What the data centre and AI boom could mean for the energy sector

Investment in new data centres has surged over the past two years, driven by growing digitalisation and the uptake of artificial intelligence (AI), which is expected to continue accelerating. Much of the spending is concentrated in the United States, where annual investment in data centre construction has doubled in the past two years alone, although other major economies, such as China and the European Union, are also witnessing an increase in activity. In 2023, overall capital investment by Google, Microsoft and Amazon, which are industry leaders in AI adoption and data centre installation, was higher than that of the entire US oil and gas industry – totalling around 0.5% of US GDP.

Average data centres are quite small in power terms, with demand in the order of 5-10 megawatts (MW). But large hyperscale data centres, which are increasingly common, have power demands of 100 MW or more, with an annual electricity consumption equivalent to the electricity demand from around 350 000 to 400 000 electric cars.

It is important to put this in perspective: Global sales of electric cars will reach 17 million in 2024. Today, data centres account for around 1% of global electricity consumption, and annual electricity consumption from data centres globally is about half of the electricity consumption from household IT appliances, like computers, phones and TVs. However, as the sector expands, it is important to examine the consequences for the energy sector, which we analysed in the 2024 edition of the IEA’s World Energy Outlook and will continue to explore in the months to come, including through a Special Report focused on energy and AI in the first half of 2025.

Additionally, data centres will be a key focus of the IEA’s forthcoming Global Conference on Energy and AI. The event, which will take place at our headquarters in Paris on 4-5 December, will bring together high-level decision makers from governments, the tech sector, the energy industry and civil society to discuss the ways in which AI could transform energy systems in the future.

In part because of expectations for AI, the next few years will see a substantial rise in the number and size of data centres. This growth will be partially mitigated by continued efficiency improvements at both the hardware and software level. Nonetheless, electricity demand from data centres is set to grow strongly to 2030 under today’s policies settings and trends.

However, when considered in a broader context of total electricity consumption growth globally, the contribution of data centres is modest. Global aggregate electricity demand grows by 6 750 terawatt-hours (TWh) by 2030 in our Stated Policies Scenario, equivalent to more than the combined demand from the United States and European Union today. While growing digitalisation, including the rise of AI, is one factor, continued economic growth, electric vehicles, air conditioners and the rising importance of electricity-intensive manufacturing are all bigger drivers. 

In large economies like the United States, China and the European Union, data centres account for around 2-4% of total electricity consumption today. But because they tend to be spatially concentrated, their local impact can be pronounced. The sector has already surpassed 10% of electricity consumption in at least five US states. In Ireland, they now account for over 20% of all electricity consumption.

For comparison, large data centres can have a power demand equivalent to that of an electric arc furnace steel mill. However, steel plants are less likely to be clustered in the same geographic area. 

The growth of data centres could therefore lead to considerable strain on local power networks, exacerbated by the huge mismatch between rapid data centre construction times and the often sluggish pace of expanding and strengthening grids and generation capacity. There have already been instances of jurisdictions pausing new contracts for data centres due to a surge of requests. For regions or countries that are particularly affected, rising electricity consumption from data centres could make meeting their climate targets more difficult.

With the role of data centres in the electricity system set to increase, it is important that policymakers and regulators have the tools to understand this new driver of demand growth. A number of key variables merit further discussion and analysis.

Firstly, the speed and manner in which AI use will grow remains fundamentally uncertain. Early data suggests that household adoption is rapid – perhaps faster than with other transformative digital technologies. But which uses of AI become popular over time, among both households and businesses, will have implications for energy demand. For example, video generation is far more energy-intensive than text generation or AI-enabled search. The future direction of AI model development matters, too, since some approaches consume significantly more energy than others. The financial returns from AI applications could also affect data centre investment trends, since current spending is tied to expectations on profitability in the future.   

Secondly, the outlook for continued efficiency improvements in both hardware and software needs to be better understood. The efficiency of AI-related computer chips has doubled roughly every two-and-a-half to three years, and a modern AI-related computer chip uses 99% less power to perform the same calculations as a model from 2008. New cooling technologies are being developed, and AI models themselves are becoming more efficient. At the same time, the operational and energy performance of data centres is relatively opaque, making demand estimates difficult. Efforts are underway to improve transparency in some jurisdictions.

Thirdly, more effort is needed to understand the physical constraints on demand growth. Understanding of the data centre project pipeline is limited, and data is not readily available. Meanwhile, chip production may present a near-term bottleneck. The energy sector itself may slow down the growth of AI, if generation and grid capacity is not available in the places it is most needed.

Finally, the impact of AI applications in the energy sector more broadly needs to be fully assessed. Promising examples include accelerating breakthroughs in clean energy innovation, managing the electricity system to facilitate more renewables, and deploying AI to enhance the profitability and speed of electrification programmes in developing economies. These applications could potentially transform energy systems, but today, their impacts, enabling conditions and scalability are not well known.  

Overall, there is an urgent need for public-private dialogue, with policymakers, the tech sector and the energy industry coming together for discussions. The promises of AI are real – not least for clean energy innovation. But delivering responsible AI will require new partnerships to quickly emerge. The upcoming Global Conference on Energy & AI aims to provide a space to kickstart and advance these conversations.