AWS signs a two-year supply deal with Rio Tinto to access its Arizona copper mine, the US' first new source in 10+ years, as AI data centers create huge demand
Context & Ripple Effects
The deal turns a previously discussed materials constraint into a procurement response: coverage has put AI data-center copper use at 27 to 33 tons per MW, while BHP had warned that expanding data centers could deepen a looming copper shortage.
It also sits alongside AWS’s longer-duration compute commitments, including its seven-year OpenAI compute agreement. Securing access to a newly available US source connects cloud-capacity planning to upstream industrial inputs.
First-order effects
- AWS gains a two-year channel to copper from Rio Tinto’s Arizona mine, reducing its exposure to spot availability for data-center-related buildout during that period.
- Rio Tinto adds a major cloud customer to the mine’s initial supply base, linking the project’s output directly to AI infrastructure demand.
Second-order effects
- Other hyperscalers and data-center developers may face stronger incentives to secure copper earlier in their construction planning rather than treat it as a standard purchased input.
- The arrangement makes new domestic mine output more strategically relevant to cloud expansion, potentially increasing competition for supply from miners and large infrastructure buyers.
Third-order effects
- If cloud operators increasingly contract for raw materials, AI infrastructure procurement could extend beyond chips, servers, and power into upstream commodity supply chains.
- Copper availability may become a more explicit execution constraint on data-center buildouts, alongside the capacity bottlenecks already associated with AI infrastructure.
The trend: AI data-center expansion is turning upstream physical inputs—especially copper—into strategic capacity dependencies for cloud providers.