�۶���Ƶ

Electricity demand has played an important role in driving trends in overall primary energy consumption, despite often playing second-fiddle to demand from the transportation sector. Similarly, electrification is generally viewed as a structural tailwind for “green” metals like copper, lithium and nickel, leaving other industrial metals like zinc and lead “out in the cold.” It might be time to re-think these assumptions! Since the mid-1970’s the proportion of total primary energy demand from the electric power sector surpassed the transportation sector (so much for second-fiddle), reaching approximately 34% of total primary energy consumption as of the end of 2024.¹ 2024 data from the International Energy Agency (IEA) also show an acceleration in this trend. In their 2025 Global Energy Review, the group found that, while the world’s consumption of energy in the prior year rose at a faster-than-average pace, demand for electricity, specifically, rose almost twice as much.² The group also tabulates that the average pace of electricity demand growth from 2010 to 2023 was 2.7%, double the rate of total energy demand growth over the same period, due to electrification across a variety of sectors.¹ These trends are expected to intensify due to increasing demand from artificial intelligence (AI) and from the largest, most efficient data center providers – so called “hyperscalers”– which require a large amount of very stable power for cooling and computation. While the market initially crowned natural gas as the primary beneficiary of rising electricity demand, due to its importance in powering these hyperscalers, recent announcements and initiatives by these firms may require a re-think of this assumption.

Hyperscalers are the largest type of data center, generally containing at least 5,000 servers and being at least 10,000 square-feet.³ They house thousands of networked computers that process, store and share data, and the need for these buildings has increased drastically with the expansion of artificial intelligence.² As a general rule, these operations require steady, reliable and perhaps most importantly, relatively cheap, power delivery. Until recently, the primary energy source was presumed to be Natural Gas; however increased competition for U.S. gas due to the build-out of liquefied natural gas (LNG) export capacity could undercut the economics of these facilities if natural gas prices rose and power prices followed. This competition for natural gas-based power has driven hyperscalers to search for other long-term solutions to avoid becoming permanent price takers. One such solution is co-locating renewable energy with their data center projects and utilizing a captive power model, similar to the one that has been successfully deployed across other industries, such as the smelting and refining of aluminum in China. An example of this approach is Meta’s construction of the Eleven Mile Solar Center in Arizona, a project that includes 300 megawatts of solar generation and is intended to power Meta’s 2.5 million square foot data center nearby.³ While this approach solves the power cost problem, it introduces an additional complexity as power delivery from renewable sources, with the notable exception of nuclear, is often quite variable and unpredictable, a massive problem for hyperscalers. However, the hyperscalers intend to solve this problem by building on-site battery energy storage solutions (BESS) to capture energy when it is not needed and fill gaps during periods of low renewable power delivery. For example, in addition to the solar farm, Meta’s plan also includes 1,200 megawatt-hours of battery storage. In addition to lithium-ion (think laptops), BESS systems of the future may utilize a variety of battery chemistries, including those that have been around for many years, such as lead-acid (think car batteries), and those that are still emerging, such as zinc-bromide. If the automotive sector is any indication, forecasting which battery chemistry will become industry standard will likely prove quite challenging.

The use of renewables generation as the primary power source for hyperscalers may require a re-think about potential winners and losers from the roll-out of artificial intelligence. A preference for renewables could reduce future expected demand for natural gas related to data center applications. In addition, the need for BESS alongside renewable power infrastructure could improve the “green credentials” of metals like zinc and lead, with underperformance in these metals versus the rest of the sector potentially underpricing future demand, depending on which battery chemistry becomes the standard-bearer in this space.

¹ IEA: April 2025 Monthly Energy Review.
² IEA: Global Energy Review 2025. March, 2025
³ Together in Electric Dreams - For Tech Giants, Powering Data Centers is Not All About Natural Gas, RBN Energy LLC. April, 2025