Data center energy consumption is one of the biggest decarbonization challenges of our time. On one hand, there is huge demand for energy-guzzling applications within the Generative Artificial Intelligence (Gen AI) and cloud computing domains. On the other hand, these technologies require an unprecedented amount of power to run them. For example, some industry gurus have suggested that a single Large Language Model (LLM) interaction could have similar electricity consumption as leaving a low-brightness LED lightbulb on for an hour. Although data centers only account for about 2% to 4% of total electricity consumption in advanced economies today, the anticipated growth in digital technology usage will induce a greater impact on the grid in the future.
As more and larger data centers are constructed around the world to scale digital transformation projects, grid instability will only be exacerbated. ABI Research forecasts the number of public data centers worldwide to quadruple by 2030. With each new data center built comes a larger carbon footprint and the urgent need for thermal management solutions.
How Much Energy Do Data Centers Consume?
ABI Research’s latest analysis indicates that data centers will more than double their annual energy consumption between 2024 and 2030—from 683 Terawatt Hours (TWh) to 1,479 TWh. This represents a Compound Annual Growth Rate (CAGR) of 14%. Unsurprisingly, the technologically advanced U.S., Chinese, and European markets account for over half of global data center energy usage.
Data Center Energy Consumption by Region
(Source: ABI Research)
The shifting digital habits of both consumers and businesses are among the main factors influencing data center energy demand. Every Google search, crypto mining hunt, and ChatGPT query requires processing power, adding to the energy footprint of a data center. Enterprises are also rapidly investing in LLMs, which often require training and inferencing to be processed within data center infrastructure. With more computing power needed to run these applications, Information Technology (IT) equipment generates more heat. Cooling this equipment currently represents 37% of a data center’s energy consumption. Computing power accounts for another 42% of electricity use at a data center.
Data Center Energy Demand by Load (2025)
(Source: ABI Research)
Hyperscale and AI Data Centers Raise Serious Energy Concerns
Hyperscale and Artificial Intelligence (AI) data centers are the backbone of cloud-based services, AI innovation, and High-Performance Computing (HPC). Hyperscale data centers—operated by major cloud providers like Amazon Web Services (AWS), Microsoft Azure, and Google Cloud—are growing rapidly. North America and Europe will continue to be data center hubs, but hyperscalers are showing significant interest in the Asia-Pacific and Middle East regions. For example, AWS recently committed to a US$6.2 billion investment to build data centers in Malaysia. Moreover, Google announced a US$1 billion investment to expand its data center cloud infrastructure to Thailand. Meanwhile, Saudi Arabia’s grand aspirations to become a manufacturing hub will attract hyperscalers to the region. These investments will only be replicated throughout the decade, making energy consumption a glaring issue.
Data Center Energy Consumption by Type
(Source: ABI Research)
Hyperscale data centers are larger than the typical data center and require massive amounts of energy to power and cool server racks and other infrastructure. Our analysts report that electricity depletion in hyperscale data centers is projected to increase from 200 TWh in 2023 to 381 TWh by 2030. The demand for energy-efficient solutions has never been more urgent. To meet this challenge, hyperscalers are focusing on renewable energy and heat dissipation solutions. For example, data center operators are increasingly partnering with nuclear power plants to offset their carbon footprint. Heat reuse/recycling is also gaining traction, with many Western European data centers diverting generated heat to local homes and offices. Such initiatives are critical for the top 20 hyperscalers to increase their low carbon energy consumption from 88% in 2023 to 100% by 2030. However, harnessing clean energy is just one aspect of creating a green data center.
Cooling systems are another essentiality as they mitigate the heat produced by high-performance workloads. Most data centers still rely on traditional air-cooled systems. However, this is changing as hybrid cooling technologies, such as adiabatic chillers and liquid cooling systems, are gaining traction. By 2030, these advanced cooling systems are expected to make up more than 55% of the market. These technologies are crucial for managing the higher thermal loads driven by Gen AI, Machine Learning (ML), and other demanding applications.
Number of Data Centers Connected to Heat Reuse
(Source: ABI Research)
Dedicated AI data centers are also growing quickly. Although they are still in the early stages, their number is expected to roughly double from 604 in 2024 to 1,204 by 2030. These technologically advanced facilities, often part of hyperscale data center campuses, focus on the intense computing requirements of AI workloads. However, AI’s increasing demand for computational power places additional strain on the electric grid and IT infrastructure. For instance, it is estimated that 10% to 20% of a data center's electricity use stems from AI applications.
Both hyperscale and AI data centers must continue to innovate. The growing need for data storage and processing power, driven by AI and high-performance applications, requires data centers to balance efficiency with environmental responsibility. The future of the industry will depend on how effectively it can manage energy consumption, while continuing to scale and support the next generation of computing.
Confronting the Data Center Energy Challenge with Cooling Technologies
Mitigating the environmental impact of data center expansion requires operators to deploy thermal management solutions, such as air and liquid cooling. Companies like Google, Intel, Iceotope, Microsoft, NVIDIA, and Schneider Electric have been testing these cooling techniques to improve Power Usage Effectiveness (PUE) scores. For example, Google employs a combination of water-based and air-based cooling technologies in its data centers. These systems are selected based on sustainable assessments of local watersheds to ensure environmental responsibility. In tandem with renewables, Google has achieved better-than-average PUE scores across its fleet of data centers.
To take a deep dive into the most promising data center cooling technologies that are proven to reduce power usage, download ABI Research’s whitepaper, Accelerating Data Center Efficiency with Advanced Cooling.
