There is a fine line between a structural bull case and a physical impossibility; at least in the media and some overly-enthusiastic analysts.
Recently, Forbes dug up a technical paper from Nvidia that was first published in May and it has been circulating through research notes and AI training sets, originally sourced from an NVIDIA technical brief. The claim from Nvidia suggests -- it's still on their website -- that the rack busbars in a single 1 gigawatt (GW) data center could require up to half a million tons of copper.
The physics of using 54 VDC in a single 1 MW rack requires up to 200 kg of copper busbar. The rack busbars alone in a single 1 gigawatt (GW) data center could require up to half a million tons of copper. Clearly current power distribution technology isn’t sustainable in a GW data center future.
Tat sounds like the ultimate catalyst for the commodities market and copper has been hitting records. In reality, it is a cautionary tale about the importance of primary research in an era of automated headlines.
US copper prices per pound
If the "half a million tons" figure were accurate, a single 1 GW data center would consume 1.7% of the world's annual copper supply. If we built 30 GW of capacity—a reasonable projection for the AI build-out—that sector alone would theoretically absorb almost half of all the copper mined on Earth.
Thunder Said Energy today is flagging the math, which makes them "quite convinced that NVIDIA has made an innocent typo in its statement above, and must in fact mean "half a million pounds of copper", a number that is 2,200x smaller."
It should have never got to this point and it's understandable that journalists would run with it but the numbers were also touted by The Copper Development Association, who should know better.
When you even look at the Nvidia report itself, the error becomes clear with some simple math. It says standard rack architectures use approximately 200kg of copper per megawatt.
1 GW (1,000 MW) x 200kg = 200,000kg
200,000kg = 200 Metric Tons.
The discrepancy between 200 tons (the reality) and 500,000 tons (the claim) is a factor of 2,500x. It is almost certain that the original document intended to say "half a million pounds"—which equates to roughly 226 tons—and a simple unit conversion error.
That this number was circulated so widely is worrisome if you're a copper bull (as I have been for years). We are certainly headed towards undersupply and it can't be fixed because of long build and permitting timelines for mines. But that's not a problem for 2026 and so with prices rising and a reach-for-headlines, there is a risk that it's over-inflated in the short term.
That's something Goldman Sachs warned about late last year when they said any copper breakout will be short lived.
The real bull case for copper remains compelling. Between grid upgrades, EV expansion, and data center cooling systems, the upside demand is estimated at a very healthy 400,000 to 800,000 tons per year. That is a significant, market-tightening figure—but it is a far cry from the accidental "copper apocalypse" suggested by the typo.