Mario Stifano is chief executive of Galantas Gold, a gold and copper company listed on the TSX Venture Exchange in Toronto and on AIM in London. He has more than 25 years of experience across exploration, development and producing mining companies. Galantas is developing the Andacollo gold project in Chile, a former producing mine, and the Indiana gold-copper project.

We asked him how the AI and data-centre build-out is changing demand for gold, copper and new mines.

Gold's role in technology

Gold has traditionally been viewed as a store of value, but it also plays an important role in advanced electronics and computing. As AI and digital infrastructure continue to expand, how do you see gold's role in the technology sector evolving?

Most people picture gold in a vault or in jewellery. Fewer think about the gold inside the devices they use every day. Engineers have relied on it for decades because it conducts electricity well, doesn't corrode, and can be made into extremely thin wires and coatings. When a connection has to work every time, those properties matter.

I expect that role to continue as computing becomes more demanding. But I wouldn't assume every increase in AI investment means the same increase in gold demand. Manufacturers are always looking for ways to use less of an expensive material.

You can see that tug-of-war in the numbers. The World Gold Council reported about 323 tonnes of technology demand in 2025, essentially flat on 2024. AI applications supported demand, but thinner wires, leaner designs and weaker consumer electronics offset that growth. So gold can become more important in specific high-end components without a dramatic jump in total tonnage.

Data centres and metals demand

The rapid expansion of AI data centers is driving demand for copper and other essential materials. How is this infrastructure boom reshaping demand for metals, and where does gold fit into the broader picture?

Start with the electricity. A data centre needs reliable power every hour it operates. That means looking past the servers to the substations, cables, transformers and generating capacity behind them.

The IEA's latest analysis found that data-centre electricity use grew 17% in 2025. It expects that use to roughly double by 2030, with AI-focused facilities growing faster still. That gives you a sense of the physical build-out that companies are planning.

Copper runs through that whole electrical system. And AI is adding demand at the same time as utilities, transport and industry are investing in electrification. The mining industry has to plan for all of it together.

Gold plays a different role. It's used selectively in electronics, where reliability justifies the cost, while copper's exposure is spread across the physical infrastructure. For gold, technology matters, but it is still one part of a market driven mainly by investment, central banks and jewellery.

Supply chain bottlenecks

As tech companies invest billions in AI infrastructure, could growing demand for copper and other critical materials create supply chain bottlenecks that ultimately slow down technological progress?

Yes. The technology industry and the mining industry work on very different timelines. A commitment to spend billions on AI infrastructure doesn't mean the materials will be there when construction teams need them.

S&P Global found that mines starting production between 2020 and 2024 took almost 18 years on average from discovery. Before construction even begins, a company has to drill the deposit, prove its economics, complete engineering, obtain approvals and secure financing.

For copper, the pressure goes well beyond AI. Data centres compete for supply with electricity networks, electric vehicles and other industries. The IEA's 2026 outlook estimates that, under today's policies, existing and announced projects could leave primary copper supply about 25% short of requirements in 2035.

Closing that gap takes more than new discoveries. It means financing projects that can actually deliver, expanding and restarting suitable operations, and recovering more from ore already being processed. Otherwise, technology companies face higher costs, longer equipment lead times and delayed projects.

That's why I put so much weight on work that has already been done. Our Andacollo gold project in Chile is a former producing mine. We still have refurbishment, technical and financing work ahead of us, but we're not starting from a blank page.

AI in gold exploration

AI is increasingly being used to analyze geological data and identify mineral deposits. How could these technologies transform gold exploration, particularly when it comes to discovering resources that traditional methods may have overlooked?

One opportunity that really interests me is getting more out of work that has already been done.

An established mining district can hold decades of drilling, surveys and geological records. Different companies collected that information at different times, often looking for different things. There's real value in bringing it together and asking questions that were hard to investigate before.

Canada is backing exactly this idea, with a federal and territorial pilot in the Northwest Territories that uses AI to scan and analyse drill core already sitting in government storage.

The majors are doing it too. Agnico Eagle works with GeologicAI, combining AI with drill-core scanning to make its resource models more complete and reliable.

At Galantas, our team in Chile spends a lot of time reading drill core and surface geology, and we share some of that work through our Field Notes video series. Technology can help, but experienced geologists still have to understand the core.

And you still have to drill. A model can suggest where to look; it can't prove you have an economic deposit. For me the test is simple: does it help the team make a better decision about the next drill hole? Exploration capital is limited, so better targeting makes a real difference to how far that capital goes.

New uses for gold in AI hardware

Gold is valued in high-reliability electronic components for its conductivity and resistance to corrosion. Could advances in AI hardware, semiconductors, and high-performance computing create new opportunities for gold, or will its role remain relatively specialized?

Specialised doesn't mean unimportant. A component can contain very little gold and still depend on it to perform reliably.

There are new applications worth watching. One is co-packaged optics, which places optical connections right next to the processor so AI systems can move data faster and use less power. Some of the optical components and high-frequency electrodes in those designs use gold.

The World Gold Council sees it as a potentially small but meaningful new source of demand if it's widely adopted in AI data centres. It's still early.

I'd leave the engineering judgments to the chip makers. From a mining perspective, I want to know three things. Is the technology being adopted commercially? How much material does it need? And can something cheaper do the same job?

There will always be pressure to use less gold because of its price. But where it offers a clear reliability advantage, it earns its place. That points to a specialised role with room to grow, not a story where AI suddenly transforms the whole gold market.

Grids, copper and realistic timetables

Building the next generation of AI data centers requires massive investments in electricity grids and power distribution systems, where copper plays a critical role. Are technology companies underestimating their dependence on physical resources, and what happens if supply cannot keep pace?

I wouldn't assume the technology companies don't understand the dependency. The better question is whether everyone in the chain is working to a realistic timetable.

A server can be ordered. A grid connection may need new lines, approvals and equipment that can't be delivered right away. The IEA has estimated that, unless these constraints are addressed, around 20% of planned data-centre projects worldwide could be at risk of delay. That's mainly a grid problem rather than a copper shortage, but it shows how physical the AI build-out really is.

The sulphuric acid problem

The supply chain also runs deeper than most people realise. In its 2026 outlook, the IEA warned that the near-term copper supply picture has worsened considerably, partly because of tight supplies of sulphuric acid. Many copper operations need that reagent to process their ore, and Chile, where we operate, is one of the major producers that use the process. Most people in tech have never thought about sulphuric acid, but it can affect how much copper reaches the market.

If supply can't keep pace, companies will delay projects, pay more or build somewhere else. Long-term purchasing commitments and earlier conversations with suppliers help those suppliers expand with confidence. In mining, we know a good asset is only part of the equation; you also need power, infrastructure and financing. Digital infrastructure is no different.

Machine learning and mining economics

How are machine learning, automation, and predictive analytics changing the economics of gold exploration and production, and could these technologies help smaller companies compete more effectively with established industry players?

The biggest opportunity is getting more value from the money we already spend. Drilling is expensive. Equipment downtime is expensive. Losing recoverable gold during processing is expensive. Those are specific problems technology can help address.

In exploration, better analysis of historical drilling and geological information can help a team decide where to drill next. It doesn't guarantee a discovery, but better targeting can help a company make more effective use of its exploration budget.

Recovery and maintenance

In production, consider recovery: the percentage of metal in the ore that the plant successfully extracts. In a simplified example, if the material processed contains 100,000 ounces of gold, improving recovery from 90% to 92% means recovering another 2,000 ounces from the same feed. You still have to account for the cost of achieving that improvement, but you haven't had to mine additional tonnes to access those ounces.

There are practical examples of this approach. BHP and Microsoft have described using real-time plant data and AI recommendations at Escondida to help operators adjust processing conditions with the aim of improving copper recovery.

Predictive maintenance tackles another cost. Identifying a developing equipment problem can allow a repair during a planned shutdown instead of an unexpected failure that interrupts production.

For smaller companies, access to specialist software and service providers can put better analytical tools within reach. But the tools still need reliable data and experienced people. Large operators retain advantages in both.

I would start with a problem we can measure, test the technology and compare the benefit with its full cost. A smaller company doesn't need to match a major miner's technology budget. It needs to spend its own budget effectively.

Recycling and new supply

As the technology industry looks for more sustainable supply chains, could advances in electronic waste recycling, AI-driven material recovery, and more efficient production reduce the need for newly mined gold and copper?

Yes, to a degree. Recovering more metal from products we've already made makes sense. The question is how much can be collected and recovered economically, and when it becomes available.

The UN's Global E-waste Monitor found that only 22.3% of the e-waste generated in 2022 was documented as formally collected and recycled. That's measured by weight, not the recovery rate for any one metal, but it shows how much room there is to improve.

AI-assisted sorting and better processing can help. They still need collection networks, suitable facilities and a business model that covers the cost.

Timing matters too. Recycling depends on products reaching the end of their lives, and copper installed in a grid today may stay in service for decades.

So I see recycling and mining as complementary. More recycling improves supply security and reduces how many new mines we need. It doesn't remove the need to invest in mining.

What matters most by 2030

Looking toward 2030, which developments do you believe will have the greatest impact on the future of gold and critical metals: AI-powered exploration, rising demand from data centers, or the challenge of bringing new mineral supplies online?

For the period to 2030, I'd put delivering supply first. I also think it's the one the market gives the least weight.

A discovery made today might be very important, but it's unlikely to produce metal before 2030. That makes the projects already being permitted, financed and built far more relevant this decade. So is what existing operations can add through expansions, better recovery and sensible restarts.

AI will help with those decisions. I expect much of its near-term value to come from understanding known assets better and helping operations run more consistently.

But the asset still has to work. It needs the right geology and metallurgy, a realistic development plan, financing and the ability to operate responsibly. Better software doesn't change that.

I'd also separate gold from copper. Data centres are part of copper's wider infrastructure story. For gold, technology plays a role, but monetary conditions, investment demand and central-bank buying will matter more.

The question I keep coming back to is simple: which projects can actually deliver, at what cost, and when?