Limited functionality available
Consumers’ insatiable hunger for internet video is driving double-digit growth in the content delivery network (CDN) market. But will legacy CDN providers win out over the growing number of cloud, media, and telecom players in this space?
Content delivery networks (CDNs) are the workhorses of the internet, designed to improve media quality, speed, and reliability by bringing content physically closer to the user. We predict that the global CDN market will reach US$14 billion in 2020, up more than 25 percent from 2019’s estimated US$11 billion. Further, the market will more than double to US$30 billion by 2025, a compound annual growth rate of more than 16 percent.1
Download the infographics
Watch the video on this year’s TMT Predictions
Download the Deloitte Insights and Dow Jones app
Subscribe to receive related content
This growth is being driven primarily by consumers’ ever-increasing hunger for streaming video over the internet, now amplified by the migration of more broadcast and cable TV onto direct-to-consumer over-the-top (OTT) internet delivery networks. Live video streaming and the emergence of streaming video games may further spur growth, as well as prompt more technical innovation. However, it’s not certain exactly who will cash in on all this growth, as bigger media, cloud, and telecom players develop or expand their own content delivery networks. Their success may threaten profits for existing CDN providers even while increasing demand for the hardware and software capabilities necessary to deliver so much media.
CDNs have played a critical role in the internet’s expansion and evolution. Originally, when bandwidth was less available, CDNs emerged as a way to make the more data-intense parts of a website, such as images with large file sizes, load more quickly. Early CDNs like Akamai set up regional hubs that held copies (caches) of all the images on a website, bringing content closer to people browsing the internet around the world. With so many points of presence (PoPs), even though a user in San Francisco might be browsing a popular website hosted on servers in Europe, their browser could fetch the images from a nearby CDN hub holding copies of the media.
As more people accessed the internet, and as bandwidth availability grew, Web-based sites and services were able to become richer and more capable, placing further demands on specialized CDNs. They evolved to support software downloads, accelerated mobile content, and richer media such as video. The build-out of these more sophisticated networks and capabilities created even greater demand for CDNs to accelerate the user experience, supporting the CDN industry’s growth. Top global CDN providers today are in a position to spend billions to expand their massive infrastructure underneath the growth of the Web—and they might need to: CDNs are expected to carry 72 percent of all internet traffic by 2022, up from 56 percent in 2017.2 And it’s not just existing CDN providers that are getting into the act: A growing number of media and telecom companies are developing CDNs as well.
A typical CDN will place media storage and playback “appliances” at the edges of its network, in internet service providers (ISPs) or internet exchange points (IXPs) in major metropolitan areas, to get closer to end users. These appliances are physical boxes with large storage capacity and software capabilities, sometimes referred to as micro data centers. Netflix, for instance, maintains thousands of these boxes across its global OpenConnect CDN, each holding up to 80 percent of Netflix’s entire media catalog.3 The boxes are custom-designed, assembled by suppliers, and delivered to regional ISPs and IXPs for installation. The Netflix core manages the transcoding of new content and regularly pushes updates to its edge appliances. The system is continuously monitored for loads and faults, and failing appliances can be quickly taken offline and capacity shifted to other redundant PoPs.4
The largest driver of growth for the CDN market has been, and will likely continue to be, the rising demand for video over the internet. The internet, of course, isn’t the only way to watch videos; many other video delivery technologies, ranging from old-school on-air broadcasting to cable, DSL (copper phone lines), and private internet protocol television (IPTV) networks, are still in use today.5 However, to keep up with evolving consumer behavior amid an explosion of mobile devices, as well as the advertising dollars that follow large numbers of eyeballs, digital media delivery has been expanding from private IPTV networks into OTT networks that run over the internet. This expansion into OTT has at once enabled and been a response to much higher internet usage and greater broadband penetration.
According to Sandvine, OTT streaming video uses over 60 percent of global internet bandwidth.6 Netflix already uses an estimated 15 percent of global downstream internet traffic,7 and a single top streaming video service can consume up to 40 percent of downstream traffic on some regional operator networks.8 Fueling all that video traffic is continued growth in streaming video on demand services (SVOD). In 2018, global subscriptions to streaming video services overtook those for cable television for the first time, reaching over 613 million people, 27 percent more than the year before.9
With several major new SVOD services launching, global subscriptions to streaming services could grow significantly. The resulting increase in OTT video traffic, however, won’t necessarily drive greater revenues for specific CDN providers. Much of the growth in video traffic comes from the largest SVOD services, social networks, and other hyperscale digital media companies that already operate their own CDNs.10 Furthermore, the largest media companies entering the SVOD space may eventually build their own networks or, in some cases, use the networks they already control. For example, some major telecoms that have acquired media properties also have their own CDNs. This could give them greater control of delivery from cloud to consumer.
Although the current SVOD expansion is being led primarily by US-based media, some see global networks as a path to unlock global markets.11 Analysts have predicted a rapid expansion of demand in Asia-Pacific markets; Asia-Pacific is expected to account for 51 percent of all video streaming traffic by 2024, almost twice as much as in 2018.12 Some US CDN providers offer CDN services in Asian markets, and some leading Chinese CDNs now host over 1,000 PoP nodes across the country, with additional networks in other Asian countries.13 However, content delivery in Asian markets can be challenging due to sparse coverage and limited mobile networks. For broadcasters eyeing expansion into Asia, working with CDNs that understand these regions’ nuances can reduce some of these challenges.
Video files are particularly large, and tremendous technological wizardry is needed to compress them, break them apart into distributed pieces, and then dynamically reassemble and stream them on demand to hundreds of millions of requesters, all with high resolution and minimal latency. The growing quantity and sophistication of OTT video content means more traffic, more routing, and a greater need for management, optimization, and prediction across the CDNs responsible for delivering speedy and reliable content.
From a technical standpoint, live and on-demand video streams move a lot of bits to render high-resolution images. The stream travels in only one direction, and the content isn’t directly interactive, but they still contain a great deal of data. Streams need buffering and caching to avoid lags and dropouts. Legacy protocols such as real-time messaging protocol (RTMP), developed over a decade ago to encode video and move it across networks to clients, will likely be displaced by newer solutions such as secure reliable transport (SRT), designed to further decrease latency and meet the demands of live and on-demand streaming. CDN giant Akamai, for instance, has announced a partnership with Bitmovin to develop “next-generation encoding capabilities” that include SRT.14
Live video streaming, in particular, can pose significant challenges for CDNs. Live video streams support real-time events like concerts and sports games, but they also include social streaming services that let anyone stream video, a global phenomenon that is growing quickly. After launching its social streaming platform TikTok, China’s ByteDance saw almost 1 billion downloads within the first two years.15 Many people are streaming their lives to the world, further fragmenting audiences and driving viewers toward free niche content. By some estimates, live video streams in 2018 accounted for 11 exabytes (EB) out of a total 58 EB of CDN video traffic; by 2024, live streaming is expected to account for as much as 238 EB out of a total of 453 EB—or half of all video streaming.16
As the audiences for live streaming services grow, so too should CDNs. CDNs delivering live streaming media should be able to encode video in real time, copy it out to PoPs, and manage demand for the content. Planned events like big concerts and championships can have network resources allocated to them ahead of time, but other events can be unexpected. Major influencers and celebrities can suddenly draw an enormous viewing audience to their streams, as can breaking news events. CDNs must be able to quickly detect such spikes in traffic and respond dynamically to support the demand, whether regional or global.
Apart from issues around timing and resource allocation, CDNs also have to deal with the challenges of operating across diverse and dynamic networks, carriers, devices, and audiences. To do this effectively requires ongoing analysis and responsiveness, and depends crucially on continuous visibility into network performance. For CDNs that are also moving content across telecom carrier networks, routing intelligence is critical to ensuring high-performance streams. These systems can dynamically shift traffic across carriers if one is experiencing low throughput and placing the stream at risk. Evaluating carrier performance, too, can help a CDN determine where faults may have originated. Beyond the edge of the CDN network, streams must transit last-mile networks that include wireline and wireless carriers as well as a myriad of access points, such as home Wi-Fi. But although CDNs can see across these long tendrils of the network topology, they are typically unable to meaningfully manage them.
Ultimately, for media companies and CDN providers in an increasingly competitive landscape, ensuring a high-quality experience for the end user is the top priority. Low-quality video and buffered streams can quickly damage a CDN operator’s reputation, making it imperative to invest in sophisticated monitoring and analytics capabilities. Software layers deliver status dashboards, extract and operate on data analytics, and automate network-scale load balancing, fault detection, and demand prediction. More sophisticated data science and machine learning solutions can predict failure rates for digital content and hardware assets before deploying them to the end user.17 Likewise, they can predict spikes in demand within specific regions, and then preallocate resources directly to those PoPs, even copying content such as a newly released movie to specific regional appliances.18
Even as networks expand to accommodate growth in OTT video, much richer and more dynamic content may soon be vying for their infrastructure. Some of the world’s largest cloud providers have announced plans to stream multiplayer video games. Doing that, however, is quite a different challenge from streaming video.
Rendering a game on a player’s device isn’t markedly different than rendering a movie. However, in multiplayer video games, the “movie” continuously changes in response to player commands—run, turn, shoot—and every player has a different view of the game based on their position in the game world. Hence, data must flow both downstream to the player’s device and upstream back to the game engine, and there can be no noticeable latency between a player’s action and the game engine’s resulting response. Latencies above 50 milliseconds or so can quickly make many top video games unplayable.19 Rendering speed, which depends on bandwidth, is also a challenge. Continuously drawing and redrawing images on a small mobile screen is one thing, but many gamers play on large 4K screens, and some TV makers promise 8K rendering.
Considerable computational power and network messaging goes into instantaneously synchronizing actions between the game and its players, particularly if synchronization is required across the globe. The most popular multiplayer games today host more than a million players simultaneously, but they only manage this by partitioning users into smaller groups in separate instances of the game (for example, by creating 10,000 different “worlds” that host 100 players each). The newer streaming video game providers claim that their services will allow a single game instance to potentially support thousands of players, placing greater demands on CDNs to manage the necessary synchronization.
Google is tackling these challenges head-on with its Stadia service for streaming video games. To do so successfully will likely require infrastructure, expertise, and significant capital. Google’s capex grew by 80 percent in Q4 2018 to reach US$6.8 billion, although much of this was in the opaque category of data centers and servers.20 Google has also been growing its fiber network, which has hundreds of PoPs and several thousand edge processing nodes. The company operates a mature CDN to support YouTube, leveraging its expertise in data science and machine learning to manage demand and delivery; in fact, Stadia is positioned as an extension of YouTube. Google is thus likely well placed to capture game traffic on its own CDN—a reminder that overall CDN revenues may accrue more to leading platform services running their own networks that support services higher up the stack.
Among the many technical and market challenges ahead for streaming video games, it’s worth noting that several leading game publishers have announced their own interest in developing cloud streaming for their top franchises. They too may look to CDN providers for the assist. With a global video game market serving over 2.5 billion players and estimated to generate US$150 billion by 2020, game publishers and content delivery networks could align to build the next generation of streaming media.
People across the world now spend the majority of their internet time streaming video, and the internet itself is segmenting into cloud-based services and CDN edge capabilities. Growth in live streaming video and the promise of streaming video games may further impact competition and innovation. These shifts have implications for a range of players: media companies, telecoms, and CDN providers.
Media companies may view the current landscape as an incentive to move more content onto CDNs. However, this is by no means a simple decision. Broadcast and pay TV are still strong, although slowing in some major geographies.21 In addition, deconstructing IPTV networks and set-top boxes is not easy, nor is shifting an audience to potentially risky OTT services. Most media companies will likely run their OTT efforts in parallel with their broadcast and pay TV services, allowing the market to determine which will draw the most subscribers.
For many media companies adding OTT to their strategic initiatives, the question becomes: How? Should they partner with a CDN provider? Do they expand on their existing relationships with networks that they have developed for streaming pilots? Or do they commit to building their own CDN? The equation becomes more difficult when considering the potential competitive advantages media companies may find in controlling delivery and owning the value chain right down to the end consumer. Major media companies may choose to build their own networks to control the entire delivery pipeline—and to secure ownership of the data their audiences generate. However, only a few will likely have the capital to do so. The rest will probably rent capacity from CDN providers who, in turn, will pay telecoms to run traffic on their networks. The cost of renting distribution over CDNs is now very low, which can make renting an attractive option—as well as increase the pressure on CDN providers to expand margins by offering other services or applying event-based peak pricing models.22
For their part, many telecoms already operate and lease access to CDNs. However, with margins on CDN rents narrowing, leasing CDN access to content providers may become less profitable. This may provide an incentive for telecoms that have purchased media catalogs to invest more in delivery capabilities around their own content. Yet telecoms are also building out appliances on cell towers and selling edge services. Can they sell CDN and edge services and still invest in differentiating delivery for their own content? Also worth considering is that both CDN and edge services allow telecoms to offer their customers more sensing and analytics capabilities that can be translated into performance and innovation. Is there a clear path from controlling a private CDN to establishing competitive advantage around delivery and analytics?
Telecoms also control the last mile of content delivery to users. Can they parlay this last-mile control into value? Cisco predicts that, by 2022, 12 percent of mobile traffic will be over 5G, with the average 5G connection using 21GB of traffic per month. If 5G becomes widely implemented and adopted, will it shift more power to the telecoms that control last-mile delivery? Will it open up room for richer media we haven’t yet imagined?
As for CDN providers themselves, they will likely face more challenges with capacity, quality of service, load balancing, and demand prediction as more audiences shift onto streaming services. Similar to what has happened among cloud providers, competition could drive CDN rental prices down further. Top CDNs may also face greater market pressure to secure delivery of intellectual property against intrusion and theft. To address these needs, CDNs could put in place additional data analytics, machine learning, and potentially content blockchains to manage growing complexity—all of which would add to their operating costs. Computation storage and routing will likely prompt investments in hardware, while load balancing, stream optimization, demand prediction, and security may demand more software capabilities.
On the other hand, while CDNs may face growing competition and lower margins for delivery, they may also expand as more cloud services migrate to the edge. Just as telecoms are dabbling in CDNs, CDNs are spinning up edge computing services,23 not just for media, but also for IoT.24 Savvy observers should pay attention to CDN provider announcements that might give them greater competitive advantage at the edge, as this new shift reshapes the businesses of cloud providers, telecoms, and CDNs alike.
The rising demand for more and more internet video, along with new offerings such as streaming video games, should continue to shape the internet to meet users’ needs. Networks of all stripes will likely keep coevolving with hardware innovations and human behavior. In the near term, the CDN market is poised for strong growth. What this means for the CDN providers, media companies, and telecoms jockeying for position in that market remains to be seen.