a16z Podcast

Olivia Moore & Josh Elman on consumer AI's power-user game

Olivia Moore & Josh Elman· Partners, a16z consumer team at Andreessen Horowitz (a16z)
·~49 min·English·a16z
AI CompanyAgentsBusiness StrategyLLM
TL;DR

In a16z's seventh top-100 consumer AI report, Olivia Moore and Josh Elman show that consumer AI today is a concentrated, subscription-funded power-user market, and argue the next phase broadens it through agents that take action, ad and transaction business models, and products whose moat is accumulated context rather than the model.

01Core Mental Model

Consumer AI's Power-User Game

<strong>Consumer AI spend follows a brutal power law:</strong> only about 4.5% of US consumers pay at all, and within them the top 1% spends $903 a month while the median payer spends $25.

The top 1% user is spending $903 per month personally on their personal credit cards on AI.

— Olivia Moore, a16z Podcast
Key Insight
About half of Americans report using AI, but only around 4.5% pay for a subscription, and within that sliver the spend is wildly skewed. A product optimizing for the median payer at $25 a month is chasing a very different customer than the one spending $903, so the real revenue opportunity is far narrower than the usage headlines suggest.
02The Big Shift

From a Tool to Something That Acts

<strong>The biggest new trend is personal agents</strong> — AI moving from a tool you prompt to software that actually gets things done for you.

the most interesting big new trend is in personal agents.

— Josh Elman, a16z Podcast
Key Insight
Moving from a tool to an agent that acts also moves the trust burden. Once the software touches your email and your credit card, the failure mode is no longer a bad answer you can ignore but a bad action you cannot undo — which is why consumer adoption keeps lagging the capability.
03Business Models

The Subscription Inversion

<strong>An investor in the space argues the subscription-first model is an inversion</strong> that has to give way to ads and transaction fees before consumer AI can reach everyone.

I actually don't necessarily want to see that 4.5% of people paying for AI products directly expand.

— Olivia Moore, a16z Podcast
Key Insight
Moore is arguing against her own short-term interest as a consumer-AI investor. Her point is that a subscription is a toll only the committed few will pay, so the market can broaden only if ads and transaction fees carry the cost instead — which makes the business-model question the real gate on mainstream adoption.
04Advertising

$1B in Ads, Years Faster Than Before

<strong>OpenAI reached a roughly $1B annual ad run-rate far faster than any company before it</strong> — on the back of about 1.2B weekly active users and the deep personal context it has on them.

Like that would previously take a company years and years and years to get to even after launching an ads product.

— Olivia Moore, a16z Podcast
Key Insight
The speed is not just about reach — personalization is the multiplier. The same intimate context that makes an assistant feel trustworthy is what makes its ads convertible, so the privacy tension and the advertising upside are the same asset seen from two sides.
05Competitive Dynamics

Claude Passed Gemini on Paid Subscribers

<strong>Claude has passed Gemini on paid subscribers</strong> despite Gemini's far larger user base, a sign that monetization strategy is beating raw reach.

Claude has actually passed Gemini in terms of number of paid subscribers, which is kind of crazy, given Gemini is a much bigger install base and user base overall

— Olivia Moore, a16z Podcast
Key Insight
Gemini has Google's distribution and still trails Claude on paid subscribers. That is evidence that in consumer AI, willingness to pay is set by the product and the brand around it, not by how many people a company can reach for free — the lever the incumbents assumed would win.
06Where Value Accrues

More Than a Shim on the Model

<strong>These products are much more than a shim on the model</strong> — the real moat is the accumulated context and playbooks about who you are, not the model underneath.

Town has built these playbooks of who I am, what I sound like, that no other product has, and that I cannot easily take and migrate over to something that comes along tomorrow.

— Olivia Moore, a16z Podcast
Key Insight
If the moat is accumulated context rather than the model, then the labs' own model lead does not automatically win the consumer layer. The defensible asset is the playbook a product has built about you over time — which is exactly why 'wrapper' is the wrong mental model for these companies.
07The White Space

Save Time vs. Spend Time

<strong>Consumer AI has only won the 'save time' categories so far;</strong> the larger 'spend time' categories like social, dating and entertainment remain wide open.

Almost everything we've seen in consumer AI is save time. And that spend time, those end up often actually being amongst the biggest companies, if not the biggest ones.

— Olivia Moore, a16z Podcast
Key Insight
Every category consumer AI has won so far helps people do a known task faster. The 'spend time' categories — social, entertainment, dating — are both historically the largest consumer businesses and the ones today's models cannot yet serve well, so the biggest consumer outcomes are still gated on model capability, not distribution.