Mobile spectrum and network evolution to 2025 slides coleago - 24 mar 21

Mobile Services, Spectrum and Network
Evolution to 2025
March 2021
Stefan Zehle, MBA
Tel: +44 7974 356 258
stefan.zehle@coleago.com
CEO, Coleago Consulting Ltd
Nick Fookes, MSc, CFA
Tel: +44 7710 350 816
nick.fookes@coleago.com
Managing Consultant, Coleago Consulting Ltd

Mobile Spectrum and Network Evolution to 2025
1. Mobile services and consumption in 2025
2. The business case for 5G
3. Impact on society
4. Evolution of mobile networks and technology
5. Spectrum demand 2020-2025
6. Spectrum management and pricing
7. Concluding remarks
Appendix: Overview of UMT bands
1
Contents
© copyright Coleago Consulting, 2021, all rights reserved
Note: this presentation accompanies Coleago’s report ‘Mobile Services, Spectrum and Network Evolution
to 2025’, March 2021. Full credits for images and references to external materials shown in this
presentation are provided in the report.

What will the world look like in 2025?
Mobile services and consumption
2
1 © copyright Coleago Consulting, 2021, all rights reserved

What will the world look like in 2025?
Total global mobile traffic in 2025 will be between 3-10x the levels
reached in 20202
3
In 2025, 1.5x more
data will be created
each year than is
contained in the entire
Digital Universe
today1.
‒‒
Operators will shift
from selling GBytes to
selling data speeds
and performance.
‒‒
More mobile spectrum
will be added in the
next 5 years than in
the last 30.
‒‒
5G capabilities could
allow lower-income
countries to narrow
the digital gap with
advanced economies.
 Video will continue to dominate (already accounts for
~80% of total fixed and mobile traffic)
 Ultra HD will account for over 20% of all video traffic
 Cloud gaming will account for 25% of total 5G traffic
 Mobile already accounts for half of all web-traffic, and its
share will continue to grow
 Immersive 360-degree video and gaming will be
commonplace, adding to demand for low latency
communications (to avoid motion sickness)
 5G network slicing will open up network access to a very
broad range of specialised service-providers catering for
key industries and verticals
 Globally, there will be 5.1 billion IoT devices with
embedded cellular connectivity (~20% of all IoT), 3.4x
more than in 2019
Sources: Coleago, IDC, Ericsson, GSMA and ITU forecasts (excluding M2M traffic)
The main drivers of the continued explosion in mobile data usage
growth are: (1) increased video usage combined with higher
resolutions; and (2) mobile cloud gaming
 Increased adoption of mobile data services will have significantly less
impact on total consumption than the above
 The explosion in cellular IoT will have a significant impact on society,
but will not contribute much to mobile traffic and connectivity revenues
– However, IoT yields opportunities for operators to create value
beyond connectivity, by moving up the value chain (discussed later)
Data consumption per 1
hour of streaming
4x global 2020 traffic
2.8 ZB annualized
4% of total data created
Ericsson view,
global mobile
9.5x global 2020 traffic
6.5 ZB on an annualized basis
10% of total data created in 2025
Upper bound,
global mobile
Up to 8x 2020 developed market traffic
Coleago view,
developed
markets
3x global 2020 traffic
1.9 ZB annualized
3% of data created
Lower bound,
global mobile
9.5x
8x
4x
3x
1 IDC estimates the Digital Universe at 44 zettabytes (5.6 TB per capita), roughly doubling in size every 2 years
2 Upper bound based on ITU (2015); lower bound based on GSMA (2020)

Video and cloud gaming are the main applications driving the explosion in mobile data use
4
A third of US
consumers tried
eGaming or viewed
eSports for the first
time during 20201.
‒‒
Mobile executives
believe that cloud
gaming may reach
25% of 5G traffic by
20222.
‒‒
There are already 2
billion mobile gamers
worldwide today3.
iPhone 12 Pro video settings
 The above shows how easily data usage can be boosted by
simply flicking a switch
 Taking and sharing one minute of 4k HD video at 60 frames per
second (FPS) consumes almost 10x more data than 720p HD
video at 30 frames per second, the lowest quality setting on the
iPhone 12
 One hour of HD 2K video streaming consumes 4x more data than
an hour of SD video. By 2025, 8K video and 24K 3D VR will be
the new ‘HD’
(Lowest quality on iPhone12)
1.4x the lowest quality
Impact on data usage:
4DReplay Brings 360-Degree Coverage to KBO League (May 2020)
5G Mobile and Fixed Wireless Access will enable UHD cloud gaming and VR
M2Consumer and M2Home: 360-degree video, gaming and VR
1,3 Deloitte research ; 2 MOVIC Livecast, April 2019

Mobile Private Networks and 5G network slicing
M2Business: Mobile Private Networks
Private Network (MPN) provides dedicated connectivity for an
enterprise’s specific sites and locations, supporting:
 Mobile Edge Computing (MEC), bringing processing power
and control close to the user for low latency and high security
 Applications, such as Enterprise-to-Enterprise (E2E) IoT
solutions which run on the network.
5
5G network slicing will
drive service diversity.
‒‒
Private enterprise
networks will explode
between 2020 and
2025, becoming a
battleground between
telcos and cloud
companies1.
‒‒
5G network slicing
could allow mobile
operators to at least
retain a portion of the
connectivity piece,
from where they could
also seek to move up
the enterprise applica-
tions’ value-chain.
Mobile and IoT2Government: emergency services and city security
Over the next 5 years, blue-light services (police, fire and ambulance
services) will rely increasingly on public mobile networks, given:
 The high cost of maintaining a dedicated private network
infrastructure for a relatively small base of users
 The high site redundancy within public networks (overlapping
coverage provided by capacity sites, especially in urban areas)
 The possibility to establish secure, private virtual networks over
public mobile infrastructure and systems
Emergency Services will have prioritized access to mobile bandwidth
both for their critical communications and IoT applications
MEC
Factories Ports Utilities
Mines Digital offices
Oil rigs
• Sensor
connectivity
• Connected
workers
• Automated
guided
vehicles
• Remote
controlled
vehicles
• Video
surveillance
• Sensor
connectivity
• Predictive
maintenance
• Push-to-talk
• Remote
controlled
and
autonomous
vehicles
• Location
tracking
• On-site
connectivity
and data
processing
• Sensor
connectivity
• Smart
buildings
• Video
surveillance
and analytics
Wireless camera
Easy deployment, no
coverage holes
Patrol car 2
Integrated command
system
Patrol car 1
5G vehicle-mounted
terminals + cameras
Command center
Convenient video viewing
Handheld
terminal
Video transmission
(UL/DL)
UAV
3D surveillance and
real-time video
Wireless
bodycam
Live-stream video
and detect gunshots
1 Key GSMA prediction for 2025

IoT: huge growth in connected devices
IoT2Government, Business and Consumer: smart health
>50% of people aged 55+ in high-income countries will have a
connected health device. This will provide a tremendous boost to
public-health management capabilities and efficiency3.
IoT2Consumer and IoT2Business: smart vehicles
6
Cellular IoT devices
will grow by 3.4x
between 2019-2025.
‒‒
5G will be the first
mobile technology that
impacts more on
industry than on
consumers1.
‒‒
76% of enterprise IoT
projects are described
as mission-critical,
with entire businesses
depending on IoT in
8% of cases2.
Unless operators move beyond simple data conveyance, they will
enable rather than participate in the huge value creation from IoT
 Cellular IoT will account for a growing share of machine-type
connections. Ericsson projects around 5.1 billion cellular IoT
devices by 2025, accounting for a fifth of all IoT connections (up
from 12% in 2019)3
 While cellular IoT will deliver very high incremental value to society,
the connectivity element is likely only to have a moderate impact on
operator revenues
 Although the total number of cellular IoT connections is set to be
large, most devices will only consume very small amounts of data;
however, many cellular IoT applications will add to demand for:
– Ubiquitous mobile coverage, both indoors and in remote
locations; and
– Ultra-low latencies to support critical IoT communications
Performance requirements for cellular IoT
 Much of the growth in cellular IoT will come from low-cost asset
trackers, smart vehicles, sensors, cloud robots, cloud AR and VR,
and advanced cloud gaming
IoT category Throughput Latency
Enhanced Machine-Type communications (eMTC) 1-6Mbit/s 20ms
Ultra-reliable, low latency communications (uRLLc) 2-6Mbit/s 10ms
1,3 Key GSMA predictions for 2025 ; 2 Vodafone IoT barometer, 2019; 3 Ericsson

of total unique data connections (2020, 2025)
Key mobile service categories and global adoption1
Key mobile service categories in the 5G era
Global cellular IoT adoption
7
eMBB will remain the
prime driver of mobile
connections, traffic
and revenues.
‒‒
Cellular IoT will
approach eMBB
penetration, but will
contribute a fraction of
traffic and connectivity
revenues.
‒‒
Dedicated FWA will
account for a fraction
of data connections,
but may drive 20% of
traffic.
‒‒
FWA will likely have a
bigger impact in
emerging markets2
Global eMBB adoption
Global FWA adoption
8%
11%
18%
24%
33%
45%
55%
19%
22%
28%
35%
44%
55%
64%
0%
10%
20%
30%
40%
50%
60%
70%
2019 2020 2021 2022 2023 2024 2025
As
%
global
population
Massive MTC BB and Critical MTC
Total excluding legacy Total including legacy
43%
51%
57%
60%
64%
67% 69%
73% 73% 74% 75% 76% 76% 77%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
2019 2020 2021 2022 2023 2024 2025
%
population
Unique eMBB adoption Total unique mobile, excl. FWA
0.7%
0.8%
1.0%
1.2%
1.4%
1.6%
1.9%
0.0%
0.2%
0.4%
0.6%
0.8%
1.0%
1.2%
1.4%
1.6%
1.8%
2.0%
2019 2020 2021 2022 2023 2024 2025
%
population
1 Based on Ericsson global forecasts
2 Where fixed broadband infrastructure is typically weaker
%
81% 54%
1% 2%
18% 44%
(eMBB) (FWA)

Contributions to total traffic1
eMBB usage per capita and per unique eMBB user
Cellular IoT data usage per capita and per IoT connection
8
eMBB will continue to
account for the bulk of
mobile data usage
and growth, followed
by FWA (20% in
2025).
‒‒
IoT is not a big deal as
far as total traffic is
concerned
‒‒
But it will be huge for
society and it could be
a big deal for operator
revenues if they can
move up the value
chain (discussed
later).
FWA usage per capita and per FWA connection
Relative contributions to total mobile traffic
1 Based on Ericsson global forecasts
4.4
6.4
8.8
11.3
14.7
18.4
22.7
10
13
15
19
23
27
33
0
5
10
15
20
25
30
35
2019 2020 2021 2022 2023 2024 2025
GBytes
per
month
eMBB usage per capita eMBB usage per unique eMBB user
0.8 1.2
1.8
2.6
3.7
5.1
6.4
124
151
183
222
267
311
337
0
50
100
150
200
250
300
350
400
0
1
2
3
4
5
6
7
8
2019 2020 2021 2022 2023 2024 2025
GBytes
per
month
per
connection
GBytes
per
month
per
capita
FWA usage per capita FWA usage per connection
0.1 0.3
0.5
0.9
1.4
0.2
0.3
0.6
0.8
1.2
1.7
2.3
0.0
0.5
1.0
1.5
2.0
2.5
2019 2020 2021 2022 2023 2024 2025
GBytes
per
month
IoT traffic per capita IoT traffic per device
88% 88% 85% 81% 78% 76% 75%
11% 11% 14% 17% 19% 20% 20%
3% 4% 5%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2019 2020 2021 2022 2023 2024 2025
% eMBB % FWA % IoT

Ratio of 5G to 4G AUPU
5G versus legacy AUPU1
5G versus legacy AUPU in Korea (2019)
Mobile subscriptions by technology (Ericsson)
9
Early 5G adopters are
invariably far heavier
data users on
average.
‒‒
As 5G penetrates the
wider base, the 5G vs.
4G AUPU multiple
should decline.
‒‒
By 2025, we expect
the global 5G AUPU
multiple to converge to
around 2x the 4G
AUPU.
‒‒
In many markets, 4G
is currently overloaded
while 5G capacity is
underutilised.
5G versus legacy AUPU (September 2020)
1 AUPU = Average Usage per Unit in Gbytes per month;
data from MSIT, Strategy Analytics, Open Signal and Ericsson
15
10
5
20
25
30
0
GBytes
per
month
 Korea: in 2019, 5G traffic accounted for 21% of
the total from 6.8% of connections, implying a 5G
AUPU multiple > 3x legacy AUPU
– By September 2020, the AUPU multiple fell to
2x as 5G started to penetrate the wider base
 Globally, we expect the 5G AUPU multiple to
converge to 1.9x by 2025
– With higher multiples in low-income countries
(driven by FWA and the small proportion of
more affluent 5G users)
 4G is overloaded in many markets (especially
S.E. Asia) while 5G capacity is underutilised
– This creates challenges for operators
(discussed later)

Total mobile traffic and regional split
Total global data usage per capita1 (incl IoT)
 The GSMA’s forecast (taken as our lower bound) implies China
traffic in 2025 will be 2.5x that in 2020, while the multiple in sub-
Saharan Africa will be 7.4x
– Scaled Ericsson forecasts imply multiples of 3.4x in China and
10x in sub-Saharan Africa
– Scaled ITU forecasts (taken as our upper bound) imply multiples
of 13.5x in China and 18x in sub-Saharan Africa
 Globally, 5G is projected to reach almost half of all data usage
 But given the explosion in total data traffic, capacity required in
2025 from legacy technologies will be at least 2x that utilised in
2020 (legacy capacity investment is discussed in later slides)
10
Depending on the
country, GSMA
projects mobile data
traffic between 3 to 7
times 2020 levels.
‒‒
Even Ericsson may be
conservative: the
ITU’s forecast implies
2.3x higher traffic.
‒‒
Almost half of all
global traffic will be 5G
by 2025.
‒‒
Yet 2x as much
capacity from legacy
technologies will be
needed in 20252
(discussed later)
Regional usage per capita (GSMA, Ericsson ‒ excl IoT)
1 Ericsson (2020) with added IoT traffic projection (Coleago, GSMA)
2 Based on Ericsson global forecast; impact of 4G capacity need discussed later
5
8 7 6 8
3 5 3
0.4
20
43
30 29
24
19 19 18
3
29
61
44
41
34
27 27 25
5
0
10
20
30
40
50
60
70
Global
(GSMA)
North
America
Europe
CIS
China
MENA
Asia
Pacific
Latin
America
Sub-Sahara
GBytes
per
capita
/
month
2019 2025 2025, scaled Ericsson
26%
32% 30% 29%
20%
35%
26%
34%
49%
34%
40%
38% 37%
28%
44%
34%
42%
59%
0%
10%
20%
30%
40%
50%
60%
70%
Global
(GSMA)
North
America
Europe
CIS
China
MENA
Asia
Pacific
Latin
America
Sub-Sahara
Total
mobile
data
traffic
GSMA Scaled, Ericsson
Regional scaling
of Ericsson’s
global forecast
assumes that the
GSMA is correct
on the regional
split.
CAGR 2019-2015
5.3
7.6 9.6 11.2
13.5 15.3 16.3
0.1
1.2
3.0
5.5
9.0
14.3
5.3
7.7
10.8
14.2
18.9
24.4
30.6
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
2019 2020 2021 2022 2023 2024 2025
GBytes
per
month
Total legacy traffic per capita total 5G per capita

Speed-experience targets add substantially to the site design
load (i.e. capacity need)
 e.g. twice as much capacity is needed to deliver 100MB each to
two users in the same second than one second apart
 To deliver 4x more traffic with a 100Mbit/s speed target with 99%
probability, a site may require 8x its current capacity in 2025.
Impact of speed targets on the site design load (capacity need)
 We assume that the 100Mit/s speed target (with 99% probability)
drives capacity by an average factor of 2x across the network
Operators will shift from selling data volumes (GBytes) to selling speeds (Mbits/s)
Network performance requirements
Capacity need in 2025 as a multiple of capacity utilised in 20202
11
A bit of high-speed
data is more costly to
produce than a slow
bit.
‒‒
In an era of unlimited
data plans, quality of
the data-experience
becomes the point of
differentiation. The
days of ‘best effort’1
service are numbered.
‒‒
Meeting both the
projected demand and
the ITU’s IMT 2020
Requirements
presents huge
challenges for
operators.
Throughput Latency Area capacity Mobility
eMBB 50-100Mbit/s 20ms - -
Gaming at 1080p, 60FPS 50Mbit/s <20ms - -
Gaming at 2k, 60FPS 100Mbit/s <10ms - -
Gaming at 4k, 60FPS 200Mbit/s <5ms - -
FWA up to 1Gbit/s <5ms? - N/A
IMT Advanced (ITU) 10Mbit/s 10ms 0.1Mbit/s/m2 350Km/hr
IMT 2020 (ITU) 100Mbit/s 1ms 10Mbit/s/m2 500Km/hr
0
2,000
4,000
6,000
8,000
10,000
12,000
0 500 1,000 1,500 2,000 2,500 3,000 3,500
Site
design
load
in
Mbits/s
Site traffic in Busy Hour (GBytes/Site/hour)
100Mbps, 99% probability 100Mbps, 95% probability
50Mbps speed target 20Mbps speed target
Best effort
3.7x best effort
1.5x best effort
3.2x
4.0x 3.7x 3.6x
2.5x
4.5x
3.2x
4.3x
7.4x
8.3x
10.7x
9.9x 9.9x
6.8x
11.5x
8.3x
11.3x
17.8x
16.5x
21.3x
19.7x 19.7x
13.5x
23.0x
16.6x
22.6x
0x
5x
10x
15x
20x
25x
30x
Global
North
America
Europe
CIS
China
MENA
Asia
Pacific
Latin
America
Sub-Sahara
Capacity
in
2025
as
multiple
of
2020
GSMA best effort ITU (2015) best effort
GSMA 100Mbps target ITU 100Mbps target
35.5x
Data speed
target
combined with
Total
consumption
drives
Capacity need
1 ‘Best effort’ data provision means delivery of a given quantity of GBytes
during the busy hour without regard to data-speed experience requirements.
2 ITU global forecast apportioned pro-rata by region
based on GSMA forecast of relative regional traffic.

To meet the 2025 demand for capacity, operators will need to:
The massive growth in capacity-need by 2025 presents huge challenges for operators
12
Even the conservative
GSMA forecasts imply
huge growth in
network capacity ‒
whether or not data
speed targets are
explicitly taken into
account.
‒‒
This makes for
challenging operator
business cases
(discussed next).
Acquire and deploy far more
spectrum, and re-farm existing 2G/
3G spectrum (initially) to 4G1
And deploy technology
enhancements
And roll out many more
sites
 Need low band for coverage quality
and to improve cell-edge
performance
 Higher bands for capacity
 Sectorisation: extra capacity
 Massive MIMO: boost capacity and
overall network performance
 Macro sites, micro sites and small
cells
1 Albeit sufficient legacy bandwidth needs to be maintained to support legacy devices. 2 Globally: doubling of 4G capacity need
2020-2025, as discussed in previous slides. 3 Unlike 3G investments in the past, which did not offer a smooth transition to 4G.
Big issue: new higher mid-bands and mmWaves are 5G only, so these do not address the need for far
more 4G capacity between 2020 and 20252
 In the absence of (significant) extra 4G spectrum, operators need to densify the network and deploy
technology enhancements, just for 4G
 The good news is that these will not become stranded 4G-only assets3: new radio equipment and MIMO
antenna systems offer a smooth upgrade path to 5G (via software upgrade)
 Therefore, near-term deployments for 4G will also support future 5G capacity requirements: the extra sites
and massive MIMO will in any case still be needed

Mobile network operator perspective
The business case for 5G
13

Current state of the industry1
Mobile market returns (in real terms, median values)
 Real-term revenues and returns have fallen over the last 10 years
 Global revenues fell by a further -2.7% in real terms in 2020; the
majority of markets experienced declines in nominal terms ‒ only
North America showed growth in real terms (around 1%)
14
5G investments come
at a stage when
mobile revenues in
most markets are
declining in real terms.
‒‒
Most operators in our
sample are achieving
returns (ROIC) below
than the cost of capital
(WACC).
‒‒
This is indicative of an
industry under
pressure, and is
unsustainable in the
long run.
The sustainability of the industry in its current form hinges on
whether 5G will allow industry to buck this downward historical
trend and cover the heavy investments that lie ahead
Returns (ROIC) versus cost of capital (WACC)2
 Only 4 out of the 10 operators in our sample are currently earning
their cost of capital (meaning ROIC is greater than the WACC)
 These are mostly leading players, with strong positions within their
key markets and may not be representative of industry as a whole
 The situation for later entrants (the market challengers) will likely be
worse, hence blended industry ROIC in most markets may be lower
than suggested above
1 Based on data from Bank of America Merrill Lynch and gurufocus.com
2 ROIC: Return on Invested Capital; WACC: Weighted Average Cost of Capital
0%
20%
40%
60%
80%
100%
120%
140%
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
2010
=
100%
0%
5%
10%
15%
20%
25%
30%
35%
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
%
of
service
revenues
Low and lower-middle income markets
Upper-middle income markets
High income markets
-20%
-15%
-10%
-5%
0%
5%
10%
15%
Vodafone
Group
MTN
America
Movil
Bharti
Cellular
Vodafone-Idea
Telefonica
Verizon
T-Mobile
USA
Deutsche
Telecom
Vodacom
Median
ROIC WACC ROIC minus WACC
EBITDA minus Capex as % service revenues
Mobile market revenue indices
Industry
returns
have fallen
most in
lower-
income
countries

5G investment requirements
Cumulative capex by region 2020-2025 (in $ millions)1
15
Average annual
mobile capex between
2020 and 2025 is
projected to exceed
17% of annual mobile
revenues, almost 80%
of which will be for 5G.
‒‒
Heavy investments
against a historic
backdrop of declining
returns would normally
be harder to justify.
‒‒
The key question is
whether 5G will be
transformative for
operator returns in a
way that 4G was not.
To cater for the explosive growth in data consumption, operators
are continuing to invest large sums in 4G and 5G radio access
networks and backhaul infrastructure
 The GSMA projects cumulative world-wide investment by mobile
operators between 2020 and 2025 to reach $1.1 trillion, almost 80%
of which will be in 5G
 The vast majority of this investment is in the radio access network
(RAN), notably cell sites, 4G / 5G radios, and backhaul
 Most 4G RAN investment currently taking place is software
upgradable to 5G
 Preparing for the launch of 5G, several operators started to deploy
Massive MIMO in combination with three-carrier aggregation,
delivering Gbit/s speed capabilities
 In the 5G era, networks are anticipated to have many more small-
cell sites, because much of the 5G traffic will be carried on higher
frequencies which have a shorter range
– Small cells will account for between 3x and 10x the total current
outdoor site count, not including indoor coverage solutions
– e.g. Frontier Economics indicates a total need of 300,000 small-
cell sites in the UK, roughly 10x the current aggregate site count
– Between 800,000 and 2 million small cells may be required to
make 5G a reality in the US2
 On the positive side, operators will find some savings as they move
to virtualised networks and increased infrastructure sharing
– However, operating a mobile network with a factor increase in
the number of cell sites remains a major network operating-cost
challenge
1 GSMA
2 FCC Commissioner statement (800,000), IDC estimate (2 million)
Emerging neutral host models offer opportunities for far more
extensive sharing of assets. These would help address the
massive capex and opex challenges facing operators.

The GSMA projects continued real-term decline in mobile revenues
 The GSMA Intelligence Consumer Insights Survey 2019 indicates that
a majority of early adopters in all regions may be willing to pay a
premium for 5G
– Globally, 41% of respondents who intended to upgrade would be
willing to pay up to 10% more then currently
– A further 14% would be willing to pay up to 20% more
 Notwithstanding this, the GSMA’s own global projections do not appear
to reflect any significant upward inflection in the revenue trend
– A premium for 5G ‘in-itself’ cannot be sustained as the technology
matures: operators have an incentive to drive users to the more
efficient technology, allowing rapid re-farming of legacy bands to 5G
– Maintaining a premium for mere technology adoption would run
counter to this
 Indeed, most operators did not gain additional revenue from 4G
compared to 3G
– For example, when Vodafone India launched 4G, customers with
4G devices and a 4G SIM received 2 GBytes of data for the same
price that 3G customers pay for only 1 GByte of data
The early evidence from 5G tariff plans suggests that not only will
consumers not pay more for eMBB, but they will also get larger data
buckets and faster speeds
 A degree of caution is indeed warranted, and regulators need to be
mindful not to increase the financial burden on operators
Package
Type
5G 4G
Tariff
KRW
Data
pack
Limit after
out of
pack
Tariff
KRW
Data
pack
Limit
after out
of pack
LGU+ Entrance 55,000 9GB 1Mbps 55,900 6.6GB 3Mbps
Middle 75,000 150GB 5Mbps 74,800 16GB 3Mbps
High 85,000 Unlimited Unlimited 88,000 30GB 3Mbps
Premium 95,000 Unlimited Unlimited 110,000 40GB 3Mbps
SKT Entrance 55,000 8GB 1Mbps 50,000 4GB 5Mbps
Middle 75,000 150GB 5Mbps 69,000 100GB 5Mbps
High 95,000 Unlimited Unlimited 79,000 150GB 5Mbps
Premium 125,000 Unlimited Unlimited 100,000 Unlimit. n/a
KT Entrance 55,000 8GB 1Mbps 49,000 3GB 1Mbps
Middle 80,000 Unlimited Unlimited 69,000 100GB 5Mbps
High 100,000 Unlimited Unlimited 89,900 Unlimit. 5Mbps
Premium 130,000 Unlimited Unlimited n/a n/a n/a
Future industry revenue prospects (1)
GSMA view on global mobile revenue evolution
5G vs. 4G data pricing in Korea, shortly after the launch of 5G
16
Early signs suggest
that as with 4G, 5G
customers will get
larger data allowances
and better speeds
without paying more.
‒‒
However, there may
be other opportunities
for revenue growth in
the 5G era.
92%
94%
96%
98%
100%
102%
104%
106%
108%
110%
2020 2021 2022 2023 2024 2025
Index
(2020
=
100%)
Developed markets, nominal Developing markets, nominal
Global, nominal Global, real
1.2% nominal growth per year
versus 6.2% nominal GDP growth (IMF)
-0.4% real decline per year
versus 4.1% real GDP growth
Nevertheless, we believe that there are areas of potential revenue
growth that may offset increases in network costs (discussed next)

Network slicing allows for the provision of ‘Carrier-grade’
performance ‒ i.e. with guaranteed Quality of Service (QoS)
 Will power critical communications needs for Enterprise and key
industry verticals
 Enable competition between operators and cloud companies
 Allow premium-revenue generation from Quality of Service
provision to Enterprise (other forms of QoS-based pricing
discussed later)
5G FWA represents a potential new revenue stream, however
this is unlikely to contribute more than 10% to mobile revenues
 eMBB will likely drive a high degree of fixed (wired) broadband
substitution, but hard to monetise in an ‘unlimited data’ context
 Dedicated FWA connections ~3% of global connections in 2025
(Ericsson); need to price competitively (we estimate max ~2.5x
current revenues per unique mobile user to compete with FBB1)
But Enterprise solutions offer substantial new opportunities
 ‘Enterprise eMBB’ accounted for within general eMBB forecast ‒
subject to similar deflationary pressures as consumer eMBB
 ‘Enterprise IoT’ accounted for within cellular IoT forecasts
(discussed next slide)
17
FWA is unlikely to
drive significant
operator returns, but
Enterprise represents
a major opportunity.
‒‒
Nokia CEO view:
private 5G spend will
outpace traditional
networks in next 10
years3.
Retaining a share of
this through Mobile
Private Networks
could be very valuable
for operators.
1 Note also that revenues per Gbyte for FWA are likely to be lower than
for eMBB, and profit margins from FWA are likely to be lower too.
Total operator network resources
Shared
resources
B2C, B2B
Dedicated
resources
Enterprise
 No Quality of Service
guarantee
 Lower prices for access
to shared resource
 Network slices with
SLA2
 Dedicated resources
enable QoS guarantee
2 SLA: Service Level Agreements.
3 At Shanghai MWC, February 2021.
Mobile
Private
Networks
Enterprise
IoT
Enterprise
eMBB
Enterprise
eMBB Cellular IoT
Already exists;
deflationary (as
with all eMBB)
New revenue
streams

Unless operators move up the IoT value chain, operators will be
mere enablers of (rather than participants in) the value created
by cellular IoT
GSMA perspective on the global IoT value chain ($ billions)
 Cellular IoT connectivity will account for ~5% of global mobile
revenues and traffic in 2025
 But operators could be well-placed to move up the value chain, as
few players can assume the burden of telecoms/IoT regulations
18
The total IoT value
chain is worth roughly
the same size as total
current global mobile
market revenues ($1.1
trillion).
Case study: Vodafone IoT
 Pursuing global coverage, through partnerships with rival
operators including China Mobile and America Movil, providing
IoT connectivity in “some of the most complex regulatory markets”
 Strategic partnership with ARM, which “removes cost and
complexity for OEMs developing connected products, and
solutions that deliver high-value business outcomes, such as
stolen vehicle tracking and assisted living”
 Active IoT communications campaign (‘Let’s talk IoT’, Jan 2021)
Relative positions within the IoT space (Gartner)
Approximately = global mobile revenues

19
Speed and priority-
based pricing would
allow operators to
better meet the
specific needs of
individual customers,
and to generate
revenues that are
closer to customers’
willingness to pay.
‒‒
This could drive
significant growth in
average revenues per
connection.
QoS1-based pricing is good for operators and consumers alike
 It is more expensive to produce a fast, reliable bit than a slow one
 Uniform pricing with ‘best-effort’ service is deeply inefficient
 QoS-based pricing solves freeriding issue: those not willing to meet
the cost of high-speed data would adjust their usage, such that
neither operators nor fellow customers would bear the cost
Speed-based pricing already exists in fixed broadband
Average EU fixed broadband prices (€ per month)
Another comparable is the airline industry, where overbooking
is common in all classes
 ‘Gold’ or ‘Platinum’ cardholders are always the last to lose their
seats if all ticket-holding passengers do actually turn up
The 3GPP standards for 4G and 5G (release 12 and above)
already provide for prioritized access to bandwidth
 For example, emergency services/PPDR use of public mobile
networks would always be prioritized over regular customers
 Customers could subscribe to ‘gold’, ‘silver’ and ‘best-effort’
packages, each with distinct tiered prices; in the event of
congestion, ‘gold’ subscribers get first call on network resources
What would happen if Ferraris and Fiat Pandas
were sold at the same average price?
 No new Ferraris would be produced, because costs would exceed prices
 The excess demand for Ferraris could not possibly be fulfilled
 Even those willing to pay the true value for a Ferrari could not get one
 Demand for Pandas would be depressed due to their excessive price
Total seller revenues and returns would be reduced
and customers would be unhappy
Images from eBay and Mini Model Shop
1.1x
1.5x
1.8x
0.0x
0.2x
0.4x
0.6x
0.8x
1.0x
1.2x
1.4x
1.6x
1.8x
2.0x
€0
€5
€10
€15
€20
€25
€30
€35
€40
12-30Mbps 30-100Mbps 100-200Mbps >200Mbps
Multiple of 12-30Mbps price
There is evidence from the 2G era of customer willingness to pay for quality and reliability: incumbents with a coverage advantage
could charge a premium, even though their customers would rarely (if ever) venture outside the footprint of rivals operators
1 QoS: Quality-of-Service

20
Industry should target
revenue-growth in line
with the annual growth
in GDP.
‒‒
Nevertheless, policy-
makers should not
assume that 5G will
automatically deliver
returns for operators
that exceed their cost
of capital.
The risk remains high
that general price
erosion caused by
intense competition
will dominate,
offsetting gains from
IoT and QoS-based
pricing.
Price paid
today
Low
data-speed
user
Low and
middle
income
High
data-speed
user
Willingness
to pay
Price paid
today
Willingness
to pay
Current price
Speed-based pricing
(reflecting relative cost
of supply)
Price and Value of
total consumption
$
Same total consumption in GBytes
QoS-based pricing allows operators to capture a greater share of customer willingness to pay
Synthesis:
(1) Cellular IoT + (2) Enterprise solutions + (3) QoS-based pricing = potential for revenue growth in real terms
Note: it may be argued that
Mobile Private Networks
(delivered through network
slicing) are themselves a form
of QoS-based pricing: high
SLAs are expensive to deliver,
hence they command a price
premium.
Note: while the example focuses on
speeds, operators may also charge
a premium for lower latencies1
1 Lower latencies are especially relevant for cloud-gaming applications, as well as for many Critical IoT applications

Operators need to bring down the cost per bit
21
To create a positive
5G business case, the
industry needs to
maintain sustainable
profitability margins ‒
especially if real
revenue growth fails to
materialise.
‒‒
The policy choices of
regulators have a
deep impact on the
5G business case for
operators. They also
bear heavily on social
welfare and economic
development.
Real-term revenue per bit index
 The bottom curve (pessimistic case) is based on the ITU’s global
traffic forecast (taken as our upper bound) and the GSMA’s global
mobile revenue forecast (taken as our lower bound)
 The top curve (optimistic case) is based on the GSMA’s global
traffic forecast (taken as our lower bound) and global mobile
revenues assuming these grow in line with real GDP growth
(taken as our upper bound for revenues)
 The resulting range for 2025 is 8% (pessimistic case) and 38%
(optimistic case) of real revenues per bit in 2020
by 2025, simply to maintain current industry margins, total
annualised costs1 per bit need to fall to between 8% (pessimistic
case) and 38% (optimistic case) of 2020 levels
0%
20%
40%
60%
80%
100%
120%
2020 2021 2022 2023 2024 2025
Index
(2020
=
100%)
Low revenue high traffic High revenue low traffic
Total capacity per dollar spent needs to increase dramatically
(1) Densifying the network
 Site TCO2 for individual operators can be reduced through
increased network and spectrum sharing
 Beyond this, to increase capacity per dollar from site densification,
the total capacity delivered per site needs to increase much faster
than site TCO (achieved with new spectrum at higher efficiency)
(2) Deploying extra spectrum
 Capacity per dollar from new spectrum depends on total licence
fees as well as on the efficiency with which it is deployed ‒
spectrum cost being heavily influenced by policy (discussed later)
(3) Increasing the efficiency of spectrum use
 Technology enhancements such as high order MIMO and
sectorisation (discussed later) typically yield more capacity per
dollar than building new sites in a basic configuration
Lower unit costs and network overheads will also help
(4) Reducing network equipment costs and driving performance
 Maintaining high competition in equipment-supply markets is
essential to drive innovation, increased performance and lower
equipment prices ‒ benefiting operators, consumers and society
– Public policy impact: denying market access to key
international suppliers may hamper this
 Network virtualisation and open RAN offer further opportunities to
reduce total network costs, improving returns for operators
1 Annualised costs = total opex + annualised capex; annualised capex should be calculated such that the Net
Present Value of annualised costs over the life of the asset is equal to the original investment in the asset.
2 TCO: Total costs of Ownership
= annualised total costs

Whether or not 5G is good for the industry as a whole, the
calculation for individual operators is different
 ‘Nash equilibrium’: optimal strategy for individual operators is the
same whatever rivals do ‒ an outcome in which none of the
operators invests is unlikely
 However, the magnitude of investments depends on expectations of
returns
Individual operator perspective
22
Unless there is a
resumption of real-
term mobile revenue
growth and increased
asset sharing across
the industry, overall
industry returns are
unlikely to improve.
 Even if 5G does not improve industry returns, there is ‘defensive
value’ in it: failure to invest (when others do) puts stranded
capital at risk, since the business would have a limited future
Similar dynamic observed in historic spectrum auctions
 While total industry bandwidth and costs of spectrum ownership
have risen over time, returns have fallen in real (if not always in
nominal) terms
 In other words, spectrum has added value to consumers, but not
to the industry itself
 Yet, while spectrum might not increase returns, failure by one
operator to invest in spectrum would see its net returns fall even
further back
But ‘defensive value’ cannot sustain investments indefinitely:
unless returns improve and cover the cost of capital,
premature market consolidation becomes more likely
 A poor outlook for returns would also reduce investment
incentives, leading to consumer harm
 Policy-makers need to be mindful of the long-term viability of the
industry in its current form, and avoid adding to the financial
burden on operators (discussed further in later slides)
Operator invests
While its rivals do not
Operator invests
And so do its rivals
Operator does not invest
And neither do its rivals
Positive
Negative
Value
Operator does not invest
But its rivals do
OR
Which is better
depends on the
potential to drive
revenues and/or
decrease total
costs per bit
The mobile ‘prisoners’ dilemma’

Managing the transition from 4G to 5G (1)
23
Many 4G networks are
currently overloaded,
while available 5G
capacity is under-
utilised.
Moreover, global 4G
capacity needs are set
to grow substantially
between 2020-2025.
‒‒
The experience of 4G
customers today will
define operator brand
perceptions for many
years.
Hence operators need
to take account of
legacy 4G capacity
needs in their
investment plans.
Global 4G capacity needs are set to double between 2020-2025, and many 4G networks (especially in South East Asia) are severely
overloaded, while available 5G capacity lies idle ‒ the transition from 4G to 5G therefore requires careful management
 Operators need to include substantial 4G capacity expansion in their investment plans, and not just focus on 5G roll-out
The experience of 4G customers today will determine their perception of operator brands for many years
 Even if the 5G network is fantastic, an underserved 4G customer may switch to a rival network before (or when) upgrading to 5G
 Higher value customers tend to be more sensitive to data experience ‒ lack of 4G capacity may have a disproportionate impact on revenues
0
50
100
150
200
250
300
350
Top
5%
2nd
5%
3rd
5%
4th
5%
5th
5%
6th
5%
7th
5%
8th
5%
9th
5%
10th
5%
11th
5%
12th
5%
13th
5%
14th
5%
15th
5%
16th
5%
17th
5%
18th
5%
19th
5%
20th
5%
ARPU
LCY
ARPU by 5 percentile
Data Users
All Users
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Top
5%
2nd
5%
3rd
5%
4th
5%
5th
5%
6th
5%
7th
5%
8th
5%
9th
5%
10th
5%
11th
5%
12th
5%
13th
5%
14th
5%
15th
5%
16th
5%
17th
5%
18th
5%
19th
5%
20th
5%
%
of
Revenue
Revenue by 5 percentile
Data Users
All Users
Typical ARPU distribution by 5 percentiles Typical revenue distribution by 5 percentiles
Customers at risk from a
below market average (4G)
mobile broadband experience
Those customers may
account for a disproportionate
share of total revenues
The investment challenge: need to ensure good 5G availability1 to encourage faster customer migration to 5G ‒ while significantly
increasing 4G capacity until the legacy-traffic peak is reached
1 Availability means both coverage and capacity. Allocating low-band resources such as 700MHz to 5G could mean less capacity for 4G (while
higher mid-band 5G is under-utilised); but without low-band 5G resources, 5G would be (relatively) less attractive due to weaker 5G coverage.

Managing the transition from 4G to 5G (2)
24
With the exception of
C-band deployment,
one cannot strictly
speak of 4G versus
5G investments.
‒‒
For example, 700MHz
is a 5G candidate
band in ITU Region 1,
but may be used for
4G low-band capacity
using DSS1,2.
‒‒
In developing
countries, the scope to
re-farm 2G/3G to 4/5G
quickly is limited by
legacy devices
(including lack of
VoLTE).
5G equipment (RF units as well as MIMO antenna-systems) supports both 5G and legacy technologies, so expanding 4G capacity with
5G-ready infrastructure does not entail deployment of assets with a curtailed economic life (i.e. no ‘stranded investments’)
 Dynamic Spectrum Sharing (DSS) provides great flexibility in the allocation of bandwidth between 5G and legacy technologies: it is no longer
necessary to re-farm spectrum in chunks of 2x5MHz across the network ‒ once peak-legacy traffic is reached, DSS can be used to gradually
shift capacity in legacy bands from 4G to 5G, allowing operators to closely match demand across each technology as time progresses
 Nevertheless, 5G remains more efficient, and operators have an incentive to rapidly migrate customers to 5G to make best use of the
available resources ‒ basic connectivity over 5G should not be more expensive than over 4G, as this could hinder upgrades to 5G
 Another key factor in 5G migration is device availability and cost (discussed in later slides)
Projected legacy-
traffic peak:
‒ N.America: 2022
‒ W.Europe: 2023
‒ N.E.Asia: 2024
‒ After 2026 in all
other regions
1 DSS: Dynamic Spectrum Sharing.
2G, 3G
4G
5G
5G
C-band
(e.g. 3.4-3.8GHz)
Legacy
bands
 5G under-utilised in near-term
 C-band: 60% of total capacity
across all technologies
 4G currently overloaded
 Need 2x more capacity in 2025
than utilised in 2020
 Near-term: accelerate re-
farming to 4G in chunks of
2x5MHz
 Expand capacity (densification
and higher-order MIMO)
 Long-term: software upgrade to
5G (no need to swap equipment)
 Long-term: use DSS1 to
allocate bandwidth to residual
users of legacy technologies
Same
equipment
(RF
units,
antennas/MIMO)
The good news is that 4G investments today will
also serve 5G needs tomorrow
2 Strong existing global device support for 700MHz in 4G ‒ so 700MHz can be used to
provide 5G coverage layer as well as extra 4G capacity.

Managing the transition from 4G to 5G (3) ‒ 2G and 3G switch-off
25
2G switch-off is
generally favoured in
Asia Pacific, while 3G
is being shut down
first in much of
Europe.
1 DSS: Dynamic Spectrum Sharing. 2 Strong existing global device support for 700MHz in 4G ‒ so 700MHz can be used to
provide 5G coverage layer as well as extra 4G capacity.
Legacy sunset strategies vary internationally ‒ with either 2G or 3G already or imminently closed in many markets
 In many Asia-Pacific markets, 2G has (or is being) switched off before 3G
 In most European markets, 3G is being switched off first (exception: Switzerland), with Vodafone switching off 3G in 2025 in some markets,
and TIM closing in 2029 in Italy
 Operators in the US and Taiwan have or are closing both 2G and 3G networks
The scope to close either 2G
or 3G in developing markets
may be more limited due to the
lack of VoLTE.
A thin 3G layer could be
supported in the longer term
through Dynamic Spectrum
Sharing, providing a circuit-
switched fall-back option.

Key 5G spectrum-acquisition considerations
26
Shareholder value is
created from spectrum
acquisition if operators
pay less for spectrum
than it is worth.
Operators need to
develop a clear
business case for
spectrum acquisition.
The prospect of under-utilised 5G resources while 4G demand continues to grow
(especially in emerging markets) may discourage some operators from paying high fees
for 5G-only spectrum
 However, failure to invest in 5G would mean that customers have no-where to go (outside 4G)
 5G is more spectrally efficient ‒ over-emphasising 4G for too long (at the expense of 5G)
means that operators will face an even deeper capacity crunch in the medium term
 Early 5G availability allows heavier data users to upgrade, relieving pressure on 4G
 Failure by an operator to acquire C-band spectrum today could allow rivals to secure all
available bandwidth ‒ in which case it would not be available when it is really needed
Spectrum acquisition is economically justified for operators if its value exceeds the cost,
albeit operators may be subject to fixed budget constraints
Case study: India 700MHz auctions
 2016 auction: 700MHz left unsold due
to excessive reserve prices
 March 2021 multiband auction: 700MHz
still left unsold
– Still too expensive, despite a 43%
reduction in prices from 2016
 Walking away from spectrum is rational
if prices exceed value
Everyone loses out as a result:
1. Less efficient network capacity
expansion for operators
2. Scarce national resources left idle for
over 5 years
3. Foregone consumer benefits from
spectrum deployment in a key
coverage band (lost opportunity to
improve indoor and wide-area coverage
quality)
4. Indirect impact on economic
development (see next slides)
Net Present Value with
NPV without
Value
Business value with
the spectrum
Business value
without the
spectrum
Value of the
spectrum
Price paid for the
spectrum
Value
created
Price
$
Value is created as long as
acquiring and deploying new
spectrum is cheaper per Mbit/s of
extra capacity than all alternatives
(such as network densification and
capacity-enhancements in other
bands).
If opportunities to secure new sites
(macros or small cells) is limited,
however, new spectrum may be the
only route to increase network
capacity to the required levels.

Socio-economic benefits of mobile and of additional mobile spectrum releases
Impact on society
27

There are very few parts of society and of the economy that are not touched by mobile communications. Mobile allows us to stay
informed, drives the wheels of the economy, and keeps us entertained and safe.
The mobile industry is a key industry with powerful externalities ‒ policy makers need to treat it with care
Widespread benefits of mobile
28
It is hard, now, to
imagine what the
world would look like
without mobile
communications.
If mobile networks
were to disappear, the
economy would suffer
a heavy loss in
productivity.
Impact on society
 Ubiquitous communications between private citizens, within businesses and with customers
 Instant transfer of information and knowledge ‒ drives economic productivity and efficiency
 Increased participation in the knowledge economy, bridging the digital divide ‒ for many, mobile is the
sole access to the Internet
 Entertainment, gaming
 Massive source of welfare (Consumer Surplus), convenience and safety
C2C, B2B, B2C1
1 C2C: Consumer-to-consumer communications; B2B: Business-to-business; B2C: Business-to-consumer.
M2M: Machine to machine
 Key driver of the 4th industrial revolution, yielding massive productivity gains (e.g. smart manufacturing)
 Unprecedented convenience (e.g. smart home, smart vehicles)
 Increased public and personal health and safety (e.g. smart health, emergency services and public
protection, smart vehicles)
 Energy efficiency, reduced waste and pollution (e.g. smart buildings, smart agriculture, smart vehicles)
M2M (IoT)
Very large absolute contribution
to the economy
Mobile
externalities
Large marginal impact on
economic development

Mobile
ecosystem
2019 2025
 $4.1 trillion global EVA1 contributed/enabled
(4.7% of global GDP)
 30 million jobs, of which 16 million direct
 $0.5 trillion in public funding (through taxes)
 ~$5 trillion EVA enabled
(4.9% of global GDP)
‒ 60% of growth from productivity gains
5G
ecosystem
 $2.2 trillion contribution to the global
economy between 2024 and 2034
‒ 1/3 third from manufacturing and utilities
‒ 1/3 from professional & financial services
 Enable $13.2 trillion in annual global
output (9% of real GDP)
 Generate $3.6 trillion in economic output
from the global 5G value chain
 Fuel $2.7 trillion in cumulative real GDP
growth (adding 0.2 percentage points to
real annual GDP growth)
 Support 22.3 million jobs
GSMA
view
IHS
Markit
view
 $235 billion average annual investment by Global 5G value chain between 2020 and 2035
Very large absolute contribution to the economy
29
By 2025, the mobile
ecosystem will
contribute nearly $5
trillion in economic
value added (4.9% of
global GDP).
‒‒
The bulk of value
created by mobile
accrues to society
rather than to the
operators.
Overtaxing the
industry (e.g. though
high spectrum fees)
threatens the wider
societal value.
Impact on society
Around a quarter of the
wider output enabled by
the 5G ecosystem
1 EVA: Economic Value Added; the surplus above the cost of capital enabled by mobile communications
EVA enabled around 4x global
mobile operator revenues
12x total mobile operator
revenues worldwide

3) Chalmers University of Technology (2012)
Doubling (average) broadband speed = 0.3% extra GDP growth
30
1) Qiang and Risotto (World Bank, 2009)
10 percentage-point increase in penetration = increases in GDP
growth of:
 0.60 percentage-points in high-income countries
 0.81 percentage-points in low- and middle-income economies
2) Deloitte (2012, 2013)
 10% higher mobile penetration = 4.2 percentage point increase in
Total Factor Productivity (TFP) in the long run
 10 pctge-point 3G penetration = 0.15 pctge-point increase in GDP
 3G: 1GByte/month/connection = 0.6 pctge points in GDP growth
Studies based on 2/3G era data
4) Capital Economics (UK study for EE, 2014)
Eventual productivity gains between 0.5% and 0.7% of GDP from
introduction of 4G
5) Goodridge et al (Imperial College Business School, 2017)
10% increase in MBB adoption = 0.6-2.8% increase in GDP
Studies spanning the 4G era
6) Frontier Economics (2018)
10% rise in M2M connections = annual increases of:
 0.7% of GDP
 0.3% in services Gross Value Added (GVA)
 0.9% in industry GVA
7) Goodridge et al (2019)
10 percentage points increase in the growth of IoT connections = 0.23
percentage points increase in TFP
 30% IoT growth rate = TFP growth of 0.69% p.a. = $0.6 trillion
contribution
Impact of IoT penetration
Both older and more
recent econometric
studies show that
moderate changes in
mobile adoption and
data usage have a
large impact on pro-
ductivity and GDP
growth.
Large marginal impact on economic development (1)
Impact on society
1) Based on 1980-2006 growth data from 120 countries; 2) Based on 1995-2010 data from 74 countries and 2008-2011 data from 93 countries; 3) Based on 2008-2010 data from the 34 OECD countries;
4) Based on 2002-2014 data from GSMAi; 5) Based on 2012-2015 data from 27 EU and OECD countries; 6) Based on 2010-2017 data from 82 OECD and non-OECD countries
Doubling usage doubles the growth attributed to 3G usage

Surprising consistency in the overall scale of impact identified by successive
studies spanning the 2G to 4G eras
 Successive studies suggest a sustained impact of mobile on GDP growth, but driven
by different factors over time
 Simple mobile adoption is the driver in the 2/3G era, mobile data adoption and use in
the 3G era, and mobile broadband adoption in the 4G era
 Each delivers a different boost to economic productivity along its own lifecycle
Large marginal impact on economic development (2)
31
Each new mobile
generation introduces
new capabilities
driving further
economic growth.
In the 5G era, cellular
IoT is likely to be a
significant contributing
factor.
Impact on society
0.81
0.15
0.60
10-percentage
point increase in
Mobile adoption
Low and
middle
income
0.60
High
income
economy
10-percentage
point increase in
3G adoption
10% increase
In total
MBB adoption
0.60
Reaching 1GB
per SIM of monthly
3G usage
Based on
2/3G era data
(to 2006)
Based on
3G era data
(2008-2011)
Based on
3/4G era data
(2002-2014)
Percentage-point
increase
in
GDP
growth
Lower-
bound
estimate
%
increase
in
GDP
0.50
0.70
4G era
(2014 study)
Lower-
bound
estimate
Upper-
bound
estimate
Eventual
impact of
4G
Economic development highly
sensitive to adoption and usage of
mobile services
Adoption and usage sensitive to
network coverage and capacity
(investment) and to retail prices
Mobile investment and retail prices
sensitive to financial position of
operators and ROIC1 prospects
Financial position of operators and
ROIC prospects heavily influenced
by public policy
Hence socio-economic
development highly sensitive to
public policy towards mobile
Especially spectrum and pricing policy
(discussed later)
1 ROIC: Return on Invested Capital

Mobile already yields high levels of consumer welfare ‒ extra
spectrum would boost Consumer Surplus (CS) even further
 Broadly flat total mobile revenues per customer against a
backdrop of steeply declining prices per unit of consumption
suggests an isoelastic demand curve1
– Halving of unit prices = doubling of AUPU2
 In these conditions, the increase in CS per user from a halving of
effective prices per GByte is roughly equal to 75% of the
customer’s total spend
Impact of halving prices per unit on Consumer Surplus (CS)
Extra spectrum is needed to allow the capacity expansion
needed for substantial increases in AUPU (and reductions in
prices per Gbyte)
Lowest (yet still very large) impact: China
 GSMA project a tripling of data usage in China between 2019 and
2025, corresponding with a reduction in unit prices by 2/3rds
 This entails extra CS of almost 1.2x total consumer spend,
amounting to around 1.4% of GDP by 2025 on top of the
unquestionably high existing CS from mobile consumption
Highest impact: sub-Saharan Africa
 GSMA: 11-fold increase in data usage between 2019 and 2025
 Extra CS would be almost 2.6x total consumer spend
 For Nigeria, this would entail added CS of $17 billion by 2025, or
almost 4% of GDP
 Much of this extra welfare would be foregone if insufficient
spectrum is made available to operators
In addition to incremental CS, extra spectrum would drive GDP
growth by enabling further adoption and use of mobile data
(as per previous slides)
Impact of spectrum deployment on economic growth and welfare
32
Far more spectrum is
needed to meet future
mobile data demand.
It follows that extra
spectrum availability
bears heavily on
mobile adoption and
use, hence on welfare
and socio-economic
development.
‒‒
For a country like
Nigeria, additional
spectrum releases
could fuel $17 billion
in increased CS by
2025 ‒ nearly 4% of
GDP.
Impact on society
1 Isoelastic curve: price elasticity coefficient of 1, implying constant revenues at all prices per unit of consumption
2 AUPU: Average Usage per User in Gbytes/month
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0
Price
per
unit
Quantity consumed (units)
Increase in Consumer Surplus
Demand curve

Impact of large spectrum allocation at low prices (illustrative)
Ensuring the socio-economic gains materialise
33
Policy-makers can
ensure the socio-
economic gains
materialise by
releasing as much
spectrum as possible,
as fast as possible
and as cheaply as
possible.
Impact on society
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0
Price
per
unit
Quantity supplied or consumed (units)
Demand curve
Ability and willingness to supply
without extra spectrum
Ability and willingness to supply
with extra spectrum at high TCO
Willingness to supply
with extra spectrum
at low TCO
Current
demand-
supply
equilibrium
Potential
demand-
supply
equilibrium
Massive increase in CS plus
substantial GDP boost
The quantity of spectrum is important,
but so is the price paid by operators
 Operator supply curves (willingness
to supply as a function of price per
unit) are affected by the scope to earn
returns (ROIC) that match the cost of
capital (WACC)
 Spectrum TCO1 is a significant line-
item in operator costs and has a
significant impact on the income and
capital requirements of operators ‒
hence on their ROIC
 Policymakers have direct control over
annual spectrum fees and the
conditions on which spectrum is
awarded (discussed in more detail
later)
1 TCO: total (annualised) cost of ownership

Meeting the demand for capacity and performance in 2025
Evolution of mobile networks and technology
34

Opening the RAN introduces additional vulnerabilities (i.e.
security concerns)
 The ‘O-RAN Alliance’ supports a wider range of functions and
interfaces than 3GPP
 The greater the number of open interfaces, the greater the
potential to diversify the vendor base and achieve cost benefits
 But the greater the ‘Threat Surface’ becomes (i.e. the greater the
security concerns)
Approaches to Open RAN under the 3GPP standards-setting
organisation versus the O-RAN Alliance
We believe that the 3GPP approach may gain greater global
acceptance (hence scale), and offer a better balance between
security and opportunity to reduce RAN TCO.
Commoditising the physical layer will allow access to a greater
diversity of (software) component suppliers and will help reduce
total network costs ‒ but this will require radical steps
 RAN Virtualisation: migrating 4G core from ‘physical domain’ to
5G ‘virtual domain’
 Implementing ‘Open RAN’; currently, there are two competing
approaches (which could lead to a fragmentation of standards):
– 3GPP approach
– ‘O-RAN Alliance’ approach
Key elements of RAN evolution
RAN architecture evolution
The 3GPP approach
to Open RAN is less
far reaching than that
of the O-RAN alliance,
so may have less of
an impact on overall
cost reduction.
But the 3GPP
approach may raise
fewer security
concerns and could
achieve greater global
acceptance ‒ hence
scale.
Evolution of networks and technology
Open RAN Security
RAN Slicing
RAN
Virtualisation
Radio
Spectrum
Licencing
1 CP: Connection Point; CU: Central Unit; DU: Distributed Unit; LLS: Lower Layer Split; RIC: RAN Intelligent Controller; SMO:
Service Management & Orchestration; UP: User Plane.

Relative bandwidth and capacity by band
Huawei MIMO evolution roadmap
Increasing the efficiency of spectrum use (1)
36
The 3.5GHz band
accounts for around a
third of current
bandwidth, but with
massive MIMO, it
could deliver over
60% of capacity
between 700 MHz and
3.8GHz.
‒‒
Massive MIMO also
allows the trading-off
of some of the extra
capacity for improved
coverage in metro
areas.
Allows higher mid-
band to behave like
lower mid-bands.
35.9%
60.0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
% MHz % Capacity
%
of
total
3500MHz 2600MHz 2300MHz 2100MHz 1800MHz
1500MHz 900MHz 800MHz 700MHz
Massive MIMO:
huge capacity
increases in
higher-mid bands
Increased sectorisation and higher order MIMO drive higher data
throughputs (in Mbit/s per MHz per site)
 Adding a 4th sector to a 3-sector site may extend the effective site
capacity from a given band by around 40%
 Conservative rule of thumb: each doubling of the MIMO order
above 4x4 MIMO increases capacity by 1.3x
– For example, 64x64 order MIMO (‘massive MIMO’) can
generate over 3.3x more capacity per MHz than 2x2 MIMO
 More optimistic views (e.g. Huawei): 32x32 MIMO in FDD yields 5x
the throughput of 2x2 MIMO ‒ 64x64 MIMO in TDD yields 3.7x the
throughput of 8x8 MIMO
 Because of the size of low-band antennas, increased sectorisation
and high-order MIMO is easier to implement in mid-to-high bands
– 4x4 MIMO available in sub-1GHz (8x8 will be standard in 2025)
Commercial LTE 900 MHz: 4x4 MIMO versus 2x2 MIMO (Huawei)
62% gain in capacity and improved cell-
edge performance. Important: low band
scarce and carries ~30% of all traffic today1
TDD
FDD
1 Based on data from Tutela, low bands currently carry around a third of all 4G traffic, despite accounting for a
small proportion of total deployed 4G capacity in many markets.

Wide-band deployments are more efficient than CA
 Notwithstanding the benefits of CA, it is still better to deploy
wider RF channels across larger chunks of contiguous spectrum
in fewer bands
 Combining a higher number of narrow RF channels (yielding the
same total MHz) but relying more heavily on CA, is less efficient
 For example, aggregating separate blocks (narrow channels) of
spectrum in the same band would lead to loss in total capacity
per MHz due to Band-Edge-Mask (BEM) filter restrictions
Wide-band deployment versus CA (example)
37
Wider logical channels
yield better speed
performance.
‒‒
But it is more efficient
spectrally rely on
consolidated spectrum
holdings, where
possible, than on CA.
Creating wider logical channels through carrier aggregation (CA)
Example: CA over three bands
In this example: 560Mbit/s delivered with CA versus 300Mbit/s if
‘camped’ on 3500MHz without CA, or 90Mbit/s if camped on 900MHz
 The delivered data rate to the device is less than the maximum for
FDD spectrum, due to propagation changes across the cell
 For TDD, the delivered rate is the same as the maximum, because
TDD either works perfectly or not at all
 The resulting performance benefits require no additional to RAN
infrastructure, albeit there is an incremental cost of CA software
within each cell-site or baseband site
Frequency Total bandwidth Max downlink data
rate1
Delivered data
rate (RF path)
900 MHz 2x10 MHz 100Mbit/s 90Mbit/s
1800 MHz 2x20 MHz 200Mbit/s 170Mbit/s
3500 MHz 50 MHz 300Mbit/s 300Mbit/s
Total delivered data rate to the handset/device 560Mbit/s
100 MHz 50 + 50 MHz
Complexity Single carrier Needs intra-band CA
Channel utilisation 98.3% 95.8%
Physical layer signalling 6.3% overhead Approx. 12% overhead
Physical layer
configuration
A single 100 MHz carrier offers more flexibility than
2x50 MHz carriers to configure sub-bands within the
carrier
Carrier activation /
deactivation delay
2ms 10ms
BS implementation Requires one radio unit
only
May need two radio units
Spectrum management Guard bands may be
required if networks are
unsynchronised
Two additional guard
bands if networks are
unsynchronised
UL support No CA required in the UL Uplink CA may not be
supported by all UEs
UE consumption 30mA additional power
consumption for the
second CC (50-90% RF
power increase over the
non-CA case)
1 Assuming a gross throughput of 10Mbps per MHz. In FDD mode, half the bandwidth is available for downlink (DL) and half for
uplink (UL). For TDD, we assume a DL/UL ratio of 4:1 (DDDSU scheme) meaning that 3 out of 5 timeslots are allocated to DL
Note that CA only benefits speed performance in available bands
 CA does not increase total capacity, nor improve coverage

Allocation of more than 100MHz per operator (2025-2030)
Cost per bit with per operator allocation of more than 100MHz
 Equipment suppliers efforts aim at allowing their 5G radios,
including those implementing massive MIMO and beamforming, to
operate with the widest possible channel bandwidth (“instantaneous
bandwidth”) and to make that “tunable” in the widest possible
frequency range (“operating bandwidth”)
 5G radios that are now deployed in 3400-3800 MHz band are
starting to operate at an “instantaneous bandwidth” of 100 MHz
within a 400 MHz “operating bandwidth”
 The ongoing research (e.g. for filters and power amplifiers) will
allow larger instantaneous and operating bandwidths by 2025-2030
 If 300 MHz is deployed in a single radio, the cost per MHz deployed
is 43% lower compared to a deployment in only 100MHz
Allocation of up to 100MHz per operator in one band
Cost per bit depending on channel bandwidth (up to 100MHz today)
 From a network cost perspective, the wider the channel that is
deployed in a single radio, the lower the cost per MHz deployed,
and therefore implicitly the cost per bit
 Deploying technology enhancements such as sectorisation and
higher-order MIMO is also much more cost effective over wider
allocations ‒ as the cost of these enhancements is broadly the
same, whether the channel is narrow or wide
 A 100MHz deployment in a single band yields a cost per bit that is
around a quarter of that for a 20MHz deployment
38
It is also far more
cost-effective to
deploy wide-band
channels than to rely
on CA, especially
when massive MIMO
is used.
0
20
40
60
80
100
120
0
10
20
30
40
50
60
70
80
90
100
20
MHz
30
MHz
40
MHz
50
MHz
60
MHz
70
MHz
80
MHz
90
MHz
100
MHz
TCO
index
Cost
per
bit
Index
20
MHz
=
100
Cost Index Cost per bit
0
20
40
60
80
100
120
140
160
180
0
10
20
30
40
50
60
70
80
90
100
100 MHz 200 MHz 300 MHz
Total
cost
of
ownership
index
Cost
per
bit
Index
20
MHz
=
100
Cost Index Cost per bit

Spectrum consolidation is also highly desirable
in lower mid-bands
 In 1800MHz and 2100MHz, for example, most
operators hold no more than 2x20MHz in either
Future fragmentation may be avoided by
packaging spectrum in larger, contiguous
chunks when awarding new usage rights
 In an auction for 3.4-3.8GHz spectrum, for
example, allocations of less than 40 MHz could
be excluded from any winning configuration ‒ to
ensure all available resources in this band can
be used efficiently
 Allowing the pooling and sharing of incremental
usage rights is a further option
Note that spectrum consolidation does not
necessarily exclude CA
 there could still be room to improve efficiency by
aggregating holdings across bands that contain
less than 100MHz
39
It is far more efficient
to deploy wider
channels across fewer
bands than the
opposite.
‒‒
This is why it is
important that policy-
makers favour wide-
band allocations
and/or encourage
spectrum trading
and/or pooling.
Spectrum consolidation through multi-lateral trades or spectrum pooling
would allow lower cost and higher-performance deployment
Low-band consolidation in a 3-player market (illustration)
 The shared 9.6MHz in this example could be used to serve residual 2G and 3G
M2M connections and handsets
20MHz at 700MHz 20MHz at 800MHz 23.2 MHz at 900MHz
40MHz at 700MHz 23.2 MHz at 900MHz
40MHz at 800MHz 23.2 MHz at 900MHz
60MHz at 700MHz
60MHz at 800MHz
60MHz at 900MHz
Shared
9.6 MHz
at
900MHz
Operator A
Operator B
Operator C
Operator A
Operator B
Operator C
Operator A
Operator B
Operator C
+
Initial assignments (fragmented)
Partial low-band consolidation
Full low-band consolidation
Barring widespread spectrum consolidation,
CA will likely be required in mid-bands to
improve speed performance

Split 2600MHz band-plan is inefficient
 2600MHz band assigned as FDD (Band 7) in several
markets, with TDD centre gap (Band 38)
 Deploying MIMO enhancements of a given order across
both FDD and TDD portions generates less of a
performance and capacity uplift and is far less cost-
effective than deploying the same MIMO order over
wider allocations in TDD only
40
Converting paired
(FDD) spectrum to
TDD could lead to
significantly improved
spectral efficiency and
would allow higher
order MIMO
deployments.
‒‒
Reassigning the full
2600MHz band to
TDD would yield the
added benefits of
more cost-effective
MIMO deployment.
It would also remove
the need for a 5MHz
guard-band.
Reorganising mid and low-band paired spectrum to TDD band-plans could
deliver a significant boost in overall capacity
The case is especially strong in the 2600MHz band
FDD (UL) TDD FDD (DL)
2500
MHz
2570
MHz
2615
MHz
2620
MHz
2690
MHz
TDD
2500
MHz
2690
MHz
Split 2600MHz band-plan (e.g. ITU Region 1)
Unified 2600MHz band-plan (e.g. china)
70 MHz
190 MHz
45 MHz 70 MHz
Guard-band
Regulators now looking at licencing 2600 MHz band as
TDD only (Band n41)
 In addition to China, the USA, Philippines and Saudi
Arabia have committed to this and regulators in
Thailand, Myanmar, Sri Lanka, India, Nepal are looking
at this option
 Would providing a continuous 190MHz wide band and
remove the need for the 5MHz guard-band

Subscribers
in
billions
Device ecosystem
Band support in a sample of 5G smartphones (GSA)
 While the number of available frequency bands has increased
significantly with each successive generation of mobile technology,
band-support within devices is fast becoming a non-issue
 Of the 198 5G smartphones listed by the GSA in January 2021,
28% explicitly support MIMO 4x4 and over half explicitly support
VoLTE (Voice over LTE)
Speed of technology adoption: 4G versus 4G (Ericsson)
41
Band-support within
individual devices is
fast becoming a non-
issue.
‒‒
It will take 2.5 years
from 5G launch to
pass the 500m
adoption mark ‒ twice
as fast as for 4G1.
The price of compatible devices is invariably an important factor
influencing the rate of adoption of new mobile technologies
Price evolution for a sample of Smartphones (pricespy.co.uk)
While current prices are comparatively high for 5G smartphones (e.g.
$1,350 for an iPhone 12 Pro Max with 128GB memory):
 Prices of new devices tend to drop rapidly after launch (see above),
with the high-end devices of today quickly joining the mid-range
devices of tomorrow
 Secondary markets yield cheaper entry-points for consumers
 5G dongles and routers are typically cheaper than smartphones
and the form-factor of routers makes it easier to include antennas
supporting higher-order MIMO
Smartphone model Launch Nr of 4G bands No of 5G bands
iPhone 12 A2403 (RoW) Q4 2020 27 17
Samsung Galaxy Z Fold2 5G Q4 2020 21 10
Huawei P40 Pro 5G (ELS-N04) Q2 2020 22 9
Nokia 8.3 5G Q4 2020 18 13
Google Pixel 5 Q4 2020 29 13
£0
£200
£400
£600
£800
£1,000
£1,200
£1,400
Nov 2016 Apr 2018 Aug 2019 Jan 2021
Device
price
iPhone11 64GB
iPhone8 64GB
iPhone7 32GB
Samsung Galaxy S20 Ultra 5G
Samsung Galaxy Note 20 Ultra 5G 245GB
Samsung Galaxy S8 64GB

Network densification: macro sites versus small cells
42
Between 800,000 and
2 million small cells
may be required to
make 5G a reality in
the US.
‒‒
In Europe, site
planning approvals for
small cells remain
onerous ‒ this may
explain why Europe
lags the US on small
cell deployments.
If planning restrictions
for small cells ease in
Europe, small cells
are likely to become a
bigger part of network
strategies.
300m Rogers Telus
Rogers: 5 bands
Telus: 2100MHz
5 bands
Rogers: 5 bands
Telus: 5 bands
2100MHz
700MHz
While small cells are central to the
strategies of some operators in North
America, they only form part of a tiny
proportion of the total site count in
Europe
 In most European countries, gaining
planning approval for a small site
remains a lengthy and costly process
– This reduces the net relative
benefit of small cells (vs Macros)
 In 2018, the US ‘5G Fast Plan’ was
introduced to further ease planning
and permitting issues:
– Reducing State and local
government response periods to
60 days to review a proposed
collocation and 90 days to review
an application for a new structure
– Exclusion of small cells from the
Environmental Policy Act and the
Preservation Act
– Restrictions on the fees charged
by local government for access to
public assets
 ‘Similar’ initiatives being developed in
the UK and EU, but less
interventionist (planning provisions
remain devolved at more local level)
Densification may be achieved by rolling out additional macro sites, small cells, or a
combination of the two
Example: Telus versus Rogers networks (Vancouver snapshot, January 2021)
 The unit cost per small cell is typically far lower than that for a macro site, but far more small
cells are required to deliver capacity across a given area
 One FCC commissioner recently estimated that the US needs 800,000 small cells to make 5G a
reality, while the International Data Corporation (IDC) expects over two million by 2021

From traditional network sharing (through bilateral network JVs) to an Active
Neutral Host model
 Because Neutral Hosts can accommodate multiple MNOs, the net savings
tend to be higher than in traditional (bilateral) network JVs
 In the above example: Neutral Host model reduces total sites from 6 to 4
Evolution of mobile asset sharing
43
Mobile asset sharing
increases the
likelihood that
operators earn their
costs of capital.
It may also enable the
extension of mobile
coverage into remote
areas that would
otherwise be
uneconomic.
‒‒
Active neutral hosts
could pave the way for
more extensive asset
sharing, while
avoiding the deep
constraints of bilateral
network JVs.
MNO 1 MNO 2 MNO 3 MNO 4
MNO 1
MNO 2
MNO 3
MNO 4
Common coverage area
Common coverage area
Neutral Host
sites
Network JV1
Network JV2
Emerging Neutral Host models allow greater collective
savings
 Active Neutral Hosts are independent network providers
offering capacity and coverage solutions to MNOs1 on a
shared basis, both indoors and outdoors
 Usually carry the upfront investment for both active and
passive infrastructure, and charge a recurring monthly fee
to MNOs
– Usually breaks even with 2 tenant MNOs, and charge a
lower fee when a site is shared by more than 2 MNOs
 Likely to play a big part in the Small Cells domain, a key
area of site expansion
Neutral Hosting can be arranged at individual site level
 Allows for competition at network level between hosts
 Avoids challenges and constraints of bilateral network JVs
Extra savings possible through neutral spectrum hosting
 Allows reduction in total band deployments on a site
 e.g. Dense Air secured 3.5GHz spectrum in Ireland,
Portugal, Australia, and 2.6GHz spectrum in Belgium and
New Zealand. This allows it to offer coverage and capacity
solutions to operators using its own spectrum
 Shared spectrum solution can also be achieved through
local spectrum pooling (e.g. localised roaming or RAN
network slicing at individual site level)
1 MNO: Mobile Network Operator

From sub-6GHz spectrum to microwave
Spectrum demand 2020-2025
44

 Low frequency bands generally provide both wide-area and deep indoor coverage, and support mobility when users travel at higher speeds
 Higher bands provide extra capacity where demand is more densely concentrated
Mix of spectrum to meet the IMT 2020 Requirements
45
A mix of spectrum
spanning low to high
bands are needed to
meet the IMT 2020
requirements specified
by the ITU.
Dense Urban Urban Suburban Rural
Upper mid-band 3.3-4.2, 4.5-4.99, 6GHz
city-wide speed coverage layer, 5G only
Sub-1GHz band 600-900 MHz
deep indoor and rural coverage layer, legacy technologies and 5G
Lower mid-band 1.4 – 2.6 GHz
basic capacity layer, legacy technologies and 5G
High-band 26GHz & other
mmWave, Super high capacity hot-
spots, 5G only
Area traffic capacity
of 10 Mbit/s/m2
User experienced 100
Mbit/s DL and 50
Mbit/s UL rate
IMT 2020
Requirements
Mobility: mobile at
500 Km/hour
Increasing user mobility (travel speed)
Latency: 1ms

Mobile Spectrum and Network Evolution to 2025 46
With the exception of
the C-band and
mmWaves, one
cannot really speak of
5G versus 4G
spectrum: over time,
all bands will migrate
to 5G.
‒‒
While 3GPP
standards currently
provide for up to 100
MHz wide channels,
they allow a maximum
bandwidth of 400 MHz
in carrier aggregation
mode.
Key band categories
1 FWA: Fixed Wireless Access
Low band
600-900MHz
Lower mid band
1400-2600MHz
Upper mid band
3.3GHz-6GHz
High band (mmWaves)
> 6GHz
Decreasing propagation (range) ‒ so less good for coverage
Increasing amounts of available bandwidth ‒ good for capacity
Greater scope for massive MIMO ‒ good for capacity and cost per bit
More scope for wide-band deployment ‒ good for speed performance and lower cost per MHz deployed
e.g. 600, 700, 800, 900MHz
Superior propagation make
these best for wide area and
deep indoor coverage
 But most scarce spectrum
 More limited scope to extend
capacity per MHz through
higher order MIMO
deployment
 700MHz: 5G coverage layer in
ITU Region1; used for 4G
elsewhere, but will be
refarmed to 5G over time
 600MHz: 700MHz-equivalent
in US/Canada
e.g. 1800, 2100, 2300, 2600MHz
Good compromise between
propagation characteristics and
capacity potential
 Historically, 1800MHz used for
2G, AWS and 2100MHz for 3G
 2300MHz and 2600MHz were
early 4G bands where available,
adding to 1800MHz spectrum
re-farmed from 2G to 4G
 2600MHz and 2300MHz: 5G
candidate bands in some
countries
e.g. 3.3-4.2, 4.5-4.99, 6 GHz
Newer to IMT and much more
plentiful ‒ key 5G capacity
resources, allowing highly
efficient, wide-band deployment
 3GPP standardised radios and
terminals available for the C-
Band (band n77, 3.3-4.2GHz)
 C-band (400 MHz in Europe) is
the first mid-band in which a
channel width of 100MHz can
be used – a 5G innovation
 Rolling out 5G in the C-Band is
an overriding policy objective;
3.4-3.8GHz used in most
countries
 Good combination of
propagation and capacity for
cities
e.g. 26, 28, 39 GHz
Effective at addressing areas
with very high traffic density
with extreme peak data rates ‒
FWA1 is a prime use case
 Not suitable for contiguous
wide area coverage given the
large number of sites this would
require
 Adding mm wave spectrum will
increase the spectrum used by
mobile operators by up to
6000MHz (dwarfing the amount
of spectrum deployed by mobile
operators as of 2020)
 66GHz is not as yet
harmonised as an international
IMT band, but is being
discussed as a potential further
5G candidate band

Mobile Spectrum and Network Evolution to 2025 47
Deployment priority
needs to take account
of the need to address
4G demand while
driving migration to
5G.
The strength of the
device and equipment
ecosystem for given
bands also influences
timing.
‒‒
Providing a good
coverage layer is
essential both for the
4G and 5G experi-
ence.
700MHz can be used
for either or both in the
near term2.
Spectrum award and deployment priorities
1 SDL: Supplementary downlink. 2 700MHz can be used for extra 4G capacity but this would weaken 5G coverage making it (relatively) less attractive.
3 700MHz SDL: 20MHz in FDD centre gap awarded in some European markets ‒ low-band capacity and cheap to deploy for existing 700MHz users.
Legacy Low band
e.g. 800, 900MHz
Legacy lower mid band
e.g. 1800, 2100, 2300, 2600 MHz
C-band (#1 for capacity)
e.g. 3.4-3.8GHz
High band (mmWaves)
> 6GHz
Low-band (#1 for coverage)
e.g. 600, 700MHz2
Additional Upper mid band
e.g. 3.3-4.2, 4.5-4.99, 6 GHz
Basic
coverage
layer
Basic
capacity
layer
Super
capacity
layer
Enhanced
capacity
layer
Phase I
Should be complete, else
accelerate (need for 4G)
Phase II
Should be complete or near
completion ‒ if not, accelerate
Phase III
By 2025
1
2
4
3
8
7
4G (5G over time) 4G (except C-band) or 5G 5G
L-band
1400/1500MHz FDD,TDD and SDL1
5
Supplemental Low-band
e.g. 700MHz SDL3
6
 In ‘Phase II’, the C-band and 5G candidate low bands should ideally be awarded at the same time; the urgency is generally slightly greater for
700MHz, because this can immediately be used to expand 4G capacity and improves indoor coverage and cell-edge performance
 In ‘Phase III’, 700MHz SDL might be deemed slightly more urgent than High Band, because it is relatively cheap for existing 700MHz holders
to deploy, and it helps relieve low-band congestion; but High Band will be important too, to serve very high traffic density areas and FWA

Spectrum allocations and roadmap in different regions
Typical low and mid-band spectrum allocated to mobile in Asia
Typical spectrum used by mobile in Europe by 2023
48
With new spectrum for
5G, the amount of
spectrum used by
mobile operators to
satisfy the growth in
mobile data will
double between 2020
and 2025.
Spectrum roadmap in Canada
 Up to 525MHz of spectrum has already been released to operators
in some markets
 By 2021, once spectrum in the C-band, in 2300 MHz and 2600 MHz
is assigned, the spectrum used by mobile operators in those
markets will have increased to 1,155MHz
– i.e. more than double the amount used in 2019
 In the EU, on aggregate, mobile operators will typically hold 190
MHz of low-bands spectrum, 460 MHz in lower mid-bands plus 400
MHz in upper mid-bands by the end of 2023, with some variation
between countries
220
340
480 480
620
970
1070
0
200
400
600
800
1000
1200
1400
1600
2005 2010 2015 2019 2020 2021 2024
MHz
850/900 1800 2100 2600 2600 TDD 700 2300 3.4-3.7
Upper mid-bands
3.5GHz 400 MHz
Low bands
700MHz 2x30 MHz
800MHz 2x30 MHz
900MHz 2x35 MHz
Total 190 MHz FDD
Lower mid-bands
1800MHz 2x75 MHz
2100MHz 2x60 MHz
2600MHz 2x70 MHz
2600MHz 50MHz
Total 410 MHz FDD, 50 MHz
TDD
High bands
26GHz 1000 to 3000 MHz TDD
Legacy bands New “5G” bands
Based on typical situation in Europe in 2021
Area traffic capacity of
10 Mbit/s/m2
User experienced data
rate 100 Mbit/s
Next 4
Years
35 Years
0
1000
2000
3000
4000
5000
6000
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
Sptrum
for
Mobile
MHz
850MHz 1900MHz AWS 700MHz 2600MHz
600MHz 3500MHz 3800MHz 26GHz 38GHz
39GHz
Indicative ’averages’
of current and future
allocations in low to
upper-mid bands in
ITU Region 3 (Asia
Pacific).
1 Taking account of the fact that certain identified IMT bands have yet to be allocated in some markets.
High
bands
Almost 5x more MHz
than low-to-mid bands

Quantifying spectrum need ‒ methodology
49
Spectrum ‘need’ (as
opposed to ‘value’)
can best be assessed
by focusing on the
busiest parts of the
network, at the point
where alternatives to
spectrum deployment
become impractical
(or prohibitively
expensive).
‒‒
We have projected
spectrum need for a
sample of 5 developed
and 4 emerging
markets.
Traffic demand and capacity supply model
Capacity supply per
km2 (Gbit/s/km2)
Population Density
Concurrent demand
from human users
Concurrent demand
from new use cases
Offload to indoor
small cells and
mmWave sites (%)
Activity factor (%)
Traffic demand per
km2 (Gbit/s/km2)
ITU Requirement
User experienced
data rate 100 Mbit/s
Macro site inter-site
distance meters
Number of outdoor
small cells relative to
macro sites
Macro site
sectorisation
Small cell
sectorisation
MHz of spectrum on
macro site
MHz of spectrum on
outdoor small cell
Spectral efficiency
bit/s/Hz macro site
Spectral efficiency
bit/s/Hz small cell
Increased network-wide demand for capacity entails higher
demand for spectrum, but the relationship between the two
is not necessarily linear
 The need for spectrum is driven by traffic density so we need
to analyse traffic demand in areas with high population
densities, i.e. cities
 There is a high degree of uncertainty over how much
simultaneous capacity will be required by different users
within any given area
– We use population density in cities as a proxy for traffic
density, to estimate the minimum or floor capacity
requirement
– While traffic generated by connected vehicles, 5G video
cameras and video-based sensors could be a multiple of
traffic generated by human users in certain areas, total
traffic intensity is likely to remain highest where people are
most concentrated
Demand-side assumptions
 The IMT-2020 requirement: DL user experienced data rate of
100 Mbit/s and 50 Mbit/s uplink
 The population density (proxy for traffic density)
 A assumption of concurrent demand from human users and
new use cases (the ‘activity factor’)
 An assumption of how much of the traffic demand would be
satisfied offloading to high bands (24GHz and above) and to
indoor small cells

Quantifying spectrum need ‒ key parameters
50
We assume that all
available low-bands,
lower mid-bands, and
upper mid-bands will
be deployed on all
macro sites.
As regards small cells,
we assume that upper
mid-bands spectrum
will be used on all
small cells.
Key 5G modelling assumptions for future urban environment
Number of macro sites
 In a typical city, macro sites use low and mid-bands, while
small cells only use upper mid-bands; the typical inter-site
distance for macro sites is ca. 400m
Spectral efficiency
 The ITU-R target for dense urban eMBB is 7.8 bit/s/Hz and
could be achieved by using 64-element MIMO; we apply a
blended average, reflecting a mix of MIMO configurations
Site sectorisation
 We assume a three-sector configuration for macro sites, and a
single sector for small cells
Design margins
 To manage interference a design margin of at least 15% is
required; i.e. 15% of the nominal capacity cannot be used
Spectrum use
 We assume that 600MHz in low and mid-bands will be FDD,
with all other spectrum resources used in TDD mode, with a 3:1
DL to UL ratio
Role of small cells
 Small cells would not provide contiguous coverage but would be
deployed to fill in “speed coverage holes”
 We assume 3 small cells per macro site
Band Category Average
inter-site
distance (m)
Number of
sectors
Average
DL/UL
spectral
efficiency
(bit/s/Hz)
700, 800, 900 MHz Macro site;
Low bands
400 3 1.8 / 1.8
1800, 2100, 2600 MHz Macro site;
Lower mid-bands
400 3 2.2 / 2.5
3.5 GHz Macro site;
Upper mid-bands
400 3 6.0 / 4.1
Additional mid-bands Macro site;
Mid-bands
400 3 6.0 / 4.1
3.5 GHz Small cell;
Upper mid-bands
n/a* 1 3.7 / 2.6
Additional mid-bands Small cell;
Mid-bands
n/a* 1 3.7 / 2.6
* For small cells this does not assume contiguous coverage because small cells are deployed to
fill in speed coverage holes rather than providing contiguous coverage. Hence the inter-site
distance is irrelevant
City High density
area (km2)
Population in
high density
area
Population density
in high density area
(pop/km2)
Karachi central 65.0 34,618 2,250,146
Paris 85.3 25,018 2,134,035
Madrid 113.1 24,246 2,741,249
Rabat central 23.0 18,394 423,056
Rome 68.6 15,839 1,086,670
Berlin 85.6 13,917 1,191,421
Amsterdam 72.3 9,788 707,220
Aman central 82.0 8,460 693,726
Khartoum 1,010.0 5,222 5,274,321

DL area traffic demand and low- to mid-band spectrum need (assuming 30% traffic offloaded
to high bands and indoor small cells)
Spectrum needed to meet the 100Mbit/s downlink speed requirement (1)
51
The higher the
population density and
the higher the
concurrent use
(‘activity factor’), the
higher the spectrum
need.
1GHz needed
low and mid-band
2GHz needed
low and mid-band
3GHz needed
low and mid-band
Karachi
central
(65km²)
Paris
(85km²)
Madrid
(113km²)
Rabat
central
(23km²)
Rome
(68km²)
Berlin
(85km²)
Amsterdam
(72km²)
Aman
central
(82km²)
Khartoum
(1010km²)
0
100
200
300
400
500
600
5000
6000
7000
8000
9000
10000
11000
12000
13000
14000
15000
16000
17000
18000
19000
20000
21000
22000
23000
24000
25000
26000
27000
28000
29000
30000
31000
32000
33000
34000
35000
Capacity
/
Traffic
(Gbit/s/km
2
)
Population Density (pop/km2)
Traffic demand 5% activity factor Traffic demand 10% activity factor Traffic demand 15% activity factor
Traffic demand 20% activity factor Traffic demand 25% activity factor
Activity factors may be expected to
increase over time in all markets
 The 5% and 10% activity-factor lines
might be representative of high-
income markets today, but will unlikely
be immediately applicable for low-
income markets
 Subject to spectrum availability
however, mobile broadband use
(including Fixed Wireless access) is
likely to expand substantially in
developing countries where fixed
broadband infrastructure is weaker
– The 5% activity-factor line might be
taken as a lower-bound for low-
income countries in 2025, with the
10% line as an upper-bound
 In areas with a population density
greater than 9,000 per km2, additional
mid-bands spectrum is required to
deliver the IMT 2020 requirements
 In areas with a population density
below 9,000 per km2, additional mid-
bands spectrum would reduce site
density
Traffic
densities likely
to increase
over time

Total low and mid-band spectrum need (MHz) to meet DL requirement
Spectrum needed to meet the 100Mbit/s downlink speed requirement (2)
52
Our analysis suggests
that additional mid-
band spectrum would
enable the 5G-NR
experienced data rate
of 100 Mbit/s anytime,
anywhere, citywide.
Activity factor
%Traffic offload 50% 30% 0% 50% 30% 0% 50% 30% 0% 50% 30% 0% 50% 30% 0%
Karachi central 1160 1420 1680 1800 2320 2840 2450 3230 4010 3100 4140 5170 3750 5040 6340
Paris 1000 1160 1350 1450 1820 2190 1910 2480 3040 2380 3130 3880 2850 3790 4720
Madrid 1000 1140 1330 1420 1780 2140 1870 2410 2960 2320 3050 3780 2780 3690 4590
Rabat central 850 990 1130 1200 1470 1750 1540 1950 2370 1890 2440 2990 2230 2920 3610
Rome 1000 1000 1040 1100 1340 1580 1400 1750 2110 1690 2170 2640 1990 2580 3180
Berlin 1000 1000 1000 1030 1240 1450 1290 1600 1920 1550 1970 2380 1810 2330 2850
Amsterdam 1000 1000 1000 1000 1020 1170 1060 1280 1500 1240 1540 1830 1430 1790 2160
Aman central 670 730 790 830 950 1080 1000 1170 1360 1140 1400 1650 1300 1620 1930
Khartoum 610 650 690 700 780 860 1000 1000 1040 1000 1060 1210 1000 1190 1390
Spectrum need
Activity factor 25%
< 1000 MHz 1000 to 1500 MHz 1500 - 2000 MHz 2000-3000 MHz > 3000 MHz
Spectrum need - to meet the DL requirement [MHz]
Activity factor 5% Activity factor 10% Activity factor 15% Activity factor 20%

Spectrum needed to meet the 50Mbit/s uplink speed requirement
53
The uplink speed
requirement could add
between 0 and 30% to
total spectrum need.
Activity factor
%Traffic offload 50% 30% 0% 50% 30% 0% 50% 30% 0% 50% 30% 0% 50% 30% 0%
Karachi central 100 170 310 290 420 700 460 670 1080 630 910 1480 800 1160 1860
Paris 50 110 210 180 280 490 310 460 760 440 640 1050 560 810 1330
Madrid 40 110 200 170 270 470 300 450 740 430 620 1010 540 780 1280
Rabat central 20 50 130 110 190 330 210 330 540 300 450 750 400 580 950
Rome 0 20 110 90 150 280 170 270 460 260 380 640 330 500 810
Berlin 0 0 80 70 130 240 140 230 390 210 320 550 280 430 710
Amsterdam 0 0 0 0 70 150 80 140 260 130 200 360 170 280 470
Aman central 0 0 20 0 40 110 50 110 220 100 170 310 140 230 410
Khartoum 0 0 0 0 0 40 0 20 100 0 70 170 60 120 220
Extra need < 1000 MHz 1000 to 1500 MHz 1500 - 2000 MHz 2000-3000 MHz > 3000 MHz
UL additional spectrum need - addressed with UL-only spectrum MHz]
Activity factor 5% Activity factor 10% Activity factor 15% Activity factor 20% Activity factor 25%
Total extra low and mid-band spectrum need (MHz) to meet UL requirement
 In areas with lower traffic densities, the uplink portion of the spectrum is sufficient to cover the UL requirement
 In areas of very high traffic density, the average uplink portion of spectrum is insufficient to meet the UL requirement, because TDD
(with a 3:1 DL-to-UL ratio) becomes a large part of the proportion of the overall mix
– In this case, extra spectrum is needed to satisfy the UL requirement in addition to the DL requirement

Total low and mid-band spectrum need (MHz) to meet the DL as well as the UL requirement
Where the estimated low and mid-band spectrum demand exceed the available supply of IMT frequencies up to 6GHz, the shortfall would
entail either:
 A failure to meet the IMT-2020 Requirements in exceptionally concentrated population areas; or
 Costly measures to overcome the shortfall, including higher than assumed network densification and/or deployment of technology
enhancements that deliver significantly higher spectral efficiency gains than projected; and/or
 An even greater reliance on traffic offloading to high frequencies and/or indoor cells
Spectrum needed to meet the both the 100Mbit/s DL and 50Mbit/s UL speed requirement
54
Bandwidth shortfalls
caused by a failure to
release sufficient IMT-
designated spectrum
could result in
substantial socio-
economic harm.
Activity factor
%Traffic offload 50% 30% 0% 50% 30% 0% 50% 30% 0% 50% 30% 0% 50% 30% 0%
Karachi central 1260 1590 1990 2090 2740 3540 2910 3900 5090 3730 5050 6650 4550 6200 8200
Paris 1050 1270 1560 1630 2100 2680 2220 2940 3800 2820 3770 4930 3410 4600 6050
Madrid 1040 1250 1530 1590 2050 2610 2170 2860 3700 2750 3670 4790 3320 4470 5870
Rabat central 870 1040 1260 1310 1660 2080 1750 2280 2910 2190 2890 3740 2630 3500 4560
Rome 1000 1020 1150 1190 1490 1860 1570 2020 2570 1950 2550 3280 2320 3080 3990
Berlin 1000 1000 1080 1100 1370 1690 1430 1830 2310 1760 2290 2930 2090 2760 3560
Amsterdam 1000 1000 1000 1000 1090 1320 1140 1420 1760 1370 1740 2190 1600 2070 2630
Aman central 670 730 810 830 990 1190 1050 1280 1580 1240 1570 1960 1440 1850 2340
Khartoum 610 650 690 700 780 900 1000 1020 1140 1000 1130 1380 1060 1310 1610
Spectrum need < 1000 MHz 1000 to 1500 MHz 1500 - 2000 MHz 2000-3000 MHz > 3000 MHz
Total spectrum need - to meet both DL and UL requirements [MHz]
Activity factor 5% Activity factor 10% Activity factor 15% Activity factor 20% Activity factor 25%
= plausible lower-bound of spectrum need in 2025

Mobile spectrum and network evolution to 2025 slides coleago - 24 mar 21

Mobile spectrum and network evolution to 2025 slides coleago - 24 mar 21

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Mobile spectrum and network evolution to 2025 slides coleago - 24 mar 21

Similar to Mobile spectrum and network evolution to 2025 slides coleago - 24 mar 21 (20)

More from Coleago Consulting

More from Coleago Consulting (20)

Recently uploaded

Recently uploaded (20)

Mobile spectrum and network evolution to 2025 slides coleago - 24 mar 21