Use of PGMs in artificial intelligence
Ruthenium plays a key role in heat assisted magnet recording technology for high-capacity hard disc drives (HDD), storing artificial intelligence data. Platinum is used in proton exchange membrane (PEM) fuel cells that backup power for data centres facing costly outages: Platinum is unmatched in its efficiency, stability and rapid startup, which is essential for reliability when backing up power for these data centres.
The AI revolution is impacting us all but one aspect that may be overlooked is the demand for precious metals.
As chips get smaller and technology must keep pace with our need for more data storage, these precious metals become strategically vital.
Ruthenium is the lynchpin for the growth of AI. It plays a key role in HAMR (heat assisted magnet recording) technology for high-capacity hard disc drives (HDDs) used in AI data storage. Thin ruthenium layers enable perpendicular magnetic recording and multi-layer media stability.
A few years ago, many expected solid state drives (SSDs) to overwhelm hard disk drives but the reality of cheap, high-capacity storage for cloud computers has driven HDD demand back after a multi-year slump.
Every time your Google or Apple account asks you to increase your storage that means gaining more access to cloud computing. This data is kept somewhere physically. Huge banks of information in global cloud storage require vast amount of hard disk drives to keep their data written and re-written. Thin layers of Ruthenium are used on the surface of the disc drive to make this happen.
How does this work? A nano-scale laser briefly heats a tiny region on the spinning disc which allows the area to be more receptive to magnetic changes, so in the magnet field of a voice coil on the hard disc drive, data can be stored. Once cooled, the information or data bit is locked in place.
What does the Ruthenium do? Ruthenium acts as an important barrier, stopping the heat from spreading elsewhere and isolates the magnetic layers giving additional recording layers in the same space. Today storage is in the range of 30-36 terabyte (TB) of data, which allows data centres to triple their storage density over conventional pre-HAMR 20 TB drives. This year we will see the rise of 40 TB HAMR based hard drives to help store more data for the AI wave. By 2030, Seagate expects to launch 100 TB HAMR drives, which will allow the creation and storage of data in real time. Total shipments by the end of this decade are expected to be 7.3 zettabyte (ZB), with 1 ZB equivalent to 1 billion terabytes. For those that remember floppy discs, 7.3 ZB has the same storage capacity as 660 trillion floppy discs!
Platinum and the rise of AI. Platinum is doing its part for AI but not in the way you would expect. Global data centres electricity use is projected to exceed 945 TWh annually by 2030. To put this number in context, this is more than Japan's total consumption today! With AI-optimized facilities quadrupling that demand, often outpacing grid capacity and risking costly outages. PEM (Proton Exchange Membrane) fuel cells provide rapid startup (<60 seconds), high reliability, and scalability from 200 kW to multi-MW, outperforming diesel generators in sustainability and maintenance according to Vertiv-Ballard. Microsoft and Ballad/Caterpillar have trialled PEM backup generators at 1,855 metres elevation and Microsoft/PlugPower have set up 3MW PEM fuel cell deployment for data centre power resilience.
Platinum is a key component in PEM fuel cells due to its catalytic performance both for splitting hydrogen at the anode and reducing oxygen at the cathode. Platinum is unmatched in its efficiency, stability and rapid startup which is essential for reliable data centre back up power.
AI helping PGMs find new applications. PGMs help AI but AI can return the favour by accelerating PGM innovation in the chemical world. More and more we are seeing data models being built in the chemical world looking for new ways to develop science. Meta and Carnegic Mellon University have released OpenDAC to help develop cheaper direct air capture molecules to support carbon capture innovation. Companies like GreenCat leverage AI for combinatorial PGM rearrangements for new catalysts discovery. Enthalpic uses generative AI for the discovery of new materials and chemistry. Iris.ai helps screen chemical literature and patents to identify PGM alloys and material innovation for clients. Citrine Informatics has an AI Platform for materials R&D, optimise catalysts and materials for decarbonisation. Combinatorial rearrangements of precious metals would be an expensive real world experiment but with AI, we can rapidly rearrange these precious elements virtually at a fraction of the cost to find the next best catalysts and finding step change benefits for the decarbonisation industry.
AI helping find PGMs and refine them too! AI can also help us find these elusive metals in the first place. Companies like Earth AI have trained their technology for remote sensing to pinpoint greenfield precious metals deposits in data poor region. Nornickel is using AI for safety monitoring and production optimisation at Kola MMC. Sibanye Stillwater is applying AI to help PGM operations for plant optimisation via its new AI policy.
As AI scales at dizzying heights, precious metals are instrumental in this transition, from data storage to power resilience and beyond. Fears of AI can be reframed as an opportunity to unlock unexpected new platforms for PGM demand in the future by innovating a greener future.