The AI Energy Debate Is Being Measured in the Wrong Unit

The AI energy debate is being measured in terawatt-hours. It should be measured in tonnes of copper.

Every data centre, and every grid upgrade built to feed it, runs on copper. A kilowatt-hour assumes the wire already exists. Building the wire is the hard part, and it starts at a mine.

So Pulse Intelligence ran it across every primary copper asset we track.

• 2,598 are in exploration
• 293 have advanced to development
• 31 are in construction
• 18 have a final investment decision

That is the whole global pipeline, by stage. From 2,598 companies looking for copper to 18 that are actually funded to build it.

Copper has run roughly 62% since the start of 2024. Demand is already pricing the squeeze. But demand cannot draw from exploration. It can only draw from the projects clearing the build gate, and right now that is 18.

The geography makes it tighter. Most of the forward pipeline sits in Australia and Canada. Tier-one, but slow to permit. Today's production leans on Chile, Peru and the DRC. We are looking for tomorrow's copper in the jurisdictions that take the longest to build.

The usual answer is recycling and substitution. Both matter. Neither changes how fast a new mine reaches production, and that is the number that is stuck.

Worth watching: how many of those 18 pour first metal on schedule, and how many of the 293 reach FID before the AI buildout needs the metal. Less searching. More strategising.™