How the AI boom might really end: 5 scenarios

Why it matters

The article lays out a personal framework for how the current AI investment boom could unwind, written as Anthropic's IPO gathers pace and chip stocks endure a prolonged losing streak amid worries about AI-related debt. The author groups the risks into two categories: AI failing to deliver or being stopped, and AI succeeding while investors fail to capture the value. The first group covers scaling walls, alignment disasters triggering political backlash, and local opposition to datacentre construction, though the author doubts any single-country ban would matter given competition between the US and China. The second group, which the author views as more likely and arguably already visible, centres on open weights models becoming good enough to collapse margins in an ongoing inference price war. Anthropic's discounted small model release, priced aggressively for short-context work, is cited as evidence that incumbent labs are now competing directly with open alternatives on price.