Over the last couple of years, analysts and commentators have been pointing out a surprising reality: artificial intelligence (AI) is set to drive much higher demand for copper than most people realise. Yet this idea is usually mentioned only briefly, without any explanation of why it matters.
For most people, the link between AI and a reddish industrial metal is far from obvious. After all, AI feels intangible. It lives “in the cloud”. Copper, by contrast, feels physical, old-fashioned, and rooted in the industrial age.
So why does the spread of AI increase demand for copper? And why is this not a short-term effect, but something structural?
To understand that we need to look at what powers AI in the real world.
The first misunderstanding is the idea that AI is mostly software. Software matters, of course, but modern AI only functions thanks to massive amounts of physical hardware.
Large AI models are trained and run inside vast data centres filled with real machines sitting in racks, drawing power from the grid. These machines consume enormous amounts of electricity, generate intense heat, and must constantly send information back and forth at extraordinary speeds. Supplying power, removing heat, and moving data all depend on the same thing: copper.
At the centre of these machines are specialised computer chips called graphics processing units, or GPUs.
Most people are familiar with a central processing unit, or CPU - the main processor in a computer that handles everyday tasks one after another. A GPU was originally designed for graphics, such as rendering images and video on a screen.
What makes a GPU different is not that it is “smarter”, or even faster at everything, but that it can work on many small jobs at the same time. Instead of doing one calculation and then moving on to the next, a GPU performs thousands of similar calculations simultaneously.
This turns out to be exactly what modern AI needs. Training an AI model means repeating the same kinds of calculations again and again - not once or twice, but billions or even trillions of times. GPUs are built for this kind of repetition at scale, which is why today’s AI data centres are filled with racks of GPUs rather than traditional CPUs.
Power: Driving the Machines
Now to the key point: all that simultaneous work requires a lot of electricity.
A single high-end AI GPU can draw several hundred watts of power - roughly the same as a small household appliance. When thousands of them are packed together in one data centre, the electrical demand quickly becomes enormous. Wherever large amounts of electricity flow, copper inevitably follows.
Copper is not used in AI because it is fashionable or rare. It is used because it is one of the most effective materials we have for moving electricity efficiently, reliably, and safely.
Every GPU needs a constant flow of power. That electricity enters a data centre from the grid, passes through transformers and switchgear, runs along thick power cables, and is distributed across boards and backup systems before it ever reaches a single chip. At every step along that journey, copper is doing the work.
As AI workloads grow, data centres are being redesigned to deliver far more power in a confined space - much more electricity per square metre than was ever required before. Meeting that demand does not happen through software updates or clever optimisation. It happens through thicker cables, additional conductors, and heavier electrical infrastructure. In other words, more copper.
This is not a marginal change. AI-focused data centres use multiples of the copper found in traditional data centres, simply because the physics of moving large amounts of electricity leaves no real alternative.
Heat: The Hidden Constraint
All the electricity flowing into GPUs has a consequence: heat.
When GPUs run continuously at full load, they generate intense heat in a very small space. If that heat is not removed quickly and reliably, performance falls, components degrade, and machines fail. Keeping GPUs cool is therefore not an optimisation problem; it is a requirement for AI to function at all.
Inside a modern data centre, heat must be pulled away from chips as efficiently as possible. It is absorbed by metal heat sinks, carried away through pipes, and, in many cases, removed using liquid-cooling systems. In each of these steps, copper is relied upon for its ability to conduct heat quickly and evenly.
As AI models grow larger and more energy-intensive, cooling is no longer a secondary consideration; it has become one of the primary constraints in data-centre design. Once again, the physics points to the same conclusion: copper sits at the centre of the solution.
Data: Moving the Information
AI is not just about raw computation. It is also about moving enormous volumes of data between GPUs, servers, and storage systems quickly, reliably, and with minimal delay.
Inside a data centre, this data does not travel through the air. It moves through physical connections linking machines together over very short distances. While fibre optics are often used to carry information over longer stretches, copper remains essential inside the data centre itself, where reliability, precision, and dense connections matter most.
As AI workloads become more complex, machines must exchange information more frequently and at higher speeds. That requires more internal wiring, tighter layouts, and increasingly dense networks of connections running between racks and within servers. Once again, copper is the material that makes this possible.
In simple terms, AI concentrates data, power, and heat into small physical spaces - and copper is the thread that ties all three together.
What matters most is that this demand is not cyclical. AI is not a one-off upgrade; it is continuously evolving. Models are growing larger, AI is being applied to more real-world tasks, and the technology is moving from novelty to critical infrastructure.
As AI expands in this way, it requires more computation, and more computation pulls in more electricity. That electricity - along with the cooling and data movement it enables - drives sustained demand for copper. This is not a temporary blip. The physical realities of AI - power, heat, and connectivity - ensure that copper will remain essential as these systems expand.
At the same time, copper supply faces real constraints. New mines take many years to develop, ore quality is declining, and environmental and permitting challenges are increasing. The result is a long-term imbalance between supply and demand - one that is unlikely to be resolved anytime soon.
AI feels abstract, but it is built on physical systems. Behind every “smart” response sits a room full of machines drawing power, shedding heat, and moving electrons. Copper is the quiet enabler of that system.
So, when you hear that AI is driving demand for copper, it is not a speculative story or a marketing slogan. It reflects a simple truth: intelligence at scale requires electricity, and electricity requires copper.
This is one of the central insights behind TCu29. AI is not just software - it is a structural, long-term driver of copper demand, and one that is likely to shape the market for years to come.
