TrendMacro conversation with Mark P. Mills about reality-checking AI -- energy and economics

Data centers for the AI hyperscalers are driving unprecedented growth in electricity demand. For these firms, price is no object, so there are many paths to powering the present build-out plans. Data centers can be located near natural gas sources, and the gas can be converted to electricity by old-fashioned reciprocating engines if need be, or by less efficient turbines that are more available than the most advanced ones. Breakthrough solutions like small modular nukes are many years off. Wind and solar are too intermittent, and mostly sourced from China -- only useful for data centers with vast arrays of batteries, also mostly sourced from China. A risk is that this drives up the price of both electricity and gas, fueling a political backlash in a world already in an "affordability" crisis after the pandemic inflation. The backlash could include a demonization of the hyperscalers, which could be blunted with differential pricing. What's more challenging is that the data centers are only the core of a layered ecosystem of related functions all of which will use more electricity, first to generate the electricity for the data centers, extending to powering even the smart-phones by which AI is accessed -- and ultimately to the entire economy, which will grow faster thanks to AI, and need more electricity as it does. The good news is that there are no physical supply barriers. The bad news is that there are political barriers.