The Copper Appetite of a 1 GW AI Data CenterIs The Same as 500,000 Cars

Build an AI megafactory and you’ll need a copper mine to match. The latest hyperscale data centers designed to handle AI workloads are orders of magnitude larger than the server farms of a decade ago.

Older facilities might consume 5–15 thousand tons of copper in wiring, busbars, transformers and cooling equipment. Next generation AI campuses can swallow up to 50 thousand tons per site (copper.org)—roughly three times the copper content of a conventional data center.

This is why AI infrastructure is becoming a materials story as much as a digital one.

This also feeds my thesis that American sourced copper is in a privileged position right now—get those local mines and supply chains built!

How Big Is 50 Thousand Tonnes?

Think about what 50 thousand tonnes actually means:

• Equivalents: That number is 50 million kilograms of copper. Using industry standard copper content per vehicle—roughly 23 kg in a conventional internal combustion car and 83 kg in a battery electric vehicle (internationalcopper.org)—that same pile of metal could build about 2.2 million gas powered cars or roughly 600 thousand EVs. In other words, one gigawatt scale data center uses as much copper as millions of vehicles.

• Global share: World mined copper output totaled about 22 million tons in 2023 (natural-resources.canada.ca). A 50 thousand tone data center therefore represents around 0.23 % of annual global production (50,000 ÷ 22,000,000 × 100). Even at that modest share, the cumulative build out of tens or hundreds of AI campuses would add a meaningful new source of demand.

• Why so copper intensive? Running AI servers at the gigawatt scale requires HUGE and stable power flows. Copper’s conductivity makes it the material of choice for high capacity cabling, busbars and grounding systems. Its thermal properties enable high efficiency liquid cooling and heat exchangers.

And although fibre optic lines carry data over distance, short run Ethernet connections inside server racks still rely on copper. As copper.org notes, the need for robust power delivery, thermal management and networking drives the jump to ≈50 thousand tons per facility (copper.org).

Context: Copper Supply and Demand

Understanding the scale of AI driven demand requires stepping back to the global market:

• Production concentration: Chile alone supplied 5 million tons of copper in 2023—about 23 % of global output (natural-resources.canada.ca). Peru and the Democratic Republic of Congo add another 12 % and 11 %, respectively (natural-resources.canada.ca). The United States produced roughly 1.1 million tons, making it a mid tier player (natural-resources.canada.ca).
• Refined output: Global refined copper production reached 26.6 million tons in 2023, with 22.1 million tons from primary ore and 4.5 million tons recycled (natural-resources.canada.ca). Recycling already supplies nearly one fifth of refined copper, but new AI and electrification projects will require fresh supply.
• Vehicle demand: Society moving to EVs (still big, though on the downslope in the West) increases copper demand. A hybrid car uses about 40 kg of copper, while a plug in hybrid needs 60 kg and a battery electric vehicle uses ≈83 kg (internationalcopper.org).

For perspective, each gigawatt data center locks in enough copper to assemble nearly half a million EVs.

Why Investors Should Care

There’s a new convergence forming a super cycle for copper—A) AI data centers, B) EVs and C) renewable energy infrastructure.

1. AI is a new demand pillar. A single 1 GW facility demands up to 50 kt of copper (copper.org). Multiply that by dozens of hyperscale projects and you have hundreds of thousands of tons of incremental demand.
2. Supply is tight. Mined output grows slowly; global production rose from 18.4 Mt in 2014 to 22 Mt in 2023 (natural-resources.canada.ca). Bringing on new mines can take 10 years or more, and ore grades are declining in established regions.
3. Competing sectors. Electric vehicles and renewable power are already major copper consumers. A wind turbine can require 8 tons of copper per megawatt (mining.com). Grid upgrades to connect AI campuses will further strain supply.
4. Geopolitical concentration. Over 45 % of the world’s mined copper comes from just three countries (natural-resources.canada.ca). Political or operational disruptions in Chile, Peru or the DRC could tighten the market quickly.

I’m playing this trend via Faraday Copper (FDY-TSX) which is run by the legendary Lundin Group of Companies—with an already established resource in Arizona, and located right beside the Kalamazoo deposit by BHP. BHP and the Lundins have worked closely together before. A lot of new drilling will result in a new PEA—Preliminary Economic Assessment—in early 2026 that has the potential to make this an even better world class asset.

But there are other near term copper producers in Canada and the USA, such as Foran Mining FOM-TSX in Saskatchewan.

Bottom Line

AI factories are not just about chips and algorithms; they are about metal. A gigawatt scale AI data center can consume as much copper as millions of cars and still represents only a fraction of global supply—yet the cumulative effect could be disruptive.

Investors monitoring the intersection of AI, electrification and commodities should keep a close eye on copper. The metal’s unique electrical and thermal properties make it irreplaceable in both digital and green infrastructure. As hyperscale computing scales up, so does the opportunity—and the challenge—for copper producers and investors alike.

EDITORS NOTE: I’ve been looking at two emerging technology plays which both have near term catalysts–within one month from today! I think both could have huge impact on these stocks–one report is ready, another is coming soon–to profit, SUBSCRIBE HERE