a16z's Olivia Moore: 'Almost all of what we think of as consumer AI is prosumer AI'

Only 2.2% of U.S. households pay for any AI product. That figure, drawn from Andreessen Horowitz's State of Markets report, anchors a new top-100 consumer AI ranking released Monday by a16z partner Olivia Moore — and frames her argument that the consumer AI market hasn't stalled. It has barely started.

Moore published her list of the top 100 consumer AI apps on Monday. ChatGPT still leads the field by a wide margin. Smaller players including Suno and ElevenLabs show what Moore called "real staying power." The ranking also catalogs roughly half a dozen consumer categories that have no AI entrants at all.

In an interview this week, Moore pushed back on the recent wave of pessimism about consumer AI economics. She said the right question is not how many people will pay $20 a month for a chatbot, but whether the industry can find ways to make money from the other 97.8% of U.S. households.

What does the revenue picture actually look like?

OpenAI's renewed focus on enterprise has fueled doubts that consumer AI can carry its own weight. Moore frames it differently. "I don't blame them, as much as I love consumer, because almost all of the AI revenue thus far has come from subscriptions and then token usage," she said. "And that is much more concentrated on the enterprise and prosumer side."

That concentration is the heart of the issue. The marginal cost of running an AI service remains far higher than the marginal cost of running a traditional internet product like Google Search or Facebook. The users willing to pay today are technical users running coding and automation workflows — exactly the workloads that demand the most expensive models.

Lower-cost paths are starting to appear. ChatGPT now offers a Go plan at $8 per month, which Moore assumes runs on cheaper models. "We don't necessarily need frontier intelligence for every task, and so that could mean cheaper models, especially for some consumer use cases," she said. "It could also mean more consumer companies building on top of open-source models."

Why is "consumer AI" actually "prosumer AI"?

The clearest argument in Moore's report is that the category label is wrong. The companies racking up consumer revenue — Cursor, ElevenLabs, Gamma, Lovable, Replit, Fal, Higgsfield, HeyGen, Manus, Fireflies AI, Granola — look consumer in name only. Their buyers are individual power users, mostly working professionals paying out of pocket or on company cards.

"Almost all of what we think of as consumer AI, I would argue, is prosumer AI," Moore said.

She calls the pattern "The Great Expansion." Pre-AI, a company like Canva took six or seven years to add team and enterprise plans after launching as a consumer product. The current cohort hits the same milestone in roughly 18 months. Moore published an essay with that title about a year ago and said the trend has only accelerated.

Three revenue clusters dominate the prosumer list:

• Product-building apps: Lovable, Replit, Fal
• Product marketing tools: Higgsfield, HeyGen
• General work management: Manus, Fireflies AI, Granola

None of these behave like a traditional consumer product. None of them live on a phone screen the way Instagram or TikTok do. The shape of the spend is closer to a freelance worker's software stack than to a teenager's app drawer.

Which categories are completely empty?

Moore's list flags several segments with zero top-100 entrants: social, dating, marketplaces, retail, travel, finance, and health. The absence is striking given how many AI wrappers have shipped in the last two years.

Moore reads the gap as opportunity, not saturation. Founders in those verticals have not yet figured out how to clear the unit-economics bar. The demand is not in question. The next six months will show whether AI-native companies can break into categories the last decade of mobile apps could not.

What has to change to unlock true consumer AI?

Moore is skeptical of the standard playbook. She does not want to see the 2.2% paying-user number climb on its own. "I actually don't know if I want to see that user number increase," she said. "I'm more interested in getting back to a world where the consumer is monetized in a way that isn't subscription dollars out of their own pocket."

The model she prefers is older and more familiar. "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," Moore said.

That framing has consequences for the model layer. Frontier systems from OpenAI, Anthropic, and Google remain too expensive to serve a billion ad-supported users. The shift toward open-source and lightweight inference — Meta's Llama family, Mistral, and DeepSeek variants — is what makes a non-subscription business model workable. Moore said a16z founders are increasingly building on those alternatives.

"It's definitely very early," she said.

The "very early" framing cuts against the consensus that consumer AI is in a slump. The data does show weakness: paying users are scarce, the cost stack is high, and OpenAI itself is steering toward enterprise deals. Moore's report reframes those problems as evidence that the consumer category hasn't been built yet — only its first prosumer proxy.

Whether the next 18 months produce a true consumer hit — an AI product that behaves like a real social, dating, or retail app — will determine whether the 2.2% figure becomes a footnote or a ceiling.