooth evolutionary path from generation to generation, partly because they have historically been dominated by one material (CMOS silicon) and improvement roadmap: reducing feature size according to Moore’s Law every 16-18 months. By contrast, the evolution of RF technology comes in larger step changes that happen roughly once a decade and with no persistent technology roadmap. New standards are implemented gradually once they are established, but the process is dominated by sharply defined “generations.” For instance, a 3G frequency band of 2.1 GHz allowed mobile phones to transmit data at 2 Mbps, fast enough to browse the internet. A 4G frequency band of 2.6GHz, allows 40 Mbps speeds, fast enough to stream video. 5G signals represent an even larger leap, to a frequency bands ranging from 24 to 100 GHz, supporting speeds up to 10 Gbps, enough for demanding low latency applications like self-driving cars. Because of this, not every RF module supplier has been able to stay apace with these technology leaps, leading many to exit. Today, only a few companies remain in this space and moreover, most rely on each other for specific components and processes. During the 2G era (starting in 1991), there were several dozen major suppliers. By the start of the 2020s, there were only five major firms left – each with some capabilities across all domains, but usually a leader only in one component or even sub-component.38 As such, to make a complete RF subsystem, several, if not all of the five major companies regularly utilize their competitors’ components and advanced design and manufacturing capabilities. Most recently, and after a decade-long struggle, Qualcomm has been able to gain traction with its more comprehensive RF Front End (RFFE) product family. Of course, mobile phones utilize many other wireless RF subsystems, such as WIFI, Bluetooth, nearfield communication, in addition to satellite connectivity, among others, each utilizing their own spectrum that is usually governed through other global standard setting organizations. However, while the standard-setting processes persist, these subsystems have gradually been digitized and encapsulated into the application processor. 38 For example, Japan’s Murata specializes in low frequency SAW filters, while the USA’s Broadcom specializes in higher frequency BAW filters. 45