AI Infrastructure & Power
AI's bottleneck is shifting from chips to power and data centers. This topic takes apart the infrastructure math: how a gigawatt data center gets built, where the electricity comes from, why the grid cannot expand fast enough, and who pays when power prices rise. Guests include power-industry practitioners. The unavoidable questions: can trillion-dollar CapEx earn its return, how should depreciation be counted, and why do grey arbitrages like token export and cyber power theft exist. Infrastructure is the most asset-heavy part of the AI narrative and the easiest to turn into a story — we try to turn it back into costs.
9 edited conversations, newest first.
EP51
Circular financing · The dishwasher paradox · The chase-down line · Long-end rates
EP48
Account pools · Multipliers · American women aged 45 to 55 · Shadow fleets
EP47
Two Three Gorges Dams · A trillion in depreciation · Finding MLCC among three thousand names · The radio was wrapped electricity
EP37
The SpaceX IPO · 5,000 Three Gorges Dams · A $30 trillion TAM · A busy summer
EP22
A grid of warlord fiefdoms · Behind-the-meter generation · Data centers as interruptible load · The legacy burden thirty years out
EP
Models with a short shelf life · Ninefold customer spending · A one-person drug company · The next coding opportunity is still coding
EP
Five-year turbine delivery times · Maintenance contracts worth more than the equipment · A decade of blade investment against the cycle · A €400 million grid connection
EP
Four parallel universes · Compute and credit risk · The courage capital provides · Making money and holding on
EP
The only spark · Two shortcuts overseas · AI bills of 200 versus 10,000 · When the old guard steps in