Artificial intelligence is driving one of the largest infrastructure buildouts the technology industry has ever seen.

Hundreds of thousands of GPUs. Massive new data centers. New power generation. Millions of miles of cabling. Fiber-optic networks. Semiconductor manufacturing.

But there’s a problem.

All of that infrastructure requires physical materials — and some of the metals required to build it are already facing serious supply constraints.

At the IPMI’s 50th Annual Conference, Ed and Rich sat down with Matt Watson, CEO of Precious Metals Commodity Management LLC, to discuss how the AI boom could reshape demand for copper, platinum group metals, fiber optics, electronics, and recycled precious metals.

AI Has a Copper Problem

When discussing commodities that could become constrained by the AI infrastructure boom, Matt doesn’t hesitate:

Copper is one of the biggest concerns.

Copper is already essential to the modern electrical system.

It is used extensively in:

• Electrical wiring
• Power generation
• Transmission infrastructure
• Data centers
• Electronics
• Motors and transformers
• Electric vehicles
• Renewable energy systems

And AI adds another enormous source of demand.

Building data centers isn’t simply about purchasing GPUs. Those facilities have to be connected to the electrical grid, cooled, wired, and supplied with enormous amounts of electricity.

That means considerably more physical infrastructure.

Data Centers Are Becoming Enormous

Modern AI data centers operate on a completely different scale than traditional computing facilities.

Matt points to massive AI installations containing hundreds of thousands of GPUs.

Those GPUs have to communicate with one another extremely quickly.

Originally, copper connections were considered for many of these systems.

But as AI clusters grow larger and performance requirements increase, fiber optics are becoming increasingly important.

That creates another commodity story entirely.

Fiber Optic Demand Could Explode

AI companies need extremely high-speed connections between GPUs, servers, data centers, and networking infrastructure.

Fiber optics can provide the bandwidth necessary to move enormous quantities of data while reducing some of the limitations associated with traditional copper connections.

According to Matt, companies in the fiber-optic industry are preparing for extraordinary growth.

Some manufacturers are evaluating whether capacity may need to increase by 10x or even 20x over the next several years.

And that matters to the precious-metals industry because manufacturing fiber-optic glass requires specialized equipment containing platinum group metals.

The Platinum Connection to Fiber Optics

Producing high-quality glass and fiber-optic materials requires equipment capable of operating in extremely demanding environments.

Platinum group metals are uniquely suited to these applications because of their resistance to corrosion, high temperatures, and chemical attack.

Matt specifically highlights platinum and rhodium used in glass-manufacturing bushings.

These components aren’t necessarily consumed immediately.

Instead, they can effectively become long-term inventories of precious metals tied up inside manufacturing equipment.

As fiber-optic production capacity expands, manufacturers may therefore require additional PGM inventory simply to build and operate the equipment needed to meet demand.

Iridium Could Benefit Too

The AI infrastructure story isn’t limited to platinum and rhodium.

Matt also discusses growing applications for iridium in optical and advanced technology systems.

Iridium is already used in highly specialized industrial applications because of its extreme durability, temperature resistance, and corrosion resistance.

As semiconductor, optical, display, and other advanced manufacturing industries expand, these niche applications can become increasingly important.

That’s particularly significant because iridium is produced in extremely small quantities compared with major industrial metals.

Even relatively small increases in demand can therefore have an outsized effect on a thin market.

AI Is Changing the Power Grid

The physical data center is only part of the AI infrastructure problem.

The bigger challenge may be electricity.

AI computing requires enormous amounts of power.

Matt argues that the scale of planned data-center development could require the United States to dramatically expand power-generation capacity over the coming years.

And some developers aren’t waiting for traditional grid infrastructure.

They’re building generation directly alongside data centers.

Data Centers Are Building Their Own Power

New AI facilities are increasingly exploring what is sometimes called behind-the-meter generation.

Instead of relying exclusively on the existing electrical grid, data-center operators can install dedicated power generation.

That could include:

• Natural-gas turbines
• Fuel cells
• Diesel backup generators
• Dedicated power plants
• Other distributed generation systems

This creates another layer of commodity demand.

Building a data center suddenly isn’t just about servers.

You’re effectively building a technology facility, electrical substation, cooling system, communications network, and sometimes an entire power plant at the same location.

The Infrastructure Buildout Is Already Happening

It’s easy to look at AI infrastructure forecasts and assume that many of the proposed projects will never actually get built.

But Matt points to construction activity already occurring across numerous sites.

Companies are moving quickly because the competitive pressure surrounding AI is enormous.

In some cases, computing equipment and temporary infrastructure are being deployed before permanent buildings are even completed.

That urgency puts additional pressure on already-constrained supply chains.

Hard Drives Are Getting an Unexpected AI Boost

AI is also creating surprising demand in older technologies.

One example is the traditional spinning hard drive.

The technology has existed for decades, and industry volumes had been declining as solid-state storage became more common.

Then AI arrived.

AI models generate and consume enormous quantities of data.

That data needs to be stored.

Large-scale data centers can therefore require millions of high-capacity hard drives.

That matters to the PGM industry because hard drives historically use metals including platinum and ruthenium.

Ruthenium Demand Goes Beyond Hard Drives

While hard drives remain an important application, Matt says the current ruthenium story involves additional technologies.

Potential growth areas include:

• Advanced semiconductor nodes
• MEMS sensors
• MRAM and other memory technologies
• Specialized electronics
• High-performance computing applications

These markets are especially important because ruthenium is a relatively small PGM market.

A meaningful new electronics application doesn’t need to consume millions of tons of material to affect the supply-and-demand balance.

Palladium Needs New Demand

One of the more complicated PGM stories involves palladium.

Historically, automotive catalytic converters have represented one of the largest sources of palladium demand.

But as vehicle technology changes and electric vehicles gain market share, long-term automotive palladium demand faces pressure.

That creates an important question for the PGM industry:

Where does the next major source of palladium demand come from?

Electronics and AI infrastructure could provide part of the answer.

Palladium in Electronics

Palladium has a long history in electronics manufacturing.

Matt highlights two areas in particular:

Lead frames and multilayer ceramic capacitors (MLCCs).

MLCCs are tiny electronic components used throughout modern electronics.

They’re found in everything from vehicles to telecommunications equipment and high-performance computing systems.

Historically, MLCC manufacturing consumed significantly larger amounts of palladium.

Manufacturers eventually reduced palladium usage and substituted other materials wherever possible.

But palladium remains useful in applications requiring extremely high reliability.

AI hardware could increase the number of those high-reliability applications.

Could Palladium Replace Gold?

High gold prices naturally encourage electronics manufacturers to look for alternatives.

Palladium has some characteristics that make it attractive for certain electronic applications.

However, semiconductor manufacturers are extremely cautious when changing materials.

Reliability and purity are critical.

A cheaper material doesn’t automatically become a substitute simply because its price is lower.

That means palladium could gain incremental electronics demand without necessarily replacing gold across the semiconductor industry.

The Palladium Surplus Problem

Even with additional electronics demand, Matt remains concerned about palladium’s long-term supply-and-demand balance.

Catalytic converter recycling will continue returning significant quantities of palladium to the market.

At the same time, future mining projects may also produce palladium-heavy PGM baskets.

That combination could create substantial palladium supply unless new applications absorb it.

For the precious-metals industry, developing additional palladium demand could become increasingly important.

What This Means for Auto Catalyst Recyclers

The changing PGM market has direct implications for catalytic converter recyclers.

Historically, recyclers could sometimes benefit from holding recovered metals while prices increased.

But holding metal is ultimately a market position.

Matt specifically warns that palladium could face substantial downward pressure if automotive demand weakens while recycled and mined supply remains available.

That means recyclers need to think carefully about the difference between running a recycling business and speculating on future metal prices.

Rhodium Is Different

Rhodium faces its own challenges.

It is heavily connected to automotive catalyst demand, but it is also one of the smallest and thinnest precious-metal markets.

That means relatively small changes in supply or demand can produce dramatic price movements.

High rhodium prices also encourage manufacturers to reduce loadings wherever possible.

However, lower prices can potentially unlock additional industrial applications.

Glass manufacturing is one example.

Manufacturers may want to increase rhodium content in certain equipment because of its performance characteristics, but extremely high prices make that economically difficult.

If prices decline enough, some of that demand could return.

Recycling Becomes More Important as Mines Age

One of Matt’s biggest long-term themes isn’t AI.

It’s recycling.

Many major PGM mines have been operating for decades.

As those mines age, maintaining production becomes increasingly difficult.

Ore bodies become harder to access.

Grades can decline.

Capital requirements increase.

Eventually, some operations reach the end of their economic lives.

Matt expects primary PGM production to face increasing pressure toward the end of the decade.

That means the industry may increasingly depend on metal that has already been mined.

The World’s Largest PGM Mine May Already Exist

It’s not an underground mine.

It’s the enormous inventory of platinum group metals already sitting inside:

• Vehicles
• Electronics
• Industrial equipment
• Hard drives
• Catalytic converters
• Glass-manufacturing equipment
• Medical devices
• Chemical processing systems

Those metals don’t disappear when the original product reaches the end of its useful life.

They become potential secondary supply.

That’s why recycling could become increasingly strategic.

AI Could Accelerate the Circular Precious-Metals Economy

The AI boom creates an interesting cycle.

AI infrastructure increases demand for specialized materials.

That demand increases pressure on limited primary resources.

Limited primary resources make recycling more economically and strategically important.

Recovered metals can then return to manufacturers producing the next generation of electronics and infrastructure.

For metals such as platinum, palladium, rhodium, ruthenium, and iridium, secondary recovery may become an increasingly important part of the global supply chain.

The Bigger AI Supply Problem

The AI race is often discussed as a competition over computing power.

Who has the most GPUs?

Who has the best models?

Who can build the largest data center?

But beneath that digital competition is an enormous physical supply chain.

AI requires copper.

It requires fiber optics.

It requires power plants.

It requires semiconductors.

It requires specialized electronics.

And many of those systems require small quantities of extremely rare metals.

Software may be digital.

The infrastructure supporting it isn’t.

The Bottom Line

Artificial intelligence could become one of the biggest technology investment cycles in history.

But scaling AI requires much more than GPUs.

It requires an enormous expansion of the physical infrastructure supporting computing.

Copper may face major supply pressure.

Fiber-optic manufacturing could expand dramatically.

Platinum and rhodium could benefit from glass-manufacturing equipment.

Ruthenium and iridium could see additional demand from advanced electronics and optical technologies.

And palladium desperately needs new applications as its traditional automotive demand faces long-term pressure.

For the precious-metals industry, that creates both challenges and opportunities.

The next major source of PGM demand may not come from catalytic converters or hydrogen.

It may come from the infrastructure being built to power the AI revolution.

Stay Informed: Explore More Insights and Exclusive Content from noble6!

Thank you for joining us in exploring the latest insights from the precious metals industry. For more in-depth articles and exclusive interviews, visit the rest of our blog. Don’t forget to subscribe to our YouTube channel for the latest videos and updates. Stay informed and ahead in the industry by subscribing today!