Dan Luu picks apart Ed Zitron's 'AI is a bubble' predictions and finds the numbers don't add up

Dan Luu scrutinizes the reasoning behind AI skeptic Ed Zitron’s widely-cited predictions, opening with a disclaimer that he holds no meaningful financial stake in AI beyond ordinary index funds. His core complaint isn’t with skepticism itself but with the quality of the argument. As a worked example, he takes a November 2024 talk in which Zitron claimed that Meta, Google, and Microsoft are dying companies that no longer know how to grow, and are therefore cramming AI into everything out of desperation. Luu points out that the revenue and profit figures for all three flatly contradict the premise: these are fast-growing, highly profitable businesses, not flailing ones.

The more damning critique is about how Zitron builds his case. Rather than cite Meta’s own metrics, he leaned on third-party Similarweb estimates of a Facebook user decline—numbers Luu considers too imprecise to support any real conclusion. For Google, Zitron blamed a single executive, Prabhakar Raghavan, for wrecking search, without establishing that this actually threatens Alphabet’s revenue, which is diversified across YouTube and Cloud. Luu concedes that Google search has genuinely degraded—he recounts how ads were gradually A/B-tested to look more like organic results over years—but notes that this trend makes Google more money, which undercuts the ‘desperate dying company’ narrative rather than supporting it.

The broader point is about credibility laundering. People cite Zitron as the guy who ‘looked at the numbers,’ lending a veneer of quantitative rigor to their own anti-AI stance. Luu argues the numbers are largely decorative: they get thrown around but rarely connect to a coherent chain of reasoning, and sometimes don’t even support the claim being made. The real draw, he suggests, is Zitron’s anger, which is effective for engagement but a poor substitute for analysis. Even a correct prediction record, Luu warns, can rest entirely on wrong reasoning.