a16z's Top 100 Consumer AI Apps vs. Stripe data: five openings for indie builders
a16z's seventh Top 100 shows consumer AI spend concentrated in power users and 9 of 15 categories empty. With Stripe's data: five openings for indie builders.
On October 5, 2026, a16z published the seventh edition of its Top 100 Consumer AI Apps. For the first time it ranks products by what U.S. consumers actually spend on their cards, using data from YipitData. The takeaway for anyone building consumer AI products on their own is blunt: plenty of people use AI, very few pay, and the money is concentrated. This piece reads that list next to Stripe’s public data on AI companies and asks where a solo developer or small team can still get in.
Key points
- As of August 2026, only 4.5% of U.S. consumers held a paid ChatGPT, Gemini or Claude subscription. The top 1% of paying users made up 19.5% of observed spend, averaging $903 a month against a $25 median (a16z, YipitData).
- a16z sorts consumer internet into 15 categories. Nine have no product in the AI Top 100: streaming, social, dating, gaming, travel, retail, finance, real estate and jobs.
- Stripe says its top AI companies grew revenue 175% in 2026, take 48% of revenue from outside their home market, and earn 18% more cross-border revenue with localized pricing.
What the two datasets measure
The a16z list and Stripe’s reports answer different questions, so they can’t be merged into one table. One caveat first: Stripe has not published a revenue-ranked list of AI companies. It shares aggregate growth figures and a handful of named revenue milestones. “Comparing” them here means comparing two views of the same market.
| a16z Top 100 Consumer AI Apps (7th ed.) | Stripe’s AI company data | |
|---|---|---|
| What it tracks | Web visits (Similarweb), mobile MAU (Sensor Tower), U.S. consumer card spend (YipitData) | Revenue, countries reached and pricing models of top AI companies paid through Stripe |
| Who is in it | Consumer-facing products, including desktop apps and agents inside messaging apps | Mostly developer and business-facing sellers, plus Link consumer data |
| Limits | Spend ranking is U.S.-only and panel-based, not total company revenue | Covers only companies that use Stripe; figures are disclosed by Stripe itself |
| Best for | What consumers will pay for | How to collect faster and sell in more countries |
Put together, a16z tells you where to build and Stripe tells you how to get paid once you have.
Who holds the money
Paying for consumer AI is rare, and the spend is lopsided. In YipitData’s panel, 4.5% of U.S. consumers had an active paid subscription to at least one of ChatGPT, Gemini or Claude in August 2026, up from 2.1% a year earlier. Among people who pay for one AI product, only 13% pay for a second.
The tail is where it gets interesting. The top 1% of payers account for 19.5% of observed spend, more than the bottom 50% combined (16.6%). They average $903 a month on AI, up 80% in 18 months, while the median payer spends $25 and has barely moved. The top 10% make up roughly half of all spend.
What the heavy spenders buy tells you where the business is. a16z says they over-index on automation and building tools (n8n, fal, Manus, Nous Research’s Hermes Agent) and on creative tools (Higgsfield, Figma, HeyGen). Of the top 50 vendors by spend, 29 appear on neither the web nor the mobile traffic list. Traffic and revenue are different things.
Stripe’s numbers point the same way on a narrower base. At Sessions 2026, Stripe said the number of Link consumers buying AI products rose from just under 6 million to over 14 million in a year, and top Link buyers now spend $371 a month. That figure can’t be set against a16z’s $903: one is Link’s top buyers, the other the top 1% of payers in a card panel. What the two share is the shape. A small group of heavy users carries the spend, and it is growing.
The gap is spending time, not saving it
In a post on X, a16z noted that 9 of 15 consumer internet categories have zero products in the AI Top 100, under the line: “Most people aren’t looking to save time, they’re looking for ways to spend their time.”
Those categories (streaming, social, dating, gaming, travel, retail, finance, real estate, jobs) hold the biggest companies of the last two consumer internet eras. Writer sleepy0x13 turned that into a claim: AI startups have spent three years betting on “saves you time,” and the next big consumer AI may look more like TikTok, a game or a social network, where users have no task and no prompt and still stay two hours a day.
Treat that as a hypothesis. The list supports only its first half: today’s top AI products are almost all tools. a16z itself notes that almost none have real multiplayer appeal.
Five openings
Each row ties to a signal in the a16z list or Stripe’s data. The “catch” column is our own judgment of the risk, not a finding from either source.
| Opening | Evidence | Catch |
|---|---|---|
| Paid tools for heavy users | Top 1% of payers average $903 a month; Cursor moved from #41 to #35 on web, Base44 debuted at #46 | Closest to the labs and to Cursor, so most crowded |
| Creative tools built on your own model or style | Suno is #19 on web traffic and #7 on spend; ElevenLabs #25 and #10 | Needs a model, voice or dataset that is truly different; Google and OpenAI image and video models are catching up |
| Products for a defined audience | OpenEvidence debuted at #47 serving physicians, an estimated 50–60% of U.S. doctors use it; Venice at #43 for private AI | Small audiences, compliance and trust work |
| Turn “spending time” into a product | 9 categories empty; excluded NSFW products would have taken over 20% of the web traffic list | Nobody has shown retention yet; you need daily stickiness, not task completion |
| Agent commerce with a take rate | Only 2% of 44 AI-native products charge transaction or platform fees; Instinct’s founder says 40% of users add a personal card within three weeks and spend $1,300 a month once they buy | Platform risk: Amazon cut off Meta’s Muse in under two weeks |
Two notes. The NSFW figure shows that traffic appears when a product gives people a reason to stay; a16z excluded those products from this edition, so it is evidence of demand, not a direction to follow. And Instinct’s $1 billion annualized transaction volume is the founder’s own claim, which a16z reports as such. Treat that row as something to watch, not a plan.
For the personal-agent line, see our comparison of Muse, Manus 2.0 and Grok Bot.
How you charge shapes the business
a16z looked at the 44 AI-native products on its web list: 84% offer subscriptions, 64% usage charges or extra credits, 14% have ads and 2% take transaction fees. Nearly all revenue comes from users paying directly, the reverse of how the last two consumer internet eras made money.
Stripe’s data is more specific about what to do. These figures come from Stripe’s Sessions 2026 talk by Maia Josebachvili, Stripe’s chief revenue officer for AI:
- Usage-based pricing is now standard. Two in three Forbes AI 50 companies use some form of it, mostly a subscription plus credits, up from under half last summer.
- Localized pricing. It lifts cross-border revenue 18%, and adding one local payment method lifts conversion by more than 7%.
- Idea to first charge. On Replit and Vercel, builders now reach their first paying customer in under six weeks.
- Global from day one. AI companies reached 42 countries in year one and 120 by year three, and top AI companies get 48% of revenue from outside their home market, up from 33% three years ago.
For an indie developer this is concrete. Don’t ship a pricing page with one monthly tier. Give heavy users a usage or credits path, show consumption before the bill arrives, and support local payment methods such as Pix in Brazil at launch. Stripe can configure all of that, and our low-cost indie stack lists alternatives for the payments layer.
Counter-arguments and limits
Several limits apply. a16z’s spend ranking covers the U.S. only and comes from a panel; a16z says it should not be read as company revenue. Its traffic lists also include incumbents that have made AI central (Canva, Notion and Figma are all in the top 15), so “nine empty categories” means AI-native or AI-centered products, not that these categories lack AI features. Stripe’s data comes from companies that use Stripe and skews toward fast-growing winners willing to share numbers. The milestones for Lovable, Cursor and Anthropic are self-reported run-rates and aren’t measured the same way.
There is a deeper objection too. Heavy users pay a lot because their output converts directly into money. People who spend time on entertainment have no such budget, so those products would have to earn from ads or transactions. a16z reaches the same conclusion: consumer AI probably needs a new business model, or an old one. OpenAI said ChatGPT advertising reached a $1 billion annualized run rate in August on 1.2 billion weekly users, a scale no indie team will touch soon.
What to test this week
Before building anything, run three cheap checks.
- Pick one row from the table and put up a landing page with a single paid tier, using the Stripe MCP server or Stripe test mode. See whether anyone leaves a card before the product exists.
- For two of a16z’s nine empty categories, find an existing AI-based experience and read the discussion on X or Reddit. Count whether people say they “use” it or “play” with it.
- Add a credits tier to your pricing page and show usage as it happens. If users pick credits over the monthly plan, your heavy users have shown up.
After that you’ll have three sets of your own data, which will answer “which row should I build” better than any ranking.