In network slicing, each slice can be tailored to support the use case to be served with distinctive 5G key performance indicators (KPIs), such as for latency, data rates, error rates, minimum resource allocation, etc., (Series, 2017; Zhang, 2019). The key technology enablers for network slicing are software-defined networking (SDN), and network function virtualization (NFV), which provide the lifecycle management of network slices by dynamically instantiating, modifying, and terminating the slices as per the end-user requirements (Afolabi et al., 2018). Network slicing is integrated with multi-access edge computing (MEC) to combine the benefits offered by SDN, NFV, and service function chaining (SFC) (Mach and Becvar, 2017). This integration helps to overcome the static resource allocation issue and convert it to dynamic resource allocation while still satisfying the network performance requirements for the slice users. The key benefits offered by this integration are dynamicity and efficient use of resources (Filali et al., 2020). This technique of capacity partitioning helps reduce the overall capital expenditure (CAPEX) and operational expenditure (OPEX) while leveraging the benefits of dynamic resource sharing and allocation (Foukas et al., 2017; Shen et al., 2020).

The working definition of an NHN is “a self-contained cellular network deployed by a service provider that builds and operates an integrated technology platform that is solely for sharing purposes” (Badmus et al., 2019). This helps to reduce the duplication of resources while providing services in the same area (Matinmikko-Blue and Latva-aho, 2017). A NHN approach allows a single physical infrastructure to be built for multiple operators acting as tenants and could use shared spectrum bands for its operation. A survey of ongoing research on neutral hosts, especially using 5G suggests that an NHN approach can enhance capacity and coverage, especially in dense small cell deployments, with the right policies in place that encourage incumbent operators to participate (Walia et al., 2017; Maeng et al., 2020; Lähteenmäki, 2021). The potential tenants could be mobile network operators (MNOs) (Matinmikko et al., 2017; Oladejo and Falowo, 2017; Colman-Meixner et al., 2019), Internet service providers (ISPs) (Pries et al., 2016; Frank et al., 2022), communication providers (CP) (Cavalcante et al., 2021), hospitals, and other private networks (Giambene et al., 2019; Zhang et al., 2019). In a NHN model, each slice tenant has an end-to-end 5G virtual network with all components of a typical wireless network (Zhang et al., 2017; Kaloxylos, 2018). Indeed, many researchers have examined the challenges associated with 5G network slicing. For example, this includes evaluating different business models, deployment options, and techno-economic feasibility levels, with an approach based on an NHN using shared spectrum, being the most cost-efficient option (Ramasetty and Masilamani, 2019; Quadri et al., 2020; Bajracharya et al., 2022). As network slicing and a NHN approach encourage an open network with dynamic resource allocation, this idea is being expanded to 6G where network slicing is one of its key enablers (Cao et al., 2022).

Infrastructure sharing may also include spectrum sharing among operators, hence necessitating the need to share both underlying infrastructure costs and spectrum licensing costs for a site across the potential tenants (Meddour et al., 2011). As national operators upgrade the network, the infrastructure will use their nationally licensed or locally shared spectrum bands for operations (Matinmikko-Blue et al., 2018, 2019). The ongoing research on NHNs helps to understand the various aspects of 5G network sharing strategies, especially in terms of technology, spectrum, security, policies, regulations, and techno-economic feasibility (Khodashenas et al., 2016; Tseliou et al., 2019; Wang et al., 2019).

Rural 5G networks can be used for a variety of purposes. For example, eMBB applications that require high bandwidth and data rates (video/voice calling, remote video monitoring, remote health care, wide-area industrial automation, video streaming, e-governance, e-commerce, and online learning). Moreover, uRLLC applications that require very low latency and low data volume (disaster management and response, control of critical services, and machine-to-machine communication). Additionally, mMTC applications that require low power and data volume (sensors and IoT devices with limitless data transmission). These applications enable the deployment of 5G in rural areas, potentially generating multiple revenue streams for infrastructure providers (InP) and ensuring network viability.

To viably deliver affordable rural connectivity there are several challenges that need to be overcome. Many issues are technological, but there are also challenges pertaining to adoption, revenue generation, and business models (Noll et al., 2018; Yusuf et al., 2021). For example, to address the digital divide, many technologies have been trialed and tested to assess their suitability, including long range Wi-Fi, 4G, unmanned aerial vehicles (UAVs), satellite systems, balloons, and TV whitespace (TVWS) (Osoro et al., 2023). However, the common shortcomings of these technologies include network performance during changing operating conditions, scalability, user roaming, high-speed data rates, reliability, data security, latency, and other performance criteria for different telecommunication operators (Fourati et al., 2022; Kumar et al., 2022; Randell-Moon and Hynes, 2022).

Moreover, to address the issue of rural connectivity, researchers have begun testing the suitability of 5G network designs to address the needs of rural use cases. Indeed, a rural 5G pilot trial that uses IEEE P2061 5G standards develops an architecture designed to support low-cost rural broadband communication using a 5G NHN without network slicing and Wi-Fi (Khaturia et al., 2020). The proposed solution utilized macro-cells to support 5G access technologies, with backhaul connectivity relying on a TVWS link and the last-mile using wireless local area networks (WLAN) (Khaturia et al., 2020). In another report by GSMA, the authors encourage sharing to reduce the digital divide using a local service provider in rural areas (Handforth, 2019). The authors find the key to minimizing the coverage gap and connecting these locations at a reasonable cost is to lower network deployment and operational costs and find newer business models. It is also important to account for risks associated with the adoption of 5G, as this has a large impact on commercial viability. The use of spectrum sharing in local 5G networks deployed by the InP is explored in Perez Guirao et al. (2017), Anderson et al. (2020). These studies highlight the benefits and ease of spectrum sharing in 5G bands, especially at higher frequencies. In another pilot trial, the 3.5 GHz band using the GSM and TV broadcast tower sites in rural areas was used for long-range coverage to provide broadband services (Lun et al., 2019). The network was able to achieve good downlink data rates even near the cell edge, however, the uplink performance was inadequate.

Many studies were carried out to investigate the costs of deploying and operating a nationwide 5G network with a duration of 10 years in different countries (Oughton and Frias, 2018; Oughton et al., 2019a,b). The conclusion was that encouraging the correct policies, technology choices, and innovations in rural telecommunication business models will drive infrastructure deployment. Ultimately, MNOs require a long time to install near-universal coverage using conventional deployment strategies since, in remote or rural locations, providing telecommunications services is a costly option with a low or negative return on investment. As a result, the research in this paper examines 5G infrastructure sharing strategies as an alternative solution for providing broadband connectivity in rural areas.

Figure 1 shows the possible upgrade sharing strategies for existing cellular sites to 5G involving many options, ranging from No Sharing to either Passive Sharing, Active Sharing, or a NHN.

Figure 1A shows the network architecture for a No Sharing strategy. In this approach, the incumbent MNO has full control over the network and its equipment from end-to-end. There is no competition over the QoS provided, as typically, this type of strategy has only one operator in a rural area. It will be expensive for another operator to deploy their infrastructure, especially in places with negative or poor ROI (Jeanjean, 2022).

Figure 1B shows the network architecture for a Passive Sharing strategy which involves sharing of backhaul, telecommunication sites, ducts, masts, towers, equipment rooms, and related power supplies, air conditioning, and security systems. The operators using this strategy would have to work toward the goal of reducing the overall cost and agree upon a common cell plan management and upgrade. The key challenge with this strategy is finding operators with similar goals in terms of deployment locations, desired site construction materials, tower heights, network protections, and backhaul capacity requirements (Oughton and Frias, 2018; Jeanjean, 2022).

Figure 1C shows the network architecture for an Active Sharing strategy. This business model involves sharing radios, base stations, backhaul, telecommunication sites, ducts, masts, towers, equipment rooms and related power supplies, air conditioning, and security systems. The spectrum bands are not shared, therefore, each operator uses their licensed bands. This method is preferred by operators who have long-term contractual agreements with each other, along with clearly defined agreements regarding operational conditions. The crucial factors affecting deployment include trust among competitors and the policies laid out by the national telecommunication regulator. The challenges with this strategy include making this a long-term commitment, network complexity, and the fact that each individual operator must relinquish their own independent decision-making e.g., for network upgrades. Similar pricing plans could act as a threat to disrupt operator cooperation (Frisanco et al., 2008).

Figure 1D shows the network architecture for a NHN strategy which involves the sharing of spectrum, core networks, radios, base stations, backhaul, telecommunication sites, ducts, masts, towers, equipment rooms, power supplies, air conditioning, and security systems. This method involves end-to-end network sharing (at all passive and active levels, including spectrum) among the slice tenants (Samdanis et al., 2016; Gomes et al., 2021; Jeanjean, 2022). Unlike the previously articulated sharing strategies, the potential operators would have a network agreement only with the 5G NHN infrastructure operator. This strategy also allows other potential slice tenants, and the operators, to co-exist on the network (Fernandez-Fernandez et al., 2021; Lappalainen and Rosenberg, 2022). In the final NHN strategy, all MNOs would lease slices from the incumbent and ideally provide services at all sites. The key challenges of this strategy are similar pricing plans and market strategies, a decline in infrastructure-based competition, management of dynamic resource allocation, and the security of data on each slice (Paglierani et al., 2020; Pápai et al., 2022). An important aspect of a successful upgrade to a 5G network using a NHN model is cooperation among the slice tenants and their corresponding resource allocation schemes (Sanguanpuak et al., 2019; Tran and Le, 2020). These barriers and obstacles to adopting the NHN strategy should be explored as future work to stimulate discussions among stakeholders. The widespread usage of this technology lies in the slicing capabilities offered by the network, and associated security aspects (Raza et al., 2019; Psyrris et al., 2021; Sciancalepore et al., 2022).

A preferable first option for operators may be No Sharing, as it would enable absolute control over network capacity resources (GSMA, 2019b; Jeanjean, 2022). However, this is not always economically viable because of the required investment, existing debt, and potential revenue. Indeed, operators in many markets worldwide have been experiencing static or declining profits while also being saddled with sizeable existing debt payments (van Kranenburg and Hagedoorn, 2008; Veligura et al., 2020; Sahoo and Sahoo, 2022). Thus, there has been a need for MNOs to seek newer 5G revenue streams, as explored in many research papers (Oughton and Russell, 2020; Bajracharya et al., 2022). Therefore, operators may choose to explore other sharing strategies (Frisanco et al., 2008; Oughton et al., 2022b).

With a weak economic outlook for MNOs but also the need to invest in new infrastructure, the willingness for operators to share assets and spectrum bands is increasing (Matinmikko et al., 2017; Oproiu et al., 2018; Colman-Meixner et al., 2019). The MNO business models utilizing 5G network sharing strategies have been found to be cost-efficient in different deployment scenarios (Atherley, 2020; Psyrris et al., 2021; Kenechi and Stefano, 2022). In the NHN case, the approach supports MNOs, private networks, ISPs, and other potential tenants to co-exist without interfering with each other's operations. Studies have shown that horizontal slices support use cases while vertical slices support multi-tenancy (Kaloxylos, 2018; Lee et al., 2019; Shruthi et al., 2021).

In recent years, techno-economic assessment (TEA) frameworks have been trying to include additional simulation parameters which better match real-world deployment conditions (Bouras et al., 2020; Oughton and Russell, 2020; Frank et al., 2022; Schneir et al., 2023). Table 1 summarizes the different TEA models and their adopted methods, parameters, and the key findings of those studies. Consumer and government pressure to provide enhanced telecommunication infrastructure, with higher data throughput per user and better overall QoS, has been encouraging operators to upgrade their networks and expand coverage (Oughton et al., 2021; Pryce, 2022). Many of the studies presented in Table 1 focus on urban deployment scenarios or specific vertical use cases. Hence, there is a need to define a generic theoretical framework of assessment to enable the techno-economic feasibility evaluation of infrastructure sharing strategies and business model options.

Typically, each rural location has a unique set of network feasibility conditions. These depend upon a range of factors, including population density, per capita income, the adoption rate, local business composition, fiber backhaul availability, and existing competition among operators (Walia et al., 2017; Kumar et al., 2022; UN, 2022). As a result, in this study, we explore the suitability and the techno-economic viability of different rural network sharing strategies. We also examine how the input parameters of the developed model affect the feasibility of 5G infrastructure sharing.

The capacity assessment helps to estimate the current level of available data traffic that existing assets are capable of transporting (Oughton et al., 2022b). Initially, incumbent operators need to assess the sites that require upgrading and their parameter requirements, such as spectrum, bandwidth, latency, 5G KPIs, network congestion during busy hours, and throughput (Ioannou et al., 2020; Oughton and Lehr, 2022). The number of site upgrades necessary can be estimated based on the number of potential future subscribers of the network, the number of concurrent users, and other possible slice tenants' applications (Duan et al., 2013; Oughton et al., 2022b). The number of sites that would require upgrading varies depending on the sharing strategies, demand assessment, and combined area of coverage.

In reality, the incumbent network operator would conduct a survey in the region of interest and list the location of each telecommunication site, its existing backhaul capacity, operating frequency bands (licensed and unlicensed bands), latency, bandwidth, throughput, data rates, busy hour traffic capacity, the population that it serves, the coverage area, the technologies supported, user plane and data plane management, servers, and other network performance indicators. With this information, it is possible the incumbent operator can analyze existing assets in detail such as cells that show high traffic congestion and hotspots where the demand is very high (Zulfadli, 2022). It is important to maximize the use of existing infrastructure during rural network upgrades to keep costs down (GSMA, 2019b; Frank et al., 2022; Kenechi and Stefano, 2022).

Let A km² be the area of study region for the network upgrade. Assume, that there are N incumbent operators, each having x_{mc, i} macro cells and x_{sc, i} small cells, such that i ϵ N in the region of interest.

The overall site density in the region of interest, ρ_site, is given as

The average coverage area per site, β km² is estimated as follows:

The theoretical data throughput for a 5G site is calculated using the equation given below (Lim, 2020):

where, PRB is the physical resource blocks (PRBs), J is the sum of 5G carriers in carrier aggregation, ν^(j) is the number of layers that a gNodeB transmitter streams to a piece of user equipment (UE), Qmi is the modulation order (shown in Table 2), f^j is the scaling factor, and R_max is a number equal to 9481024. Finally, NPRBBW(j),μ is the resource block allocation that is determined by the sub-carriers depending upon μ numerology and BW in bandwidth, T_s is the symbol time, and O_h is overhead.

Next, to understand the practical throughput implications of multiple 5G sites in close proximity, we need to utilize a 5G new radio (NR) link budget. Via stochastic geometry, spectral efficiency values can be estimated producing a distribution of capacity among different UEs at varying distances from a gNodeB (Lim, 2020; Jang, 2022). The NR link budget estimation considers a standard deviation of 6 dB for a rural scenario and different propagation models for rural areas (Lim, 2020; Oughton and Jha, 2021), in line with the literature (Oughton, 2020a,b). The analysis also considers interference from nearby base stations. The NR link budget per UE (in dBm), shown in Equation (4), is described in the 3GPP 38.901 standard (Lim, 2020). The link budget depends on a range of factors including climatic conditions, foliage, building clutter, distance between the transmitter and the receiver, indoor or outdoor location of the receiver, atmospheric conditions, and frequency of operations.

The 5G pathloss equations (PL_{propagationModel}) for the rural macro cell scenario as defined in 3GPP 38.901 standards for a line of sight (LOS), PL_RMALOS, is as given below (Ghosh et al., 2019; Jang, 2022; Lin, 2022):

where c is the speed of light, d_2D is the ground distance between BS and UE, h_BS and h_UT are the height of the base station and UE, respectively, and f_c is the center frequency in Hz. For PL₁ has a shadow fading of, σ_SF = 4, h_BS = 35m, h_UT = 1.5m, while for PL₁ has a shadow fading of, σ_SF = 6. These formulas are valid for 10m ≤ h_BS ≤ 150m and 1m ≤ h_UT ≤ 10m.

The signal-to-noise ratio (SINR), γ = 10LinkBudgetRx,dBm, values are used in Equation (7) to calculate the capacity and spectral efficiency per user. The actual channel capacity per site, C bits/sec of the existing infrastructure, is estimated using bandwidth B, channel utilization χ, SINR γ, and spectral efficiency, μ (Capozzi et al., 2013). Generally, realistic channel capacity C, is lower than theoretical channel capacity C_5G (Abozariba et al., 2019; Lin, 2022).

Network modeling focuses on capturing congested peak demand periods, with the network consequently able to handle traffic during less congested periods. Moreover, traffic estimation helps to understand network traffic demanded by 5G users and the ability to meet user requirements. Additionally, the network should be flexible enough to accommodate future growth in 5G demand. To understand the demand per site, 5G NR traffic modeling as well as scheduling is utilized (Benoist, 2018; Zainal, 2022).

As the number of active users on the network grows, there is a need to increase the number of gNodeB assets deployed to meet the minimum user data rate requirements (Comşa et al., 2018; Nor et al., 2022). The number of sites that are required to be upgraded is estimated using link budget analysis, traffic management, and user scheduling (Oughton et al., 2023). The incumbent operator would select the outcome which provides the maximum number of towers for the upgrade, to account for future demand from end-users and their applications (Lee et al., 2014; Amer and Puttaswamy, 2019).

Data traffic demand is estimated by determining market share, anticipated smartphone users or other business subscribers, population distribution, active users exchanging traffic at peak times, the amount of traffic per user, and then the amount of traffic handled per site (Sciancalepore et al., 2017). Here, we follow the demand method applied in the digital infrastructure costing estimator (DICE) (Oughton et al., 2023). This model was developed for the analysis of universal broadband studies with the aim of quantitatively modeling the various factors described above to estimate traffic demand, which is broadly commensurate with other modeling approaches commonly found in the literature (de la Torre et al., 2020; Oughton et al., 2022b; Oughton, 2023).

Rural areas tend to have a small number of settlements, although there are a few outliers (Yaacoub and Alouini, 2020). The demand estimation also includes business subscriber data and throughput requirements for potential end-user applications, including Internet of Things (IoT) devices or other technologies for health, energy, transportation, etc. (Musacchio et al., 2006). Another major unknown parameter that affects network feasibility is the expected average revenue per user (ARPU). In theory, if an operator expects the existing ARPU to increase following the deployment of new services, then there would be a higher appetite to invest, for example, in upgrading to 5G services (Kenechi and Stefano, 2022). This situation, however, has considerable uncertainty, which requires scenario analysis (Oughton et al., 2022b). Finally, compared to consumers, business subscribers are typically expected to pay higher subscription rates (which may translate into a more reliable service, and revenue stream) (Lappalainen and Rosenberg, 2022).

In this step, the incumbent operator would estimate the potential 5G subscribers and their use cases. There would be a survey/discussion with the potential slice tenants about their application requirements that the network would need to satisfy. The incumbent operator would tabulate the demand assessment model's outputs and estimate the ARPU that end-users would be willing to pay for their services. The end-users could be business-to-business (B2B) or business-to-consumer (B2C) (Psyrris et al., 2021; Schneir et al., 2022). The number of small and macro cells that require an upgrade is dependent on this analysis.

To estimate the traffic demand that should be supported by the network over a period of T years (say, T is the study period), there is a need to include the data obtained from the demand assessment model. Let the expected average user traffic be given as δ_t GB/user/month, such that, t ϵ T. Then, the data consumed per day per user, δ_{t, day} MB/day (Oughton and Frias, 2018; Oughton and Jha, 2021). The minimum data speed required per user ζ in Mbps, during the busiest hour of the day (B_HF) using the conversion value of 1 Byte (B) with 8 bits (b), and 1 h with 3,600 s (Oughton et al., 2022b), is calculated as:

Then the population density, ρ_pop for the study area A with population P. Typically, x% of the population density, ρ_pop for the study area A of the P, would be the number of subscribers for a service.

Finally, the area traffic ι_area is estimated as (Oughton and Russell, 2020; Oughton, 2023),

This assessment includes the estimation of the cost incurred in deploying and operating the different business model options. The expenditure for a particular 5G upgrade is calculated first per site and then aggregated to a local statistical area level. A rational incumbent MNO designs and deploys a forward-looking network that accounts for future traffic demand over the next 10–20 years. The discounted total cost of ownership (TCO) ω is estimated as the sum of CAPEX ω_c and OPEX ω_o over this time horizon (Chiaraviglio et al., 2017; Ioannou et al., 2020; Oughton and Lehr, 2022). CAPEX includes the cost of the radio equipment upgrade ω_RAN, the backhaul upgrade (wired as well as wireless) and any labor ω_b, a small edge cloud site ω_edge, spectrum ω_s, and any core network upgrades necessary ω_core. OPEX includes the cost of power, administrative operations, core network maintenance, routine maintenance of radio equipment, operational spectrum, and edge cloud maintenance.

Generally, there is a need to upgrade the backhaul capacity to support 5G data rates. Additionally, unlike other studies which exclude current asset debts, ω_d, this assessment also includes a nominal existing debt payment per site which is closer to what is experienced in reality (Cheng et al., 2003). The debt payment factor ω_d, is not included in OPEX because it lacks the traits required for ongoing routine operations and maintenance. As a result, they are included in CAPEX, which is in accordance with the discussion with operators related to the inclusion of debt in their estimates. By adopting this parameter, analysts can more accurately reflect the level of debt owed to each operator and its impact on brownfield telecommunications deployment. Furthermore, as per standard industry practice, the backhaul cost is split between CAPEX and ongoing OPEX (Oughton et al., 2022a), while the existing debt payment is factored into CAPEX (Brach, 2016). Therefore, the modified TCO for this study is (Yaghoubi et al., 2018; Chiha et al., 2020; Oughton and Lehr, 2022).

The 5G network upgrade assessment estimates the infrastructure requirements for future assets. This model includes details about site locations, additional backhaul capacity, macro and small cell quantities, future spectrum bandwidth, expected spectral efficiency, usage of the network traffic, and slice requirements of various potential tenants. The tenants may find it desirable to obtain the resources they lease on a near-real-time basis (Sanguanpuak et al., 2019; Jeanjean, 2022). In addition, the incumbent may also need to account for upgrades to support potential future tenants.

Table 3 shows the number of physical components for upgrade per cellular site for each sharing strategy in a 4-operator scenario. The total number of sites required for the upgrade is subject to the existing coverage areas and network sharing strategy. From Table 3, it can be observed that in a No Sharing (baseline) deployment, most sites and base stations of the incumbent would need to be upgraded and no physical sites are shared. In a Passive Sharing deployment, the resources required for upgrading are reduced compared to the baseline scenario as the physical site locations and other passive components are shared among all operators, whereas the radios, spectrum, hardware, and core are not shared. Furthermore, in an Active Sharing approach, the operators deploy a lower number of radios and hardware components compared to Passive Sharing. Finally, in the NHN deployment, the total resources for a 5G network upgrade reduce further as end-to-end components are shared by all operators.

Let κ be the number of towers that require a network upgrade and ω be the TCO of the network upgrade. The key optimization equation to lower the TCO associated with 5G brownfield deployments in rural areas, while satisfying the aims to increase coverage (β) and data rates, (C) but minimizing the number of towers that require an upgrade, is stated as follows (Duan et al., 2013; Shruthi et al., 2021):

This framework performs a feasibility analysis of the possible infrastructure sharing strategies for each rural 5G business model. The results of the network upgrade requirements, specifically, the number of necessary upgrades needing to be made, are then fed forward to be combined with the potential costs of each component from the cost model, thus producing the key assessment result metrics.

The revenue per year R_i such that i ϵ T, and the total revenue, R, over the period T for ARPU Ψ_5G for 5G services are calculated as (Ioannou et al., 2020; Oughton et al., 2021):

where Ψ_old shows the ARPU for existing infrastructure, ρ_new5G are the additional new subscribers joining the network who require 5G KPIs for their applications, and ρ_{up, sub} is the existing subscribers who upgrade their services to 5G technology who can be charged more than existing technologies. The cash flow for year i, α_i such that i ϵ T, is estimated as (Besanko and Braeutigam, 2020):

where ω_i is the cost per year toward the upgrade. The incumbent operator would upgrade the network sequentially to match the network demand and earn higher revenues. For analyzing the profitability of the network upgrade to 5G using different sharing strategies, the net present value (NPV), Υ, method is used with a discount factor r and is calculated as (Ye and Tiong, 2000; Besanko and Braeutigam, 2020):

when Υ = 0, then it is a no profit-no loss scenario for the infrastructure provider. This helps in the estimation of the minimum ARPU, Ψ_min, at which the infrastructure provider would consider deploying the 5G network. The minimum ARPU would in turn help in determining the pricing range which the operators should charge to the end-users (Shruthi et al., 2021).

As incumbent operators shift their 5G business models to rural areas, they are driven both by a desire to reduce overall TCO by maximizing current and future resources and to sell new vertical services to increase revenue (Partners, 2019; Gómez et al., 2022; Pryce, 2022). Hence, the appraisal outputs focus on population coverage in rural areas: the minimum provided speed per user in busy hours, the percentage of subscribers with 30+ Mbps peak speed, an NPV feasibility analysis, and a sensitivity analysis for uncertainties. The various costs for network upgrades for an area with an existing 2G/3G network tend to be higher than upgrades from existing 4G networks. For example, existing 4G hardware can support future infrastructure upgrades with relatively minimal software updates, lowering the upgrade cost significantly (Oughton et al., 2021; Kenechi and Stefano, 2022; Lappalainen and Rosenberg, 2022).

Consider a generic rural study area (A) of 500 km² with interspersed low population density villages (<300 people per km²) for the time period of 2023–2032. In this study, the population density for the base scenario is 36 people per km². Assume there are four national MNOs, and each wants to be omnipresent within their operating country (Saha, 2020). Table 4 provides a summary of the simulation parameters and the inputs for different models. The data for the study are obtained from various sources from the literature (Oughton and Frias, 2018; Grijpink et al., 2020; Ofcom, 2020; Oughton and Jha, 2021; 5G-New-Thinking, 2022; ITU, 2022; Lappalainen and Rosenberg, 2022). Typically for rural areas, the operating frequencies are in the range of 700 MHz and 3,800 MHz (Partners, 2019). Note that a higher coverage is offered by 700 MHz (at lower throughput rates), while higher throughput rates are offered at 3,800 MHz band (but with lower coverage) (Wahyudin et al., 2021; GSA, 2022). For example, during the network upgrade the 3,800 MHz band could be used to serve a small rural hamlet with enhanced mobile broadband services, while the 700 MHz band could be used for greater wide-area coverage across the cell (e.g., for highly mobile users in vehicles). The network could use either time or frequency division duplex (TDD/FDD) as a modulation scheme.

In this evaluation, for the existing technology with a take-up rate (x), 50% of P, consider that 30% of those users would upgrade to 5G services, and an additional 20% of P would join the network for 5G services. Therefore, the rural region of interest in this study, especially sensitivity analysis, is treated as having between 1,000 to 25,000 mobile subscribers along with thousands of devices for private networks and IoT applications (Oughton and Mathur, 2021). These subscribers could belong to local businesses or corporate companies. Today, the ARPU from wireless broadband around the world ranges from $2 to $45 (ITU, 2021). Therefore, the incumbent MNO would expect a higher ARPU from the 5G services and higher data rates along with exponentially improved performance, say +20% compared to the ARPU of the 4G services (Suryanegara, 2018; Yang, 2022). But over time, the price per GB would drop, which would encourage operators to share the network, which would further encourage infrastructure sharing among the operators (Hunukumbure et al., 2022). Next, the annual increase in the subscriber base is considered to be around 4% although the rural population growth is below zero (Bank, 2021), this is because improved wireless broadband will increase the demand for connectivity for use cases such as remote job opportunities, personal communication, e-governance, farming, and healthcare facilities.

In this paper, we follow a long run incremental cost (LRIC) approach where a “hypothetical MNO” is modeled representing an entity with an average market share, average spectrum portfolio, average set of existing sites, etc. In a perfectly competitive market, each of the four MNOs would acquire roughly 25% of the entire 5G services market share, adding up to 100% of the market share. For this study, we consider the busy hour factor (B_HF) to be 0.15, and maintain low latency (<50 ms). An aspirational target for governments globally is to provide a minimum 30 Mbps average data speed per user and to increase the data usage to a minimum of 30 GB/Month per user (UN, 2022). Therefore, in this study, we consider a minimum average data speed of 30 Mbps per user and data consumption of 50 GB/month per user. Table 4 shows the modeling conditions for the study location and its simulation parameters, along with the cost for various components required for the 5G upgrade.

The average cost per site for the macro-cell and small-cell varies considerably. In this study, we estimate that a brownfield macro-cell upgrade is approximately $45,000 including RAN upgrades, core, spectrum ($0.7 per MHz at 700 MHz, and $0.03 per MHz at 3,800 MHz), and other parameters, while the greenfield small-cell deployment is around $12,000 (Osio, 2021; Oughton and Jha, 2021). The variation in the costs depends on the selected strategy and the upgrade required for the existing hardware and software components such as the core, radios, site, gNodeB, and processing units. Next, the cost of an average backhaul upgrade would be approximately $10,000 for macro-cell and greenfield deployment would cost around $5,000 for small-cell. The OPEX for the small cell could be as low as $800 per year, while the same for the macro cell would be around $2,500. The brownfield deployment already has an existing debt payment to be cleared, which is about 5% of the TCO of each site (Bhatia, 2022). Note that the debt payment is for the existing infrastructure (2G, 3G, and 4G), not the new 5G network. The study considers that the CAPEX and backhaul depreciate at a 3% rate, while the OPEX and debt payment appreciate each year at 5 and 2%, respectively (Schneir et al., 2019).

Generally, there would be around 60 active users at any instance in time per site, except during congestion periods when the capacity can be expanded using virtual and cloud resources via network slicing (Chiha et al., 2020). As the number of active users grows, a corresponding rise in the number of demand requests for scheduling users takes place. Furthermore, as demand rises, SINR values decrease because resources are shared among many users (Benoist, 2018; Zulfadli, 2022).

Figure 3 reports the stochastic geometry analysis of the SINR, spectral efficiency, and channel capacity estimate of the realistic channel throughput at the cell edge for a 95% confidence interval, for UEs operating on a 5G cell site. Indeed, there is a clear relationship between increasing UE distance from the cell site, and a decreasing SINR leading to lower spectral efficiency and thus poorer channel capacity. For the 5G network upgrade, both small- and macro-cell strategies are considered. We treat:

The minimum theoretical peak cell throughput is around 177 Mbps and 1,700 Mbps at 700 MHz and 3,800 MHz frequency bands respectively, with power levels below 4 W (Biradar and Hallur, 2022; Vinogradov, 2022; Shruthi, 2023). Small cells offer higher peak data rates than macro cells but require higher quantities to provide services in the same geographic area (Wang et al., 2014). Moreover, the densification of the network by the deployment of small cells at millimeter-wave frequencies depends on the number of subscribers and the potential ROI (Wahyudin et al., 2021).

The stochastic and traffic modeling results for the base scenario of 36 people per km² suggest that each MNO would aim to upgrade an average of 4 macro-cells, and additionally deploy 8 small-cells, to provide maximum coverage at 30 Mbps in the brownfield rural area 5G deployment. Furthermore, the overall increase in data rates is evident only when the backhaul capacity is above 5 Gbps, specifically to support users and applications that demand more resources.

In this section, we explore the different infrastructure sharing strategies and their techno-economic implications. The TCO is estimated for all the 5G network upgrade sharing strategies and shows that OPEX becomes higher over time due to increased network complexity, breakdowns, repairs, and inflation. MNOs are likely to prefer different sharing strategies depending on existing demand and resource utilization. In order to meet 5G security and dynamic slicing requirements, an NHN has greater equipment requirements that must be met because multiple network operators are using the network infrastructure simultaneously (Raza et al., 2019; Sciancalepore et al., 2022). Hence, the single site cost for upgrading to the NHN strategy is the most expensive. In reality, an upgraded 5G network in rural areas helps to provide eMBB-related applications while supporting other 5G rural applications relating to vertical sectors, such as health and transportation.

Figure 4 shows the overall upgrading cost for all the sites for each year in the period of 2023-2032. The estimated costs shows that the TCO for No Sharing (baseline) is approximately $1,996,791, for Passive Sharing $1,459,224, for Active Sharing $994,446 and for NHN $659,864. Figure 4 and Table 3 show that for an incumbent MNO, the cost of upgrading to a rural NHN per site is higher compared to the incumbent MNO's upgrade to 5G per site, by 6–20% against other 5G network sharing strategies.

Moreover, Figure 4 presents the financial cost savings possible from 5G infrastructure sharing strategies. Passive Sharing strategies exhibit substantial savings between 10–20% for 50 GB/Month against the baseline. Meanwhile, the Active Sharing strategy results in savings between 20–35% for 50 GB/Month against the baseline. Lastly, a rural NHN provides impressive cost savings of around 35–50% against the baseline scenario.

Additionally, Figure 4 shows that for each network sharing strategy, the cost per year increases due to various factors, including inflation, the loan interest rate, and operating costs. Indeed, the cost increases by 7.6, 6, 5.6, and 5.5% for No Sharing, Passive Sharing, Active Sharing, and NHN, respectively. Also, Figure 4 shows that in the four sharing strategies, the CAPEX to OPEX ratio is around 1.9 in the first year and falls to almost 0.95 in the final year of assessment.

Figure 5 shows the minimum investment required per user toward the 5G network for a sustainable business case. It can be observed that as the number of users increases, the investment required per user decreases. For subscribers below 500, only Active Sharing and NHN strategies are profitable for the operators with per user investment less than $60, which is very high for a monthly ARPU (Taylor, 2023). As the number of subscribers increases, the minimum investment required per user falls below $10, increasing viability. When the number of subscribers is above 10,000 in the region of interest, then the minimum investment required per user is below $1.5 per subscriber. In the present scenario, the deployment is self-sustaining since the network is profitable for the operator while being affordable for end users.

Figure 6 shows the sensitivity of the NPV by varying the revenue from −93% to +100% of the baseline value at a population density of 36 people per km² (=18,000 people in the study area), with a subscriber growth rate of 4% per year. The results show that the increase in subscription demand leads to a commensurate rise in revenue, which overall provides an improvement in the viability of the 5G deployment across the sharing strategies. Also, the estimates in Figure 6 illustrate that the NHN business case is superior by at least 15% compared to other sharing strategies under the same revenue and demand conditions. For a network to be profitable at a low ARPU, say $10 per month, the subscriber base needs to be very high per cellular study area, e.g., above 20,000 subscribers per area. As the ARPU increases, say at $60, the required number of subscribers could reduce to as low as 3,400 subscribers in the NHN strategy for rural areas. The base scenario is calculated with a monthly APRU of $30 per subscriber. Figure 6 shows that all sharing strategies are feasible when the ARPU is higher than $40, that is, viable business models.

These results demonstrate that the techno-economic feasibility in rural areas is extremely sensitive to the number of subscribers and ARPU for the network, along with the number of towers requiring upgrading. The estimates also demonstrate the difference in the ROI for each 5G sharing strategy. Figure 6 shows that at $30, the ROI is negative for No Sharing and Passive Sharing, whereas the ROI is positive for Active Sharing and NHN.

Figure 7 shows the sensitivity analysis for the different 5G network sharing strategies. It can be observed that for a 20% increase in the ARPU, the NPV increases twice in Passive Sharing, 4 times in Active Sharing, and 5 times in NHN compared to the NPV of the baseline scenario (No Sharing). Similarly, when the existing infrastructure increases by 10%, the NPV doubles in Passive Sharing, trebles in Active Sharing and trebles in NHN approach compared to the NPV of the baseline scenario. The network is least sensitive to the debt repayment amount. The NPV hardly changes from the base NPV even when the debt payment parameter changes by 60% for all sharing strategies.

The results obtained in this study are compared against the existing rural connectivity studies in Oughton and Jha (2021), Laitsou et al. (2022), Oughton (2023). According to a recent study, Active Sharing and Passive Sharing strategies for 10–30 GB/month in an African scenario would reduce the cost by 48–78% and 10–44% respectively, while the same strategies in this study for 50 GB/month would reduce the cost by 20–35% and 10–20% for Active Sharing and Passive Sharing, respectively (Oughton, 2023). Meanwhile, when considering developing countries with a case study of the Indian context, the cost per user of 4 × 4 active sharing with 5–50 Mbps QoS is 70% lower compared to that of traditional LTE deployments (Oughton and Jha, 2021). These figures are slightly higher than the reduction in the investment per user of 65% estimated here. Similarly, according to a recent study for European countries, it was shown that greenfield 5G standalone fixed wireless access without infrastructure sharing is expensive in rural areas having a cost per user over $100 per month, whereas wireless broadband using infrastructure sharing could potentially be feasible for less than $60 per month (Laitsou et al., 2022).

Studies in the literature on the successful NHN trials and initiatives have shown that a 5G NHN approach is beneficial to MNOs in places where the population density is very high or low, especially if there are multiple use cases targeted (Lähteenmäki, 2021). In this study, the cost savings offered by using a 5G NHN are approximately 35–50% compared to the baseline No Sharing approach. In a city-wide NHN deployment for an ultra-dense urban area, such as in London, the cost-savings are expected to be around 40% (GSMA, 2018; Schneir et al., 2019). Similarly, cost savings for different deployment scenarios, such as a 100-story commercial building in urban indoor and outdoor settings, were estimated to be 70 and 80%, respectively (Allawi et al., 2022). Whereas, for a university campus a NHN approach is estimated to result in a 20% saving (Walia et al., 2017). Finally, for an industry vertical, such as a seaport at Hamburg, a NHN strategy could result in up to a 50% cost saving (Schneir et al., 2022).