a16z's Olivia Moore Sees a Consumer AI Market Beyond Subscriptions and APIs

a16z partner Olivia Moore, who covers the firm's consumer AI products, talks in an interview with TechCrunch about a "huge opportunity" in consumer AI. The condition is simple: the industry needs revenue sources beyond subscriptions and API fees. Today's market standard is a plan at around $20 a month, like ChatGPT Plus, while top-tier plans cost up to $200. The available excerpt of the interview contains only this thesis, so what follows is an analysis of the logic behind it, not a retelling of Moore's words.

Why subscriptions hit a ceiling

Subscriptions were designed for software with a near-zero marginal cost of serving each user. Generative AI is different: every request costs money in inference, and the more active the user, the more expensive they are to serve. The price, meanwhile, is fixed. The most engaged people end up being the least profitable, and free tiers have to stay in place for the sake of reach.

There is also a psychological limit. Only a limited share of the audience is willing to pay monthly for a single AI service. Very few will agree to pay for a second or third. If the consumer market rests on this model alone, it grows more slowly than the number of users.

Which revenue sources remain

API revenue works differently: developers and companies pay, not ordinary people. That is a business market, and it relates to consumer products only indirectly. Consumer AI needs models where the money comes from the use case itself. These are the options worth considering:

• Subscription: $20 a month as the standard, up to $200 in premium plans. The ceiling is set by the share of people willing to pay.
• API: developers pay for tokens. The customer is a business; the consumer product is merely a storefront.
• Advertising: monetizing the free audience, especially on search and recommendation queries.
• Transaction commissions: an agent books, orders or picks a product and takes a cut of the deal.
• Affiliate payouts: built-in recommendations paid per click-through or purchase.

None of these options is new to the internet. What is new is that an AI assistant sits much closer to the decision point than a search results page or a feed. It knows the context, the budget and the preferences, and an agent can also carry the action all the way through to payment on its own. That is why commissions and ad formats look to investors like a logical continuation of the subscription model.

Risks of the new approach

Ads and commissions have a price, and trust pays it. If an assistant recommends a product because someone paid for it, the value of the answer drops. Users are already used to treating a chatbot as a neutral interlocutor, and the first notable scandal over promotion inside an answer could quickly damage a product's reputation. Regulators will also have to decide how to label paid recommendations inside a conversation.

The second risk: agentic transactions require reliability. It is one thing when a model makes a mistake in text, and quite another when it buys the wrong ticket. The market for transaction revenue will open up once the rate of such errors becomes comparable to the human rate.

What teams building consumer AI should do

Calculate unit economics by activity level, not by the average user. Find out how much inference costs for the top 10% most engaged users and whether the subscription price covers it. Check whether your use case has a moment where value is measured in money: a purchase, a booking, a hire, payment for a service. That is the easiest place to put a commission, and it irritates users less than a subscription price hike.

If you plan to run ads, separate them visually from the answer and set the rules in advance. And if you are an investor or a product manager, watch who is first to launch paid agentic scenarios with real transactions: that data will show how right Moore's thesis about the market's potential is.