Copper underpins every layer of electrical infrastructure, from wiring and transformers to data centers and transmission lines. The article argues that while most discussions on AI focus on models and compute, the supply chain for copper often goes unnoticed; yet it’s increasingly under stress. A single large data center may require thousands of tonnes of copper, and forecasts suggest that global copper demand for infrastructure expansion could reach 37 million tonnes by 2031, tells IEEE Spectrum.
The core challenge is that the accessible, high-grade copper deposits are depleting. More than 70% of known reserves now lie in ores that are complex and expensive to process using conventional mining and smelting techniques. Meanwhile, billions of tonnes of material lie in waste heaps and marginal deposits. Conventional extraction methods consume energy, generate emissions, and struggle to scale economically to these lower-grade sources.
Enter bioleaching, or microbe-assisted extraction. The article profiles the company Endolith, founded by geoscientist Liz Dennett, which deploys specialized microbes to dissolve copper from ores in situ. Rather than grinding rock and using acid baths or high-temperature smelting, microbes attach to mineral surfaces and slowly mobilize copper ions, often in a gentler, more targeted manner.
What makes this promising now is that Endolith combines microbial science with machine learning. The team models which microbial strains and conditions are optimal for different ore types, adapting the mix over time. In pilot tests, they’ve worked with major copper producers, showing that microbe-based recovery can operate in field conditions and on challenging ores such as chalcopyrite.
Bioleaching isn’t a silver bullet. Historically, it has struggled with speed, yield, and scaling, especially for stubborn ores. But the article suggests that recent advances in genomics, heap engineering, and AI optimization are closing the gap. If successful, microbe-enabled recovery could unlock overlooked resources, reduce environmental impact, and ease the metal constraints on AI infrastructure.