Technology

(Image source: 123RF)

Every technological boom in history has strained against inconvenient physical limitations. With railways, it was the cost and time of laying track; with motor cars, it was the need to build usable roads; and the internet suffered from slow connectivity until the mid-2000s. For AI, the constraint is datacentres.

This constraint is every bit as real as a lack of railroad tracks between two cities. AI requires staggering quantities of computing power and, should the current rate of uptake continue, that capacity will quickly run short. This has driven the hyperscale firms to invest trillions of dollars in building these digital behemoths.