Really insightful framing on energy as the new moat in AI. The Fervo investment makes total sense when thinking about how compute arbitrage will play out over the next decade. I've been watching datacenter colocation decisions and its wild how much power pricing variance exists even within the same grid region. The geothermal angle is interesting becuase it sidesteps transmission losses entirely. Wondering if this creates second-order effects where model training geographyy becomes as strategic as the architecture itself.
Really insightful framing on energy as the new moat in AI. The Fervo investment makes total sense when thinking about how compute arbitrage will play out over the next decade. I've been watching datacenter colocation decisions and its wild how much power pricing variance exists even within the same grid region. The geothermal angle is interesting becuase it sidesteps transmission losses entirely. Wondering if this creates second-order effects where model training geographyy becomes as strategic as the architecture itself.
I agree that this dynamic is already emerging.
As energy becomes the binding constraint training location stops being an ops choice and starts looking like a strategic one.