Beyond the chips: The materials behind Singapore’s AI infrastructure

AI infrastructure depends on group metals that advance data storage and semiconductors.

Artificial intelligence (AI) may be digital, but the infrastructure behind it is physical. Data must be stored, semiconductors must process it, and the hardware supporting these systems must perform reliably as computing requirements continue to rise.

For Singapore, with its established semiconductor and advanced manufacturing base, this physical side of AI is an important part of the technology story. Greater computing capability depends not only on chip design and system architecture, but also on advances in the materials used to manufacture the components behind them.

Amongst these materials, precious metals — and particularly platinum group metals (PGMs) — perform highly specialised functions in data storage, semiconductor manufacturing, and testing. Their magnetic, electrical, chemical, and mechanical properties allow them to meet demanding requirements that change as technologies advance.

Their value, however, extends beyond what they enable during use. PGMs can also be recovered and refined after manufacturing, creating a broader materials lifecycle that connects technological performance with the effective use of valuable resources.

Looking at AI through these materials' lenses reveals an important part of the infrastructure that often goes unnoticed.

More data, more demanding materials
AI applications generate enormous amounts of data, making storage such as hard-disk drives (HDDs) an essential part of AI infrastructure.

As HDD manufacturers increase storage density, the requirements placed on magnetic recording materials also evolve. This can be seen in the transition from perpendicular magnetic recording (PMR) toward technologies such as heat-assisted magnetic recording (HAMR).

Precious metals perform specific functions in these recording technologies. In PMR media, ruthenium is used in layers within the recording medium, whilst platinum is a component of magnetic alloys such as CoCrPt used in recording layers. These magnetic alloys can also contain oxides that contribute to the granular structure of the recording medium.
HAMR introduces different material requirements. 

As recorded areas become smaller, magnetic materials need sufficient stability to retain information reliably. HAMR uses localised heating during writing, enabling materials with greater magnetic stability to be used at these smaller dimensions. Iron-platinum (FePt) is an important material for this approach.

The evolution from PMR toward HAMR therefore illustrates a broader relationship: as AI and digitalisation drive greater storage requirements, recording technologies advance, and the materials supporting those technologies must advance with them.

Turning advanced materials into functional thin films
The performance of these materials also depends on how they are incorporated into devices. In magnetic recording media, sputtering targets provide an important link between material design and the thin films that ultimately perform the required function.

During sputtering, material from a solid target is deposited onto a substrate to form a thin film. As recording technologies evolve, targets may combine magnetic alloys, precious metals, and oxides in increasingly specialised compositions.

Producing such targets requires precise control of their composition and material distribution. Sintering is one method used to manufacture these targets, consolidating powders containing the required materials into a solid target. This is particularly important for targets combining precious-metal alloys and oxides, where material uniformity contributes to stable sputtering and consistent film formation.

The important point extends beyond any generation of HDD technology. As device technologies evolve, materials suppliers must translate changing performance requirements into the composition, structure, and form of the materials they provide.

More computing, new semiconductor materials
Growing computing requirements are driving advances in semiconductor devices. This is where precious metals can support semiconductor manufacturing at different stages — from materials being evaluated for advanced device structures to established materials used in semiconductor testing. Ruthenium, for example, is considered for applications including advanced interconnects and electrodes.

At another stage of semiconductor manufacturing, precious metals have established roles in testing. Probe pins must repeatedly make reliable electrical contact with semiconductor devices whilst maintaining durability through repeated physical contact. Palladium- and rhodium-based materials can provide combinations of electrical conductivity, hardness, and wear resistance suited to these applications.

The materials story continues after manufacturing
PGMs’ importance in advanced manufacturing extends beyond their use to what happens after manufacturing.

Not every gram of precious metal introduced into a manufacturing process becomes part of the finished product. During thin-film deposition processes such as sputtering, precious metals can remain in used targets or accumulate on components and fixtures inside deposition equipment.

Although these materials may no longer serve their original manufacturing purpose, the precious metals they contain still have value. They can be collected, recovered, and refined for further use.

For advanced electronics applications, however, recovery alone is not enough. Semiconductor and data-storage applications can require tightly controlled purity and composition. Recovered precious metals therefore need to be appropriately refined, analysed, and quality-controlled before they can be considered for another demanding application.

Precious-metal recycling in advanced manufacturing is therefore as much a materials challenge as a waste-management challenge. The amount recovered matters, but so do the quality and purity of the refined metal and whether it can meet the requirements of its next application.

Performance and circularity: One materials story
These examples show how closely AI infrastructure is tied to advances in materials.
Growing demand for computing and data storage is raising hardware requirements, driving advances in recording technologies and semiconductors — and, in turn, the materials that enable them.

For high-value materials such as PGMs, beyond manufacturing, recovering, and refining these materials create the possibility of returning valuable resources to productive use, provided they can once again meet the demanding purity and quality requirements of advanced applications.
This perspective is particularly relevant to Singapore and the wider region as AI, semiconductor manufacturing and digitalisation become increasingly interconnected.

As AI infrastructure continues to develop, material performance and resource efficiency will need to be considered together. The precious metals used in advanced computing applications must meet increasingly demanding technical requirements, whilst their value makes effective recovery and reuse equally important.