Andreessen Horowitz's Olivia Moore says most consumer AI money comes from prosumers

A ChatGPT user who wants the product without the top price can now pick the Go plan, at $8 a month ^[14]. Olivia Moore, the Andreessen Horowitz partner behind the new top-100 ranking, assumes it runs on cheaper models ^[1][14]. "We don't necessarily need frontier intelligence for every task," she told TechCrunch ^[21]. Here is what teams tell themselves users do: a big audience of individuals adds up to a consumer business. Here is what users actually do, going by Moore's description of the report's revenue list. The spenders build products in Lovable, Replit and Fal, make ads in Higgsfield and HeyGen, and manage work in Manus, Fireflies AI and Granola ^[8]. "Almost all of what we think of as consumer AI, I would argue, is prosumer AI," Moore said ^[6]. The users driving revenue are doing coding and technical automation, where she said they probably need frontier intelligence ^[15]. The report credits Suno and ElevenLabs with real staying power, TechCrunch reported ^[5]. The interview does not include retention figures, nor does it name the half-dozen consumer categories the report finds untouched by AI ^[7]. ElevenLabs also appears on another of Moore's lists: companies, alongside Gamma and Cursor, that start as consumer and become majority-enterprise within 18 months ^[17]. Canva, she said, took six or seven years just to add teams or enterprise plans ^[16]. The AI companies made the move in between a quarter and about a fifth of that time ^[20]. Moore said "almost all of the AI revenue thus far has come from subscriptions and then token usage," money she described as "much more concentrated on the enterprise and prosumer side" ^[10]. For that reason she does not blame OpenAI for moving back toward enterprise this year ^[18]. "To me, it's not necessarily a pivot because they're still launching a lot in consumer. It's more of an expansion," she said ^[11]. Just 2.2% of U.S. households pay for AI, according to the State of Markets report cited in the interview ^[3]. That leaves 97.8% paying nothing ^[19]. Moore is not trying to raise the paying share. "I actually don't know if I want to see that user number increase," she said ^[12]. Her preferred model is ads: "I think most people would actually rather have free access to something and see some ads, and then they can decide if they want to subscribe or not to make the ads go away" ^[13]. Cost is the harder part. TechCrunch's interviewer put the marginal cost of AI services well above classic internet services such as Facebook or Google Search ^[4]. Moore pointed to cheaper models and open-source ones, an approach she said a16z has heard a lot about from its founders lately ^[22]. We think a team shipping a consumer AI product should settle two questions before it designs a pricing page. The first is who pays: the user out of pocket, or someone else, such as an employer or an advertiser. The second is whether the core task needs a frontier model. Coding and automation need frontier models, sit with enterprise and prosumer buyers, and hold today's revenue ^[10][15]. By Moore's account, those products tend to turn majority-enterprise within 18 months ^[17]. A frontier product paid for out of pocket is selling to the 2.2% ^[3]. A cheaper model on a personal subscription looks like the $8 Go plan ^[14]. The last corner, a cheap model funded by ads, is Moore's bet, for products "where the model is not the product" ^[13][23]. Our recommendation is to build there only if the product works on a model cheap enough to give away, since a free AI user costs more to serve than a search user, on TechCrunch's account ^[4].