Copper's AI Supercycle

There's a metal you probably don't think about.

It's in your phone. Your car. Your house. The data center that just served you this article.

It's not sexy. It's not a rare earth element with a mysterious name. It's just copper.

And right now, it might be the single most important commodity on Earth.

Here's why: every AI data center, every electric vehicle, every solar panel, every wind turbine, every grid upgrade—all of them run on copper.

And we're running out.

Not in a "we'll figure it out eventually" way. In a "the math doesn't work" way.

Global copper demand is projected to hit 30.6 million metric tons by 2026. Supply? About 26.5 million metric tons. That's a 4.1 million metric ton deficit—the largest supply gap in modern history.

And it's about to get worse.

Because AI isn't just using copper. It's devouring it. And unless something changes—fast—copper scarcity will become the bottleneck that stops the entire AI infrastructure boom in its tracks.

Let's start with a number that should make you pause: one AI-optimized data center uses approximately 2,000 to 3,000 metric tons of copper.

That's for a single facility.

Now multiply that by the hundreds of data centers being built globally to support AI workloads. Microsoft alone is planning to add 20 gigawatts of data center capacity by 2030. Amazon, Google, Meta—they're all building at similar scale.

Here's the math:

When you add it all up, global copper demand could reach 50 million metric tons annually by 2035.

Current annual production? About 26 million metric tons.

That's a 24 million metric ton gap.

There is no plan to close it.

You might be thinking: "Okay, so prices go up, miners produce more, problem solved."

If only it were that simple.

Here's the reality of copper mining in 2026:

1. Ore grades are declining.
The average copper ore grade has fallen from 1.5% copper content in 1990 to 0.6% today. That means miners have to dig up 2.5 times more rock to get the same amount of copper. More energy. More cost. More environmental impact.

2. New mines take 10-20 years to develop.
From discovery to first production, a new copper mine requires:

Even if we started a dozen new mega-projects today, they wouldn't produce meaningful copper until 2035 at the earliest.

3. Capital costs are exploding.
Building a new copper mine now costs $5-10 billion. And that's before delays, permitting battles, and cost overruns. Many projects simply aren't economically viable at current copper prices—even with prices near all-time highs.

4. Geopolitical concentration.
Over 50% of global copper refining capacity is in China. Chile and Peru control 40% of global mine production. Any disruption—trade war, political instability, labor strikes—ripples through the entire global supply chain.

5. ESG constraints.
Copper mining is water-intensive, energy-intensive, and often located in ecologically sensitive or indigenous lands. New projects face years of legal challenges and community opposition. In some jurisdictions, new copper mines are effectively impossible to permit.

The result? Supply is structurally inelastic. Demand can spike 50% in five years. Supply can't.

Copper prices hit $11,104 per metric ton in May 2024—a record high.

And then they kept climbing.

By March 2026, spot prices are hovering around $10,500-$11,000, with analysts projecting prices could reach $12,000-$15,000 by 2028 if supply constraints persist.

Why? Because the deficit is structural, not cyclical.

This isn't a temporary supply disruption that will resolve when a mine comes back online or a strike ends. This is a decade-long mismatch between exponential demand growth (driven by AI, EVs, and renewables) and linear supply growth (constrained by geology, permitting, and capital).

Goldman Sachs calls copper "the new oil". Bank of America projects a cumulative supply deficit of 10 million metric tons by 2030.

And here's the kicker: higher prices won't fix this quickly.

In oil markets, higher prices bring new supply online within 6-12 months (shale drilling, OPEC spare capacity). In copper, higher prices take 10-15 years to translate into meaningful new production.

The lag is everything.

Let's bring this back to data centers.

If copper prices double from $10,000 to $20,000 per metric ton, what happens?

But here's the deeper problem: at some point, you can't just pay more.

If copper becomes scarce enough, it doesn't matter what you're willing to pay. You physically can't get it.

We're already seeing early signs:

This is what scarcity looks like in real time.

Here's my framework again:

Code is abundant. Atoms are scarce.

AI models can scale infinitely. You can train GPT-7, GPT-8, GPT-100. The software has no physical limit.

But every additional model, every additional inference, every additional user—requires atoms.

And copper is the ultimate atom bottleneck.

You cannot:

The digital economy is hitting a hard physical constraint.

And unlike software bugs, you can't patch geology.

This isn't just a supply chain story. It's a wealth transfer story.

Winners:

1. Copper miners (Freeport-McMoRan, BHP, Southern Copper, Glencore)
If copper prices stay elevated for a decade, these companies will print money. Their existing mines become cash machines. Their undeveloped reserves skyrocket in value.

2. Copper-rich countries (Chile, Peru, Democratic Republic of Congo, Zambia)
Resource nationalism is already rising. Expect higher royalties, export taxes, and state ownership stakes. Copper is the new oil—and producing countries know it.

3. Recycling and urban mining companies
As primary copper becomes scarce, secondary copper (from recycling) becomes more valuable. Companies that can efficiently extract copper from e-waste, old buildings, and retired infrastructure will thrive.

4. Substitution technology developers
Aluminum wiring, graphene conductors, advanced ceramics—any technology that can reduce copper intensity in data centers, EVs, or grids will attract massive investment.

5. First-movers in AI infrastructure
Companies that secured copper supply chains early (or built data centers before 2024) have a structural cost advantage. Latecomers will pay 2-3x more for the same infrastructure.

Losers:

1. Late-stage data center developers
If you're trying to build a hyperscale data center in 2026-2028, your copper costs just doubled. Your competitors who built in 2023-2024 have a permanent cost advantage.

2. Utilities in slow-moving states
If you can't secure copper for grid upgrades, you can't connect new load. Data centers will go elsewhere. Economic development follows.

3. EV manufacturers without secured supply chains
Tesla, BYD, and other EV leaders locked in copper supply years ago. Smaller EV startups? They're competing for scraps in a tight market.

4. Developing countries
Rich countries will outbid poor countries for scarce copper. Infrastructure projects in Africa, South Asia, and Latin America will be delayed or canceled as copper flows to AI data centers in the U.S., Europe, and China.

5. Consumers
Higher copper prices flow through to everything: electricity bills, EV prices, home construction costs, consumer electronics. Inflation isn't over.

Copper scarcity doesn't just raise prices. It changes behavior.

Here's what I'm watching:

1. Data center design shifts
Engineers are already redesigning data centers to use less copper per megawatt. Expect:

2. Onshoring of copper refining
The U.S. currently refines almost no copper domestically. Expect massive investment in domestic refining capacity to reduce dependence on China.

3. Geopolitical competition for copper assets
China has spent the last decade buying copper mines in Africa and Latin America. The U.S. and Europe are now scrambling to catch up. Expect more state-backed investment in mining.

4. Acceleration of deep-sea mining
The ocean floor contains vast copper deposits. Environmental concerns have blocked deep-sea mining for years. But if land-based supply can't meet demand, expect political pressure to open the ocean floor.

5. Strategic copper reserves
Just as countries hold strategic petroleum reserves, expect governments to start stockpiling copper. It's too critical to leave to market forces alone.

This is a macro trend that will play out over years. But here's what you can do now:

1. If you're an investor:

2. If you're building infrastructure:

3. If you're in policy:

4. If you're a consumer:

AI is not a software revolution. It's a materials revolution.

And the most critical material—the one that makes everything else possible—is copper.

We need 24 million more metric tons per year by 2035. We have no plan to produce it. New mines take 15 years. Recycling can't close the gap. Substitution is limited.

The math doesn't work.

So what happens?

One of three things:

1. Prices spike until demand is rationed.
Some projects don't get built. Some countries get outbid. AI infrastructure growth slows.

2. We find radical new supply sources.
Deep-sea mining. Asteroid mining (seriously). Breakthrough extraction technologies. Geopolitical deals that unlock stranded resources.

3. We redesign everything to use less copper.
New materials. New architectures. New transmission technologies. Efficiency gains that reduce copper intensity by 50-70%.

My bet? All three happen simultaneously.

Prices will spike. New supply will come online (slowly). Innovation will reduce copper intensity.

But in the meantime—the next 5-10 years—copper scarcity will be the defining constraint on AI infrastructure.

Not compute. Not capital. Not talent.

Copper.

The atoms matter. And we don't have enough of them.

Precision in a world of noise.

Analysis by Slone Sterling