Nvidia, Meta, Samsung back materials foundry to escape rare-element dependency

The consortium framing matters more than the materials science. When Nvidia, Meta, and Samsung co-own a foundry, they are not just hedging supply risk, they are potentially setting the terms on which future competitors access the same inputs.

This story sits in a cluster of moves where AI's largest players are pulling critical infrastructure closer to their own balance sheets rather than relying on shared markets. The recent coverage of OpenAI's anxiety over open-weight models ("OpenAI is scared of open-weight models") is a useful contrast: that piece traced how closed-model economics face pressure from freely distributed alternatives. Here the dynamic runs in the opposite direction. Vertical integration of materials supply is a mechanism for entrenching cost advantages that smaller labs and open-weight projects cannot easily replicate. If the consortium controls yield improvements at the substrate level, the compute cost gap between members and non-members widens over time, which is a structural pressure on the open model economics that the earlier piece described.

Watch whether any foundry output is licensed to third parties or kept exclusive to consortium members within the next 18 months. Exclusive access confirms this is a competitive moat play; open licensing would suggest the primary goal is genuinely supply diversification.