Perspective

Controlling energy use: the

role of AI-based solutions
March 2023
Caroline Gabriel, Michela Venturelli and Grace Langham
 Controlling energy use: the role of AI-based solutions | i

Contents

1. Executive summary 1

2. Rising energy costs and usage are forcing CSPs to focus on efficiencies 2
5G was designed to be energy efficient but its use cases risk a rise in consumption 4

3. CSPs can adopt a variety of strategies when it comes to tackling energy usage 6

4. AI-based solutions can enable CSPs to achieve short-term energy and cost savings 9
Selecting the optimal business and delivery model 10
Overview of an ideal AI-based energy saving solution 12
Potential benefits of an AI-based solutions 13

5. Recommendations and conclusion 15

6. About the authors 17

List of figures
Figure 1.1: Overview of an ideal AI-based energy-saving solution ................................................................. 2
Figure 2.1: Wholesale electricity prices, selected countries worldwide, July 2018–July 2022 ........................ 3
Figure 2.2: Maximum power consumption of a base station that supports multiple mobile generations, by
component ........................................................................................................................................................ 5
Figure 2.3: Energy increase, generic developed country that is similar in size to the UK, 2022 ...................... 5
Figure 3.1: Energy management solutions for CSPs’ networks ....................................................................... 6
Figure 4.1: Key benefits of SaaS deployments ............................................................................................... 11
Figure 4.2: Features of an ideal AI-based energy saving solution .................................................................. 12
Figure 4.3: Examples of major CSPs making cost and energy saving using AI-based energy saving solutions
........................................................................................................................................................................ 13
Figure 4.4: Maximum percentage of power reduction achieved using various AI-based energy-saving
features ........................................................................................................................................................... 14

This perspective was commissioned by Nokia. Usage is subject to the terms and conditions in our copyright
notice. Analysys Mason does not endorse any of the vendor’s products or services.

© Analysys Mason Limited 2023 Contents

 Controlling energy use: the role of AI-based solutions | 1

1. Executive summary

Building and running energy efficient networks is becoming a strategic imperative for all telecoms CSPs.

Commitment to reducing energy consumption is a common issue in many businesses, driven by increasing
energy prices as well as by regulatory demands to control and be more transparent about energy management
and carbon emissions. However, 5G deployment, which risks increasing CSPs’ energy use, is intensifying the
need to improve energy efficiency in the telecoms industry. Even though 5G technologies were designed to
reduce energy consumption by 90% compared to 4G on a like-for-like basis, the greater density of 5G means
that, without intervention, adding 5G to existing networks will increase CSPs’ energy use.

CSPs have plenty of opportunities to reduce energy consumption in the mobile network without affecting user
experience. CSPs have a vast choice of energy-saving measures, each with different implementation time-scales
and potential savings, that go beyond just the natural deployment of more-modern networks. Measures to reduce
energy consumption can address all elements of the network and can be grouped into four categories: network
modernisation, intelligent power-saving features, efficient use of assets and alternative ways of sourcing energy.

CSPs need initially to prioritise the solutions that can be implemented in a short time-scale and that can deliver
results quickly. AI-based software solutions offer CSPs the opportunity to implement energy-saving measures as
soon as possible. Most importantly, the unique advantages that AI can deliver to energy management efficiency
can be enhanced by the technical features of 5G connectivity.

AI can expand the potential of traditional energy-saving features. For example, AI can be applied to static
switch off/on networking equipment to enable it to adjust for dynamic changes in customer behaviour and
usage. It can also help to tap into new energy-saving opportunities across the network. AI can be used to predict
traffic patterns and fluctuations, forecast network utilisation and weather impacts, provide maintenance and fault
management data, and suggest the most actionable approach to energy management in even the most complex of
network scenarios. Energy savings will be dynamically adjusted to align with network performance, ensuring
that CSPs’ goals and KPIs are met but customer experience is not compromised.

However, it is extremely important that CSPs select a business and delivery model for AI-based energy-saving
solutions that is best-suited to address the urgency of CSPs’ current cost challenges. The software-as-a-service
(SaaS)-based business model, as an alternative delivery mechanism to on-premises and hosted cloud
deployments, can greatly reduce time to value. CSPs can quickly have access to SaaS solutions because they are
kept as standardised as possible, often including a set of blue-print services, which allows for minimum
customisation and reduced time spent on installation and configuration. Similarly, continuous and automatic
updates and upgrades can be efficiently managed and delivered by the software vendors. As a result, CSPs
benefit from faster engagement with vendors, meaning shorter time to value.

We have provided an overview of an ideal AI-based energy saving solution (Figure 1.1) that effectively aligns
energy consumption reductions with network performance requirements. This solution will allow CSPs to
realise quantifiable energy and cost-saving benefits across the entire mobile network; and can be achieved in the
short term with limited upfront investment.

© Analysys Mason Limited 2023 1: Executive summary

 Controlling energy use: the role of AI-based solutions | 2

Figure 1.1: Overview of an ideal AI-based energy-saving solution

2. Rising energy costs and usage are forcing CSPs to focus

on efficiencies

Energy management and energy savings in mobile networks are becoming a clear priority for most CSPs. The
recent surge in the cost of electricity, the deployment of 5G networks and commitments to environmental,
sustainability and governance (ESG) goals have brought power consumption and management to the top of
CSPs’ agendas.

For a mobile operator, energy costs account for up to 7% of their operational expenditure (opex) and depending
on the type of networks they run, CSPs can spend from 3% to 20% of their total revenue on energy bills. The
mobile radio access network (RAN) accounts for a large part of their energy costs and for over half of their
energy consumption (70% in a typical 5G macro network). In a typical 5G macro network, the core accounts for
8% of the total energy consumption, while the digital baseband units and the radio account for 16% and 76%,
respectively. Around 55% of the RAN energy is consumed by passive components such as air conditioning, fans

© Analysys Mason Limited 2023 2: Rising energy costs and usage are forcing CSPs to focus on efficiencies
 Controlling energy use: the role of AI-based solutions | 3

and power systems. The RAN’s share of energy usage is continuing to rise and will continue to put pressure on
opex. In the current climate, this pressure is further intensified by the increase in energy prices. Even if an
operator can keep its energy consumption flat, costs will still be increasing.

Recent geopolitical events have put pressure on the global supply of natural gas, resulting in a significant
increase in the wholesale price of gas and electricity in 2022 (Figure 2.1). This has resulted in many CSPs citing
high inflationary environment and increased energy prices as factors behind growing operating expenses. Those
same factors are also pushing CSPs to take additional action to maintain the cash flow needed to support
network investments (for example, BT in the FY23 half year results announced an increase in its cost savings
target from GPB2.5 billion to GPB 3.0 billion by the end of FY25).

Figure 2.1: Wholesale electricity prices, selected countries worldwide, July 2018–July 20221

Customers, governments, standards/specification bodies and policy makers are also urging CSPs to actively
reduce emissions and energy consumption. Legislation, such as the Paris Agreement and the EU Green deal
action plan, is pushing towards more-sustainable management by mandating disclosure of carbon emissions and
energy-saving strategies.2,3 As a result, sustainability is becoming one of the biggest priorities in the
transformation strategies of many CSPs and an important trigger to scrutinise energy consumption. It is vital for
CSPs to maintain up-to-date datasets on energy efficiency and communicate their progress and vision
transparently. Some major CSPs have already outlined their carbon-neutral targets; for example, Orange aims to
be carbon-neutral by 2040, with 50% of its energy coming from renewables by 2025; and Vodafone plans to
eliminate all carbon emissions from its own activities by 2030.

1 For more information see Analysys Mason’s Managing telecoms energy consumption using inventory solutions.
2 Paris Agreement, https://unfccc.int/process-and-meetings/the-paris-agreement/the-paris-agreement.
3 EU Green Deal, A European Green Deal | European Commission (europa.eu).

© Analysys Mason Limited 2023 2: Rising energy costs and usage are forcing CSPs to focus on efficiencies
 Controlling energy use: the role of AI-based solutions | 4

5G was designed to be energy efficient but its use cases risk a rise in consumption
The deployment of 5G networks is another element that is driving CSPs to consider energy consumption and
energy management as a top priority. While 5G New Radio (NR) was designed to support greatly improved
energy efficiency compared to previous generations of mobile technologies, rising numbers of cell sites and
antenna elements compared to 4G mean that energy usage can actually increase significantly. Furthermore,
because mass adoption of 5G is ongoing and the number of 5G subscriptions is still a small percentage of all
mobile subscriptions, the connectivity and bandwidth of the network are not used to their full extent and the
energy used to power 5G networks is partly wasted.

5G technology is more efficient than 4G when considering single base stations with a study concluding that 5G
new radio access technology can decrease the energy consumption by more than 50% while providing around
10 times more capacity.4 The architecture of 5G NR was designed to support a higher capacity than 4G and to
deliver greater energy efficiency for low-to-medium traffic levels. This results in lower average load in terms of
the total capacity that a given traffic volume uses and higher percentage of 5G traffic falling into the ‘low-to-
medium’ category, which is subjected to the greatest efficiencies. The design of 5G cells also supports power-
efficient features such as micro-sleep functions, which puts radios into sleep mode whenever there is a gap in
transmission and hence helps to save energy.

In general, the architecture of 5G base stations reflects one of the key objectives set for 5G by the International
Telecommunications Union (ITU) to reduce energy consumption by 90% compared to 4G, on a like-for-like
basis (same type and number of base stations with the same traffic and signalling load).

However, the use of increasingly high-frequency spectrum for 5G mobile networks together with the
implementation of high-capacity and ultra-reliable low latency (URLL) use cases, will affect the density and the
type of cell sites required and can cancel out the energy efficiencies of 5G NR.

5G uses a variety of spectrum bands, including mid-band (3.5–4.2GHz) and millimetre wave (26/28GHz and
39GHz) ranges that are unique to 5G. Base stations in higher-frequency bands have a shorter signal range and so
coverage is reduced when compared, for example, to low-band spectrum. This requires CSPs to either increase
the number of cells or to deploy large arrays of antennas (massive MIMO (mMIMO)), which can maximise
effective signal reach through beamforming technology. In both cases, CSPs’ energy consumption will increase:
on the one hand, operator will be required to build denser networks with a higher number of elements of passive
and active equipment, while on the other hand, the deployment of mMIMO will increase the power consumption
of RF circuits due to the high levels of compute processing required (Figure 2.2).

4 Tombaz S., Frenger P., Olsson M. and Nilsson A. (2016), ‘Energy performance of 5G-NX radio access at country level’, 2016
IEEE 12th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), pp.1–6, doi:
10.1109/WiMOB.2016.7763183.

© Analysys Mason Limited 2023 2: Rising energy costs and usage are forcing CSPs to focus on efficiencies
 Controlling energy use: the role of AI-based solutions | 5

Figure 2.2: Maximum power consumption of a base station that supports multiple mobile generations, by component5

Therefore, adding 5G without mitigating steps can add significantly to energy usage. In Figure 2.3, the energy
increase incurred by the introduction of 5G is calculated for an operator in a developed country in a market
similar to the UK.

Figure 2.3: Energy increase, generic developed country that is similar in size to the UK, 2022

5G is not the only network that CSPs should consider when focusing on energy efficiency
In devising energy-efficiency strategies, CSPs also need to consider older technologies within their networks
and take a holistic approach to energy usage. Integrated fixed–mobile players need to consider both the

5 For more information, see Analysys Mason’s Driving down energy usage across telecoms networks: 5G RAN and beyond.

© Analysys Mason Limited 2023 2: Rising energy costs and usage are forcing CSPs to focus on efficiencies
 Controlling energy use: the role of AI-based solutions | 6

challenges and the synergies arising from running a fixed network alongside a mobile one. On the mobile side,
5G usually has to coexist with older technologies, namely 2G, 3G and 4G, greatly increasing an operator’s total
power consumption. Applying intelligent management solutions to already deployed base stations and
equipment is not trivial and, in most cases, will require significant investments (which, at times, are not worth
the energy saving efforts because the networks will soon be disconnected).

CSPs will need to consider how to efficiently run a multi-technology network. Decommissioning older
generation networks is the biggest step that most CSPs can take to reduce their energy usage. Alternatively, they
should consider the modernisation of the legacy 2G, 3G and 4G base stations with energy-efficient Single RAN
software and hardware, in which multiple legacy networks can be replaced with a set of multiple radio access
technology (multi-RAT) base stations, while rolling-out 5G radios.

It is worth noting that regulators are offering the possibility to refarm legacy 2G, 3G and 4G spectrum for 5G
use. This can incentivise CSPs to take steps in transitioning from legacy technologies and/or in encouraging the
investment in energy-saving solutions.

3. CSPs can adopt a variety of strategies when it comes to

tackling energy usage

There are plenty of ways to reduce energy consumption in mobile networks and adopt new energy management
solutions. Measures to reduce energy consumption can be taken in all elements of the network, but some will
have a greater impact than others. The core accounts for only 8% of the total energy consumption in a 5G
network (excluding transport), and the majority of the RAN’s energy (83%) is used by the cell site equipment
rather than the digital baseband. A large proportion of the energy used to power the network is wasted in
operating cooling systems (55%), running idling equipment and in powering amplifiers and other unused
network elements. Only around 15% of mobile network energy is used for transmission, which is the only
activity that generates revenue within the cell site.

Some other savings lie in architectural and network transformations, including network modernisation and the
introduction of intelligent power-savings features, some in more efficient usage of network assets and some in
alternative ways of sourcing energy. We have classified them depending on the timescale across which the
solution can be implemented, the timing when benefits can be achieved and the drawbacks of the solution
(Figure 3.1).

Figure 3.1: Energy management solutions for CSPs’ networks

Approach Solution Timescale Description Solution drawbacks

Alternative Renewable Short-term This is a quick and effective Renewable energy relies on external
ways of energy sources way to reduce an operator’s factors such as climate, regulatory
sourcing environmental footprint and environment, availability and the
energy secure power supply by price of renewable energy through
reducing reliance on non- the grid. At times, green energy is
renewable energy sources. offered by traditional utility provider
at a premium price.

© Analysys Mason Limited 2023 3: CSPs can adopt a variety of strategies when it comes to tackling energy usage
 Controlling energy use: the role of AI-based solutions | 7

Approach Solution Timescale Description Solution drawbacks

Alternative Self-generation of Long-term Increasing the use of self- The extent to which a company can
ways of power supplies produced energy to reduce scale such a model will depend on
sourcing dependence on external the suitability of the site locations for
energy power grid supply and prices. alternative power source, such as
solar and wind, availability of space,
and cooperation of the site’s
landlord.

Alternative Establish power Short-term PPA is a multi-year contract PPAs are temporary solutions. After
ways of purchase taken out by CSPs with energy the agreement expires (it usually
sourcing agreements suppliers that guarantees a lasts a few years), CSPs need to
energy (PPAs) more stable price for CSPs renegotiate or seek other ways to
and facilitates the lender in control energy prices.
planning the necessary energy
capacity.

Network Network Long-term Shutting down 2G/3G The decommissioning of networks

modernisation decommissioning networks and moving to requires large resources (labour and
4G/5G single RAN could help capital) and is a long-term
CSPs to save 571GWh/year. investment, meaning cost reductions
Without the straightforward and energy savings will not be
retirement of legacy networks, realised until a much later date.
5G simply adds to energy
consumption. The early
decommissioning of 3G could
reduce mobile energy
consumption by 10%.

Network Liquid cooling Medium- Traditional air-cooling systems Liquid cooling is more expensive
modernisation term account for 55% of the energy than traditional cooling methods.
consumption of the RAN. This, The installation can be complex and
as well as requiring regular if not fitted correctly could result in
maintenance, filters changes leaks and damage to electornic
and re-gassing. Liquid cooling components.
solutions convert heat into
liquid and remove it from the
site by liquid circulation. CSPs
have claimed a 30% reduction
in energy costs by switching
from air cooling to liquid
cooling.

Network AI-driven network Long-term Customising RAN, processors Only applicable on new networks and
modernisation design as well as passive ideal for those sites where a large
infrastructure specifications set of data about traffic and usage
site by site, depending on load are available.
conditions.

Network The use of Long-term Thermal interface materials Not convenient for legacy networks.
modernisation thermal transfer heat energy away The installation and maintainence
management from the base station requires a high level of expertise,
materials for components and into a heat labour and expenses.
base stations sink. Thermal gels and
polymer-based materials are
optimal for thermal
management and can
dissipate heat at high rates.

© Analysys Mason Limited 2023 3: CSPs can adopt a variety of strategies when it comes to tackling energy usage
 Controlling energy use: the role of AI-based solutions | 8

Approach Solution Timescale Description Solution drawbacks

Network IoT-enriched Medium- IoT sensors are used to track The fitting of IoT sensors can be a
modernisation networks term parameters, such as energy resource- and time-intensive task. If
consumption or site not supported by an adeguate IoT
temperature, giving CSPs platform, data analysis and AI-
more control over their powered tools, the solution can lose
resources and enabling part of its value.
consumption optimisation.

Intelligent AI- driven Short-term Smart energy management The solution relies on data analysis,
power-savings shutdown such as switch off/on of base therefore would be suitable for CSPs
features solutions stations. AI refines and that have tools already in place for
expands the potential for network analysis. It might be
energy-saving opportunities by challanging to apply these solutions
analysing real-time traffic to legacy networks.
patterns and network-resource
availability and taking
automated decisions about
which parts of the system can
be put into sleep mode or shut
down.

Intelligent AI-driven Short-term AI-powered equipment can It might be challenging to apply

power-savings equipment couple smart shutdowns with these solutions to legacy networks.
features control cross-equipment optimisation
by adjusting coverage or
power consumption
dynamically.

Efficient use Reducing FWA Long-term Minimising the use of FWA For integrated CSPs only. The
of assets usage or steering and steering substitutable incentive to make this change is
mobile traffic to mobile traffic to Wi-Fi (where often outweighted by shorter-term
Wi-Fi the network usage per line is commercial demand.
about 40 times lower) to help
reduce the rate of energy-
costly upgrades to the RAN.

Efficient use Weigh the Long-term Opt to scale back/delay While the impact on energy usage
of assets opportunities for deployment of power-hungry will be only delayed, CSPs face the
energy mMIMO to limit and control risk of being less competitive in the
consumption power usage. market in terms of connectivity and
against technology offered.
commercial
priorities

Efficient use Engaging in Long-term Sharing a network with other Establishing network sharing
of assets mobile network CSPs optimise the use of agreements is a lenghtly process
sharing network equipment and avoid that require regulatory approval. It
agreements duplication, enabling CSPs to has a big impact on CSPs’ strategy
achieve substantial savings in so the rational of the agreement
energy. should go beyond energy savings.

Source: Analysys Mason

The potential savings that each strategy can deliver will vary by operator and country, and will also depend on
engagement with third parties, such as regulators and government policy makers. These bodies must provide
CSPs with clear guidance on ESG policies. This could include making it easier for CSPs to remove old
technologies and switch users to more-energy efficient options. Often the incentive for CSPs to reduce energy

© Analysys Mason Limited 2023 3: CSPs can adopt a variety of strategies when it comes to tackling energy usage
 Controlling energy use: the role of AI-based solutions | 9

consumption across the mobile network is outweighed by shorter-term commercial incentives to preserve user
experience and to reduce longer-term TCO; and so, this is where policy interventions will be key to driving
change.

Transitioning to more energy-efficient networks is extremely important for CSPs because those savings can
reverberate across the whole value chain, including end users such as enterprises and retail customers. By
reducing energy costs and protecting themselves from energy price spikes, CSPs could increase profitability
and, ultimately, be able to pass those savings to customers, gaining a competitive advantage.

4. AI-based solutions can enable CSPs to achieve short-

term energy and cost savings

Energy management solutions are not a new concept for CSPs, but to date the solutions have been inflexible and
unintelligent. Most old generation equipment deployed by CSPs has energy-saving features, such as automatic
switching on/off in networking equipment and sleep modes during periods of low traffic. However, these
solutions work within pre-defined and static parameters that often do not align with real usage patterns, which
can compromise both potential savings and quality of service. The lack of optimisation of power saving features
may discourage CSPs because of the risk of a negative impact on network performance.

AI-based energy management could help CSPs to optimise their approach to energy management across both
active and passive elements of the network. The ability to analyse large amounts of data relating to real-time
traffic patterns and network-resource availability allows for real-time decisions on how different elements of the
network can be operated while optimising energy usage. The unique advantages that AI can deliver to energy
management are particularly relevant now because CSPs urgently need to cut energy consumption and are
deploying 5G networks.

• AI-based solutions are quick to implement and set up due to their software nature, which means that energy
cost savings are achieved within a short period (weeks). This feature addresses the urgency with which
CSPs need to act to tackle increasing energy bills.

• The efficiency of AI-based solutions can scale according to the increase in the amount of available data.
The growth in data consumption expected with 5G, where customer data usage on 5G can be 5–10 times
greater than 4G, means that CSPs can tap into a larger pool of data that can help to deliver more accurate
traffic behaviour prediction.

• The efficiency of AI-based solutions can also be enhanced by the technical features of 5G connectivity. 5G
can support a massive number of devices that simultaneously access the network, which is far more than
previous generations can handle. As this feature enables IoT applications, these applications can be
purposed at monitoring energy and network equipment usage, which can be fed into AI-based solution. The
low latency that 5G networks can deliver could also play a fundamental role in increasing the efficiency of
AI because it can guarantee real-time data analysis and real-time decision-making, optimising both the
energy savings and the overall network performance, including quality of experience (QoE).

AI can expand the potential of traditional energy-saving features (for example, AI can be applied to traditional
fixed switch off/on networking equipment that is largely not effective given the dynamic changes in customer

© Analysys Mason Limited 2023 4: AI-based solutions can enable CSPs to achieve short-term energy and cost savings
 Controlling energy use: the role of AI-based solutions | 10

behaviour and usage) as well as tap into new energy-savings opportunities across the network. AI can be used to
predict traffic patterns and fluctuations, forecast network utilisation and weather impacts, in order to provide
data-driven recommendations to help CSPs to effectively optimise their network configurations by pinpointing
the most power-hungry network components and seeing which changes will have the greatest impact on costs.
Energy savings will then be dynamically adjusted to align with network performance and customer experience;
ensuring that it meets CSPs’ goals and KPIs.

Other examples of AI-based energy management solutions include the following.

• Intelligent power up/down management can improve energy efficiency of network elements, including g
base stations. AI-based algorithms and machine learning can be used to dynamically turn off and on
(partially or wholly) passive and active network elements in response to changes in traffic load. AI can
modify energy saving windows compared to static schedules and avoids disruption to network performance.
This solution can help to offload macrocell traffic to a low-power small cell during a quiet time, so that the
macrocell can stay asleep for a longer period. Intelligent power management can also be applied to Air
conditioning (AC) elements, where savings can be achieved by adjusting the number of AC working hours,
particularly at low traffic levels, according to the characteristic temperature of different computer rooms
and the features of the equipment in use.

• Equipment anomaly detection provides a holistic view across active and passive network domains for
real-time detection of service-impacting incidents and abnormal energy consumption such as wastage,
leakage and theft. Intelligent anomaly detection can save energy across the network through faster
localisation of energy anomalies. Some of these anomalies can be solved remotely, reducing the number of
on-site visits that personnel must make to troubleshoot network issues and thereby cutting costs for CSPs.

• Power source alterations. AI can be used to actively choose the most efficient source of power during
peak traffic hours. For example, AI can forecast and change the power source of the air conditioning units
to either electricity grid or batteries depending on which is most energy-efficient during peak/off-peak
periods.

Selecting the optimal business and delivery model

It is important that CSPs select a business and ownership model for AI-based energy saving solutions that
addresses the urgency of their cost challenges.

The software-as-a-service (SaaS)-based business model (Figure 4.1), as an alternative delivery mechanism to
on-premises and hosted cloud deployments, can support a short deployment time. SaaS-based solutions radically
differ from traditional on-premises models of software delivery because, with a SaaS solution, CSPs pay a
regular fee for accessing software that is completely managed by the SaaS solution vendor, as opposed to
owning, maintaining and deploying the solution on premises. With SaaS delivery, software vendors are
responsible for the whole lifecycle of the software, including IT and software maintenance, updates, and
security. The software is based on the vendor’s own platforms and applications and is hosted in the public cloud.
This reduces the level of maintenance and in-house expertise required from CSPs and enable them to focus
internal resources and staff elsewhere. SaaS also allows CSPs to flexibly scale their operations up and down,
with minimal service disruption. In the CSP market, SaaS solutions need to support the specific processes and
requirements of telecom networks, and can be regarded as a distinct category of ‘telecom SaaS’.

All the different costs associated with staff training, updates and upgrades, maintenance and security that CSPs
had to incur in traditional ownership models are replace by subscription fees to SaaS service providers that take

© Analysys Mason Limited 2023 4: AI-based solutions can enable CSPs to achieve short-term energy and cost savings
 Controlling energy use: the role of AI-based solutions | 11

care of delivering and maintaining the software, cancelling out heavy upfront payments and reducing the need
for lengthy contract commitments. This means that CSPs can be shielded from fluctuating hardware prices and
do not need to consider additional hardware purchases. Such changes can enable them to reduce, and better
control, the TCO of a new solution.

Moreover, CSPs can quickly get started with a telecom SaaS solution because they are kept as standardised as
possible, often including a set of blue-print services, which allows for minimum customisation and reduced time
spent on installation and configuration. Similarly, continuous and automatic updates and upgrades can be
efficiently managed and delivered by SaaS service providers. As a result, CSPs can benefit of faster time to
value and can achieve their targeted results quicker. CSPs will have immediate control of their spending and
improved transparency and clarity regarding the level of service they are getting from the vendor. The agility
offered by telecom SaaS-based solutions also grant a higher level of control to CSPs that want to add or modify
features or services delivered by a SaaS provider. They can easily add features and benefit from rapid
procurement, simply by modifying the subscription, and the quick deployment process.

SaaS-based models have increased in popularity in recent years; with operator spending on SaaS expected to
increase from 11% in 2023 to 22% in 2027 growing at a CAGR of 17%. The growing adoption of this delivery
model is also due to the maturing of this technology, which has increased CSPs’ confidence in adopting it. One
of the key challenges around SaaS was related to the level of security and privacy that could be guaranteed;
vendors have increasingly addressed this challenge and now often add layers of security on top of public cloud
infrastructure.

Figure 4.1: Key benefits of SaaS deployments

© Analysys Mason Limited 2023 4: AI-based solutions can enable CSPs to achieve short-term energy and cost savings
 Controlling energy use: the role of AI-based solutions | 12

Overview of an ideal AI-based energy saving solution

We have provided an overview of an ideal AI-based solution (Figure 4.2) that can be used for achieving energy
and cost reductions across the entire mobile network. The overview incorporates important features that will
help maximising the success of the solution for both vendors and CSPs.

Solution providers should consider the importance of providing a single analytical tool in a network supporting
a wide range of vendor equipment, particularly as we move towards Open RAN concepts. The solution should
be able to provide holistic energy management, helping to reduce the large amount of energy waste occurring in
both active and passive equipment of the mobile network. AI technology should be dynamic and mature and
must be able to successfully align high network performance and quality of service with reductions in energy
usage. A SaaS-based cloud deployment will enable CSPs and enterprises to have greater control of costs and
outcomes, provides immediate use and allows for services to be flexibly customised, while also delivering
sizable results over a much shorter space of time.

Figure 4.2: Features of an ideal AI-based energy saving solution

Features Description
Multi-vendor The solution should support a wide range of vendor RAN equipment by establishing a
common set of standardised KPIs cutting across different vendors in the market. By
removing the dependency on vendor-specific equipment, the solution can be rapidly and
quickly deployed to optimise power consumption of RAN equipment across all layers in
multi-vendor mobile networks.

Dynamic and holistic energy The solution should be able to analyse energy trends and optimise energy use of both
management passive and active components across the entire mobile network. Too often the
impact of passive elements have been overlooked when analysing energy
consumption. The solution should also be able to identify opportunities for energy
optimisation in both complex and simple network scenarios, such as site level
analytics like weather data and asset databases, as well as being able to deal with
different base stations with different load requirements, such as urban or rural
environments.

Guaranteed network It is important that quality of service is not compromised at the expense of energy
performance and quality of savings. AI technology should be used to precisely and dynamically adapt energy
service consumption to traffic levels by predicting power off/on timings and optimising the use
of resources such as wide wake windows and stand-by modes to shorten wake-up
times. AI-based solutions can help to maintain high network quality and hence prevent
poor customer experience while optimising power savings.

SaaS delivery model CSPs should be able to consume and access the solution through flexible delivery
models. Selecting a SaaS AI-based solution will enable CSPs to quickly engage with
vendors, meaning they can benefit from faster time to value. The software nature of
the solution also guarantees minimal disruption, short installation time and quick
scalability across the network. CSPs can have more control over the solution and the
benefits achieved.

Flexible cloud environment CSPs should adopt a solution that guarantees maximum flexibility of deployment
support options. This allow CSPs to implement one unique solution whether they choose to
deploy it on public or private cloud or on hybrid deployments, giving CSPs the benefit
of short delivery timescales and the flexibility of not being locked to a specific
infrastructure.

Delivery of sizeable results It is important that CSPs are able to rapidly quantify and assess the efficiency of the
solution. Understanding the energy savings that have been achieved since the AI
solution was implemented can help CSPs to tailor its performance as well as plan
medium- to long-term energy savings strategies.

© Analysys Mason Limited 2023 4: AI-based solutions can enable CSPs to achieve short-term energy and cost savings
 Controlling energy use: the role of AI-based solutions | 13

Features Description

Modular solution Selecting a solution that can analyse the energy performance of the whole RAN
equipment is important. Equally important for CSPs is selecting a solution that has
modular features, meaning that CSPs can benefit from greater business agility by
picking and chosing, depending on their budgets and requirements, which features
they want to deploy.

Source: Analysys Mason

Potential benefits of an AI-based solutions

AI-based energy saving solutions can provide multiple benefits for CSPs, including:

• reduced energy costs

• efficient use of energy and network resources
• more sustainable patterns of energy use and realisation of carbon-neutral targets.

Several global Tier-1 CSPs have already implemented AI-based energy solutions and have reduced their overall
mobile network energy consumption (Figure 4.3). AI-based energy solutions have significantly reduced the
power consumption of various elements of the network, with one operator reporting a 74% power reduction in
air conditioning units, which accounted for nearly half of the energy consumption of a radio site. This was
achieved by using AI to collect temperatures of different rooms within and outside the cell site and to regulate
the air conditioner in the computer room while maximising the active exchange and cooling of the hot and cold
air. Another operator achieved cost savings because the deployed AI-based energy-efficient solution was able to
identify anomalies in the energy billing systems, highlighting an overpayment of 34% in 7% of its sites. It is
important to highlight that these savings were achieved without disruption to network performance and quality
of service; and at no point was customer experience affected.

Figure 4.3: Examples of major CSPs making cost and energy saving using AI-based energy saving solutions

AI-based energy saving features can be applied to various components of the network (Figure 4.4) and it is
Reduction in AC
important that CSPs consider deploying multiple features to achieve the greatest reduction in power. Ultimately,
consumption
AI-based energy saving solutions can lead to both short-term savings, with CSPs recording overall savings of
more than 5%, and medium- to long-term savings of up to 30%.

© Analysys Mason Limited 2023 4: AI-based solutions can enable CSPs to achieve short-term energy and cost savings
 Controlling energy use: the role of AI-based solutions | 14

Figure 4.4: Maximum percentage of power reduction achieved using various AI-based energy-saving features
Type of AI-based energy Maximum percentage Description
feature feature power reduction (kWh)

Hardware Hard switch off 14% Turning on and off (partially or wholly) base
station modules (RRM/RRU/AAU) in response to
changes in traffic loads using AI.

Software Soft switch 15% Using AI to constantly adjust the relative timing of
elements within the switching system. These
savings on top of base stations energy features.

Hardware Intelligent AC control 74% Automatically alters the number of AC working

hours, particularly at low traffic levels.

Software Equipment anomaly 2% AI identifies abnormal equipment consumption.

detection

Hardware Intelligent fresh air 31% AI reduces the time that air conditioners are
ventilation operating by using fresh air to cool down the site.

Source: Analysys Mason

Anomaly detection and the activation of different types of shutdown functions through AI-powered software can
help CSPs to detect malfunctioning or improperly configured equipment, restarting cell sites and reducing the
unnecessary use of equipment. These can increase the longevity of the network and hence the deployment of AI-
based solutions can provide a better return of investment (ROI) for CSPs.

Opex reduction is ranked as one of the top priorities for CSPs and energy is expected to be one of the only major
opex elements that is expected to increase within the next 3 years. AI-based energy saving solutions can help to
mitigate the effect of increasing energy prices on opex and hence free up resources for other business-critical
needs and commitments. For example, a South African operator has recently trialled an AI-based energy-saving
solution and indicated that power reduction equated to a USD1.6 million saving each year based on 5000 sites
after the solution was implemented.

Examples of where CSPs can achieve opex savings using an AI-based energy solutions include the following.

• The need for fewer on-site maintenance and engineer call outs. AI provides predictive maintenance and
enhanced troubleshooting and can immediately detect network disruptions or faults. In most cases, these
can be resolved remotely and without manual intervention.

• A reduction in unnecessary energy costs due to wasteful mobile network equipment. AI software can detect
inefficient use of power across both passive and active network components and can reduce the run hours of
equipment during low traffic periods.

• Less frequent replacement or upgrade of network components. AI can be used to optimise equipment
lifetime.

• A reduction in the use of uneconomical power sources, particularly during peak hours. AI can be used to
actively choose the most efficient energy source depending on network conditions and external factors.

• Minimised mobile network outages and widespread business disruption.

© Analysys Mason Limited 2023 4: AI-based solutions can enable CSPs to achieve short-term energy and cost savings
 Controlling energy use: the role of AI-based solutions | 15

4.4 AI-based models will be essential in reducing CSPs CO2 emissions.

AI will play a crucial role in reducing greenhouse gas (GHG) emissions in mobile networks, where idling
network equipment and energy-intensive cooling systems are the two biggest contributors to carbon emissions.

AI-driven power management solutions provide insights about immediate actions for shrinking carbon
emissions, helping CSPs as they move towards a low-carbon future and ramp up their efforts to support the Paris
Agreement’s 1.5-degree climate change pledge.

AI-powered algorithms can be used to monitor, predict and reduce CO 2 emissions by collecting data across the
entire mobile network and aligning it with CSPs’ KPIs and reduction targets. AI-based simulations can
recommend how to correct, upgrade or modernise an CSPs’ network and will forecast how the proposed
changes will affect emissions. This will help CSPs to be more aware and in control of their carbon emissions
and will be an essential tool for planning how to reach CO2 reduction targets.

The use of AI-based solutions can also help to reduce carbon emissions by detecting anomalies in energy usage
across the network, which can point to network disruptions or faults. The role of AI in this specific case can help
CSPs to detect and identify network faults, facilitate remote troubleshooting and avoid unnecessary on-site
visits, which minimise the impact that a fleet has on CO2 emissions.

5. Recommendations and conclusion

Focusing on building and running energy efficient networks is becoming a strategic imperative for all CSPs.
This requirement is driven by increasing energy prices, as well as by regulations that mandate the reduction of
energy usage and carbon emissions. We provide the following recommendations to CSPs that want to accelerate
their progress in transforming their existing networks and operations into energy-efficient systems.

• CSPs should actively introduce energy-saving measures that go beyond just the implementation of
more-modern networks. It is essential to deploy 5G in a way that mitigates the energy impact of increased
numbers of cells and higher spectrum bands, it is even more important to adopt energy efficiency measures
that work across the whole network, including legacy infrastructure such as 2G and 3G. In many cases, the
older networks are the most energy-inefficient so a holistic approach is important to maximise savings.

• CSPs should consider AI-based software solutions as a way of implementing energy-saving measures
as soon as possible. CSPs cannot afford to postpone dealing with the issue of rising energy costs; they need
to take advantage of technologies that can be implemented in a short time-scale and that can deliver results
quickly. It is important that CSPs choose suitable technology partners that can provide highly optimised
solutions that:

— support a wide range of vendor equipment

— optimise both active and passive equipment
— do not compromise the quality of the service at the expense of energy savings
— can be consumed via SaaS so that CSPs can benefit from faster engagements with vendors and quickly
achieve the targeted results
— can deliver a sizeable outcome to CSPs in a timely manner.

© Analysys Mason Limited 2023 5: Recommendations and conclusion

 Controlling energy use: the role of AI-based solutions | 16

• CSPs should plan to adopt a variety of energy management measures that add up to a fully holistic
approach. CSPs need to have a comprehensive view of their systems when considering the overall
potential energy savings within their networks. Their approach must go beyond energy consumption in the
mobile network, and should extend to other assets, including fixed networks, or to more radical
architectural and network transformations to make savings, as well as alternative ways of sourcing energy.
Adding AI capabilities on top of existing or new solutions can help CSPs to develop a holistic approach for
optimising energy consumption.

• CSPs should assess regulatory and commercial environment to establish business priorities and the
best timing to implement different energy saving measures. CSPs should consider the impact of energy
price rises and increased usage on their own business, for example by analysing energy waste in different
elements of their network and by understanding where major gains can be achieved, in order to investigate
and consider the suitability of a potential solution for the specific business goals they are targeting, in terms
of energy savings and implementation time.

© Analysys Mason Limited 2023 5: Recommendations and conclusion

 Controlling energy use: the role of AI-based solutions | 17

6. About the authors

Caroline Gabriel (Research Director) leads Analysys Mason’s Networks research practice, as
well as leading many 5G-related research activities across multiple programmes including
Next-Generation Wireless Networks and Transport Network Strategies. She is responsible for
building and running Analysys Mason’s unique research base of mobile and converged
operators worldwide. She works directly with Analysys Mason’s research clients to advise
them on wireless network trends and market developments. She has been engaged in
technology analysis, research and consulting for 30 years, and has focused entirely on mobile and wireless since
2002.

Michela Venturelli (Senior Analyst) is a member of the Networks research team in London,
and contributes to various research programmes, focusing on Operator Investment Strategies
and Wireless Infrastructure Strategies. Michela holds a PhD and an MSc in Physics from
University College London and Statale di Milano university, respectively.

Grace Langham (Research Analyst) is based in the London office and is part of the Networks
research practice. She holds an MChem degree in chemistry from the University of York.
Grace's final year research project was based on pharmaceutical drug discovery, and she also
worked as an analytical chemist.

Analysys Mason Limited. Registered in England and Wales with company number 05177472. Registered office: North West Wing Bush House, Aldwych,
London, England, WC2B 4PJ.
We have used reasonable care and skill to prepare this publication and are not responsible for any errors or omissions, or for the results obtained from the use of
this publication. The opinions expressed are those of the authors only. All information is provided “as is”, with no guarantee of completeness or accuracy, and
without warranty of any kind, express or implied, including, but not limited to warranties of performance, merchantability and fitness for a particular purpose. In
no event will we be liable to you or any third party for any decision made or action taken in reliance on the information, including but not limited to investment
decisions, or for any loss (including consequential, special or similar losses), even if advised of the possibility of such losses.
We reserve the rights to all intellectual property in this publication. This publication, or any part of it, may not be reproduced, redistributed or republished
without our prior written consent, nor may any reference be made to Analysys Mason in a regulatory statement or prospectus on the basis of this publication
without our prior written consent.
© Analysys Mason Limited and/or its group companies 2023.

© Analysys Mason Limited 2023 About the authors