Ben Thompson argues in Stratechery that Meta should invest in AI for its advertising business

Ben Thompson argues in Stratechery that Meta should invest in AI for its advertising business

Ben Thompson’s piece was published July 7, 2026. It is an imagined script for Meta CEO Mark Zuckerberg to deliver on an upcoming earnings call, not a record of Zuckerberg’s remarks. Thompson uses the script to argue that Meta’s case for heavy AI investment rests on advertising: AI could improve ad targeting and content recommendations, expand the advertising inventory Meta can sell, and help connect businesses with customers. The script presents Meta’s past product missteps as lessons in favor of building around what users choose and what the company does well. It also argues that renting out some compute could bring in revenue and impose a hurdle rate—a benchmark for whether using that capacity itself would earn more. Thompson’s proposed speech rejects competing to sell business productivity tools as Meta’s central purpose, instead emphasizing connection, entertainment and commerce. These are arguments and forecasts in a fictional speech, not evidence that the predicted benefits will occur or that Meta has announced these plans. The open question is whether the proposed rental benchmark would meaningfully discipline Meta’s investment while leaving enough compute for its own business.

Publication: d33edf70-59f1-b494-7a70-d9cbc591a339

attribution — pass: The reading correctly attributes the piece to Ben Thompson of Stratechery. It makes clear that the Zuckerberg speech is Thompson's imagined script, not Zuckerberg's actual remarks. The source text says so explicitly.

What follows isn’t actually me: it’s what Ben Thompson of Stratechery thinks I should say on this call.

summary — pass: The summary identifies the actor, the fictional earnings-call format and the central ads-driven case for AI capex. It explains hurdle rate in plain language and notes the rejection of productivity tools. It is somewhat flattened: it omits the existential-risk argument and the mea culpa voice. The library's Benchmark definition (about tests) does not fit here, but the inline explanation suffices. Overall it remains faithful and understandable.

every single digital company on earth faces an existential threat from AI, and we are no exception

date — pass: originalPublished is 2026-07-07, which matches the dateline Tuesday, July 7, 2026 (the weekday is consistent). The reading calls the early-August earnings call upcoming and presents no outcomes. The trailing link titles look like later listings, and the reading does not use them. The page was edited after publication, but the cited passages are plainly part of the article body.

Tuesday, July 7, 2026

support — pass: Each claim is grounded in quoted text from the article body. The forecasts are framed as fictional-speech arguments, and the Escape Meridian interpretation is labeled and carries its excerpt. The open question about how the benchmark would work follows from the source; the script itself notes a claw-back-at-any-time condition. A post-publication edit is flagged, but the cited text is plainly original article content.

we are going to sell access to a portion of our compute infrastructure on a short-term basis, with the ability to claw back that compute at any time

media — pass: No media was selected and no verified media candidates were supplied, so there is nothing to misidentify.