Forecasting capacity of network edge computing

We have updated this forecast. Check the latest report here

Telco edge build has been slower than expected

Telecoms operators have been planning the deployment of edge computing sites for at least the last three years.

Initially, the premise of (mobile) edge computing was to take advantage of the prime real estate telecoms operators had. Mobile operators, in particular, had undergone a process of evolving their network facilities from sites which housed purpose-built networking equipment to data centres as they adopted virtualisation. The consolidation of networking equipment meant there would be spare capacity in these data centres that could easily host applications for enterprises and developers.

That evolution has now been accelerated by the advent of 5G, a mobile generation built on a software-based architecture and IT principles. The result will be a proliferation of edge data centres that will be used for radio access network and core network hardware and software.

However, the reality is that it has taken time for telcos to deploy these sites. There are multiple reasons for this:

  1. Cost: There is a cost to renovate an existing telco site and ensure it meets requirements common for world-class data centres.
  2. Demand: Telcos are hesitant to take on the risk of building out the infrastructure until they are certain of the demand for these data centres.
  3. 5G roll-out: Mobile operators have been prioritising their 5G RAN roll-out in the last two years, over the investment in edge data centres.
  4. Partnership decisions: The discussion around who to partner with to build the edge data centres has become more complicated, because of the number of partners vying for the role and the entrance of new partners (e.g. hyperscalers) which has slowed down decision-making

Enter your details below to request an extract of the report

Early adopters have taken significant strides in their edge strategy in 2021

2020 and 2021 have been seen as inflection points as a number of leading telecoms operators have launched edge sites: e.g. AT&T, Verizon, Cox Communications, SK Telecom and Vodafone. Arguably, this was triggered by AWS announcing partnerships on AWS Wavelength with four telecoms operators in November 2019, with more recently announced (e.g. Telstra in 2021).

Going forward, key questions remain on the trajectory of telco edge build:

  • How many edge data centres will telcos build and make available for consumer/enterprise applications?
  • How much capacity of telco edge computing will there be globally?
  • How much of telco edge computing will be used for distributed core network functions vs. consumer/enterprise applications?
  • What proportion of telco edge data centre capacity will be taken up by hyperscalers’ platforms?

This report seeks to forecast the capacity at telecoms operators’ edge data centres until 2025 and provide clarity on the nature and location of these sites. In other words, how many sites and servers will be available for running applications and where will these sites be located, both physically and logically in the telecoms operators’ networks.

Before reading this report, we would recommend reading STL Partners’ previous publications on telco edge computing to provide context for some of the key themes addressed, for example:

The report focuses on network edge computing sites

Edge computing comprises of a spectrum of potential location and technologies designed to bring processing power closer to the end-device and source of data, outside of a central data centre or cloud. This report focuses on forecasting capacity at the network edge – i.e. edge computing at edge data centres owned (and usually operated) by telecoms operators.

The initial version of the forecast models capacity at these sites for non-RAN workloads. In other words, processing for enterprise or consumer applications and the distributed core network functions required to support them. Future versions of the forecast will expand to RAN.

Forecast scope in terms of edge locations and workload types

The report covers two out of three scenarios for building the network edge

Table of content

  • Executive summary
  • Introduction
  • There are 3 key factors determining telco edge data centre build out
  • Logically, most network edge will be in the transport aggregation layer
  • Geographically, we will see a shift in the concentration of network edge data centres
  • The limited capacity at network edge DCs will largely be used for edge applications
  • Most telecoms operators are taking a hybrid approach to building their edge
  • Conclusions and next steps
  • Appendix: Methodology

Enter your details below to request an extract of the report

Fixed wireless access growth: To 20% homes by 2025

=======================================================================================

Download the additional file on the left for the PPT chart pack accompanying this report

=======================================================================================

Fixed wireless access growth forecast

Fixed Wireless Access (FWA) networks use a wireless “last mile” link for the final connection of a broadband service to homes and businesses, rather than a copper, fibre or coaxial cable into the building. Provided mostly by WISPs (Wireless Internet Service Providers) or mobile network operators (MNOs), these services come in a wide range of speeds, prices and technology architectures.

Some FWA services are just a short “drop” from a nearby pole or fibre-fed hub, while others can work over distances of several kilometres or more in rural and remote areas, sometimes with base station sites backhauled by additional wireless links. WISPs can either be independent specialists, or traditional fixed/cable operators extending reach into areas they cannot economically cover with wired broadband.

There is a fair amount of definitional vagueness about FWA. The most expansive definitions include cheap mobile hotspots (“Mi-Fi” devices) used in homes, or various types of enterprise IoT gateway, both of which could easily be classified in other market segments. Most service providers don’t give separate breakouts of deployments, while regulators and other industry bodies report patchy and largely inconsistent data.

Our view is that FWA is firstly about providing permanent broadband access to a specific location or premises. Primarily, this is for residential wireless access to the Internet and sometimes typical telco-provided services such as IPTV and voice telephony. In a business context, there may be a mix of wireless Internet access and connectivity to corporate networks such as VPNs, again provided to a specific location or building.

A subset of FWA relates to M2M usage, for instance private networks run by utility companies for controlling grid assets in the field. These are typically not Internet-connected at all, and so don’t fit most observers’ general definition of “broadband access”.

Usually, FWA will be marketed as a specific service and package by some sort of network provider, usually including the terminal equipment (“CPE” – customer premise equipment), rather than allowing the user to “bring their own” device. That said, lower-end (especially 4G) offers may be SIM-only deals intended to be used with generic (and unmanaged) portable hotspots.
There are some examples of private network FWA, such as a large caravan or trailer park with wireless access provided from a central point, and perhaps in future municipal or enterprise cellular networks giving fixed access to particular tenant structures on-site – for instance to hangars at an airport.

Enter your details below to request an extract of the report

FWA today

Today, fixed-wireless access (FWA) is used for perhaps 8-9% of broadband connections globally, although this varies significantly by definition, country and region. There are various use cases (see below), but generally FWA is deployed in areas without good fixed broadband options, or by mobile-only operators trying to add an additional fixed revenue stream, where they have spare capacity.

Fixed wireless internet access fits specific sectors and uses, rather than the overall market

FWA Use Cases

Source: STL Partners

FWA has traditionally been used in sparsely populated rural areas, where the economics of fixed broadband are untenable, especially in developing markets without existing fibre transport to towns and villages, or even copper in residential areas. Such networks have typically used unlicensed frequency bands, as there is limited interference – and little financial justification for expensive spectrum purchases. In most cases, such deployments use proprietary variants of Wi-Fi, or its ill-fated 2010-era sibling WiMAX.

Increasingly however, FWA is being used in more urban settings, and in more developed market scenarios – for example during the phase-out of older xDSL broadband, or in places with limited or no competition between fixed-network providers. Some cellular networks primarily intended for mobile broadband (MBB) have been used for fixed usage as well, especially if spare capacity has been available. 4G has already catalysed rapid growth of FWA in numerous markets, such as South Africa, Japan, Sri Lanka, Italy and the Philippines – and 5G is likely to make a further big difference in coming years. These mostly rely on licensed spectrum, typically the national bands owned by major MNOs. In some cases, specific bands are used for FWA use, rather than sharing with normal mobile broadband. This allows appropriate “dimensioning” of network elements, and clearer cost-accounting for management.

Historically, most FWA has required an external antenna and professional installation on each individual house, although it also gets deployed for multi-dwelling units (MDUs, i.e. apartment blocks) as well as some non-residential premises like shops and schools. More recently, self-installed indoor CPE with varying levels of price and sophistication has helped broaden the market, enabling customers to get terminals at retail stores or delivered direct to their home for immediate use.

Looking forward, the arrival of 5G mass-market equipment and larger swathes of mmWave and new mid-band spectrum – both licensed and unlicensed – is changing the landscape again, with the potential for fibre-rivalling speeds, sometimes at gigabit-grade.

Enter your details below to request an extract of the report

Table of contents

  • Executive Summary
  • Introduction
    • FWA today
    • Universal broadband as a goal
    • What’s changed in recent years?
    • What’s changed because of the pandemic?
  • The FWA market and use cases
    • Niche or mainstream? National or local?
    • Targeting key applications / user groups
  • FWA technology evolution
    • A broad array of options
    • Wi-Fi, WiMAX and close relatives
    • Using a mobile-primary network for FWA
    • 4G and 5G for WISPs
    • Other FWA options
    • Customer premise equipment: indoor or outdoor?
    • Spectrum implications and options
  • The new FWA value chain
    • Can MNOs use FWA to enter the fixed broadband market?
    • Reinventing the WISPs
    • Other value chain participants
    • Is satellite a rival waiting in the wings?
  • Commercial models and packages
    • Typical pricing and packages
    • Example FWA operators and plans
  • STL’s FWA market forecasts
    • Quantitative market sizing and forecast
    • High level market forecast
  • Conclusions
    • What will 5G deliver – and when and where?
  • Index

Gigabit Cable Attacks This Year

Introduction

Since at least May, 2014 and the Triple Play in the USA Executive Briefing, we have been warning that the cable industry’s continuous improvement of its DOCSIS 3 technology threatens fixed operators with a succession of relatively cheap (in terms of CAPEX) but dramatic speed jumps. Gigabit chipsets have been available for some time, with the actual timing of the roll-out being therefore set by cable operators’ commercial choices.

With the arrival of DOCSIS 3.1, multi-gigabit cable has also become available. As a result, cable operators have become the best value providers in the broadband mass markets: typically, we found in the Triple Play briefing, they were the cheapest in terms of price/megabit in the most common speed tiers, at the time between 50 and 100Mbps. They were sometimes also the leaders for outright speed, and this has had an effect. In Q3 2014, for the first time, Comcast had more high-speed Internet subscribers than it had TV subscribers, on a comparable basis. Furthermore, in Europe, cable industry revenues grew 4.6% in 2014 while the TV component grew 1.8%. In other words, cable operators are now broadband operators above all.

Figure 1: Comcast now has more broadband than TV customers

Source: STL Partners, Comcast Q1 2015 trending schedule 

In the December, 2014 Will AT&T shed copper, fibre-up, or buy more content – and what are the lessons? Executive Briefing, we covered the impact on AT&T’s consumer wireline business, and pointed out that its strategy of concentrating on content as opposed to broadband has not really delivered. In the context of ever more competition from streaming video, it was necessary to have an outstanding broadband product before trying to add content revenues. This was something which their DSL infrastructure couldn’t deliver in the context of cable or fibre competitors. The cable competition concentrated on winning whole households’ spending with broadband, with content as an upsell, and has undermined the wireline base to the point where AT&T might well exit a large proportion of it or perhaps sell off the division, refocusing on wireless, DirecTV satellite TV, and enterprise. At the moment, Comcast sees about 2 broadband net-adds for each triple-play net-add, although the increasing numbers of business ISP customers complicate the picture.

Figure 2: Sell the broadband and you get the whole bundle. About half Comcast’s broadband growth is associated with triple-play signups

Source: STL, Comcast Q1 trending schedule

Since Christmas, the trend has picked up speed. Comcast announced a 2Gbps deployment to 1.5 million homes in the Atlanta metropolitan area, with a national deployment to follow. Time Warner Cable has announced a wave of upgrades in Charlotte, North Carolina that ups their current 30Mbps tier to 200Mbps and their 50Mbps tier to 300Mbps, after Google Fiber announced plans to deploy in the area. In the UK, Virgin Media users have been reporting unusually high speeds, apparently because the operator is trialling a 300Mbps speed tier, not long after it upgraded 50Mbps users to 152Mbps.

It is very much worth noting that these deployments are at scale. The Comcast and TWC rollouts are in the millions of premises. When the Virgin Media one reaches production status, it will be multi-million too. Vodafone-owned KDG in Germany is currently deploying 200Mbps, and it will likely go further as soon as it feels the need from a tactical point of view. This is the advantage of an upgrade path that doesn’t require much trenching. Not only can the upgrades be incremental and continuous, they can also be deployed at scale without enormous disruption.

Technology is driving the cable surge

This year’s CES saw the announcement, by Broadcom, of a new system-on-a-chip (SoC) for cable modems/STBs that integrates the new DOCSIS 3.1 cable standard. This provides for even more speeds, theoretically up to 7Gbps downlink, while still providing a broadcast path for pure TV. The SoC also, however, includes a WLAN radio with the newest 802.11ac technology, including beamforming and 4×4 multiple-input and multiple-output (MIMO), which is rated for gigabit speeds in the local network.

Even taking into account the usual level of exaggeration, this is an impressive package, offering telco-hammering broadband speeds, support for broadcast TV, and in-home distribution at speeds that can keep up with 4K streaming video. These are the SoCs that Comcast will be using for its gigabit cable rollouts. STMicroelectronics demonstrated its own multigigabit solution at CES, and although Intel has yet to show a DOCSIS 3.1 SoC, the most recent version of its Puma platform offers up to 1.6Gbps in a DOCSIS 3 network. DOCSIS 3 and 3.1 are designed to be interoperable, so this product has a future even after the head-ends are upgraded.

Figure 3: This is your enemy. Broadcom’s DOCSIS3.1/802.11ac chipset

Source: RCRWireless 

With multiple chipset vendors shipping products, CableLabs running regular interoperability tests, and large regional deployments beginning, we conclude that the big cable upgrade is now here. Even if cable operators succeed in virtualising their set-top box software, you can’t provide the customer-end modem nor the WiFi router from the cloud. It’s important to realise that FTTH operators can upgrade in a similarly painless way by replacing their optical network terminals (ONTs), but DSL operators need to replace infrastructure. Also, ONTs are often independent from the WLAN router or other customer equipment , so the upgrade won’t necessarily improve the WiFi.

WiFi is also getting a major upgrade

The Broadcom device is so significant, though, because of the very strong WiFi support built in with the cable modem. Like the cable industry, the WiFi ecosystem has succeeded in keeping up a steady cycle of continuous improvements that are usually backwards compatible, from 802.11b through to 802.11ac, thanks to a major standards effort, the scale that Intel and Apple’s support gives us, and its relatively light intellectual property encumbrance.

802.11ac adds a number of advanced radio features, notably multiple-user MIMO, beamforming, and higher-density modulation, that are only expected to arrive in the cellular network as part of 5G some time after 2020, as well as some incremental improvements over 802.11n, like additional MIMO streams, wider channels, and 5GHz spectrum by default. As a result, the industry refers to it as “gigabit WiFi”, although the gigabit is a per-station rather than per-user throughput.

The standard has been settled since January 2014, and support is available in most flagship-class devices and laptop chipsets since then, so this is now a reality. The upgrade of the cable networks to 802.11ac WiFi backed with DOCSIS3.1 will have major strategic consequences for telcos, as it enables the cable operators and any strategic partners of theirs to go in even harder on the fixed broadband business and also launch a WiFi-plus-MVNO mobile service at the same time. The beamforming element of 802.11ac should help them to support higher user densities, as it makes use of the spatial diversity among different stations to reduce interference. Cablevision already launched a mobile service just before Christmas. We know Comcast is planning to launch one sometime this year, as they have been hiring a variety of mobile professionals quite aggressively. And, of course, the CableWiFi roaming alliance greatly facilitates scaling up such a service. The economics of a mini-carrier, as we pointed out in the Google MVNO: What’s Behind It and What Are the Implications? Executive Briefing, hinge on how much traffic can be offloaded to WiFi or small cells.

Figure 4: Modelling a mini-carrier shows that the WiFi is critical

Source: STL Partners

Traffic carried on WiFi costs nothing in terms of spectrum and much less in terms of CAPEX (due to the lower intellectual property tax and the very high production runs of WiFi equipment). In a cable context, it will often be backhauled in the spare capacity of the fixed access network, and therefore will account for very little additional cost on this score. As a result, the percentage of data traffic transferred to WiFi, or absorbed by it, is a crucial variable. KDDI, for example, carries 57% of its mobile data traffic on WiFi and hopes to reach 65% by the end of this year. Increasing the fraction from 30% to 57% roughly halved their CAPEX on LTE.

A major regulatory issue at the moment is the deployment of LTE-LAA (Licensed-Assisted Access), which aggregates unlicensed radio spectrum with a channel from licensed spectrum in order to increase the available bandwidth. The 5GHz WiFi band is the most likely candidate for this, as it is widely available, contains a lot of capacity, and is well-supported in hardware.

We should expect the cable industry to push back very hard against efforts to rush deployment of LTE-LAA cellular networks through the regulatory process, as they have a great deal to lose if the cellular networks start to take up a large proportion of the 5GHz band. From their point of view, a major purpose of LTE-LAA might be to occupy the 5GHz and deny it to their WiFi operations.

  • Executive Summary
  • Introduction
  • Technology is driving the cable surge
  • WiFi is also getting a major upgrade
  • Wholesale and enterprise markets are threatened as well
  • The Cable Surge Is Disrupting Wireline
  • Conclusions
  • STL Partners and Telco 2.0: Change the Game 
  • Figure 1: Comcast now has more broadband than TV customers
  • Figure 2: Sell the broadband and you get the whole bundle. About half Comcast’s broadband growth is associated with triple-play signups
  • Figure 3: This is your enemy. Broadcom’s DOCSIS3.1/802.11ac chipset
  • Figure 4: Modelling a mini-carrier shows that the WiFi is critical
  • Figure 5: Comcast’s growth is mostly driven by business services and broadband
  • Figure 6: Comcast Business is its growth start with a 27% CAGR
  • Figure 7: Major cablecos even outdo AT&T’s stellar performance in the enterprise
  • Figure 8: 3 major cable operators’ business services are now close to AT&T or Verizon’s scale
  • Figure 9: Summary of gigabit deployments
  • Figure 10: CAPEX as a % of revenue has been falling for some time…