NVIDIA CEO Says AI's Future Isn't Just Copper

One of Wall Street's favorite commodity narratives is the link between copper and AI demand. The demand logic flow is simple: AI requires enormous data centers; data centers require enormous amounts of copper; therefore, copper demand will keep rising. On the other hand, rising capex, lower grades, and a lack of new large discoveries are hurting the supply side . It looks like a double whammy that guarantees success, yet reality is far more complicated. While forecasts from S&P Global suggest copper demand from data centers and related infrastructure could rise from 1.1 million metric tons in 2025 to 2.5 million by 2040, those projections often assume a relatively static technological environment. In practice, AI infrastructure is evolving rapidly, shifting toward less copper-intensive solutions. Two Key Narrative Flaws The bullish argument begins with a valid observation of how metal-intensive AI facilities are. According to S&P Global, a crypto data center requires roughly 21 metric tons of copper per megawatt of installed capacity, while an AI training center can require as much as 47 metric tons. The difference reflects denser wiring, larger cooling systems, and dramatically higher power requirements.  However, the difference between the announcements and actual deployment is considerable. The concept of "bragawatts" is increasingly present in 2026, as capacity figures look impressive in ... Full story available on Benzinga.com