How mining is feeding the AI data-centre boom

The latest quarterly report from the World Platinum Investment Council (WPIC) has highlighted AI infrastructure as an emerging source of platinum demand, as higher-performance data-centre hardware drives growth in glass, electrical and semiconductor applications. 

In its Q2 2026 Platinum Quarterly, the WPIC expects industrial platinum demand to rise 5% year on year in 2026 to 2.385 million ounces (Moz), with its forecast upgraded partly because of AI-related applications.  

Within the industrial category, platinum demand for glass applications is forecast to increase by 23%, or 98,000 ounces, while electrical demand is expected to rise 19%, or 19,000 ounces. 

The connection is not that data centres themselves contain large quantities of platinum. Rather, PGMs are used further upstream in the manufacture of components and materials needed to build and operate increasingly powerful computing infrastructure. 

Edward Sterck, director of research at WPIC, said the emerging AI application was broader than platinum. “It’s not just platinum; it’s ruthenium, iridium, and to a lesser degree palladium as well, being used in this sort of AI infrastructure.” 

The development has also emerged relatively recently as a potential source of PGM demand. “That’s something that’s only really kind of come out of the woodwork towards the end of the second quarter. So for us, it’s quite new,” Sterck said.  

E-glass fabric 

One of the clearest links is through printed circuit boards (PCBs), which connect electronic components in AI servers and other data-centre equipment. 

Increasing demands on AI infrastructure are pushing PCB designs towards higher speeds and frequencies, increasing the importance of materials that can transmit signals quickly while limiting signal loss. 

Electrical-grade fibreglass, or E-glass, is an important component of PCB housing, providing mechanical stability and electrical insulation. Newer low-dielectric-constant formulations can also help reduce signal loss and support faster transmission. 

The platinum connection comes earlier in manufacturing. Molten glass is drawn through perforated platinum-rhodium fixtures to form fine fibres, with the alloy able to withstand the extreme temperatures involved.  

The result is an indirect chain of demand: greater AI computing requirements drive investment in higher-performance servers and networking equipment, increasing demand for more advanced PCBs and, in turn, the glass materials and manufacturing equipment used to produce them. 

HDDs, semiconductors and sensors 

Electrical demand is forecast to increase by a further 19,000 ounces, with AI also contributing to the growth. 

Platinum is used in electrical applications including hard disk drives (HDDs), semiconductor manufacturing and sensors. WPIC said all three were benefiting from AI, although through different mechanisms. 

HDDs remain important to large-scale data storage because they offer a lower cost per gigabyte than solid-state drives. Their spinning storage disks, known as magnetic platters, use platinum-alloy coatings to improve thermal and magnetic stability, while heat-assisted magnetic recording has increased storage density. 

Platinum is also used as a sputtering target, with the metal deposited as an ultra-thin film onto components used in semiconductor and sensor applications. WPIC identifies thermal cameras, ultrasonic, photocell, inductive, radar and motion sensors among AI-related applications.  

Global data-centre expansion 

The scale of the expansion provides the backdrop to the emerging demand opportunity. 

JLL expects global data-centre capacity to increase from 103GW in 2025 to around 200GW by 2030, with almost 100GW of new capacity added. It estimates that up to $3tn (£2.21 trillion) could be invested when data-centre construction and tenant spending on GPUs and networking equipment are included. 

AI accounted for around a quarter of data-centre workloads in 2025, according to JLL, but could represent half by 2030.