THREAD 06
Externalities & Infrastructure
Power, cooling, silicon, water, and the physical constraints behind training and inference at scale.
Open question
Which physical constraint becomes binding first as AI demand grows?
How we investigate
Public-data analysis, transparent estimation, and infrastructure literature review.
3 min readWhere the Tokens Actually Go in Long-Context Inference
A transparent look at KV-cache memory, bandwidth pressure, and why a large context window is not the same as reliable recall.
Language ModelsRead
4 min readWhat a Data Centre Actually Does With Water
Cooling is a heat transfer problem with several solutions. Which one a facility uses decides whether it consumes water, electricity, or land.
ExternalitiesRead
4 min readWhy AI Water Estimates Differ by 2000x
Published figures for water per AI prompt range from 0.26 millilitres to over 500. Both are defensible. The gap is entirely in what each one counts.
ExternalitiesRead
4 min readWhat a $50M Cluster Actually Buys You
GPU count is the least interesting number in a cluster. Power, cooling and interconnect decide what the money actually turns into.
InfrastructureRead