This week, I explored the harsh economic realities facing consumer AI—though what looks like a crisis to some is a goldmine to others. Olivia Moore, a partner at Andreessen Horowitz who focuses on consumer AI, published a report on Monday identifying the top 100 consumer AI applications. While ChatGPT dominates by an enormous margin, emerging tools like Suno and ElevenLabs are demonstrating genuine durability. Perhaps most intriguing is what's absent: Moore's findings reveal roughly six consumer segments that AI appears to have entirely bypassed.

Moore believes consumer AI holds enormous potential, especially if companies can unlock revenue sources beyond subscriptions and API fees. During a conversation this week, we discussed the peculiar economics of consumer AI and why the sector remains in its infancy. The interview has been condensed and edited for clarity.

The timing of a consumer AI report is notable, given OpenAI's renewed focus on enterprise customers this year. This shift has fueled considerable skepticism about consumer AI's revenue prospects—myself included. Yet you seem more hopeful.

I acknowledge that OpenAI has shifted back toward enterprise. However, I wouldn't call it a pivot since they continue launching many consumer products. It's more like an expansion. I don't fault them—despite my enthusiasm for consumer, nearly all AI revenue so far has derived from subscriptions and token usage, which skews heavily toward enterprise and prosumer markets.

How might consumer AI products address this revenue challenge? The State of Markets report highlighted that merely 2.2% of American households pay for AI. Does that figure need to climb before the market becomes appealing?

Honestly, I'm not certain I want that user number to grow. What interests me more is returning to a model where consumers are monetized without subscription fees coming directly from their wallets. In Silicon Valley, where people have high incomes and corporate cards, it's simple to say, "I'd rather pay for the product." But most people would prefer free access with some advertising, then decide whether to subscribe to remove ads.

Another hurdle is the marginal cost of AI services, which remains significantly higher than traditional internet offerings like Facebook or Google Search. Are you seeing many of these services finding ways to reduce those expenses—through open-source or lighter-weight models?

Yes, improvements are happening. Even ChatGPT now offers a Go plan at just $8 monthly, which I assume uses cheaper models. We don't require frontier intelligence for every task, so cheaper models could work, particularly for certain consumer applications. It might also mean more consumer companies building on open-source models. Our founders have mentioned this frequently, and it genuinely feels like a shift is underway. The issue is that current revenue-driving users are engaged in coding and technical automation, where frontier intelligence is likely necessary. As more people build for consumers—where the model isn't the product—we'll likely see lower-cost models adopted more widely.

Discussing technical automation within a consumer service is unusual, which seems like another distinctive AI characteristic. Are the boundaries between consumer and enterprise beginning to blur?

I wrote about this roughly a year ago in an article called The Great Expansion. Before AI, companies like Canva were consumer-first and often took six or seven years to introduce teams or enterprise plans. Now companies like Gamma, ElevenLabs, and Cursor begin as consumer-focused and within 18 months become predominantly enterprise businesses. I'd argue that nearly everything we consider consumer AI is actually prosumer AI. This becomes evident in the revenue rankings, where essentially three categories drive power-user spending: product-building apps like Lovable, Replit, and Fal; product marketing tools such as AI ad generators Higgsfield or HeyGen; and general work management platforms like Manus, Fireflies AI, and Granola. Consumers initially pay for these, but I'd contend they aren't consumer in the pre-AI sense.

Then there are all these untouched categories: social apps, dating apps, marketplaces, retail, travel, finance, health. Our top 100 list contains no entrants in those areas, which is quite striking. That's what I believe we need to see in the next six months.

So perhaps we haven't truly witnessed genuine consumer AI yet.

It's absolutely still very early.