AI Is Easy. Infrastructure Is Hard.

The infrastructure build-out for AI turns out to be pricey. Datacenters filled to the brim with GPUs, plus all the energy, cooling, and connectivity to run these things, add up to “The biggest economic bet in U.S. history.” Expected annual AI infrastructure spend amounts to 3.63% of U.S. GDP over the eight-year period from 2025–32; the next most expensive infrastructure investment the U.S. has ever undertaken, the national railroad network, cost 2.24% annually over 20 years. Bain estimates the industry must make $6 trillion in annual revenue by 2031 to justify the investment. As someone who lived through the dot-com boom and bust, I find these numbers staggering – and I’m not convinced the math works out.