The essayist compares the current investment in AI to the 19th-century railways: a real financial bubble, followed by an infrastructure that will transform the economy sustainably.
Derek Thompson defends a two-step thesis: the artificial intelligence infrastructure will undergo a brutal correction, and this correction will not prevent technology from transforming the world afterwards. The precedent he called was not the 2008 crisis but the 19th-century railways and telecommunications of the 1990s — sectors where capital had been destroyed on a large scale while leaving behind networks that could be exploited for decades.
His central argument is arithmetic. Technology companies commit approximately $400 billion a year to AI infrastructure, equivalent to an Apollo program every 10 months, and projected expenditures exceed $500 billion for 2026-2027. On the other hand, US consumers spend only about $12 billion per year on AI services. The difference between the two orders of magnitude is what, according to him, does not hold.
He pointed out a second macroeconomic effect: half of US GDP growth in the first half of 2025 came from data centre spending, a concentration which he considered absurd and which diverted capital from other sectors.
Read the full article on the original source
Go to the source →