Ben Thompson argues in Stratechery that Google is testing OpenAI’s and Nvidia’s moats

Ben Thompson argues in Stratechery that Google is testing OpenAI’s and Nvidia’s moats

Ben Thompson’s article was published on December 1, 2025. He casts Google’s gains in AI as an ordeal for OpenAI and Nvidia, but argues that their risks differ. Google’s TPUs could challenge Nvidia’s margins, while Nvidia’s flexibility and CUDA ecosystem may still matter, especially to customers with varied workloads. Thompson sees OpenAI’s consumer reach as a more durable defense: changing the habits of many individual users is harder than persuading a few large buyers to change suppliers. But he argues that OpenAI is weakening its position by relying on subscriptions rather than advertising. In his account, advertising could both improve ChatGPT and strengthen its moat by drawing more users and feedback. Google, meanwhile, has experience monetizing Search and the resources to compete for consumer attention. Thompson’s conclusion is that Google’s advantages make the contest serious, but ChatGPT retains a meaningful moat because of its users. The open question is whether a scaled Aggregator can be displaced by a rival’s resources when it declines to adopt the business model Thompson considers optimal. He says he is “nervous — and excited” to see how far Aggregation Theory goes.

Publication: b2a2258f-530e-f644-fca0-d45010688310

attribution — pass: The reading attributes the argument to Ben Thompson in Stratechery, which matches the source page. This is a single essay with its own argument, not a listing. Its claims are kept as Thompson's assertions.

I still like OpenAI’s chances more. Indeed, if anything my biggest concern is that I seem to like OpenAI’s chances better than OpenAI itself.

summary — uncertain: The summary faithfully tracks Thompson's argument and keeps his voice ('nervous — and excited'). However, it relies on central terms that a general reader may not understand. 'Aggregator' and 'Aggregation Theory' carry the conclusion and the open question, but neither is explained in plain language or listed in the definitions. 'CUDA' and 'moat' also go unexplained. The smallest repair is a one-line gloss of Aggregator (a service that wins by controlling user demand) and CUDA (Nvidia's programming software), or term-library additions for them.

The hero of this site has been Aggregation Theory and the belief that controlling demand trumps everything else

date — pass: evidence.originalPublished is 2025-12-01T15:03:58Z, and its UTC calendar day matches the source dateline of December 1, 2025. The evidence kind is publisher-page, so no archive disclosure is required. Trailing links to other issues are not used. The reading presents no hindsight framing, and Thompson's views are kept as arguments rather than outcomes.

Monday, December 1, 2025

support — pass: Each claim is stated in the source, including: - TPU pressure on Nvidia's margins - Nvidia's flexibility and CUDA advantages - concentrated buyers versus 800 million+ individual users - advertising improving ChatGPT and deepening its moat - Google's experience monetizing Search - the closing open question The question follows directly from Thompson's argument and does not assume an outcome.

changing the habits of 800 million+ people who use ChatGPT every week, however, is a battle that can only be fought individual by individual.

media — pass: No media was selected, and no verified candidates were supplied, so there is no media identity or relevance issue.