First-principles estimate of the personal AI assistant market
The author estimates a roughly $140B revenue pool for personal AI assistants in 2030, and argues that split among 5+ players it does not justify Meta's current expectations.
these $META shenanigans bc of muse/ personal AI assistants are getting out of hand, so thought helpful to ground this discussion in some first principles thinking way way too aggressive of a forecast, but let's entertain the following for the sake of the argument 2030, global -> ecom $5tn x 30% agent share x 3% take rate = $45bn rev pool -> 4bn users x 10% paying x $20/m = $96bn rev pool -> total ~$140bn this will be a market with perfect competition and extremely low switching costs (and massively depend on cooperation from large merchants like $AMZN, but let's also ignore that for a moment). I assume 5+ players -> $GOOG, $SPCX, at least one of OAI and Anthropic (if they don't go bankrupt) -> $AAPL offering for free and private, run locally on hardware (which is also the biggest bear case, but let's ignore that as well) -> tbh only maybe $META bc people don't really trust them today (and I don't see how they'd change that) -> global/ regional third party providers/ protocols like instinct/ openclaw (at the end of the day if you believe in AI, this is just a compilation of scripts and maybe memory architecture ie there's really no MOAT here) so let's divide the pool by 5 to be generous = that's ~$30bn rev by 2030 in the blue sky bull case for reference, consensus 2030 rev for $META is $470bn (agents cannibalize their own ad spend, which is why I haven't added ad rev on top) and consensus 27 CAPEX alone is $200bn the math ain't mathing