Energy consumption has become one of the most substantial operational expenses (OPEX) for mobile network operators (MNOs) in today's rapidly evolving telecommunications industry. As 5G networks continue to densify and more advanced technologies are being deployed, radio access network (RAN) energy costs are rising in step. They now account for a growing share of an operator's budget, making energy efficiency a top priority.

Considerations When Taking the Traditional Approach

Traditional energy-saving mechanisms, such as Transmission (TX) Path Shutdown and Carrier Shutdown, have been widely adopted across the industry to reduce energy consumption. TX Path Shutdown deactivates the transmission path of a cell during periods of low traffic, while Carrier Shutdown focuses on shutting down specific carriers within a cell. Both methods rely on static thresholds and predefined rules, which can be quite effective, but may not always be able to adapt to real-time, dynamic traffic conditions.

This consideration can pose challenges in modern 5G networks, where the deployment of frequency range 2 (FR2) small cells, multiple LTE, and 5G carriers and overlapping coverage areas introduces added complexity. As a result, these static approaches may not always be able to optimize resource use, especially during low-traffic hours, when they cannot dynamically adjust to actual traffic loads.

For instance, during low-traffic periods, static thresholds may keep certain carriers or transmission paths active unnecessarily, potentially leading to missed opportunities for cost savings. Conversely, during peak traffic hours, these mechanisms may not activate resources quickly enough to meet demand, which could affect the user experience. These considerations highlight the importance of an adaptive, intelligent approach to optimizing energy consumption in modern networks as traffic patterns grow more dynamic.

The Enhanced Path Forward

The answer lies in moving away from rules and toward dynamic reasoning. Dynamic energy optimization replaces static thresholds with real-time traffic analysis and machine learning algorithms to adapt to varying network conditions and user behavior. Unlike static mechanisms, dynamic approaches continuously learn and optimize energy consumption while maintaining Quality of Service (QoS) for the end user. This adaptability ensures that resources are used efficiently, reducing unnecessary energy consumption during low-traffic periods while meeting demand during peak hours.

The financial case is clear. When a network can automatically power down underused cells or transmission paths during quiet hours — and spin them back up the moment traffic picks up — operators stop paying for capacity they’re not using. At scale, across thousands of sites, those savings add up fast.

Samsung’s Efficiency Solution

To address this need, Samsung developed its AI-powered Energy Saving Manager (AI-ESM), a solution hosted on its RAN Intelligent Controller (RIC) platform. The AI-ESM is an intelligent application that dynamically optimizes RAN behavior by leveraging network data, policies and predictive analytics. This intelligent approach minimizes energy consumption without compromising user experience. In fact, one operator that trialed the AI-ESM reported an average energy cost reduction of 15% across its network, with peak savings of 30% per sector measured during a 5-hour duration when the network is least utilized. When the peak savings demonstrated in the trial are scaled across a network of 50,000 macro-RAN sites, the annual savings can reach as much as USD 13 million.

What sets the AI-ESM apart from traditional tools is that it doesn’t work from a fixed rulebook. Instead, it learns — continuously analyzing historical and real-time data to understand how each site behaves across different times of day, days of the week, and in various traffic scenarios. The AI-ESM also analyzes overlapping coverage boundaries to maintain network continuity during energy saving execution. That knowledge and informed insight feeds directly into its decision-making, so the system is always acting on current conditions rather than yesterday’s assumptions.

The AI-ESM essentially serves as a “round-the-clock” manager of the network, running tasks like:
⦁ AI/ML-Driven Predictive Analytics: By analyzing historical and real-time network data, the AI-ESM identifies various site environments, learns traffic patterns by location and time of day, and evaluates the extent of impact on network performance to find the optimal threshold value. This predictive capability allows for more efficient resource allocation, leading to energy savings.

⦁ Dynamic Cell or TX Path Activation and Deactivation: The AI-ESM sends crucial data to the base station, which informs the activation and deactivation of cells or TX paths based on real-time traffic conditions. Based on this information, the base station automatically switches cells or TX paths off during low-traffic periods to conserve power and turns them back on when traffic increases. This adaptive approach ensures energy is only consumed when necessary, generating cost savings during low-traffic hours.

⦁ Policy-Based Optimization: The AI-ESM incorporates operator-defined policies to ensure that energy-saving measures do not compromise network performance or user experience, tailoring the solution to meet the specific needs of each network.

⦁ Deep-Sleep: In this advanced energy-saving state, the AI-ESM is able to drive significant energy reduction as the Power Amplifier as well as the RUs' analog and digital front ends are turned off.

⦁ Operation Window for Energy Savings: The AI-ESM dynamically determines the ideal hours of the day to transition cells into an energy saving state, while still adhering to predefined, necessary policies.

Looking ahead, Samsung will continually evolve the AI-ESM, with plans to support enhanced savings modes like Super-Sleep — an even more advanced state beyond Deep-Sleep — and slice-based energy savings services, offering operators the flexibility to define specific policies per network slice. This capability will mean that services like broadband or premium user traffic can be prioritized and managed differently, ensuring a tailored approach to energy efficiency while maintaining service quality.

The AI-ESM's ability to adapt across different network configurations — from dense urban deployments to rural macro sites — means it can be tuned to each operator’s specific environment. And because it sits on the RIC platform, it fits naturally into open RAN architectures, making it a practical choice for operators building toward more disaggregated networks. The result is a solution that will pay for itself in reduced energy spend while laying the groundwork for more intelligent, automated network management going forward.

The Future of Energy Savings

Energy costs are not going to decrease on their own, but implementing Samsung’s AI-ESM is a direct step toward realizing improved energy efficiency without impacting network performance. By helping to reduce power consumption and operational costs, the AI-ESM not only benefits operators but also contributes to a more sustainable future for the industry.

With its innovative approach to energy optimization, Samsung is at the forefront of this shift, helping operators turn energy management from a blunt, reactive process into a smarter, leaner, and far more cost-effective one.