Digitalization is one of the most powerful enablers of the global shift toward cleaner, smarter, and more resilient energy systems. Yet today’s electricity grids were not built for renewable energy. They were built for fewer, centralized sources such as coal-fired plants. Renewable energy is also often distributed across regions, and the power output fluctuates over time, making predictability a challenge.
To decarbonize and phase out fossil fuels, we must electrify heavy industries and the transport sector. We also need more electricity generated from clean sources such as renewables and to eliminate power produced from fossil fuels.
This is why digitalizing electrical grids and the whole electricity sector is vital to create control mechanisms and to optimize grid management that will support these greener energy goals. If we also digitalize all end users and increase data sharing between users, distributors and electricity producers, we can create a true digital energy ecosystem.
In many ways, this is what we already do within the telecommunications sector when we predict network traffic to optimize network management. The same digitalization principles are needed for electrical grids, for which data is key. And collecting the right and necessary amount of data from electrical grids will require advanced connectivity and artificial intelligence (AI).
In recent years, however, many have feared that the information and communications technology (ICT) sector will require large amounts of electricity from the total consumption due to the rapid adoption of AI and the related expansion of data centers. It is important to show fact-based numbers on energy consumption, which is why Ericsson has done extensive research on the electricity consumption of the ICT sector, AI, and data centers.
As global leaders meet in Brazil for the latest COP U.N. climate talks, the transition and decarbonization of energy, manufacturing and transport sectors – as well as the role of renewable energy deployment – will be key priorities. Countries will present their Nationally Determined Contributions (NDC), which are climate action plans reaching from 2025 until 2035 to achieve the Paris agenda. These will set the pace for the climate transition for the coming years.
Will digitalization further increase energy consumption?
Understanding the true energy impact of digitalization is essential for informed policymaking and investment. As digitalization accelerates exponentially, the ICT sector, particularly telecommunication networks and data centers, is expected to consume even more electricity. Yet this rise will likely be far smaller than many reports suggest. What remains uncertain is how fast AI services and technologies will evolve.
Looking back, most ICT studies have greatly overestimated future electricity usage because they assumed that larger data volumes would proportionally require more electricity to transmit and process the data. However, Ericsson’s own research reveals that while total data traffic has increased by about 80 times between 2007 and 2023, the ICT sector used only 1.4 times more electricity. This means that data transmission and electricity usage are clearly not directly proportionate.
Ericsson has also calculated the estimated electricity usage and potential carbon footprint for the ICT sector, up until 2030. It is expected that overall electricity consumption will continue to increase to approximately 1,245 terawatt-hours (TWh), from about 1,070 TWh in 2024. In recent years, data centers have been identified as the major contributors to the ICT sector’s growing energy consumption, whereas networks have contributed to this growth less.
How much electricity does AI consume?
AI’s overall energy demand is a frequent concern. Current systems use substantial electricity for training purposes and require increasingly powerful hardware. It remains difficult to predict how AI services and technologies will evolve or how energy efficient they will become.
However, while the energy demand from AI is significant, concerns surrounding exponential increases in usage are mitigated by AI’s advancements toward energy efficiency. Ericsson estimates that up to 12 million AI graphics processing units (GPUs) have been in operation at the end of 2023, consuming approximately 21 TWh. This corresponded to approximately 8 percent of the total energy needed for all data centers, and less than 1 percent of the whole ICT sector. By 2028, it is estimated that AI alone will account for approximately 25 percent of the total data center electricity usage. This highlights the urgent need to prioritize energy efficiency in data centers.
Despite the ongoing and predicted rise, AI hardware supply limitations will prevent an exponential increase in energy usage. Additionally, AI chips and other energy-related equipment are becoming more efficient, while new-generation network system standards are gradually incorporating measures related to AI and energy efficiency. Furthermore, deploying new generations of more energy-efficient networks and server systems will allow companies to phase out older and wasteful technologies.
Finally, we need to assess how reliable AI electricity forecasts are. According to the International Energy Agency (IEA), AI-related chips have doubled their efficiency every three years, and recent models use 99 percent less power to perform the same calculations as a similar chip from 2008. In addition, the IEA predicts that both servers and AI GPUs will advance with more efficient processors entering the market before 2030.
Regarding energy consumption, there is a considerable difference between generative AI—large language models (LLMs) such as ChatGPT and Google Gemini—and dedicated AI services. Generative AI is trained on large data sets, requiring huge amounts of energy, while more dedicated models with a specific purpose only need smaller data sets for training, meaning a smaller energy footprint and, consequently, a more energy-efficient version of AI.
Let’s look at the European energy sector as an example. Utilities need dedicated and energy-lean AI solutions for grid management and control that could run either through edge computing or in local data centers without relying on the large computing power required for generative AI. Similar solutions and machine learning algorithms are used for network control in telecommunication systems, consuming much less electricity than LLMs.
Why 5G Standalone matters
Decisions on digitalization-related investments and future regulatory policies depend on accurate information and assumptions. The ICT sector’s electricity consumption is an aspect that needs to be considered by utility companies and governments alike, which is why reliable predictions are an important basis for decision-making. These choices are crucial to the successful digital transformation of the energy sector.
As mentioned above, data is key to obtaining the information needed to optimize energy systems and provide functionalities such as AI for grid control and management. Energy-related data needs to flow between end users, distributors, and producers to provide, for example, forecasting, grid management, and flexibility solutions.
Connectivity is critical to obtain this data—and this is where 5G Standalone (SA) becomes so important. In non-standalone 5G, the radio components, such as the base stations, use 5G, while the core relies on legacy 4G technologies. Although this leads to higher speeds and larger bandwidth than 4G, the main advantages of 5G cannot be fully realized
In contrast, 5G SA utilizes 5G base stations and a full 5G core to increase speed and bandwidth. This way, it can support advanced services crucial for the energy sector’s digital transformation and the creation of a future-proof digital ecosystem, including but not limited to network slicing, edge computing, 5G voice, RedCap devices, time-critical communication (TCC), network exposure, and enhanced fixed wireless access (FWA).
In this sense, 5G SA is not just a connectivity upgrade but a foundational technology for the global energy transition.
Conclusion
With COP30, many people now realize that energy and AI should be top of mind. In fact, the host country of COP30, Brazil, decided to present its largely renewable-based energy sector as an opportunity for tech giants looking to establish data centers in the country. This underscores the critical role of connectivity and digital solutions in supporting the green transition of utilities, power grids, transport systems, buildings, and industries. Digital infrastructure needs to be prioritized to help phase out fossil-fuel-based energy sources and decarbonize our societies.
Stable and long-term policy frameworks are crucial for enabling these technologies to reach their full potential, providing the certainty necessary for large-scale investment and innovation. Additionally, the technology sector must continue to develop increasingly energy-efficient data centers, AI solutions, and network architectures, as well as uphold their Net Zero commitments across the entire value chain to keep emissions down.
Digital and connectivity solutions must be seen as equal pillars alongside clean technologies, grids, and renewables. Only through connectivity and digitalization can we achieve a decarbonized energy system powered by efficient data centers, responsible electricity consumption of AI, and large-scale renewable deployment. In short, digitalization is not a byproduct of the energy transition but its backbone.
Further reading
Explore the challenges of and solutions for industrial decarbonization
Learn how telecom, data centers, and AI revolutionize ICT energy usage
See the opportunities 5G Fixed Wireless Access presents
Discover Ericsson's vision for a greener 6G future
Discover how 5G core unleashes a new era of scalable and service-rich cloud-native networks.
Explore 5G core solutions that make it possible for you to deploy, scale and evolve your operations across multiple deployment scenarios.
Explore modern network management approaches that give you greater visibility and control across today's increasingly complex networks.
Learn how network slicing enables customized connectivity with assured performance for enterprise, consumer and industry-specific services.
