I have often wondered how much of Innovation is constrained by physical materials. Semi conductors can only get so small. There is only so much silver. Etc. This is a good example of that. It will be interesting to see if an alternative can be developed or if certain sciences are finally hitting metal ceilings.
Yes, this is the “physics meets economics” moment. When materials get tight, the constraint becomes the teacher: prices rise, engineers get creative, and innovation shifts from “smaller” to “different” through new materials, new architectures, and smarter reuse. The real question is which sciences can reroute, and which ones truly hit a hard ceiling.
Great paper and extremely relevant with the increase in silver prices being a hot topic of discussion. At what point do material constraints seriously limit the growth of AI?
Thank you, Jonathan. Great question! I do not have the complete answer, but I think about it through supply and demand elasticity.
When key inputs for AI have inelastic supply in the short run, a jump in demand shows up mostly as higher prices rather than much more quantity. That does not stop AI, but it makes scaling more expensive and pushes innovation toward efficiency, smaller models, and better deployment.
Constraints become truly binding when demand for compute stays strong even as prices rise while supply remains tight. Then the main impact is cost pressure and concentration among those who can pay. Over time, supply can become more elastic through capacity expansion and substitution, but that adjustment is slow.
I have often wondered how much of Innovation is constrained by physical materials. Semi conductors can only get so small. There is only so much silver. Etc. This is a good example of that. It will be interesting to see if an alternative can be developed or if certain sciences are finally hitting metal ceilings.
Thank you for your comment, Anton.
Yes, this is the “physics meets economics” moment. When materials get tight, the constraint becomes the teacher: prices rise, engineers get creative, and innovation shifts from “smaller” to “different” through new materials, new architectures, and smarter reuse. The real question is which sciences can reroute, and which ones truly hit a hard ceiling.
Great paper and extremely relevant with the increase in silver prices being a hot topic of discussion. At what point do material constraints seriously limit the growth of AI?
Thank you, Jonathan. Great question! I do not have the complete answer, but I think about it through supply and demand elasticity.
When key inputs for AI have inelastic supply in the short run, a jump in demand shows up mostly as higher prices rather than much more quantity. That does not stop AI, but it makes scaling more expensive and pushes innovation toward efficiency, smaller models, and better deployment.
Constraints become truly binding when demand for compute stays strong even as prices rise while supply remains tight. Then the main impact is cost pressure and concentration among those who can pay. Over time, supply can become more elastic through capacity expansion and substitution, but that adjustment is slow.