At the heart of the infrastructure buildout for artificial intelligence, there is a paradox: For the data center sector to keep growing, the energy sector also needs to expand — but both are increasingly relying on many of the same supply chains.
The latest International Energy Agency’s latest World Energy Outlook, released yesterday, notes that “as data centres increase in number and as their needs develop, their supply chains for materials and energy-related technologies will increasingly overlap with supply chains critical to the energy sector.”
This overlap risks creating a bottleneck to scaling both industries, as sectors that depend on one another could end up competing for the same resources, slowing down technology advancement and pushing up prices. “Monitoring how the material footprint of AI and data centres evolves will be critical to anticipating some key energy security risks,” the report adds.
Amanda Van Dyke, founder of the critical minerals intelligence platform Critical Minerals Hub, says critical minerals are where the overlap of the AI and energy supply chains becomes particularly evident.
“You need critical minerals for AI infrastructure, critical minerals for transmission infrastructure, and critical minerals to build the power generation infrastructure,” she told Latitude Media, adding that in a data center dedicated primarily to AI, roughly half of the physical materials go into the supporting power infrastructure.
The refinement of those minerals, the World Energy Outlook highlights, is overwhelmingly concentrated in China, which poses a potential challenge to energy security. The same day the IEA report was released, the U.S. House of Representatives committee issued a 50-page report accusing China of manipulating global critical minerals prices for decades.
Copper is the ‘canary in the coalmine’
There are at least 20 different minerals in a single GPU. Chips manufactured on high-purity silicon wafers are used in data centers, EV power electronics, and energy storage, for instance, while gallium is needed both for AI hardware and scaling up renewables. But, Van Dyke said, not all of the supply chains are equally valuable: “When it comes to power, you need copper more than anything else,” she said. Copper is essential to AI data centers, as well as renewable energy transmission and distribution and EVs. And as a result, copper demand has been soaring in the past couple of years. Mining output, however, has not kept up, with the IEA predicting a potential 30% gap between projected copper supply and demand by 2035.
“The extent to which progress is made in closing the projected demand and supply gap for copper will have a material impact on both data centres and the energy sector, since both are projected to be a key source of copper demand growth,” the report notes.
According to Van Dyke, copper is the “canary in the coalmine” of future shortages. Ramping up mining capacity is expensive and time-consuming; it can take up to 15 years to open a new mine. “It is a very good example of other shortages that could occur with competing demands for resources, from power, AI, EVs, and renewable generation,” she added.
Critical minerals and materials aside, the energy industry and the AI sector are also competing for a skilled workforce, the shortage of which is identified as a “critical bottleneck” in the IEA report. Electricians, grid line workers, pipefitters, and welders have been in short supply for a while now, and the issue has been exacerbated by the AI boom, which is flooding them with work and limiting their ability to take on new clean generation projects.
At least in the short term, data centers, which are flush with cash, are likely to come out on top in a head-to-head competition for resources with the energy industry. But as the industries become increasingly intertwined, and shortages, whether of materials, minerals, or workers, won’t affect one without the other.
“In the short term, shortages like next year’s copper deficit are going to affect the rollout of everything,” van Dyke said.
