What If AI Is a Bubble? (www.theatlantic.com)

A provocative Atlantic Radio episode and transcript asks whether the current AI boom is a speculative bubble: trillions of dollars are flowing into AI (global AI spending projected at ~$375B in 2025 and close to half a trillion in 2026), Nvidia has hit a $5 trillion valuation, and investors are pouring hundreds of billions into data centers whose power needs could rival major U.S. cities. Yet commercial returns lag: many generative-AI products haven’t translated into matching revenue, OpenAI and others have disputed how they’ll monetize future breakthroughs, and a McKinsey-cited finding suggests ~80% of surveyed companies saw no “significant” bottom-line impact from AI so far. The episode stresses why this matters: the investment is deeply woven into markets via the “Magnificent Seven” and index funds, so a correction would be systemic. Technical and economic pressure points include massive compute and electricity demands, sustained capital expenditure on data-center buildout, and disputed back-of-the-envelope break-even figures (estimates range from ~$160B to $320–480B in required revenue to justify 2025 capex). The piece frames two hard outcomes—either AI eventually pays off (potentially at great environmental and infrastructural cost) or the hype outruns economics, creating a bubble whose burst would echo past crashes but with uniquely concentrated tech and energy risks.