The opinions expressed in this report are those of World Platinum Investment Council (WPIC) and are considered market commentary. They are not intended to act as investment recommendations. Full disclaimers are available at the end of this report.

Executive summary

Artificial intelligence (AI) is rapidly emerging as the defining technology of the modern economy, with projections indicating a dramatic expansion in market size as its rollout scales. The United Nations Conference on Trade and Development forecasts that the global AI market will increase from U.S. $189 billion in 2023 to U.S. $4.8 trillion by 2033 – a 25-fold increase in just a decade – while total available computing capacity from AI chips across all major designers has grown by approximately 3.3x per year since 2022, effectively doubling every seven months (figure 1). 

This transformation is not only digital but physical, relying on a complex chain of critical materials and industrial processes to enable computation, storage, connectivity and power.

Figure 1: Total available computing capacity from AI chips across all major designers has grown by approximately 3.3x per year since 2022

Among these critical materials are platinum group metals (PGMs) due to their unique properties such as thermal stability, corrosion resistance, catalytic efficiency and electrical reliability. While AI is not yet classified as a discrete demand category within platinum markets, PGM demand across AI spans a range of existing industrial uses and proven technologies. AI applications are an emerging end market for PGMs. 

The main segments of PGM industrial demand (figure 2) that are benefiting from AI proliferation include: chemical (platinum catalysts in silicone production); electrical (hard disk drive platters and sputtering targets for thin film deposition on sensors and semiconductors); glass (E-glass fabrics); hydrogen (proton exchange membrane fuel cells for back-up power); and crucibles for crystal growing (categorised under ‘other’ demand).

Figure 2: Platinum demand end-use shares 2025 vs. 2026

It is too soon to quantify global PGM demand growth potential from AI, but WPIC data* indicates that platinum demand growth is already in evidence from areas such as silicone and fibreglass production, semiconductor fabrication and hard disk drives for data storage.

PGM applications enabling AI rollout

Silicone production (chemical demand)

Platinum-cured silicone is known for its high-quality, flexibility, durability and resistance to extreme temperatures. Its properties make it ideal for thousands of applications requiring a reliable, safe and long-lasting material. 

Industries ranging from automotive to healthcare rely on platinum-based silicone products like seals, gaskets and even baby bottles. Notably, given its high purity, it is especially suitable for medical applications, where chemical stability, safety and non-toxicity are paramount.

Platinum catalysts are integral to the manufacture of silicones because they ensure a highly efficient and consistent curing process. For example, the chemical company WACKER uses platinum as a catalyst for the curing of crosslinking silicone release agents and for addition-curing silicone products. This efficiency translates to lower production costs and improved product quality.

The manufacturing process begins with silicone monomers, small molecules that link together to form polymers. When a platinum catalyst is mixed with the silicone monomers, it facilitates a reaction that leads to the formation of long silicone chains. This process is crucial because it determines the final properties of the silicone, including strength and flexibility.

In China, AI-enabled humanoid robots represent the fastest-growing emerging application for platinum-cured silicone. Thanks to its purity and biocompatibility, it is a key material for human-like interactive experiences, primarily being used in facial ‘skin,’ hand coverings and joint cushioning, where realistic touch and temperature matter. 

Global chemical demand for platinum is expected to grow in 2026, increasing 6% year-on-year to 612 koz, supported by modest growth in the silicone sector. 

Hard disk drives and sputtering targets for coating semiconductors and sensors (electrical demand)

AI applications generate and process vast volumes of data, making storage infrastructure critical. Despite competition from solid-state drives (SSDs), hard disk drives (HDDs) remain indispensable due to their cost efficiency at scale. 

Platinum plays a crucial role in HDD manufacturing, particularly in magnetic media coatings that enhance thermal and magnetic stability. The evolution of Heat-Assisted Magnetic Recording (HAMR) technology marks a significant advancement. By enabling higher areal densities, HAMR allows more data to be stored without increasing device size or power consumption. 

This shift toward higher-capacity drives has increased platinum loadings per unit. As AI-driven data centres expand, HDD production capacity is increasingly committed to meeting demand for high-performance computing and cloud storage applications. 

Sputtering is an established technology which enables a thin film to be applied to an underlying layer (substrate) by eroding or ejecting particles from a source material known as a sputtering target. The process is repeatable and can be scaled up from small research and development projects to production batches involving medium-to-large substrate areas of different shapes and sizes. Sputtering allows films that are only a few atomic or molecular layers thick to be deposited onto a surface. 

By using thin film technology, materials can be produced with specific, customized properties that are often difficult or impossible to achieve with other technologies. Thin films have assisted with the development of microelectromechanical systems (MEMS) and nanotechnologies within the field of electronics. 

In addition, thin film technology can often help to save material and costs, as only very small quantities of expensive or rare materials are required. The semiconductor industry uses platinum sputter targets as platinum’s conductivity and stability make it an ideal material for creating the thin films necessary to ensure efficient electron flow within a semiconductor.

Semiconductors are central to AI, enabling data processing, storage, and transmission. As AI systems become more sophisticated, semiconductor manufacturers are advancing toward smaller process nodes and more complex architectures. The growth of AI is driving strong demand for semiconductors, with significant investment in fabrication facilities and advanced manufacturing processes. This expansion directly supports increased demand for platinum in thin-film deposition technologies. 

Platinum sputtering targets are also used to produce platinum thin films for sensors in a wide variety of uses, including AI, where models need to be trained with large amounts of data. By combining AI models with sensors, such as thermal cameras, ultrasonic sensors, photocells, inductive sensors, radar sensors and motion sensors, the amount of data needed to train a model can be reduced. 

In 2026, platinum electrical demand is projected to increase 20% year-on-year to 119 koz, supported by AI-related HDD and semiconductor growth.  

E-glass fabric (glass demand)

Platinum glass demand is forecast to increase by 83% year-on-year to 377 koz in 2026, recovering from a low base of 206 koz in 2025 due to the cyclical absence of incremental new fibreglass capacity expansion last year. Growth will be driven by renewed expansion in fibreglass applications, supported by the return of capacity additions and stronger demand from higher value end-uses, including those linked to AI.

A unique feature of platinum is that it is a metallic element that is exceptionally resistant to heat and wear. In fibreglass production, platinum-rhodium bushings are used to draw molten glass into fine fibres at extremely high temperatures. Fibreglass made this way is used to produce electrical-grade glass (E-glass) yarn, which can be woven to create different types of E-glass fabric.

Known for its high strength and excellent electrical insulation, E-glass fabric is an important component in a printed circuit board (PCB), an electronic assembly that uses copper conductors to create electrical connections between components, allowing advanced semiconductors to communicate with each other at very high data rates. In a PCB, E-glass fabric is impregnated with epoxy resin and layered to form a substrate, the structural base of the PCB. Its job is to provide mechanical stability while ensuring the board can withstand heat and electrical stresses.   

Glass composition, thickness specification and weaving technology can all be altered during the E-glass manufacturing process to create materials with different characteristics. E-glass fabric in PCBs has been developed to exhibit low dielectric constant/low loss (low-Dk) properties to meet the demands of new technologies. Among other benefits, low-DK materials provide fast signal transmission and reduced signal loss. 

AI servers and data centre equipment cannot function without PCBs and, as AI workload increases, PCB designs are evolving to support higher speeds and frequencies, especially through the integration of low-Dk materials. 

Fuel cells (hydrogen demand)

According to the International Energy Agency, worldwide electricity demand from data centres is set to more than double by 2030 to around 945 terawatt-hours, more than the entire electricity consumption of Japan today. AI will be the most significant driver of this increase, with electricity demand from AI-optimized data centres projected to more than quadruple by 2030.

Yet the existing grid is struggling to cope with this additional demand, and the pace of AI adoption is driving a need for new power solutions, with a focus on energy resilience (including uninterrupted power supply) and low or zero-carbon energy options.

Data centre operators are developing their own tailor-made energy systems designed to deliver power faster, and in a way that is both reliable and sustainable. Platinum-based proton exchange membrane (PEM) hydrogen fuel cells are increasingly being incorporated into these future-ready models. 

Hydrogen fuel cells convert hydrogen gas into electricity through a chemical reaction with oxygen, producing only water as a by-product. PEM fuel cells are particularly well-suited for use in data centres because of their quick start-up times and high power density. They can provide reliable back-up power and are efficient in managing fluctuating energy demands.

For example, global provider of critical digital infrastructure, Vertiv Infrastructure Solutions, has partnered with Ballard to develop, supply and install a zero-emission, uninterrupted power supply system for data centres using Ballard’s PEM fuel cells. This back-up power application is scalable from 200kW to multiple megawatts.

Ballard has also worked with leading manufacturer of construction and mining equipment, Caterpillar, together with Microsoft, to successfully conclude a project to demonstrate the viability of using hydrogen fuel cells. The demonstration provided valuable insights into the capabilities of fuel cell systems to power multi-megawatt data centres, ensuring uninterrupted power supply to meet requirements.

It was conducted in a challenging environment and validated the hydrogen fuel cell power system’s performance at 1,855 m above sea level and in below-freezing conditions. The project simulated a 48-hour back-up for an external power loss event at Microsoft's data centre in Wyoming, United States. Elsewhere, Microsoft has collaborated with PEM fuel cell-provider Plug Power on a three-megawatt fuel cell designed for data centre use. 

Meanwhile, cloud computing firm Equinix has evaluated the use of hydrogen and hydrogen fuel cells for sustainable data centre back-up power generation in a study with the National University of Singapore’s College of Design and Engineering. Their analysis highlights PEM fuel cells as offering promising back-up power solutions.

Global platinum hydrogen demand is forecast to increase by 7% year-on-year to 69 koz in 2026.

Crucibles to produce industrial crystals (other demand)

In materials science, crystal growth is the process of arranging atoms or molecules into highly-ordered solid structures. For example, the Czochralski and Kyropoulos methods are used to produce crystals for a wide range of end uses. The crystal structures produced give materials predictable electrical, optical and mechanical properties – which is why crystals sit at the heart of technologies we use every day.

Industrially-grown crystals are found in smartphones, computers, LED lighting, medical imaging equipment and advanced sensors. Using the Czochralski method, silicon crystals form the wafers used in semiconductors. Here, a tiny seed crystal is introduced into a molten feedstock and slowly pulled upwards with the crystal growth occurring on the basal surface.

With the Kyropoulos method, sapphire crystals are grown for scratch‑resistant screens and optical windows. The seed crystal grows downwards into the molten raw material bath as temperature is slowly reduced. Both methods involved heating raw materials to exceptionally high temperatures. The molten materials can be highly corrosive and reactive, which is why specialist materials and instruments are needed.

With its high melting point (1,768°C), platinum is non‑reactive and stable, retaining its strength at extreme temperatures. These properties make it ideal for the crucibles used to hold and shape molten materials during crystal growth. Importantly, platinum does not contaminate the growing crystal, helping achieve the ultra‑high purity the end applications demand. 

At least four types of crystals that are grown using platinum crucibles are directly used to make optical interconnects for AI computing, covering functions from optical sources to signal transmission (akin to ‘nerve’ pathways). AI-related crystal growth is expected to increase demand for platinum (and iridium) crucibles.

Conclusion

AI represents a transformative force, but its success depends on physical infrastructure built from advanced materials. PGMs, with their unique properties, are integral to this foundation, playing a multifaceted role in the AI ecosystem with positive implications for PGM demand. 

*WPIC Platinum Quarterly Q1’26, May 18, 2026.


IMPORTANT NOTICE AND DISCLAIMER: This publication is general and solely for educational purposes. The publisher, The World Platinum Investment Council, has been formed by the world’s leading platinum producers to develop the market for platinum investment demand. Its mission is to stimulate investor demand for physical platinum through both actionable insights and targeted development, providing investors with the information to support informed decisions regarding platinum and working with financial institutions and market participants to develop products and channels that investors need.

The research for the period since 2019 attributed to Metals Focus in the publication is © Metals Focus Copyright reserved. All copyright and other intellectual property rights in the data and commentary contained in this report and attributed to Metals Focus, remain the property of Metals Focus, one of our third-party content providers, and no person other than Metals Focus shall be entitled to register any intellectual property rights in that information, or data herein. The analysis, data and other information attributed to Metals Focus reflect Metals Focus’ judgment as of the date of the document and are subject to change without notice. Any data contained in this report and the spreadsheets available on the World Platinum Investment Council website that is the copyright of Metals Focus and its affiliated companies must not be extracted and reproduced, distributed, transmitted or communicated to any third party without express written consent from Metals Focus. For the avoidance of doubt, this does not limit the ability to distribute the report itself or to download and use the relevant data for internal purposes.

The research for the period prior to 2019 attributed to SFA in the publication is © SFA Copyright reserved. 

This publication is not, and should not be construed to be, an offer to sell or a solicitation of an offer to buy any security. With this publication, neither the publisher nor its content providers intend to transmit any order for, arrange for, advise on, act as agent in relation to, or otherwise facilitate any transaction involving securities or commodities regardless of whether such are otherwise referenced in it. This publication is not intended to provide tax, legal, or investment advice and nothing in it should be construed as a recommendation to buy, sell, or hold any investment or security or to engage in any investment strategy or transaction. Neither the publisher nor its content providers are, or purports to be, a broker-dealer, a registered investment advisor, or otherwise registered under the laws of the United States or the United Kingdom, including under the Financial Services and Markets Act 2000 or Senior Managers and Certifications Regime or by the Financial Conduct Authority.

This publication is not, and should not be construed to be, personalized investment advice directed to or appropriate for any particular investor. Any investment should be made only after consulting a professional investment advisor. You are solely responsible for determining whether any investment, investment strategy, security or related transaction is appropriate for you based on your investment objectives, financial circumstances, and risk tolerance. You should consult your business, legal, tax or accounting advisors regarding your specific business, legal or tax situation or circumstances.

The information on which this publication is based is believed to be reliable. Nevertheless, neither the publisher nor its content providers can guarantee the accuracy or completeness of the information. This publication contains forward-looking statements, including statements regarding expected continual growth of the industry. The publisher and Metals Focus note that statements contained in the publication that look forward in time, which include everything other than historical information, involve risks and uncertainties that may affect actual results and neither the publisher nor its content providers accept any liability whatsoever for any loss or damage suffered by any person in reliance on the information in the publication.

The logos, services marks and trademarks of the World Platinum Investment Council are owned exclusively by it. All other trademarks used in this publication are the property of their respective trademark holders. The publisher is not affiliated, connected, or associated with, and is not sponsored, approved, or originated by, the trademark holders unless otherwise stated. No claim is made by the publisher to any rights in any third-party trademarks.

©2026 World Platinum Investment Council Limited. All rights reserved. The World Platinum Investment Council name and logo and WPIC are registered trademarks of World Platinum Investment Council Limited. No part of this report may be reproduced or distributed in any manner without attribution to the publisher, The World Platinum Investment Council, and the authors.

All examples in this report are hypothetical interpretations of situations and are used for explanation purposes only. The views in this report reflect solely those of the author and not necessarily those of CME Group or its affiliated institutions. This report and the information herein should not be considered investment advice or the results of actual market experience.