The 500,000-ton typo: Why data center copper math doesn’t add up

A single gigawatt data center requiring half a million tons of copper? That's the startling figure that has been circulating through research notes and AI training sets, originally sourced from an NVIDIA technical brief. The claim, which suggests the rack busbars alone in a single 1 GW data center could consume 500,000 tons of copper, paints a picture of an impending physical impossibility—and a potential catalyst for a copper supercycle. But a closer look at the underlying math reveals a far more mundane explanation: a likely typo that has been amplified without scrutiny.

The original claim stems from a technical paper first published by NVIDIA in May. The document states that standard rack architectures use approximately 200 kg of copper per megawatt for rack busbars. It then extrapolates this to a 1 GW data center, concluding that the busbars alone could require "up to half a million tons of copper." This figure, if accurate, would represent a staggering 1.7% of the world's annual copper supply for a single data center. If the industry were to build 30 GW of capacity—a reasonable projection for the AI build-out—the sector alone would theoretically absorb almost half of all the copper mined on Earth.

The problem is that the math doesn't hold up. As Thunder Said Energy pointed out, a simple calculation reveals the discrepancy. The NVIDIA report itself states that a standard rack uses about 200 kg of copper per megawatt. For a 1 GW (1,000 MW) data center, the calculation is straightforward:

1,000 MW × 200 kg/MW = 200,000 kg

200,000 kg converts to 200 metric tons. The claim of 500,000 tons is off by a factor of 2,500. Thunder Said Energy is "quite convinced" that NVIDIA made an innocent typo and must have meant "half a million pounds," which equates to roughly 226 tons—a figure much closer to the calculated 200 tons. A simple unit conversion error, from pounds to tons, appears to be the culprit.

This error should have been caught before the number was widely circulated. Yet it was not only picked up by media outlets but also touted by the Copper Development Association, an industry group that should be well-versed in such metrics. The incident highlights a growing problem in an era of automated headlines and rapid information sharing: the erosion of primary research. Journalists and analysts, often working under tight deadlines, may run with a compelling number without verifying the underlying calculations, especially when it comes from a reputable source like NVIDIA.

For the copper market, the implications are significant. Copper prices have been hitting record highs, driven by genuine structural demand from grid upgrades, electric vehicle expansion, and data center cooling systems. The real bull case for copper remains compelling, with upside demand estimated at a healthy 400,000 to 800,000 tons per year from these sectors alone. This is a significant, market-tightening figure that supports a long-term positive outlook for the commodity.

However, the inflated "copper apocalypse" narrative based on the typo risked over-inflating prices in the short term. Goldman Sachs warned late last year that any copper breakout would likely be short-lived if driven by speculative hype rather than fundamentals. The typo-induced frenzy could have contributed to unnecessary volatility, making it harder for the market to price in the real, albeit substantial, demand growth.

The real challenge for the copper market is not a sudden, massive spike in demand from a single data center, but the long-term structural undersupply. Building new mines is a lengthy process, often taking over a decade from discovery to production, and permitting timelines are notoriously slow. This means supply cannot easily respond to demand spikes, creating a tight market environment. The genuine demand from AI data centers, EVs, and grid modernization is enough to create a compelling bull case without resorting to hyperbolic figures.

In the end, the 500,000-ton typo serves as a reminder that in the rush to connect technological trends to commodity markets, accuracy matters. The data center build-out will indeed be a major driver for copper demand, but the scale is measured in hundreds of thousands of tons per year, not millions of tons per facility. For investors and analysts, the lesson is clear: always check the math. The real story of copper's future is complex and compelling enough without the need for sensational errors.

Featured image: Copper prices have been hitting records, but the real demand story is more nuanced than a single typo suggested.