AI industry demand sends ruthenium prices to a record high and raises shortage risk

Ruthenium, a metal used in the production of hard drives, has risen to an all-time price high. The main driver is the rapid growth in demand from AI infrastructure, while supply remains tight and may struggle to keep up with the market as early as 2026.

Why Demand Has Grown

In the commodities market, this is a rare situation: a metal that is not at the center of mainstream attention suddenly becomes critically important for an entire technology supply chain. That is exactly what is happening with ruthenium now. Against the backdrop of new data center construction and the expansion of storage for training and serving AI models, demand has grown for components without which such infrastructure cannot scale.

While attention is usually focused on chips and accelerators, physical data storage remains an indispensable part of the AI economy. For the industry, this is an unwelcome signal, because the growth in demand looks not like a one-off spike but a continuation of a longer trend. Companies investing in AI services are simultaneously scaling up computing, redundancy, and archives — meaning they are purchasing more storage devices and associated materials.

When one of those materials becomes scarce, pressure quickly travels up the chain: from component manufacturers to system assemblers and then to large enterprise customers.

Where the Deficit Risk Comes From

The problem is that ruthenium is almost impossible to bring to market quickly in large volumes. It is obtained not as a standalone metal but as a byproduct of mining other platinum-group metals. This makes supply inflexible: even if the price rises sharply, mining companies cannot simply open separate capacity dedicated specifically to ruthenium.

Additional risk is created by the geography of supply: the market is heavily dependent on South Africa, and any delay, disruption, or decline in output there immediately affects the global balance. Because of this market structure, a deficit can emerge not gradually but abruptly. If demand from AI and data storage persists while supply remains at current levels, even a small additional volume of consumption can sharply push the price higher.

Unlike larger commodity markets, there is almost no flexibility buffer here: if industry needs the metal right now, there is simply nothing to quickly close the gap between supply and demand.

• ruthenium cannot be rapidly scaled as a standalone mining operation
• key volumes are tied to a single region
• storage device manufacturers will compete for limited raw materials
• price increases quickly flow through into equipment costs

Who Will Pay More

The first to feel the pressure will be hard drive manufacturers and storage system suppliers, especially those serving large data centers. For them, rising ruthenium prices are not an abstract commodity story but a risk of higher costs for specific components and margin compression. If suppliers cannot pass the cost increase on to customers immediately, some contracts will become less profitable.

If they can, the costs will move up the chain — into the price of server infrastructure, backup storage, and large-scale AI projects. This does not mean the entire storage market will become uniformly more expensive overnight. But if a deficit does materialize in 2026, companies will start acting pre-emptively: revising procurement, locking in volumes through long-term contracts, and more carefully calculating the economics of storage.

For businesses building AI products, this is yet another reminder: infrastructure costs are not determined solely by GPU prices. A bottleneck can appear in a less conspicuous material and unexpectedly hit the budget.

What This Means

The ruthenium story illustrates just how dependent the AI boom is on materials that are rarely discussed outside the industry. When demand for computing grows faster than the extraction of rare materials, risk shifts away from laboratories and models to the entire supply chain — from storage devices to the economics of data centers.