Data volume is soaring. Here’s how the ICT sector can sustainably handle the surge

One hundred and eighty-one zettabytes. That’s the amount of data expected to be created, captured, copied and consumed globally in 2025 – nearly three times as much as in 2020.

Data volumes have grown substantially with the ongoing digital transformation of the economy, the rapid growth of generative artificial intelligence (AI), increasing demands on mobile data networks, and cryptocurrency mining.

Rising amounts of data require ever higher computational power, which will, in turn, raise electricity demand both in data centres and across communication infrastructure such as telecom and data networks. Electricity systems will need help to keep pace with this expansion, which will also impact greenhouse gas emissions (GHGs). The global geographical hubs of the information and communication technology (ICT) industry will face a particular challenge when it comes to obtaining enough sustainable electricity.

To address these interlocked concerns, the power sector and the ICT industry need to align, balancing the growing demand with the electricity system's capacity to deliver it.

Scaling up processing capacities – for example in data centres (DCs) – requires more electricity. A particular concern is that the additional computational power needed for the rapid growth of AI models and their applications will send electricity demand soaring.

The extra electricity is required not only for additional equipment like servers but also for essential cooling, among other systems, to deal with the immense heat build-up from data processing, according to the International Energy Agency (IEA).

Electricity demand from data centres, cryptocurrencies and AI could reach as much as 1,000 Terawatt Hours (TWh) in 2026, compared to 460TWh today, the IEA projects.

Communication networks that facilitate the transfer of data from the point of generation to the point of processing are also contributing to the DC energy load.

The IEA’s latest statistics show that DCs and communication networks account for 2-3% of global electricity consumption and 1% of GHG emissions. But this is set to rise markedly, even if expansion stays at the lowest estimate.

Over the past few years, processing has increasingly moved from smaller, often proprietary DCs to hyperscale cloud DCs, which are more energy efficient.

However, the expansive growth of Internet of Things (IoT) technologies, along with other high-speed data applications, is reliant on 5G mobile networks. While 5G hardware is inherently more energy-efficient, 5G network expansion could still increase energy consumption by as much as 140% compared to current 4G networks. This is mainly because more cell towers are required, according to the Financial Times.

Another factor is the rising number of edge data centres (EDCs). IoT and technologies requiring high-speed or local data processing utilize edge computing, which processes data closer to the network edge and connected devices. By analyzing data in small, on-site EDCs, turnaround times can be reduced significantly, but they still put further pressure on local electricity grids.

Industry insiders have expressed concerns that clean power generation would struggle to keep pace with such multifaceted growth, with GHG emissions set to rise as a result, the Financial Times reports. This would only add to the sector’s current emissions, which Accenture research suggests have been on an upward trajectory since 2016.

Providing sufficient electricity and managing GHG emissions will be particularly challenging at the world’s key DC hubs. This includes the “DC capital” of the world, Northern Virginia, and many Northern European countries. There are also significant hubs in smaller geographies such as Singapore – which benefits from some of the world’s fastest internet speeds and low-cost electricity – and Ireland, where cool weather, good global connectivity and a welcoming economic and tax regime have spurred DC growth.

Many countries with large DC hubs have already introduced restrictions on new DCs and associated infrastructure to manage the pressure on both electricity grids and national climate targets.

Singapore had a four-year moratorium on new DC builds, which only ended in 2023, and remains cautious about granting new licences in light of both energy and space constraints. The country’s regulator is also working on a roadmap for green DCs in line with Singapore’s net-zero targets.

Ireland’s energy grid also imposed a moratorium on connecting new DCs in Dublin in 2022 – to last until 2028 – amid concerns for national grid capacity and a fear of rolling blackouts. But according to the latest data from the IEA, the island’s DC energy consumption is still set to more than double, with dramatic increases in other geographies expected, too.

DC operators in particular have been driving efficiencies in power usage – not only to lower costs but also to help with environmental compliance. Strong performance improvements have helped to limit growth in energy demand from DCs globally, but given the scale and speed of expansion, more needs to be done to accommodate future data growth and meet emissions targets.

New high-efficiency cooling techniques, along with using AI to optimize DC operations, hold great potential. Google, for example, achieved a 40% reduction in cooling with the help of its Deep Mind AI.

The ongoing replacement of copper with fibre will create energy efficiencies in communication networks, while shifting processing to regions with ample green electricity could help with lowering emissions.

Other promising approaches include green coding, which aims for applications to require the lowest possible amount of processing, and the efforts of MIT’s Lincoln Laboratory Supercomputing Center to cap processing power and optimize its use in the training of AI models.

Meanwhile, small language models (SLMs) are emerging as an alternative to large language models (LLMs) such as ChatGPT. SLMs use only a fraction of the parameters compared to an LLM, which is an indication of the scale of an LLM’s capacity to recognize and extrapolate from complex data patterns. This means the processing needs are lower for SLMs, and they are far less resource-intensive – while still providing high-quality results.

MIT is also working on working on raising energy awareness by creating the equivalent of a home energy report for data processing. The goal is for users to understand how much energy their processing tasks consume, how their carbon footprints compare to others and how they could be improved. The university’s approach could become a model for the ICT industry more widely, especially as it still lacks comprehensive industry data, MIT researchers say.

Demand flexibility solutions can also help manage the impact of increasing energy requirements on grids. Demand response and load shifting, for example, allow DCs and communications infrastructures to reduce power use during peak hours and switch it to times when demand – and pricing – are lower.

While ICT companies’ actions are crucial to managing energy consumption growth, they are only one part of the picture when it comes to future energy needs. Electricity demand is also rising across areas such as transportation, heavy industries and consumer behaviours in the transition from fossil fuels to electrification.

As this momentum builds – and with it the demand for clean electricity – power companies are poised to become bottlenecks, held back by often lengthy permitting processes for the construction of new infrastructure.

Broader engagement is needed to close the widening supply and demand gap. The private sector, grid operators, planners and regulators must all collaborate to find holistic solutions. This will require the ICT sector to learn from and work with other industry sectors that are driving demand for clean electricity. To enable such a collaborative approach, industry players need to find secure ways of sharing data and DC growth forecasts among themselves and with regulators while avoiding disclosure of business-sensitive information.

And it needs to happen locally, regionally and at an international level. The latter is particularly vital when it comes to creating a consistent regulatory environment across geographies. An integrated approach will help with everything from improving energy emissions accounting and reporting across the ICT value chain to developing new energy efficiency initiatives and alternative energy solutions.

And, finally, there is a part for us all to play. ICT is power-hungry because our society is data-hungry. As individuals and organizations, we need to rein in the data we keep or consume and consider reducing our own carbon footprint.

More about managing the changing dynamics shaping energy supply and demand is available in the World Economic Forum’s Secure, Sustainable and Equitable Energy Systems – Centre for Energy and Materials and The Clean Power, Grids and Electrification initiative.