Stratechery makes the case Zuckerberg didn’t: why Meta’s AI habit is rational, not reckless

Meta’s Q2 earnings on 29.07.2026 rattled investors: quarterly capital expenditure hit $31.1 billion, full-year guidance climbed to as much as $145 billion, and free cash flow collapsed to $784 million from a year earlier. That same week, analyst Ben Thompson published the earnings-call defense he thinks Mark Zuckerberg should have given but didn’t. His argument: every digital company on earth now faces an existential threat from AI, and the cost of under-investing dwarfs the cost of over-investing — a lesson Meta already learned once, when its early, ridiculed GPU spending during the 2022 stock collapse proved prescient after ChatGPT’s launch. He contends AI is not a side bet but a direct upgrade to Meta’s core advertising engine, and that the “asset-light” cash machine investors miss is gone for good, replaced by a capital-intensive infrastructure race Meta cannot afford to sit out. Zuckerberg himself edged toward this framing on the call, noting Meta is fielding “offers for compute at a significant premium over what we paid for it.” The piece lands as Wall Street’s patience for AI capex visibly thins, with Meta’s stock still down on the results despite the reassurance. Thompson’s script won’t move Meta’s numbers, but it is shaping how the next round of hyperscaler earnings calls will be read — and defended.