The Vault

The Future of Video Distribution
White Paper / Feb 2015 / Video Distribution

An examination of the value chain that drives video is necessary to understand and examine the future of video distribution. Operators, broadcasters or platforms that either produce or aggregate content, pay for video distribution with revenues generated directly by consumers or via advertising. In the past five years, consumer expectation for content and total delivery costs have seen significant increases. These new economic pressures have made video technology, third-party standards and network infrastructure vital.

This technology focused paper discusses the impact that infrastructure and technical evolution will have on video consumption over time. It examines the demand for video content, two core approaches for video delivery: multicast and unicast, how future technology can optimize video supply and more.

We found some problems with the form:
To view this content please fill out the form below to tell us about you.
Privacy Policy
March 2015 technology.ihs.com IHS tecHnology Insight Report IHS tecHnology copyright notice and legal disclaimer ? 2015 IHS. No portion of this report may be reproduced, reused, or otherwise distributed in any form without prior written consent, with the exception of any internal client distribution as may be permitted in the license agreement between client and IHS. Content reproduced or redistributed with IHS permission must display IHS legal notices and attributions of authorship. The information contained herein is from sources considered reliable but its accuracy and completeness are not warranted, nor are the opinions and analyses which are based upon it, and to the extent permitted by law, IHS shall not be liable for any errors or omissions or any loss, damage or expense incurred by reliance on information or any statement contained herein. For more information, please contact IHS at customercare@ihs.com, +1 800 IHS CARE (from North American locations), or +44 (0) 1344 328 300 (from outside North America). Consumer Electronics & Video Technology Intelligence Service Ed Border, Principal Analyst ? Scaling unicast to meet total video demand is not yet economically viable, and will not be for at least 20 to 25 years even in developed markets. ? Total demand for unicast delivery could account for up to 50% of current viewing, but currently supply could only cost-effectively provide 15-20% of total viewing. ? Total viewing time has been on the increase for years across all key markets, but showed signs of stabilizing in 2013. 4 hours and 46 minutes of video was watched per household per day in the US, of which 19% was for non-linear content. ? The total number of connected devices is growing rapidly across the globe, at a total of 4.5 billion installed in 2013 and forecasted to reach 9.6 billion by 2018. Smartphones and tablets make up the bulk of new devices in households, with PCs on the decline. ? The majority of online streaming is overwhelmingly SD, at relatively low average bitrates of just over 1 Mb/s. Growth in HD streaming will create further strain on the network and necessitate new compression technology and network re-structuring. ? HEVC has the capability to offset some of the additional data required for higher quality video, but will take time to filter into new devices, and requires too much processing power to be currently implemented currently as a pure software upgrade in existing devices. SECTION 1: INTRODUCTION 1.1: Background To look at the future of video distribution requires looking at the value chain that drives video. Video distribution is paid for by operators, broadcasters or platforms that either produce or aggregate content, with revenues generated directly by consumers or via advertising. Consumer expectations for content have risen significantly in the last five years: higher resolutions, in-home storage, multiple devices and anytime viewing are expected features for many. Total video delivery costs have grown from $8.6 billion in 2005 to $16.8 billion the future of video distribution ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 2 in 2014, and are expected to reach over $20 billion by 2020. However, the rate has begun to slow, as video service providers strive to reduce costs and avoid pushing higher pricing rates to consumers. Video technology, third-party standards and network infrastructure have become vital in the face of these new economic pressures. They define the rates at which different forms of video can be transferred, and therefore create the bounds that effect volumes of video transport. This report is therefore primarily focused on technology rather than content. It discusses how infrastructure and technical evolution will both enable and limit video consumption over short, mid and long-term scenarios. To do this, we first look at the forces that drive video consumption, such as video platforms, connected devices and viewing methods; and then subsequently look at the technology and infrastructure that underpins these systems, and how spend and data transfer are affected under different assumptions. 1.2 Key Factors and Definitions There are two primary, contrasting approaches for video delivery: delivering multicast/broadcast or delivering unicast. Broadcast, or multicast, delivery: Point-to-multipoint technologies, where the same content is distributed once centrally and then received separately by multiple end-points (for example, distribution of content to a single satellite, with each household then having its own receiver equipment). Unicast delivery: Point-to-point transmission, in which each end-point makes a request, and the video is transported separately and uniquely to it (such as over-the top OTT catch-up over BBC iPlayer, where each user makes a unique request for the video). IHS sees the main tension in the video transport market revolving around the interplay of unicast and multicast delivery. As the cost of a single unicast stream lowers, there are several key questions this report seeks to answer: ? To what degree can on-demand or catch-up TV replace linear TV, before becoming prohibitively costly? ? Can multicast methods be introduced into aspects of OTT delivery to minimize costs? ? To what extent will newer video formats, such as HD, Full HD and Ultra HD, grow? And how sustainable is higher quality video distribution at scale? ? What future technology developments are in place to influence video distribution? And what impact will they have in the short-term? SECTION 2: DEMAND FOR VIDEO CONTENT 2.1: Viewing Habits Over the past ten years the total amount of video viewing time per household has continued to rise across all major TV markets. In 2013, the average household in the US watched 4h 46m of video per day, up from 4h 14m in 2005; meanwhile, the average household across the largest five European countries ? France, Germany, Italy, Spain and the UK ? watched 4h 17m per day in 2013, up from 3h 39m in 2005. Linear TV still makes up the vast majority of all viewing time ? 81% in the US, 89% in the five largest European markets ? but has been steadily declining since 2009. The decline in viewing time for linear TV is inversely related to two newer forms of viewing content: the first is DVR time-shifted viewing, in which a consumer records content onto either a physical, in-home hard-disk or onto the network; the second is on-demand multiscreen viewing, in which a user can select uniquely streamed content whenever they want, either via a fixed platform or over the open internet. Linear viewership amongst channels is still fairly concentrated ? analysis of BARB data for the UK market suggests that in 2013, 53.2% of all viewing came across the five main linear channels (BBC One, BBC Two, ITV, Channel 4 and Channel 5) and their associated HD and +1 feeds. This has fallen over the past decade ? the equivalent totals for 2005 and 2010 were ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 3 70.4% and 56.3% respectively. However, the decline has slowed in the past few years, suggesting these viewing habits are relatively entrenched. The second key habit is that linear viewership is highly concentrated towards specific times of the day; in 2009, BARB data again shows that approximately 70% of total average daily viewing, over a month long period, took place between the eight hour window between 3pm and 11pm. there are two key takeaways around video consumption: 1. Non-linear viewing is rising, and will continue to rise, but linear TV still makes up 80 to 90% of all viewing in major markets. 2. Roughly half of linear TV viewing is highly concentrated on a small subset of specific channels at specific times of the day. 2.2: new technologies Higher resolution video, in-home DVR storage and on-demand video are three major influences on viewing habits, and will be key to video distribution going forward. The transition from Standard Definition (SD) content to High Definition (HD) content started a little over 20 years ago. However, it wasn?t until the mid-2000s and the development of H.264 (MPEG-4) codec that HD broadcast became a viable reality. As of 2013, the HD transition is still a third of the way through ? just 32% of households with a TV are equipped to view HD, although significantly more have HD TVs but lack reception equipment for HD signals. This is expected to continue to rise, but even by 2018 only 50% of all TV households globally will have access to HD content via TV, although as of 2014 77% of households in the US are already HD-equipped. Meanwhile DVR viewing continues to grow at the expense of linear TV, particularly through pay TV. The majority of DVR is still undertaken through embedded hard-disks directly within a set-top box (STB), as networked based DVR (in which content is recorded on the network and streamed to subscribers) has run into copyright issues. This makes DVR still a 0 50 100 150 200 250 300 350 2005 2006 2007 2008 2009 2010 2011 2012 2013 Linear DVR time-shifted Pay-TV VoD Online Long-form OnlineShort-form Viewing times by type; US vs Big 5 EU Source: IHS ? 2015 IHS Vi ew in g tim e (m in ut es ) U S A E U b ig 5 ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 4 relatively expensive up-sell to a traditional STB, which in turn lends itself favorably to pay TV service providers, looking to offer higher-value services to retain higher-spending consumers. Global household penetration of DVRs has risen significantly but was still at just 12% in 2013, and is weighed heavily towards pay TV subscribers within North America and Europe. To give an example, by the end of 2014 over 50% of US households will own a DVR, almost exclusively paid subscribers. Finally there is video-on-demand (VoD), which has grown rapidly over the past decade. Paid for on-demand content or service can be most typically broken down into two categories: transactional VoD, in which a user pays a one-off fee for specific pieces of content; and subscription VoD (sVOD), in which a consumer subscribes to a content library for a monthly fee. Subscription VoD, such as Netflix, is more interesting for the purposes of this report, as it promotes longer bouts of constant viewing. As of the end of 2014, there will be 80m sVOD subscribers, of which 51m will be in North America and 13m in Western Europe; by the end of 2018, this will rise to 125m globally, with significant increases in Western Europe and South America. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Africa and Middle East Asia Pacific Central and Eastern Europe North America South and Central America Western Europe HD TV Set penetration (%) Source: IHS ? 2015 IHS Pe ne tr at io n (% ) Year 0 200 400 600 800 1,000 1,200 1,400 1,600 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 TV HD DVR SVoD Households with either a TV, HD, DVR or VoD subscriptions Source: IHS ? 2015 IHS U ni ts (M ) Households ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 5 2.3: there were 4.5 billion connected devices in households in 2013 The growth in connectable devices that can stream content is the final major influence on viewing habits. Between 2005 and 2013, the total number of connected devices installed in homes worldwide grew nearly fivefold, from under 1 billion to over 4.5 billion. This will grow further still, reaching nearly 10 billion devices globally by 2018 ? driven by factors such as increased broadband speeds, consumer replacement cycles for older devices, and the rise of online platforms and ecosystems. While all connected device categories have grown in installed base, this has not been uniform. At a device level, the landscape has shifted significantly in this timeframe ? back in 2005, 93% of all connected devices were computers (PCs and laptops), and just 6% were mobile devices (phones and tablets); by 2013, just 38% of devices were computers, with 52% mobile devices; and by 2018, just 20% will be computing devices, with 66% mobile. 0 2 4 6 8 10 12 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Western Europe South and Central America North America Middle East and Africa Central and Eastern Europe Asia Pacific Source: IHS ? 2015 IHS Connected Devices per household Global Connected Device per HH 2.45% 1.86% 6.22% 4.66% 42.29% 2.97% 26.86% 12.37% 0.24% Blu-ray Players Connected Audio/Video Receivers (AVRs) Connected soundbars Digital Media Adapters (DMAs) Free-to-air (FTA) Set-top Boxes Game Consoles Pay-TV set-top boxes PCs Smart TVs Smartphones Tablets Wireless speakers An ?average? US Household 2013 Source: IHS ? 2015 IHS ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 6 Putting this all together, we can form a picture of the capabilities of the ?average? consumer household. As of 2013, the average household across the globe has 2.3 connected devices. In developed markets, such as North America (7 connected devices) or Western Europe (4.3) this is significantly higher; whereas less developed markets, such as Middle East and Africa (1) or South America (1.7) are lower. The key devices are now computers, phones and tablets ? an average North American household will have 3 computing devices, 2 smartphones, a tablet and then either a games console, smart TV or connected set-top box. 2.4: not all devices are equal The growth in connectable devices has impacted video consumption habits, but in different ways for each device. Taking BBC iStats data for BBC iPlayer views in the UK, we can see that the total number of views per month per device with the UK is still heavily biased towards connectable pay TV set-top boxes, computers and tablets. This content is primarily longer-form programs of over 30 minutes in length. The dynamic is different for shorter-form content or user-generated content. To take YouTube as an example, it has reported that globally, 40% of its 6 billion hours of video watched per month is on mobile devices. Data from Ooyala?s ?Global Video Index ? Q1 2014? also suggests that, as of Q1 2014, 37% of all time spent viewing content on mobile phones was for video of less than 6 minutes - from which it may be estimated that approximately 75-80% of all mobile views were for short-form, sub 6 minute length, content. And looking at data from Conviva, a provider of CDN management solutions, we can see that the average viewing time and streaming rate for an iOS device, compared to either a Roku Box or Samsung Smart TV, shows us that long-form content is still focused on set-tops and TVs. iOS viewing time was 11 minutes per view at an average of 1Mb/s ? whereas Roku and Samsung were 34 minutes and 30 minutes at 2.6 and 2.5 Mb/s respectively. To summarize - more connectable devices in households has initially supplemented traditional viewing, but there were signs in 2013 that further non-linear content growth will increasingly come at the expense of linear viewing. This will mean that in the long-run, total viewing time will remain fairly flat, but costs and data transferred will rise. This is because, due to long-term contracts, the total cost and data transfer of multicast linear distribution will stay broadly fixed, irrespective of total views ? while the cost and data transfer of unicast non-linear distribution will continue to grow. SECTION 3: THE TRANSITION FROM MULTICAST TO UNICAST 3.1: long-term transition inevitable The current state of video is split between two core approaches: linear TV distributed via multicast; or non-linear on- 0 1 2 3 4 5 6 7 8 9 2011 2012 2013 Mobile Tablet Computer Games Console STB Smart TV / Other Average iplayer device views per month compared to installed base Source: IHS ? 2015 IHS A ve ra ge ip la ye r v ie w s pe r m on th ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 7 demand content delivered via unicast. This evolution has occurred based on historical communication networks, but the dynamic will change in the future. Broadly speaking, multicast is suitable for more concurrent views but less content choice, while unicast scales well for fewer concurrent views but greater choice. The effect of this is that demand for less popular linear content at non-peak viewing hours will eventually move to a unicast system; and vice-versa, more popular non-linear content will likely incorporate further multicast elements into delivery. The next logical evolution for operators, particularly those owning fixed cable or IP networks, is therefore blended delivery of content, where the delivery method depends on the content source and popularity. If we compare satellite with CDN, we can see that at average viewership figures of smaller, niche channels could already be supported more cost-effectively by unicast. For an SD channel in MPEG-4, assuming average current CDN prices, a channel in North America with less than 120,000 daily average viewers could be supported at a lower cost-per-hour via CDN than satellite. The key factor here is cost. From a cost perspective, CDN is already the largest single platform cost for third-party consumer video in several of the largest five European countries and in the US. This is despite the fact that in 2013 non- linear viewing accounted for just 4% of all viewing across the Big 5 Europe countries, and 9% in the US. Taking into account, as referenced earlier, that the majority of CDN viewing is still in SD, at around 1Mb/s, and we can already see important limitations on the potential current volumes for non-linear streaming. The potential increase in cost-per- thousand (CPM) for personalized advertising can potentially offset this, but while current signs are positive, there are few hard deployments of the technology, so it has been discounted as a factor for now. 3.2: Unicast cannot cover all video demand for decades To discuss the current limitations of unicast, take the UK distribution market as an example. In 2013, the total cost of video transport from content and service providers within the UK was $540 million, and the total amount of content viewed was 255 minutes per person per day. Within this, non-linear content accounted for 7.5% of all viewing time, but $112 million of spend on CDN transport - over 20% of revenues. As a whole, the UK watches slightly less than 4 hours of linear TV per person per day. Assuming that HD content maintains similar viewership levels, scaling CDN distribution to match this demand would mean a total of 273k petabytes of data that would need to be transported in a year ? at a total cost of $2.3 billion. To put this in perspective, the equivalent figures for satellite would be 11 petabytes of data transferred, for a total cost of around $30 million. The same argument also holds true for cable distribution, where channels will be distributed via multicast over proprietary wired networks. Therefore, by order of magnitude, the current cost of unicast CDN delivery for all viewing demand is approximately 100 times that of multicast delivery via satellite or cable, and the order of data transferred approximately 100,000 times that. 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 CDN Satellite CDN revenue vs satellite revenue Source: IHS ? 2015 IHS U SD (M ) Product ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 8 If we assume current CDN price decay rates of 15-20% per annum stay continuous, it would take between 20 and 25 years for the price of CDN to depreciate to a level at which it could cover total viewing demand as cheaply as satellite. This is slightly disingenuous ? after all, should such price decay continue, total CDN revenues would decline massively from what they are today. Such price competitiveness seems unlikely in the long-term, as the CDN provider market has already consolidated significantly. It is therefore unlikely that CDN delivery would ever be as cost-effective as satellite for the total linear demand. To do so would mean that CDN providers would have to take a significant hit on their total revenues. A more viable scenario would be for CDN price decay to fall at an average of between 5-10% per annum ? in which case it would take between 21 to 33 years for the total cost of CDN transport for all linear TV to be equivalent to current CDN revenues, assuming 4% annual inflation. 3.3: current limits of unicast The next question is: what proportion of total video demand could feasibly be covered by unicast without raising costs? As of 2014, approximately 23%, or 47 minutes per viewer per day, of current linear viewing could be transferred to unicast in the UK, before the cost of unicast delivery exceeds the total video transfer spend on all other networks. Looking forward to 2018, the equivalent figures would be 38%, or 1h 20m. Total transport spend (including all platforms) is quite a large upper bound for just unicast delivery, but provides a reasonable cap on the total amount that could feasibly be spent in the short-term. Non-linear, unicast content accounted for 9% of viewing in the US in 2013. Taking the calculations above, we can get some indication on the degree to which unicast could replace multicast viewing under current conditions ? between 15-20% of total demand at present day level, between 25-30% in 2018, and around 40% by 2023. If we consider the earlier data presented, which indicated that approximately 50% of linear viewership is concentrated on a small subset of key channels, we can begin to confirm that, long-term, viewership will fragment in two directions. The first is cost-efficient multicast broadcast for the large, popular linear channels within each country; the second is unicast delivery of less widespread, more tailored content. However, a divide also becomes clear. Approximately half of the daily content currently viewed in developed markets such as the UK or US is watched linearly on mainstream channels at primetime. The other half is more personalized, less time dependent and therefore is suitable for unicast. As shown above, the current economics of unicast would make achieving 50% of demand prohibitively expensive for several years, until sometime around 2025. The challenge therefore becomes how to match demand with supply ? developing new technologies and architecting networks to optimize the split between multicast and unicast. 3.4: the effect on current landscape The current world is still predicated on linear delivery, which accounts for 81% of all viewing time in the US. The majority of this linear viewing in the US is over paid-for networks, which account for 100m of the 120m current TV households ? and of this, 65% belongs to wired cable and IPTV operators that own a network, with the remaining 35% delivered via satellite. A long-term shift to unicast delivery changes the economics of video distribution, and will have an effect on the operator business models. Today?s pay TV operators don?t charge delivery costs separately from content costs, because such delivery costs are multicast, and independent of any individual user. One likely consequence of increasingly blended delivery would be for wired infrastructure providers to separate out the network for delivery from the content aggregation platform ? essentially provide a model similar to many broadband networks, where virtual ISPs run atop one or two core networks, and the network owners can also be ISPs should they choose. The separation of the network from the content aggregation has parallels ? Comcast owning a network through multiple high value cities could be seen as the equivalent of SES holding a satellite in a high-value orbital slot ? and would end the difficulty in using bundled pricing models for highly variable multiscreen video costs per user. This could, for example, one day see both Netflix and a Comcast content subsidiary compete for subscribers over both Comcast?s core network and Cox?s core network; or alternatively, could see a Netflix-like TV service also become a virtual ISP and begin to bundle services. ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 9 However, while increased unicast delivery will likely have a large effect on how an operator its network and content, the pay TV operator landscape itself is likely to stay relatively stable. The key is content rights acquisition, and for current paid content the expense rules out many new entrants immediately, and is difficult for even the larger-scale OTT service providers to broach. In this scenario, incumbent pay TV operators are best positioned to maintain content based services, even as the value of the network itself separates from the value of the content. The most likely threats then come from companies with large resources and in adjacent fields ? for example, a Samsung or an Apple striking global content deals that can be bundled with consumer devices. SECTION 4: FUTURE TECHNOLOGY TO OPTIMIZE VIDEO SUPPLY 4.1: network Architecture The first major way to optimize unicast supply is through re-architecting networks to incorporate elements of both multicast and unicast dynamically. This is not an uncommon scenario ? a DVR essentially creates a unicast network between the DVR and the viewer, while the DVR is fed by a multicast network. Similarly, a CDN creates a multicast network to the edge caches (storage at the edge of the network), while a unicast network is then used to get to the viewer. Unicast is required to allow the range of consumer behavior defined by ?watching anything at any time?, but that unicast element can be embedded within a network rather than comprise it entirely. The most common use of a mixed unicast and multicast submission is edge cashing using a DVR or an edge cache in the network. This way a piece of content that is likely to be watched widely, such as a feature film or a major primetime series, can be distributed widely using multicast to best utilize the network efficiencies of distributing the same thing to many people, but locally cached to allow time-shifting. By moving the edge cache or allowing different behavior from the edge cache any consumer behavior can be simulated ? for example the edge cache can be on a mobile phone simulating out-of- home unicast, or retransmitted from a set-top box DVR, or transmitted from an in-network edge cache such as an internal CDN. Essentially, the future of network structure is a question of blending multicast and unicast by shifting the location of storage, defining how deep into the network the content should be cached. Content with very little demand and sporadic viewing should be stored in the center of the network - for example, in YouTube servers in California, where it takes up little space. The limited demand for this content means it won?t consume significant amounts of bandwidth even with unicast distribution. For content with significant demand from many people within a short period of time, multicast distribution most efficiently pushes that content further out to the edge of the network; it can be locally cached on devices, DVRs or near the home within the network for unicast, time dependent consumption. The more popular the content, the closer to the home it should be stored or cached ? maximizing the multicast element of the distribution and End-user device Request for content Household storage Very popular content can be cached directly on a DVR or local devices, after being multicast to the households Local Edge Caching Fairly popular content can be cashed at the edge of the network, at nodes that split off to reach between 500-1000 households on average Global Data Centre More niche, less popular content is most effectively stored at a global data centre, as unicast streaming will be less volume, therefore cheaper ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 10 minimizing the unicast element. 4.2: Bitrates Another key development in bridging the gap between potential unicast demand and supply is optimizing the bitrate for transfer. Based on current use of the MPEG-4 codec, the average transfer speeds for video for broadcast over satellite or terrestrial are approximately 1.5 Mb/s for SD content, and 3.5 Mb/s for HD (1080p) content. These are also the theoretical speeds at which OTT video is distributed, but the reality is slightly different. Analysis of average streaming speeds from major online video providers suggests that video quality is lower on average ? HD content can be streamed anywhere between 2.5 and 3.5 Mb/s, but is often limited to around 2.8-3.0 Mb/s in practice; while SD (468p) content can be streamed anywhere between 1-1.5 Mb/s, but is typically delivered at slightly lower, 1 Mb/s quality. Moreover, lower quality video streamed to smaller-screen devices can be streamed at anything between 0.3-0.7 Mb/s, for households with lower internet speeds, or when done over mobile networks. However, IHS data suggests that, in 2013, the total average streaming speed for all online video in North America, including premium and non-premium content, was a little less than 1 Mb/s. This broadly means that HD viewing online has been still relatively infrequent ? SD is the norm, and the market for HD online views is balanced out by the higher volume, lower quality, non-premium content streams. This is also backed-up by data from video optimization provider Conviva, which suggests that at the end of 2013, the average speed per stream in the UK for viewed content was approximately 1.2 Mb/s, up from around 0.7 Mb/s in 2010. There are signs that this is beginning to change. In 2014, Conviva data showed the average streaming rate for September rose significantly, up to 1.7Mb/s. Meanwhile, Akamai reported that two of the largest three broadcasters of the winter Olympics had average streaming speeds of between 1.8 and 2.2 Mb/s, demonstrating that HD streaming is becoming more prominent for premium video. This surge in demand for HD video is absolutely key ? long-term, a doubling in average bitrates is equivalent to a doubling in the number of SD subscribers under the unicast model, and places further strain on the network. 4.3: compression technology The effectiveness of video compression is crucial, as it allows for higher volumes of demand to be met. The next stage in video compression is the development of the H.265 (HEVC) codec. HEVC is frequently marketed around UHD. HEVC UHD streaming will deliver approximately 40-50% efficiency gains on H.264, meaning an average streaming speed of approximately 15-16 Mb/s. Moreover, for HEVC-enabled devices, it will be also possible to stream SD and HD HEVC- encoded content directly. This should see efficiency gains of approximately 30-40% - meaning that HD content could be streamed at closer to 2 Mb/s than 3 Mb/s, and SD content at closer to 0.7 Mb/s. In theory, therefore, HEVC is a huge step forward in reducing bandwidth consumption, which can apply 40-50% efficiency gains for each stream. However, HEVC is too heavy for the vast majority of current processors (with the exception of UHD TVs and a small subset of current mobile devices). This means that very few existing phones, tablets, set-tops or games consoles will benefit from an upgrade to the technology. The effect of this is that in the short-term, HEVC will primarily be a technology that is present in next-generation PCs, mobile devices, set-top boxes and TV sets ? which will severely limit Avg. speeds Definition Approximate Bitrate (Mb/s) SD MPEG 2 4.0 SD MPEG 4 1.5 HD MPEG 2 10.0 HD MPEG 4 3.5 HD HEVC 2.0 UHD MPEG-4 28.0 UHD HEVC 15.0 Source: IHS ? 2014 IHS ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 11 its short-term impact on the installed base of devices. It is therefore likely that HEVC will be important for UHD, as the rollout of new devices for the two can occur simultaneously, but HEVC devices are unlikely to achieve the scale that will be necessary to have an impact on HD content for quite some time. SECTION 5: CONCLUSIONS The creation of non-linear, multi-device services has not yet had a fundamental impact on average viewing habits ? linear TV still makes up the vast proportion of viewing, even in heavily developed markets ? but it has had a fundamental impact on how video is distributed, and the business models that underpin this. Delivering unique streams, as requested by the user, opens up a world of new possibilities for personalized advertising and retail, but until these propositions develop, the services will not generate a considerable amount of revenue. Balancing the desire to reach as many consumers as possible on as many devices as possible, without prohibitively large costs of delivery, is a crucial challenge facing the wider industry. The first key point to recognize is that it is simply not possible within any medium term timeframe for unicast, point- to-point delivery to match total consumer demand for video. The total cost of delivering all content uniquely to users would be orders of magnitude above the current costs of all video distribution ? and the additional revenue, if any, from multiscreen simply cannot cover it. When the increased demand for higher quality video is factored in (at an extreme level, UHD at 60fps; at a less extreme level, more widespread HD content), it becomes clear that non-linear viewing simply cannot scale effectively enough, and that increased strain and costs will be placed on networks. HEVC will mitigate the effects of UHD to a degree, as they rollout somewhat simultaneously ? but it will be unable in the short-term to have much of an impact in reducing the load of HD distribution. The second key point is that while ?multicast versus unicast? transport is currently broadly synonymous with ?linear versus non-linear? content, this need not be true in the long-term. A more realistic scenario is that the ?multicast versus unicast? split moves more in line with ?popular versus niche? content. A large proportion of viewing is still concentrated around a small subset of key channels, times of day, and events. When it comes to current niche content that is distributed via the channel model, unicast shares the same benefits as it does with non-linear, and will be increasingly more cost- effective (particularly as CDN costs decline further). This can lead to a dichotomy in distribution: multicast for high viewership content; and unicast IP for lower viewership, more niche content. This scenario may seem inevitable, but there are challenges that must be faced in the short-term. As unicast begins to account for a larger portion of viewing, particularly that of longer-form content, the expectations and demands for higher quality video will continue to rise. Current costs for total CDN video transport are already large, and this usage is currently predicated mainly on SD content, at average speeds of just over 1 Mb/s. There is currently a discord between the degree to which consumers are viewing content that would be ideally serviced by unicast, and the costs which that unicast delivery would impose. Long-term, the business case for a shift to blended linear delivery makes sense ? but technological innovation, at all stages of the delivery chain, will be imperative to help this transition occur. One source of innovation will occur within the network, where unicast and multicast delivery becomes blended, with storage/caching moving increasingly close to the consumer for more popular content. Other solutions include third-party standards for reducing bitrates, whether mainly video specific, such as HEVC, or whether core network technologies. ? 2015 IHS IHS Technology | Consumer Electronics & Video Technology Intelligence Service The future of video distribution 12 IHS Customer Care: Americas: +1 800 IHS CARE (+1 800 447 2273); CustomerCare@ihs.com Europe, Middle East, and Africa: +44 (0) 1344 328 300; Customer.Support@ihs.com Asia and the Pacific Rim: +604 291 3600; SupportAPAC@ihs.com