Electric Power, Energy Transition, Renewables

January 20, 2025

5 signposts shaping the future of global data center power demand

Getting your Trinity Audio player ready...

The rapid expansion of data centers is now central to discussions about global energy consumption.

Worldwide data center power demand could grow 10%-15% annually over 2024-30, according to a recent S&P Global Commodity Insights report.

By 2030, data centers could represent up to 5% of all power demand globally, according to the Commodity Insights report, which outlines the implications of data centers on energy consumption and provides regional outlooks using different forecasting approaches across the technology and power sectors.

Beyond the global picture, data center impacts on power demand will vary significantly by region, Commodity Insights analysts said in the report.

Based on the report, here are the five key signposts that are expected to shape global data center power demand trends.

1. The ability of electric transmission networks and power supply to scale at pace

Modern data centers -- individual facilities and, potentially, gigawatt-scale data center campuses -- can consume as much or more electricity than a small city. The demand from new data centers will necessitate significant expansions in power supply and transmission infrastructure. Regions such as the US, Ireland, the Netherlands and Singapore have already experienced grid connection bottlenecks that delay development timelines.

The bottleneck is observed in the time required to develop power sector infrastructure, with construction timelines facing delays from local opposition and supply chains for equipment. In the US, local concerns over noise, the visual impact of new power lines, road saturation and environmental degradation have emerged in regions like Northern Virginia, Memphis and Missouri.

Supply chains are already stretched, especially for long lead time equipment like uninterruptible power supply, or UPS, generators and transformers. Timelines for new equipment are much longer than just a couple of years ago.

2. Reliance on off-grid power supply

Many data center operators prefer grid-based electricity, but delays in obtaining service are prompting some to explore off-grid options.

The largest data centers today consume around 50 MW-100 MW. As new data centers may soon reach the gigawatt scale, local grid reliability will be increasingly at risk, necessitating the adoption of on-site generation solutions.

For example, xAI has installed 66 MW of gas turbines at its Colossus data center in Memphis, Tennessee, as a temporary solution until the Tennessee Valley Authority can provide a 150-MW grid connection. Similarly, the Lancium Clean Campus in Abilene, Texas, is considering on-site renewable energy.

As new data centers consume more power, the impact on infrastructure also becomes more pronounced. In Indonesia, Microsoft announced plans to invest $1.7 billion in new cloud and artificial intelligence infrastructure, but power supply availability and grid constraints remain key challenges to investment.

3. Energy efficiency improvements in AI software and hardware

AI is still in its early commercialization phase, and while generative AI queries may consume significantly more energy than traditional searches, advancements in energy-efficient designs and custom chips could mitigate this impact over time.

Concerns about the technology sector's voracious appetite for electricity also emerged in the 2010s -- during the rise of internet services. However, from 2010 to 2018, data center services demand grew around 600%, while power demand barely increased, thanks to significant improvements in energy efficiency. The same may occur in the coming years.

On the other hand, there is the potential for AI use cases to increase computational complexity and become more energy-intensive. For example, in emerging use cases like AI agents, responding to a single query may involve calling on multiple other AI models for input. As a result, while individual models may become more efficient, using more models per prompt may increase the overall energy requirement.

4. Implications of AI on electricity consumption in other economic sectors

As data centers increasingly drive economic activity, a key question arises about their role in changing electricity consumption patterns. Data centers are vital infrastructure for the growing digital economy, influencing how people shop, work, and spend money, which in turn affects electricity use. For instance, data centers facilitate remote work, reducing the need for office space and on-premises servers. They also boost e-commerce, thereby decreasing the footprint of physical stores.

Recently, many cryptocurrency miners have announced plans to convert to traditional data centers, which will expand computing capacity but have a minimal net impact on electricity demand. Existing operators may also adapt their facilities to enhance cloud and AI services.

5. Governments' role in driving the growth of data centers

Globally, governments will be instrumental in promoting the development of data centers through policies that offer regulatory support, capital expenditure incentives, operating expense subsidies and infrastructure to accommodate data center hubs.

In South Asia, some governments offer subsidies on retail tariffs, land and building costs, and a fast approval process for data centers. In India, policymakers are helping data center development in key industrial and urban centers such as Mumbai and Chennai.

In China, the government released a national plan to develop data centers in coordination with renewable energy megabases to capitalize on the clean energy resources available in the east and the industrial computing centers in the west. However, progress has been slow as companies prefer to locate data centers near user bases in western population centers.

In North America, AI is viewed as critical to national security. Increased focus at the federal level might lead to stricter controls over where and how data can be stored and processed, limiting the potential for offshoring AI services outside the US market. Data center development worldwide will require careful management.

Further reading: Awaiting the demand surge: Impacts of datacenter growth on power demand globally



Qingyang Liu, Ben Levitt

Editor:

Barbara Caluag