Up 109% in a Year, the Best AI Trade Isn’t a Chip Stock at All

“AI-driven copper demand alone is unlikely to be a durable, multi-year driver for miners; macro cycles, cost inflation, and policy risk will cap upside.”

While the copper-as-AI-physical-layer framing is catchy, the story rides on a brittle assumption: AI compute demand will steadily tighten copper use for a decade. In reality, copper is extraordinarily cyclical, and supply/demand can swing with capex cycles, inflation, and policy shifts. A 50% demand rise by 2040 is plausible, but the path hinges on electrification and data-center build rates that are not guaranteed to accelerate in lockstep with AI hype. Short-term record copper prices may reflect bottlenecks or speculative flows rather than durable demand, and miners’ margins hinge on copper prices more than AI adoption. ESG strains and geopolitical risk add further downside risk to these equities.

“The current valuation of copper miners is overly dependent on spot price volatility, masking the execution risks of long-lead-time capital projects.”

The pivot from 'compute' to 'infrastructure' is a rational late-cycle shift, but the article conflates long-term secular demand with immediate equity performance. While SCCO’s EBITDA margin expansion to 66.6% is impressive, it is highly sensitive to LME copper spot prices, which are currently inflated by supply-side bottlenecks rather than just AI-driven consumption. Investors are chasing the 'physical layer' of AI, but they are ignoring the massive capital expenditure cycles required to bring new mines online. If global manufacturing demand outside of data centers softens, the copper price could face a sharp correction, leaving miners with high fixed costs and cooling demand.

“Copper demand from AI infrastructure is real, but the article mistakes a cyclical commodity at all-time highs for a structural AI play, ignoring mean reversion risk and supply elasticity.”

The copper thesis is real but the article conflates correlation with causation. Yes, data centers will need copper—but so will EV charging, grid modernization, and renewables. S&P's 50% demand growth by 2040 is a 16-year projection; it doesn't tell us if copper supply will tighten or if prices normalize. Southern Copper's 66.6% EBITDA margin and $0.05 cost per pound are extraordinary, but margins this wide attract new supply and capex. The article cherry-picks a one-year window where copper miners outpaced NVIDIA—but NVIDIA's 34% gain still beat the S&P 500. The real risk: copper is cyclical, and we're at an all-time price high. Mean reversion is a legitimate tail risk the article ignores.

“The miners' outperformance reflects speculative copper pricing more than durable near-term earnings power from AI.”

The article positions copper miners as the overlooked AI infrastructure play, citing SCCO's 109% gain and FCX's 64% versus NVDA's 34%, backed by S&P Global's 2040 demand forecast and data-center power needs. Yet the 50% demand growth is a distant projection, while the LME price spike to $14,533/ton and miners' margin expansion appear driven by near-term speculation. SCCO's Q2 2026 cost collapse to $0.05/lb and 66.6% EBITDA margin are striking but vulnerable to input cost rebounds or mine disruptions. Broader electrification tailwinds exist, but the piece underplays how quickly new supply or slower AI buildouts could reverse the trade.