New technologies across cellular networks promise an array of new use cases and business opportunities for communications service providers (CSPs), as well as increased means of monetisation. Chief among these, as networks develop advanced 5G and look toward 6G technologies, is the integration of AI into the network, both to help optimise the network itself and to support AI-powered use cases for CSPs’ customers. However, these AI models are generating rapidly increasing energy costs, as the amount of AI processing and the sophistication of that processing increases. As AI models become more complex and require greater computational power, the infrastructure needed to support them is expanding rapidly, with data centre capacity expected to grow two to six times by 2030. This explosive growth is placing immense pressure on energy consumption and prompting investments in both connectivity and energy access to address these demands.
In response, CSPs and data centre operators want to limit energy-based costs, primarily by reducing energy consumption. There is also a sustainability imperative in many cases, both from a regulatory and enterprise perspective. Regulatory pressure is mounting for companies to reduce emissions, and Scope 2 emissions (indirect greenhouse gases from electricity consumption) are a key part of that mission. With nationalised grid infrastructures, one of the most reliable ways to reduce Scope 2 emissions is by reducing energy consumption. This is becoming a factor in enterprise purchase decisions as well; Kaleido Intelligence’s 2024 Enterprise IoT Connectivity Survey found that 51% of enterprises consider environmental, social and governance (ESG) framework compliance among the top five most important capabilities for a CSP, with the energy and utilities sector placing particular emphasis on this. As such, being able to demonstrate low energy consumption as well as plans to minimise and reduce emissions is becoming an operational imperative and a commercial differentiator.
While 5G, in general, and AI, in particular, are often touted to increase network efficiency and thereby reduce operating costs, these gains will be offset by the increased energy usage of more powerful infrastructure, especially as AI reaches further into cellular networks and the applications they support. As a result, CSPs need to find ways to optimise their network energy consumption and output, beyond the application of 5G technologies. This is a central puzzle for network planners and will become more challenging in the coming years. It is not sufficient to build networks that can handle increasing data and computational loads; it must be balanced against the initiative to minimise energy expenditures.
Energy measurement and metrics
The best way to bring these reductions is granular measurement and analysis of energy consumption metrics across different elements of network infrastructure. This allows for an assessment of where efficiencies can be delivered, both in terms of which network components can be upgraded to reduce opex and where power supply and related infrastructure can be adjusted to meet demand. Additional metrics can help here; for example, if heat dissipation is known, then cooling infrastructure can be more accurately planned. If output power and location can be estimated at the start of a deployment, then excessive coverage can be avoided, saving on infrastructure costs.
However, this is becoming difficult, as multi-vendor environments are increasingly common, most notably the Open RAN movement, which is explicitly intended to allow combinations of equipment from multiple vendors. This can be a problem as multiple vendors often report different performance metrics, making apples-to-apples comparisons for the power consumption and outputs of different equipment difficult. In the absence of a common frame of reference for what metrics are used and how they are referred to, bespoke nomenclatures emerge, which may not fully capture emergent requirements, and have the potential to cause confusion when assets change hands (for example, through M&A activity or transitioning from a managed service to a self-service environment).
Standardised metrics for equipment make this process more transparent for all parties involved, eliminating confusion between different equipment inventories, and ensuring that in-field energy management is handled in a uniform manner that can be processed by everyone. This will help operations both understand normal network function and adjust based on available resources and current levels of consumption; and provide a framework for unified inventory management that can avoid duplication and other inaccuracies that can result from different forms of record-keeping.
With increasing interest in sustainability and reduced energy consumption being expressed by governments and regulators in many countries, demonstrating reduced energy consumption may also become a compliance issue as well as a performance one. Having a standard and independent frame of reference for these measures can help CSPs demonstrate compliance with sustainability directives and demonstrate projections for energy reduction or consumption based on the deployment of new equipment. This will, in turn, make the application of grants tied to power consumption metrics easier to acquire.
You’ve read 40% of the report. To continue reading the report, read the latest instalment of the Agile Telco magazine.
