November 02, 2025

Many energy companies in North America, including natural gas producers, are trying to capitalize on the growth of data centers, which require a significant amount of energy to operate. But is this growth sustainable?

AI and Data Center’s Power Consumption
The growth of the data center industry is primarily driven by AI workloads. However, data centers serve other purposes as well. It is important to distinguish between different types of data centers, as each has a different power consumption profile.

Figure 1 illustrates AI electricity consumption (TWh) across multiple sources. The spread widens quickly—from relatively modest starting values in 2022–2024 to several hundred TWh by the end of the decade. This spread reflects uncertainties in estimating future power consumption. At the beginning of 2025, Goldman Sachs Research estimated global data center power usage at around 55 gigawatts (GW). This total consists of cloud computing workloads (54%), traditional workloads for typical business functions such as email or storage (32%), and AI (14%). However, AI-related power consumption is expected to grow by 165% by 2030, according to the same

Figure 1. Power Consumption for AI Data Centers

Share of Electricity Consumption By Type
AI model training and deployment primarily take place in data centers. Figure 2 illustrates how data centers use energy:

• Servers — They perform computing using CPUs and accelerators such as GPUs. Servers typically draw about 60% of total electricity, though this share can vary.
• Storage systems — Centralized storage and backup systems keep data safe and accessible. They account for roughly 5% of energy use.
• Networking equipment — Switches, routers, and load balancers move traffic and maintain performance. Networking typically consumes up to ~5% of demand.
• Cooling and environmental control — These systems keep IT equipment within safe temperature and humidity ranges. Their share ranges from ~7% in efficient hyperscale facilities to 30% or more in less-efficient enterprise sites.
• UPS batteries and backup generators — These systems ensure continuity during power interruptions. They are rarely active but are critical for reliability.
• Other infrastructure — Lighting, offices, and facility control systems make up the remaining share. Exact proportions depend on facility design, workload, and efficiency.

Figure 2: Share of Electricity Consumption By Type
Source: IEA

Reducing Energy Consumption or Data Centers
There are several factors that will affect the power consumption of data centers. Figure 3 illustrates the “pendulum” of data-center power demand.

Figure 3: Factors influencing power demand for data centers

Computational algorithms for AI will continue to improve. For example, the AI model DeepSeek has raised questions about the long-term need for extremely large data centers, as newer models may require significantly less computation than previous generations. At the same time, the development of generative AI will require greater computational resources. At present, it is impossible to extrapolate current AI computing demand to accurately forecast future energy consumption.

Data centers can curb energy use by improving cooling efficiency. For example, liquid cooling and well-sealed hot/cold aisle containment reduce the power required to move and chill air. Free-air cooling (airside economization) brings in cool outdoor air to remove server heat, often bypassing chillers for much of the year and significantly reducing energy used for mechanical cooling. Several additional technologies will further improve cooling efficiency.

Server hardware, including GPU platforms, will also continue to advance. For instance, in 2024 Nvidia reported that its newer platform (Blackwell) may deliver up to a 25× reduction in cost and energy consumption compared with its predecessor. Server virtualization and consolidation allow many workloads to run on fewer, more efficiently utilized hosts, enabling the retirement of idle (“zombie”) servers. Other technologies include Dynamic Voltage and Frequency Scaling (DVFS), which allows CPUs to adjust voltage and clock speed to match real-time load. Effective power management and monitoring will further reduce power demand.

At the same time, AI applications themselves have limits. Currently, most computational demand comes from video and voice generation, analysis, and recognition. Text-based applications, such as chatbots like ChatGPT and Gemini, require significantly less power. There is growing demand for specialized AI applications in areas such as medicine, pharmaceuticals, engineering, and defense. However, Incorrys believes that many future AI applications will be associated with physical AI systems, such as robots—including self-driving cars, manufacturing robots, and other machines. These systems will perform tasks autonomously in real-world environments by processing sensor data locally rather than relying primarily on data centers.

Conclusions
Data centers will require additional power generation capacity at least until 2030. In addition, growing demand from data centers will necessitate further utility investment. Goldman Sachs Research estimates that approximately $720 billion in grid spending may be required through 2030.
However, Incorrys believes that this growth will not continue indefinitely. Improvements in computational algorithms, server hardware (including GPU chipsets), and cooling technologies will help to limit future power requirements for data centers. At the same time, the applications driving AI and other compute-heavy workloads also have practical limits. Therefore, investments in power generation dedicated to data centers are unlikely to remain a primary focus for energy companies after 2030, and potentially even earlier.

References:
“What Is a Data Center? Tiers, Types, and More.”, Nlyte Software, https://www.nlyte.com/faqs/what-is-a-data-center/ . Accessed 14 October 2025.

“NVIDIA Blackwell Platform Arrives to Power a New Era of Computing”. Nvidia press release. March 18, 2024. https://nvidianews.nvidia.com/news/nvidia-blackwell-platform-arrives-to-power-a-new-era-of-computing

“What are the standards for data center physical infrastructure?”, Cisco, https://www.cisco.com/site/us/en/learn/topics/computing/what-is-a-data-center.html . Accssed 14 October 2025.

“Energy Consumption.”, Seneca Notes, https://senecalearning.com/en-GB/revision-notes/igcse/computer-science/edexcel-igcse/6-1-2-energy-consumption . Accessed 14 October 2025.

“Green Data Center – The Complete Guide for 2024.”, Encor Advisors, October 2024, https://encoradvisors.com/green-data-center/

Williams, Alex. “Optimizing Energy Efficiency in Datacenters with Advanced Cooling Technologies.”, Communications of the ACM, April 2024, https://cacm.acm.org/blogcacm/optimizing-energy-efficiency-in-datacenters-with-advanced-cooling-technologies/

“Decoding the Power Consumption of Data Center: A Deep Dive into Energy Use Dynamics.”, hivenet, https://www.hivenet.com/post/decoding-the-power-consumption-of-data-center-a-deep-dive-into-energy-use-dynamics . Accessed 15 October 2025.

“How to Reduce AI Power Consumption in the Data Center.”, Pure Storage, June 2024, https://blog.purestorage.com/purely-technical/how-to-reduce-ai-power-consumption-in-the-data-center/

Kalra, Vani. “Four strategic steps to greener data centers.”, TCS, https://www.tcs.com/insights/blogs/sustainable-green-data-centers-strategies . Accessed 15 October 2025.

Paatela, Inca. “How to Build an Energy-Efficient AI Data Center.”, skeleton+, April 2025, https://www.skeletontech.com/skeleton-blog/how-to-build-an-energy-efficient-ai-data-center

“How AI Can Improve Energy Efficiency in Sustainable Data Centres.”, Digital Realty, https://www.digitalrealty.ie/resources/articles/sustainable-data-centre-ai . Accessed 15 October 2025.

“Cutting-Edge Technologies Enhancing Energy Efficiency in Data Centers.”, Prismecs, June 2024, https://prismecs.com/blog/energy-efficiency-in-data-centers

“16 More Ways to Cut Energy Waste in the Data Center.”, Energy Star, https://www.energystar.gov/products/data_center_equipment/16-more-ways-cut-energy-waste-data-center . Accessed 15 October 2025.

“Smart Strategies for Data Center Energy Consumption.”, Arcus Power Corp, November 2024, https://www.arcuspower.com/post/smart-strategies-for-data-center-energy-consumption

Kamiya, George. Coroamă, Vlad C. “Data Centre Energy Use: Critical Review of Models and Results.”, EDNA – IEA 4E TCP, March 2025, https://www.iea-4e.org/wp-content/uploads/2025/05/Data-Centre-Energy-Use-Critical-Review-of-Models-and-Results.pdf

“Energy and AI-Energy demand from AI.”, International Energy Agency, April 2025, https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai

“AI to drive 165% increase in data center power demand by 2030”, Goldman Sachs, Feb 4, 2025. https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030