Takeaways & Observations
Key Features of the Consolidation of the Internet Economy
This section looks more closely at some of the features of the consolidation of the Internet economy. We uncovered 5 key features during the course of the research, which raised a broad span of issues, including how consolidation might shape the user experience, its impact on innovation, to how governmental or other stakeholders might respond to it. While not exhaustive, the list of considerations below helps us understand how consolidation trends in the future might impact the Internet and its users.
- Many Internet platforms are expanding into new service and content areas, both to retain customers and to continue to grow revenues.
- Total service environments for business and innovators operate at a scale that will allow entrepreneurs to do things they couldn’t otherwise. This includes accessing a much larger customer base, as well as resources and expertise that no small business could tap on its own with limited resources or time.
- Platform environments unleash huge opportunities, but because innovation and entrepreneurship remain within the platform, competition to the dominant players will effectively be neutered.
Total service environments; the convenience of a one-stop shop
While some companies dominate their original markets such as search, e-commerce, and messaging, their increasingly diversified service offerings are overlapping and starting to compete in other markets. Search-oriented Google, for example, is trying to capture more of e-commerce, as is Facebook. The consumer-to-consumer Facebook Marketplace competes with eBay. Similarly, Google and Amazon compete for the smart home and smart device market, along with Samsung and Apple.
At the same time, how people spend their time when using the Internet is evolving. For instance, in some regions, young people may be spending less time on social media and more on and gaming platforms.1 As they expand their reach and services, many platforms will compete with each other more directly, trying to leverage their dominance in one service or application space to quickly build a presence in another.
In content provision, as it is broadly defined, Internet platforms and telecommunications and media players such as AT&T/Time Warner are likely to work hard to retain and grow their customer bases, offering increasingly tailored content and services to meet their users’ demands for content. Many Internet platforms compete with each other and with the traditional media players: from Google’s YouTube, to Amazon’s Prime and studio offerings, to Facebook’s Facebook Watch, Disney, Oath, and Netflix, alongside telecommunications/content licensing deals such as India’s Eros and China’s iQIYI,2 Alibaba’s Tmall Box Office,3 Safaricom’s Iflix partnership,4 and numerous Over-the-Top (OTT) Video on Demand partnerships all over the world.
In China, the Alibaba consumer-focused financial spin-off Ant does everything from mobile payments to life insurance, credit, investments, and savings. For many of its users, it has become a financial one-stop shop.5 WeChat, the original “app for everything,” focuses on “everything from point-of-sale purchases to accessing public services – (and) is likely the template that other social platforms around the world will emulate as they strive for more thorough integration with their users’ lives.”6 WeChat is owned by Tencent, whose platforms, which include WeChat/Weixin, QQ, Qzone, Tencent Video, and Tencent News, capture more than 55% of Chinese consumers’ mobile Internet usage.7 Tencent provides a one-stop shop with everything from “social networking to gaming, digital assistants, mobile payments, cloud storage, education, live streaming, sports, movies, and artificial intelligence”.8
With the evolution and proliferation of IoT and technologies like voice-controlled devices, these total service environments are intertwining more with our physical lives. Our homes, cars, roads, hospitals, and more will form greater connected ecosystems as ubiquitous computing and networking evolve. Users may choose the platform’s own offerings over niche providers because one-stop shops can provide a better and more convenient service, driven by scale and pooled resources.
Choice by default
The success of a total service environment is the consequence of the ability to provide a range of services that users believe or know they need. Much of this success is driven by convenience and by providing easy access to useful and targeted products and services. Platforms grow ever more “sticky, understood as their ability to retain users, with AI-tailored hyper-personalised services and business strategies that seek to capture the users of today and tomorrow.
Platform stickiness is driven by the idea of default: users often stay with a search engine, social media platform, or cloud-based business services because it’s convenient or because of necessity. Becoming a default platform increases the potential of retaining users for longer periods of time, which means the platform can gather ever more data to be used for targeted advertisement and further improving the service. This path towards becoming the default provider occurs through different ways, and largely by design:
- Default options on devices or in software might not seem difficult to change for many Internet users, but many users will refrain from taking the extra effort needed to figure out how to switch. But remaining with a particular platform reinforces that platform’s dominance. Google is the default search for Apple’s Safari, Mozilla’s Firefox (replacing Yahoo), and the Siri smart assistant. Google is also the default search service on the Android mobile operating system and platform. Google developed Android and it is now used in more than 86% of mobile phones around the world.9 Applications having Google as the default search engine reinforces its dominance and drives its advertising revenues.
- The popularity of smart voice assistant devices means there’s less of a need for people to use a smartphone or computer or use a particular service to buy something online. One can simply speak out loud and the service-provider’s fulfilment centre instantly takes an order. If users don’t adjust the settings and ask for a specific music service on an Amazon Echo, for example, it will automatically use Amazon’s own music service.10
This strategy of default and nudging users to one’s own products is not new, but it illustrates the power inherent in controlling the platform through which other services and products are accessed. With the evolution of more personalised services powered by personal data that can anticipate our needs, this power of default will become even more important. For the user, it can have great advantages, but it can also impact competition and choice in ways that will become more pronounced in the future.
As the platforms expand, they acquire companies that might have become a future competitor: the largest US tech platforms have acquired hundreds of start-ups in the last 10 years.11 Customers moving from one service to another, although somewhat different, may merely be moving to a differently branded service of the same platform:
- Customers of Facebook’s Messenger may decide to move to the instant messaging app WhatsApp and younger people are leaving Facebook — or just not joining it— in some regions, and using Instagram instead.12 But Facebook owns both WhatsApp and Instagram. Similarly, China’s Tencent owns both the original QQ social network, Qzone, and the WeChat platform.13
- In the world of online travel agents, choice is more apparent than real. Priceline, a $98 billion USD company, owns Booking.com, Kayak.com, and RentalCars.com. Expedia ($23 billion USD) owns Hotels.com, Travelocity, and Orbitz.14
- Big platforms also use their financial power to expand geographically through acquisition. Alibaba acquired Indonesian online retailers, Tokopedia and Lazada,15 and Amazon purchased in 2017 the biggest online retailer in the Middle East, souq.com.16
These types of acquisitions have led some tech investors to claim there is a kill zone around the top 5 American technology platforms. This kill zone lessens the incentives for investors to look for potential competitors to the dominant platforms, undermines competition generally, and leads to the ossification of the market through the perpetuation of the one-stop shop and default service. Ultimately, the kill zone enables the continued dominance of the current players.17
Total service environments are becoming de facto platforms for new and evolving businesses
Total service environments extend both vertically and horizontally to serve the ever-growing range of needs of business users and technological innovators. Amazon provides cloud services through its Amazon Web Services (AWS), as does Microsoft through Microsoft Azure. Google extends through its own cloud offerings and content delivery network services (CDNs). Beyond servicing large companies and government agencies, these services offer small and emerging businesses and innovators compelling access to infrastructure and expertise, along with functionality.
The ambition of the large Internet platforms does not stop there. Through digital assistants, drones, connected cars, smart homes, and smart cities – all underpinned by increasingly sophisticated AI — they are building their capacity across multiple domains, securing the footprint and developing a range of services to ensure that they remain the preferred total service environments of the future.
Platforms strive to keep innovation within their services ecosystems. The recent product launches from Amazon at their re:Invent 2018 conference18 provide several examples of this dynamic at work. The larger platforms recognise that capturing innovation within their platforms not only drives their own functionality and popularity, but also identifies and channels innovations that have the potential to be disruptive and competitive. While platform driven APIs and SDKs provide great opportunity and access to markets for developers, they also help ensure that future content, gaming, and other services remain mostly within the platform.
The trend towards a total service environment is an attractive choice for many businesses and innovators. The savings and operational benefits that accompany the services of a single provider, from productivity suites to cloud services and the ability to outsource business infrastructure and services, are considerable. But at the same time, outsourcing to a single provider builds a dependency on the platform offering the services, creating a potential lock-in and vulnerability to business and systems-wide cyber threats. A recent debate around whether the US Department of Defense should use a single cloud services provider highlights concerns related to adopting a single-vendor solution.19
AI, the competitive frontier
AI will be an important element for online businesses in the next decade, including data-driven platforms’ ability to attract and retain users, as it enables platforms to provide users and business customers with more value by continuously tailoring responsive products and services. Scale – and related access to data and resources – may be key to unlocking AI and maintaining platform “stickiness” to keep users in one-stop shops indefinitely. Examples of how AI is currently being pursued to attain a competitive edge include:
- Google, Amazon, and other large companies are acquiring AI start-ups20 and are competing to recruit new talented researchers.21
- Big cloud platforms like AWS, Google, Microsoft, and Alibaba readily offer AI capabilities to developers, which may also tie future innovation to those platforms.22
- China’s Ministry of Science and Technology has set out a detailed national strategy23 and identified its 3 consumer-focused tech giants, Baidu, Alibaba, and Tencent, as the national team to dominate AI globally by 2030.24
Are one-stop shops inevitable?
Most users can benefit from the tremendous convenience, versatility, and deep wealth of offerings of the one-stop shops. It is usually much easier to navigate within an environment of services provided by 1 player than to find everything we want or need independently. With the move to voice-controlled devices our interactions will become even smoother. That said, one-stop shops could also constrain the potential for new market entrants, limiting the opportunity and impact of disruptive innovation outside of the proprietary platform environments.
A question arises at this stage: Is the one-stop shop inevitable, or is there a path towards an Internet economy in which there’s a greater diversity of players that are more competitive and innovative?
- The nature of an open, collaborative, and interoperable Internet is influenced by a small number of large companies, where organisational scale and market share play a significant role in the development and deployment of the open technical standards on which the Internet depends.
- Large organisations can accelerate the adoption of existing but under-deployed standards like IPv6 and push the development and testing of new standards, benefitting the Internet as a whole.
- The growing use of APIs puts more of the Internet’s innovation, functionality, and interoperability into the hands of the dominant Internet platforms, whose interests may not always align with those of the broader technical community and other players.
- Future innovation, services, and applications may depend on the availability of a small set of proprietary platforms and services, rendering those applications less resilient, reliable, and capable of supporting further innovation.
A dumb network
The Internet’s architecture is unusual because of the nature of “inter-nets”: they are networks of networks, and the Internet spans the whole world. Because each network operates autonomously, (hence “Autonomous System”), there is no central authority to determine what a network might or might not do. But to make each node in each network operate tolerably with other nodes (nodes that might be in other networks), we need common protocols. It emerged that the best way to operate in such an environment is to place as much functionality as close as is practical to the “end” of communication;25 sometimes called the “end-to-end” architecture.
It is also sometimes called the smart endpoint design, and distinguished from the smart network design typical of the traditional telephone network: in the former, most of the ability to work with the network resides at the endpoint (with the network basically just carrying messages). In the latter, the ability to work with the rest of the network resides in the middle of the network, and the end points are dumb clients, which means that their functionality is fully dependent on connecting to a central server. The end-to-end architecture is one of the key enablers of the development of the Internet economy over the past 3 decades.
Another feature of the Internet’s technical architecture is its independence from pre-existing contractual relationships among all parties. Any party using the Internet can talk to any other party, without permission from intermediary networks, as long as each conforms to the relevant protocols for the network’s intercommunication. Combined with a smart endpoint design, this architecture enables permissionless innovation, because nobody else on the Internet needs to change anything in order that 2 end points use some new network application.
Just as shared languages enable people to cooperate, build communities and exchange ideas, the use of open and globally interoperable protocols has enabled the Internet to grow into the network of networks it is today.
However, the impact of a consolidating Internet economy on the development and deployment of protocols for interoperability indicates that scale is not just a source of efficiency, but also a source of power.
84% of our survey respondents identifying as “Technical Community” had noticed larger players in the Internet Economy increasingly influencing standardization for networking, technology, software, and interoperability.
Scale drives the adoption of standards
Open standards are the basis of the Internet’s success, but adoption requires investment, which can result in sluggish take-up. While many agree that IPv6 is important and beneficial to the Internet as a whole, individual organisations have hesitated to invest in the hardware, technical expertise, and possible changes to network operations that are required. Without a significant deployment of IPv6 in the Internet overall, there have been fewer incentives to adopt it because the benefits were seemingly outweighed by the costs of dual-stack deployment.26
Large organisations, including network operators and content providers, have addressed this challenge, adopting IPv6 in a coordinated way.27 Since the World IPv6 Launch began in mid-2012, IPv6-enabled connections from Internet users to Google services have increased from less than 1% to more than 25%.28 While it is difficult to determine exactly how much of that growth was stimulated by the intentional movement of very large organisations, or whether growth was inevitable due to IPv4 shortages, the effect has been significant. Since at least mid-2012, several major operators now deliver the majority of traffic from major content sources like Google, Akamai, and others over IPv6.29
In addition to directly adopting a particular standard themselves, big players can also nudge others to adopt standards that benefit the Internet as a whole. Google displays a “Not Secure” warning in the Chrome browser’s address bar if a user connects to an unencrypted website, incentivizing websites to use HTTPS30. In a similar push to service providers to do the right thing, Apple requires all apps on its iOS AppStore to be IPv6 compatible.
Scale and market share therefore imply that when big players in the Internet economy decide to support a new standard, they can accelerate and spread its adoption and ultimately change the incentives for others to do the same.
Scale drives standards development, too
In a network of networks, there is no centre of control because there is no centre: each network does what it wishes according to local needs. As such, there is no way to impose a common standard. It is only possible to invite participation in a standard protocol according to the wants and needs of others. The inter-network’s participants therefore use an open standards development process to reach consensus on communication protocols.31 Because expertise and resources are necessary for people to engage effectively in standards processes, large corporations, universities, and other influential organisations have always played prominent roles. Competitive markets that exist outside the standards process has generally meant that the interest of particular businesses in influencing particular protocols was offset by rivals. But just as scale is important for deploying standards, it is also an important and growing factor in developing standards, too.
HTTP/3 is a new protocol based on QUIC, a protocol initially developed by Google, and now under consideration for standardisation by the IETF. HTTP/3 reduces the time for establishing a connection for IP traffic without compromising security, and has the potential to benefit the Internet as a whole. Google’s role in originating the protocol was enabled by its significant resources and expertise. The company’s enormous size allowed it to test QUIC on a global scale using its browser, Chrome, and its servers for popular services including YouTube. The fact that it controls both the browser platform and the service being delivered is a tremendous benefit in analyzing Internet usage and experimenting, developing, testing, and deploying potential improvements. However, it is also an example of how its significant browser and content provisioning market share provides it with unique resources that few can match.
An example of how 2 companies could change the dynamic of a bigger technical ecosystem is “DNS over HTTPS” (DoH). Mozilla is championing DoH, a protocol for doing DNS resolution via HTTPS32, and testing it in partnership with Cloudflare.33 DNS data can be sensitive from a privacy perspective and DoH seeks to improve user privacy by encrypting DNS requests using HTTPS. On the one hand, this development appears increase the privacy of user data. On the other, it would be possible for a web browser to select a single DoH provider as the default option to handle all requests. In that case, it would concentrate a great deal of information about that browser’s users under the control of a single DoH provider. This might happen without user intervention, and if it happened in a web browser with large market penetration it could change the effective privacy properties of a large fraction of global DNS requests, while changing the trust model of the DNS itself.
APIs can operate less like the Internet and more like a monolithic system
New services and applications on the Internet increasingly depend on their ability to interoperate with third-party services made available through an application programming interface, or API. A public API is an interface to functionality of a third-party service. APIs intended for use in this fashion are publicly specified, (e.g., the Google Street View API34).
In some ways, network protocols and APIs are similar, but there is an important difference. A protocol must be agreed to by everyone who uses it. An API is under the control of the entity that owns the “A”: the application with which one interfaces.
On the Internet, APIs have implications for change control. In an open standard, the interested community has a say about changes to the interface defined by the standard. APIs can be made publicly available, (e.g., by publishing the specifications for use), but unilateral control of changes always remains in the hands of the application owner who publishes the API. Greater dependence on an API for Internet-facing functions may therefore entail greater control by the API owner as to what is possible on the Internet, as well as a corresponding reduction in the reliance on open protocols.
Will new protocols, standards, or practices championed by large organisations have positive effects for all or only some?
New protocols, standards, or practices championed by influential organisations could have significant effects. Some impacts could be positive, such as overcoming market failure in adoption and the ability to widely test protocols before full deployment. But others may be negative. This evolution might benefit small numbers of players or concentrate traffic flows, with possible adverse effects on competition or even privacy. While the Internet may benefit from specific developments, the ability of small numbers of huge organisations to be benevolent rule-makers while everyone else is a rule-taker could concentrate power on or over the Internet.
Interoperability has been fundamental to how the Internet has evolved and why its many uses and innovations have flourished. But while the Internet will continue to enable new and interoperable services and applications in the future, innovation might be concentrated where value is readily available, on a small set of proprietary platforms. Are users locked into 1 or 2 platforms really benefiting from all the opportunities the Internet has to offer, or are they merely using the Internet for transport?
This evolution raises several questions: Will new protocols, standards, or practices championed by especially large organisations have positive effects for all or only some? How will concentration affect the development of standard and non-standard protocols on the Internet?
The Internet was envisioned as a decentralised network which facilitates communication between two endpoints. However, a discussant pointed out that the landscape of network traffic is wholly different now: video streaming makes up a significant chunk of the traffic now, with Cisco estimating the 80% of the Internet Protocol traffic will be video by 2021.96 Since video hosting is dominated by a few companies such as Youtube, Netflix, and Facebook, there has been an emergence of “super-nodes” in the Internet. This has been followed by a concerted effort to be efficient at meeting such consumers’ needs, which further contributes to changing the network topology. For example, several companies employ edge caches at various Internet Service Providers (ISPs) to optimise user performance. — Regional panel report, The Centre for Internet and Society, India
- The ability of a small number of content and cloud services to invest in their own networks and deploy their servers close to the broadband network edge is amplifying the existing trend of a flattening Internet, where access networks are increasingly interconnected and have less need for international transit.
- Access networks are evolving rapidly, driven by the Internet of Things (IoT) deployments and other demands for processing on a range of user devices, including evolving technologies such as autonomous vehicles.
- Big cloud providers like Amazon, Alibaba, Google, and Microsoft are well-placed to dominate the new era of IoT and edge computing, further driving a changed Internet topology with less international transit and more complex, private, specialised networks and services. This could come at the cost of the general-purpose Internet.
A flattening Internet
As access networks increase their interconnectedness and the need for transit decreases, the Internet’s topology is flattening. Geoff Huston, Chief Scientist at APNIC,35 calls this "the death of transit."36 He is describing how the current, denser mesh of access networks is driven by Internet Exchange Points (IXPs) that facilitate peering arrangements, as well as the ability of access networks to peer directly with content delivery networks (CDNs). This flattening will continue as activity to develop access networks intensifies to facilitate IoT deployments and the processing of real-time applications.
Flattening is not an inherent consequence of consolidation trends, but is strengthened and shaped by a small group of content and cloud providers with the ability to deploy content delivery caches globally. Caches, in turn, become more effective and useful by the increasing numbers of IXPs, which tend to interconnect access networks and concentrate the points where transit needs to be provided.
67% of our survey respondents expect consolidation to significantly shape the evolution of the Internet in the next five years.
A global network of data centres
As we saw in the previous section, large content and cloud providers are investing in their own international connectivity infrastructure to serve their networks of data centres and servers across the globe. TeleGeography estimates that the capacity deployed by a small set of content providers is growing faster than all other customers of international bandwidth and may become the largest share of used international bandwidth as Internet backbone providers’ share decreases.37
Underpinning the trend of content and cloud providers investing in their own infrastructure is the dominance of a small set of providers in the application layer, predominantly Google, Facebook, Amazon, and Microsoft. Their services rely on a global network of interconnected data centres and servers to process and deliver content closer to the users.38 With the exception of Facebook, these providers are also competing in the market for cloud computing. Having a large customer base in the application layer supports investment in infrastructure to serve both those customers and users of service infrastructure.
Infrastructure investments by the large platforms are being driven by ever-increasing traffic between their data centres as cloud-based applications and services continue to grow. Cisco estimates that global traffic between data centres will grow by 32.7% annually between 2016 and 2021, a higher growth rate than the traffic between data centres and the users, which is projected at 25.2%.39 This trend has been driven by CDNs and the need to disseminate large volumes of static content closer to the user, such as images and video. In the near future, however, evolving services where processing moves closer to the end user or device will likely dominate traffic growth.
In the digital networking world, we are seeing more and more data traffic go ‘dark’. Content service operators are using their own transmission systems or slicing out entire wavelengths from the physical cable plant. This withdrawal of traffic from the shared public communications platform is now not only commonplace, but the limited visibility we have into this activity suggests that even today the private network traffic vastly overwhelms the volume of traffic on the public Internet. And the growth trends in the private data realm also is far greater than growth rates in the public Internet. — Geoff Huston, RIPE NCC, December 2018
An evolving edge
Access networks and access devices — the edge of the Internet — evolve rapidly, with many and varied devices connecting to new services, potentially using specialised networks, driven in large part by the IoT. 5G cellular network standardisation and deployment are partially driven by anticipated uses that depend on bespoke access networks with much greater processing capabilities in base stations close to mobile terminals. It is not yet clear how much of the 5G vision will be realised in practice, but it is at least possible that 5G deployments will increase the tendency for access network specialisation and the growth of edge computing. Of course, neither of these tendencies form part of an Internet approach to networking.
Some processing is better done on the device or at data centres closer to the user. Facial recognition in Apple’s recent iPhone X is processed on the device and not performed in the cloud.40 Some applications dependent on AI and machine learning need to be trained on data from local environments. In most cases, however, speed is a key concern. Real-time applications such as self-driving cars cannot afford the time it takes to have decisions processed in the cloud.
The evolution of edge computing is not a substitute for cloud computing in large data-centres, but rather a complement that facilitates the needs of some applications to use both types of computing resources. While some applications powered by AI need edge processing to make inferences on the device, they still rely on data models trained with the cloud’s abundant processing power and enormous amounts of data.
While the evolving edge of the Internet is a new frontier that could see new entrants, large cloud providers like Amazon, Alibaba, Google, and Microsoft may be in a favourable position to dominate the new era of IoT and edge computing. Ownership of the underlying infrastructure is one advantage, as is provisioning the platforms for the development of new applications. Microsoft’s new slogan, “intelligent cloud, intelligent edge”41 is an attempt to capture this trend. Services like Amazon Greengrass,42 or Google Cloud IoT Edge,43 are all intended to support developers to deploy new applications that require processing in access networks. Google has even launched a new chip, the Edge TPU, which is specifically designed to optimise machine-learning inference on the edge computing device itself.44
Will the deployment of new infrastructure support the general-purpose Internet for everyone?
The trend of localizing content, and the growing complexity of specialised networks and purpose-built services, will drive the Internet topology towards a more densely connected mesh. This trend towards a more densely connected network has in the past been positive in how it improved the delivery of content, and reduced overall traffic costs. But we must consider how the continued flattening of the Internet disrupts economic models of the past.
Several questions arise: Will traffic in access networks be driven by large content and cloud providers feeding their localised infrastructures via private networks? Does this mean that a smaller player, depending on transit and therefore subject to longer latency, will necessarily lose out? Will the deployment of new infrastructure support the general-purpose Internet for everyone, or could the Internet fade into the background as the evolving edge is captured by a small set of private networks and services designed for a few operators?
61% of our survey respondents said consolidation is likely to impact the scale and severity of cyberattacks and cybercrime in the next five years.
- Central to the Internet’s success has been its ability to evolve, as captured by the characteristic of “no permanent favorites” where no specific company or technology is above disruption, evolution and competition. This characteristic could be challenged as dependencies continue to grow.
- The development of new applications, services, and businesses across the global economy is increasingly dependent on a small number of private platforms owned by the largest Internet companies. While the risk of catastrophic failure may be very low, consequences of failure could create a domino effect for other parts of the global economy.
- As platform environments expand further, entering and often dominating more sectors and markets, combined with an exceptional economic power, there is a risk of growing societal dependencies on a handful of powerful economic actors.
The largest Internet platforms are expanding horizontally from dominance in their core areas — search, social, operating systems, e-commerce — into new economic sectors like health, automobiles, or even urban infrastructure. While the success of such horizontal expansion remains to be seen, these companies are also expanding vertically, through the application, infrastructure, and access domains of the Internet economy.
Governments rely on these companies for service procurement, and at least one, Alphabet, has a stated goal to improve traditionally public services.45 If this trend continues, future societies may become dependent on a small number of companies providing much of the backbone for daily life and becoming too big to fail.
“Too big to fail” is what happens when the potential consequences of a single organisation’s failure are so severe they create systemic risk. In one example context, it could mean risk to the economic system as a whole. This is one of the reasons why some governments stepped in to bail out or guarantee failing banks during the global financial crisis in 2007 and 2008. Could a few Internet companies become so central to how we do business, deliver public services, and sustain social and political life that the failure of one could create cascading effects throughout the economy as a whole?
Cascading effects would likely only occur if the platforms’ services were not substitutable. This would mean that when the service failed, those depending on it were not able to find and implement a replacement. While the likelihood of such a crisis is speculative, its potential consequences are vast and unpredictable. Will our growing dependence on the largest Internet companies motivate a wider concern to ensure they survive, no matter what? This could unintentionally hamper the continuing technological and business innovation the Internet supports, effectively creating permanent favourites.
A platform’s SDKs and APIs, which provide access to functionality of a service or platform, are increasingly important building blocks for technology and business innovation. APIs are often used by an application that, in turn, is used by another application, and so on. If an API fails, changes its conditions, or shuts down, it will likely harm all dependent businesses built on top of it. The resultant harm could range from negligible (such as when a service is easily substituted) to severe or critical (such as when a service is otherwise completely unavailable or when the loss of the service does not completely break another service).
For example, the ride-hailing service, Uber, has been using Google Maps’ API to connect drivers and customers to routes and directions.46 This integration was a critical part of Uber’s success, but it meant the company’s core activity depended on access to another firm’s API. The conditions for that access could change47 at any moment and abruptly alter or even hamper a company’s ability to operate. Uber’s investments and acquisitions of mapping technology is seen by some as a way to reduce the company’s dependency on the Google Maps and other APIs for its services.48
The same type of dependency can be seen in applications developed in the cloud Platform as a Service (PaaS) market. PaaS offers an easily accessible software environment for operating systems and middleware that can include critical business functions like database management or specific AI/ML capabilities. If providers change the conditions, greatly increase the prices, impose unacceptable conditions, decide to abruptly terminate services, or simply fail to continue providing the functionality their client-businesses depend on, these businesses may fail and create a further cascade of failure elsewhere in the economy.
Service failures happen all the time. The core concern is not that of vertical integrated providers, which is expected in layered technologies, but substitutability, which becomes an issue when a service is proprietary such that alternatives cannot be had at any price. If a single provider has several such popular services that cannot be substituted, then it becomes critical to a broader range of economic activities, and may be too big to fail.49
Governments may use single providers
Government departments and agencies may contract with large companies who can act as single providers of important systems. This allows governments to adopt the most up-to-date functionalities that citizens expect. The scale required often means only the biggest providers are suitable. For instance, cloud-based services from single providers like Google’s G Suite50 or Microsoft 36551 are targeted to governments looking for broad and flexible solutions.
The U.S. Department of Defense has been criticised for considering use of a single provider, likely Amazon Web Services (AWS), to move its data to the cloud.52 This could create critical dependencies. While technical responses may mitigate the associated risks involved, and private sector provision of services to government is hardly new, it raises important questions about how much public institutions should become dependent on a small set of powerful actors in the Internet economy.
Could our increasing reliance on just a few companies in the Internet economy create critical dependencies, or make them too big to fail?
An important feature of the Internet’s success has been its ability to evolve, where no specific company or technology is above disruption, evolution, and competition. As the largest companies establish dominance in more markets and sectors of the economy, our dependence on them is growing, raising questions about what extent this feature still holds true. The reality is that some of the largest Internet companies are already more valuable than most governments on earth.53 They are seemingly becoming the central locus of new innovations, services, and businesses across the global economy.
As we consider the future, several questions arise: Could our increasing reliance on just a few companies in the Internet economy make them too big to fail? Are there economic and technical dependencies on services that cannot be substituted that effectively create a set of permanent favourites?
The information superhighway cracks apart more easily when so much of it depends on privately owned infrastructure. An error at Amazon Web Services created losses of service across the web in 2017; a storm disrupting a data center in Northern Virginia created similar failures in 2012. — The New York Times, October 2018
- Responses to the negative effects of consolidation have been distributed across the applications, services, and access domains of the Internet economy; in different sectors, regions, and by different institutions.
- Countries have adopted different strategies. Some have a higher tolerance for the risk of dominance if it also delivers Internet access and services, and they may have a traditionally lower tendency to regulate. Others, such as those in the European Union, are mobilizing more concerted, cross-agency responses (often spanning competition, consumer protection, and data protection regulators).
- Similarly, different stakeholder groups have also focused on different issues as far as dominance is concerned.
Examples of government responses
As awareness grows of the potential economic, developmental, and social benefits of the Internet, as well as its challenges, governments around the world have taken increasingly interventionist, but also diverse, stances to dominant actors in the Internet economy. Some politicians publicly denounce the power of large companies or threaten to regulate, like U.S. President Donald Trump, who has accused Google’s search engine of ideological bias,54 or French President Emanuel Macron, who has warned that Americans may come to see “Big Tech” as “not just too big to fail, but too big to be governed”.55
Besides making controversial public statements, governments are increasingly asserting their authority to compel global tech companies to adhere to local laws. Examples of government initiatives in this field, with a variety of justifications, include:
- Data localisation laws (seen in Russia, Indonesia, Vietnam56)
- Data protection frameworks (e.g., General Data Protection Regulation, or GDPR, in the European Economic Area; the African Union’s Convention on Cyber Security and Personal Data Protection; and France, Germany, and the UK’s enforcement actions and investigations into data protection57)
- “Digital protectionism”58 and taking steps towards and state-encouraged consolidation (e.g., China59)
- Challenging alleged tax avoidance (e.g., the European Commission, Apple, and Ireland60)
- Investigating and/or imposing taxes and content restrictions to protect the revenues of incumbent operators (e.g., Zambia,61 Uganda,62 and Tanzania63)
55.5% of our survey respondents said consolidation is likely to trigger a response in government policy and regulation over the next five years.
Other countries, like the UK, have taken a broader approach by constituting a panel of experts to investigate and consider potential responses to digital dominance.64 The overall trend where governmental responses are concerned, however, is of much more interventionist state action to compel platforms to fulfill a variety of public objectives on competition, taxation, data protection, consumer protection, and content regulation and plurality. This trend is driven partly by consolidation, but also by the sheer size and rapidly growing influence of platforms in many countries around the world:
- The EU has opened competition cases and taken enforcement action against U.S. tech giants for abuse of market dominance. It has, for example, levied large fines against Google’s comparison-shopping services,65 and its Android66 mobile operating system, for abuse of market dominance. Antitrust authorities in the U.S., Argentina, Canada, Brazil, Israel, Taiwan, India, South Korea, and Russia have similarly opened antitrust cases against Google.67 While most competition authorities primarily use traditional models focused on consumer harm arising from dominance, some legal scholars and practitioners argue for the need to consider broader economic issues and future market consolidation.68 At the same time, the interplay between competition law, consumer protection law, and data protection has become increasingly important and complex, leading to calls for relevant regulatory agencies to better collaborate to achieve more holistic responses to digital dominance (despite potential turf wars).69
- Many states have also expressed concern over the uneven tax obligations faced by local tech start-ups and traditional firms versus global tech platforms, which are able to minimise their local tax liabilities partly due to their international structures. In March 2018, the European Commission announced proposals to allow EU countries to tax profits generated within their territory, regardless of the physical presence in the EU of the company generating the profits. It also proposed a 3% interim tax on digital activities currently not effectively taxed to generate immediate revenues for EU member states.70
- A growing number of countries are trying to deal with concerns of political radicalisation and threats to electoral political processes, driven by the platforms’ increasing role as a locus of societal debates and their information dissemination models.71 Germany now gives social media platforms 24 hours to remove content that violates existing provisions in Germany’s criminal code.72 The European Commission has called upon platforms to tackle online disinformation, as “the weaponisation of online fake news and disinformation poses a serious security threat to our societies.”73 In Bangladesh, new digital security legislation imposes jail sentences for offences like spreading misinformation or distributing “negative propaganda” using a digital device.74 Brazil has drafted bills against fake news currently that are under examination by Parliament.75 One such bill proposes fines and up to 4 years of imprisonment for users who disseminate “fake news.”76 Malaysia has outlawed so-called “fake news” entirely, the first measure of its kind in the world.77 Indonesia has established a government agency to “monitor news circulating online” and “tackle fake news.”78 There are clear risks that these measures be used punitively against opposition parties, activists, journalists, and others, and many of these laws may have extra-territorial effect. They could also prove virtually impossible to implement.
- As people grapple with the vast extent of data collection central to the business models of platforms, some regulators and governments have adopted or expanded privacy measures and data protection provisions. Concerns relate not only to the use of personal information to influence individual choices, but also to the possibility of data leaks. The European General Data Protection Regulation (GDPR) recently came into effect, with sweeping changes to consent and many other issues. Data protection authorities in the Netherlands,79 France,80 Germany,81 and Belgium82 have already launched investigations about or taken steps to enforce user-consent for advertising purposes.
It is problematic that these and other measures often focus on issues in isolation, potentially failing to understand how different activities fit into the platforms’ overall business models. Some could create unintended consequences by imposing rules with which only the largest companies can comply, further strengthening a dominant position, or by undermining the open and global nature of the Internet itself.
Traditional regulatory tools are evolving to account for the specifics of the Internet economy.83 Regulators are on a steep learning curve, and seem to rarely coordinate with other government actors or regulators with different expertise or mandates looking at the same companies. So far, no holistic responses to consolidation have emerged.
In publishing this report, one of our hopes is that policymakers and other decision-makers begin to understand that consolidation in the Internet economy is a much more complex set of issues than the popular press typically covers. Furthermore, this report demonstrates that consolidation trends have positive and negative implications at a number of different layers of the Internet, and that these implications are often interrelated.
Examples of responses from other stakeholders
Some actors in the Internet technical community have also responded to consolidation trends with targeted approaches:
- The Internet Architecture Board (IAB) has kicked-off a community discussion with a post on the IETF’s blog84 and the publication of a working document (or “Internet-Draft”) providing some perspectives on the issue.85 Further IAB activity on the topic is reportedly planned for 2019.86
- The software company Mozilla has made proposals to the U.S. Federal Trade Commission around API governance, which it calls ‘”the fundamental connective tissue of the Internet.”87
- Tim Berners Lee’s Solid project, located at MIT, "aims to radically change the way Web applications work today, resulting in true data ownership as well as improved privacy,"88 with a platform for decentralised but linked data applications under the full control of users. Other developments related to blockchain technologies have similarly led to broader enthusiasm for a broader decentralisation movement, which may yield specific and useful applications, even if a “revolution” still seems somewhat remote.89
- Google engineers have taken steps to protest their employer’s proposal for a country-specific search engine to meet China’s censorship requirements;90 Facebook engineers have complained about a lack of political diversity on the platform;91 and the www.neveragain.tech platform encourage programmers to pledge not to allow data gathering and analysis to be used to collectively harm ethnic groups in the U.S.
Finally, civil society organisations, NGOs, and intergovernmental organisations92 have been vocal about the need for transparency and oversight of influential Internet companies. They play an important role in highlighting gaps. These gaps are, most notably, in the protection of human rights in a diversity of areas ranging from concerns about the gig or platform economy and labour rights,93 and Internet users' freedom of expression and privacy on social media platforms that are increasingly important public spheres for the exchange of ideas and debates.94 These and other civil society organisations have started to develop practical standards, benchmarks, and incentives for global platforms to respect human rights everywhere they operate. This includes discernible self-organisation by consumers, (e.g., capacity-building by organisations like Consumers International), to influence business and governments on digital issues.95
37% of our survey respondents said that there are current efforts in their country or region to address concerns about market dominance in the Internet economy.
Will governance efforts remain siloed?
Many of the consolidation trends highlighted in this report result from platforms leveraging market incentives and opportunities, as well as Internet users’ enthusiasm for the quality, range of offerings, and lower prices – or seemingly-free services – that platform environments can provide. While there is nothing inherently wrong with these trends, concerns remain, including competition, choice, data and consumer protection or welfare, dependencies, and the potential for single points of failure, as well as societal challenges exacerbated by size, scope, and speed of social change wrought by platforms. How those concerns are addressed in the coming years will not only determine the success of addressing the specific question at hand, but also the future development of the Internet as a whole.
This challenge is also an opportunity to test the efficiency of collaborative multistakeholder approaches that typify Internet governance, ensuring a diversity of views inform policy and technical decisions. But will this methodology be effective and sufficient? The question remains to be answered and will largely depend on the ability of stakeholders to quickly stop working in silos.