Prosumer AI is the phrase that best sums up what Andreessen Horowitz partner Olivia Moore told TechCrunch this week about the state of consumer AI. “Almost all of what we think of as consumer AI, I would argue, is prosumer AI,” Moore said in an interview published on 9 October 2026. The people paying for AI apps today are mostly building, creating and getting work done, not chatting for fun.
The interview followed the release, on Monday 5 October, of a16z’s seventh Top 100 Gen AI Consumer Apps report. For the first time the list adds spending data from US consumer card panels, and the result is stark. Only 4.5% of US consumers had a paid personal subscription to ChatGPT, Gemini or Claude in August, while the top 1% of AI spenders averaged $903 a month.
This article sets out what Moore said, what the a16z data shows, why the prosumer AI market behaves so differently from the consumer internet of the last two decades, and which business models and app categories could change that. We also run the numbers on ChatGPT’s advertising revenue and on Instinct’s transaction volume, and explain what it all means for businesses that buy AI.
Table of contents
- What Olivia Moore Said About Prosumer AI and Consumer AI
- The a16z Top 100: Why Prosumer AI Hides in a Stable Leaderboard
- The Spending Data Shows Prosumer AI, Not Consumer AI
- Why Prosumer AI Grows Into Enterprise Revenue
- The Subscription Problem Facing Prosumer AI Apps
- Ads, Fees and Agents: New Revenue Streams for Prosumer AI
- The Whitespace: Categories Prosumer AI Has Not Reached
- What Prosumer AI Means for Businesses Buying AI
- Prosumer AI and the Ugly Economics Debate
- Prosumer AI Questions Answered
- References
What Olivia Moore Said About Prosumer AI and Consumer AI
The interview was conducted by TechCrunch’s Russell Brandom, who had written on 30 September about the ugly economics of consumer AI. Brandom’s starting point was pessimism: OpenAI has leaned back towards enterprise customers this year, and very few households pay for AI. Moore’s answer was that the pessimism measures the wrong thing.
OpenAI’s enterprise turn is an expansion, not a pivot
Moore agreed that OpenAI “has moved back towards the enterprise”, but said it is “not necessarily a pivot because they’re still launching a lot in consumer”. Her explanation for the shift was simple. Almost all AI revenue so far has come from subscriptions and token usage, and that revenue “is much more concentrated on the enterprise and prosumer side”. A company follows the money, and the money in prosumer AI sits with heavy users and their employers.
Fewer payers, more ways to pay
Brandom cited a16z’s own State of Markets report, which found that just 2.2% of US households pay for AI. Moore’s reply was the most striking line of the interview: “I actually don’t know if I want to see that user number increase.” What she wants instead is “a world where the consumer is monetized in a way that isn’t subscription dollars out of their own pocket”.
She argued that Silicon Valley’s preference for paying over seeing ads is shaped by “high-income” people “with corporate cards”. Most people, she said, “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”.
“It’s definitely very early”
Moore also named the categories where AI has barely arrived: “social apps, dating apps, marketplaces, retail, travel, finance, health”. None has an entrant in the top 100, which she called “pretty surprising” and “what I think we need to see over the next six months”. Asked whether true consumer AI had arrived yet, Moore answered: “It’s definitely very early.”
| Moore’s claim | What the a16z data says |
|---|---|
| Most consumer AI is prosumer AI | Top spenders over-index on coding, automation and creative tools |
| Revenue is concentrated | Top 1% of payers account for 19.5% of observed spending |
| Subscriptions are the bottleneck | 84% of the top 44 AI-native web products sell subscriptions; 14% carry ads |
| Cheaper models will help | ChatGPT Go costs $8 a month; founders report building on open-source models |
| Whole categories are empty | No top 100 entrant in social, dating, marketplaces, retail, travel, finance or consumer health |
The a16z Top 100: Why Prosumer AI Hides in a Stable Leaderboard
a16z ranks the top 50 web products by monthly visits, using Similarweb, and the top 50 mobile apps by monthly active users, using Sensor Tower. Generative AI must be central to the product. Three years in, the authors write that the leaderboard “is becoming familiar”: just 11 products made their first-ever appearance in this edition, the smallest number of debuts across all seven lists.
One rule changed. Products designed mainly for NSFW use are now excluded and treated as a separate category. a16z says they would have made up more than 20% of the web traffic list, so the change is not a sign of falling demand.
ChatGPT’s lead is narrowing
ChatGPT is still first by every measure. In August it had roughly twice Gemini’s web visits and six times Claude’s. On mobile its lead was 2.5 times Gemini and 14 times Claude in monthly active users. But comparing this edition with the sixth edition from March 2026 shows the gap closing on the measures that matter for revenue.
| Measure | 6th edition (March 2026) | 7th edition (October 2026) |
|---|---|---|
| ChatGPT web lead over Gemini | 2.7x | Roughly 2x |
| ChatGPT mobile lead over Gemini | 2.5x | 2.5x |
| ChatGPT US paid subscribers vs Claude | 8x (January 2026) | 3x |
| ChatGPT US paid subscribers vs Gemini | 4x (January 2026) | 3x |
| Suno web rank | #15 | #19 (and #7 by spending) |
By our reading of the two reports, ChatGPT’s lead over Claude in US paid subscribers shrank from eight times to three times in roughly seven months. That is the clearest sign in the data that paying users, the prosumer AI base, are more willing to switch than casual users are.
Claude became the clear number three, then cooled
Claude did not appear on a16z’s first web list in September 2023. It has now passed DeepSeek and Perplexity in traffic and passed Gemini on US consumer subscribers earlier this year. a16z credits new products (Cowork in January, Claude Design in April), the Fable 5 model in June, and attention from Anthropic’s February dispute with the Pentagon, when Claude reached number one among free US mobile apps.
Claude also monetises its users hardest. Some 7.3% of its consumer payers are on the Max plan, which starts at $100 a month, against 1.3% for Google and 1.1% for ChatGPT on their $100 tiers. But by mid-summer YipitData’s panels showed Claude’s daily sessions falling in July and August, while ChatGPT re-accelerated after the July launch of GPT-5.6 Sol, Terra and Luna and the debut of ChatGPT Work. Only 8% of US ChatGPT subscribers also pay for Claude.
Only seven products make all three lists
The new spending ranking is where the prosumer AI story becomes obvious, because it measures money rather than visits. Some 29 of the top 50 vendors by spend do not appear on the web or mobile traffic lists at all. Only seven companies appear on all three: ChatGPT, Claude, Suno, Perplexity, Photoroom, Canva and Notion. Photoroom, a photo editor built on computer vision, shows that a narrow tool can earn real money without a chatbot’s reach.
The Spending Data Shows Prosumer AI, Not Consumer AI
a16z worked with YipitData, whose US e-receipt and card panels track what people actually pay. The firm warns that the data is panel-based, US-only and “should not be read as total company revenue”. Even so, it gives the first clear look at who pays for AI.
The top 1% outspend the bottom half
In August, 4.5% of eligible consumers in the panel had an active paid personal subscription to at least one of ChatGPT, Gemini or Claude, up from 2.1% a year earlier. Among those payers, spending follows a power law. Only 13% of people who pay for one AI product pay for a second, and the top 10% of spenders account for roughly half of all observed spending.
What prosumer AI power users buy
The top 1% averaged $903 a month on consumer cards in August 2026, and their spending rose 80% over 18 months. They “disproportionately purchase tools for automation and product building, like n8n, fal, Manus, and Nous Research”, and over-index on creative tools such as Higgsfield, Figma and HeyGen. In the interview Moore grouped prosumer AI spending into three buckets.
| Category | Examples Moore named | What users get |
|---|---|---|
| Product building | Lovable, Replit, Fal | Apps, sites and media pipelines built from prompts |
| Product marketing | Higgsfield, HeyGen | AI-generated adverts, avatars and video |
| General work management | Manus, Fireflies AI, Granola | Agents, meeting notes and task automation |
“Those are like being paid for by consumers to start with,” Moore said, “but I would argue that they’re not consumer in the way that we would think about pre-AI.”
Our arithmetic: what one prosumer AI user is worth
The a16z figures allow two simple comparisons. First, the top 1% average of $903 a month is about 36 times the median payer’s $25 (903 divided by 25 is 36.1). Second, per head, the top 1% hold 19.5% of spending across 1% of payers, a share of 19.5 per percentage point, while the bottom half hold 16.6% across 50% of payers, or 0.33 per point. On that basis a typical top-1% payer accounts for roughly 59 times as much spending as a typical payer in the bottom half.
That is what prosumer AI looks like in revenue terms: a small group of people who treat AI as a work tool and pay for it like one. It also explains why the median payer “barely” increased spending over time, as a16z notes, while the top end kept growing.
Why Prosumer AI Grows Into Enterprise Revenue
Moore has written about this pattern before. In The Great Expansion, published in September 2025, she argued that AI had changed revenue retention for consumer software. Before AI, a consumer subscription cohort keeping 30% to 40% of its revenue after a year was considered best in class. The fastest-growing AI apps were seeing revenue retention above 100%, meaning each cohort spent more over time, not less.
From six years to 18 months
In the TechCrunch interview Moore gave the timeline that makes prosumer AI different. “Pre-AI, you would have companies like Canva that were consumer-first and often took six or seven years to even add teams or enterprise plans,” she said. “Now we’re seeing companies like Gamma, ElevenLabs, and Cursor that start as consumer and within 18 months become majority-enterprise businesses.”
The mechanism behind prosumer AI’s jump into enterprise is the corporate card. An individual signs up, finds the tool useful, then expenses it, and the employer eventually buys a team plan. That is the second half of the Great Expansion argument: consumers “bringing tools into their workplaces at unprecedented speed”.
Usage pricing changes the maths
The first half is pricing. Instead of one flat fee, prosumer AI apps mix subscription tiers with usage-based credits. A heavy user who generates more video or runs more agent tasks pays more. That is how the top 1% of prosumer AI buyers reach $903 a month, and why a16z found that 64% of the top AI-native web products already sell usage charges or extra credits.
The Subscription Problem Facing Prosumer AI Apps
The report’s fifth finding is the one Moore pressed hardest: “Most consumer AI products monetize via subscription. We need new business models.” Of the 44 AI-native products in the top web rankings, every one has some way to make money live. The methods are lopsided.
Products can use more than one model, so the shares add up to more than 100%. Applied to 44 products, the percentages work out to roughly 37 selling subscriptions, 28 selling credits, 6 showing ads and just 1 taking a cut of transactions.
An inversion of the old internet
a16z calls this “an inversion of the pre-AI consumer Internet, where consumers were the product instead of paying for the product”. The comparison with the giants of the last era is sharp.
| Company or group | Main way it makes money | Figure cited by a16z |
|---|---|---|
| Meta | Advertising | 97.6% of 2025 revenue |
| Alphabet | Advertising | 73.2% of 2025 revenue |
| Amazon | Retail; subscriptions a small line | Subscription services $49.6 billion, 6.9% of revenue |
| Top 20 global consumer subscriptions | Mostly media | 65% are media businesses such as Netflix and Spotify |
| Top 44 AI-native web products | Subscriptions and credits | 84% subscriptions, 14% ads, 2% transaction fees |
Why prosumer AI started with paywalls
There is a reason for the inversion, and Brandom raised it directly: the marginal cost of AI “is still a lot higher than classic internet services like Facebook or Google Search”. Ads and transaction fees need a large user base before they pay off. When serving a user cost almost nothing, consumer companies could afford to give the product away while they built scale. With expensive models, a16z writes, “forgoing early revenue has not been an option in consumer AI”.
Cheaper models could change it
Moore pointed to ChatGPT’s Go plan at $8 a month, which she assumes “runs on cheaper models”, and said founders increasingly build on open-source models. “We don’t necessarily need frontier intelligence for every task,” she said. The catch is that today’s paying users are doing “coding and other technical automation, where they probably do need frontier intelligence”. As more products are built where “the model is not the product”, cheaper models should spread, and the cost barrier to free, ad-supported prosumer AI and consumer apps should fall.
Ads, Fees and Agents: New Revenue Streams for Prosumer AI
If subscriptions cap the market, the obvious alternatives are the two that built the last internet: advertising and transactions. Both have started to appear in AI.
Advertising: our maths on ChatGPT’s $1 billion
a16z reports that OpenAI said ChatGPT advertising reached a $1 billion annualised revenue run rate in August, on a base of 1.2 billion weekly active users. Spread across the whole base, that is about $0.83 per weekly user per year ($1 billion divided by 1.2 billion). A single ChatGPT Go subscriber at $8 a month pays $96 a year, roughly 115 times as much.
Those numbers show both sides of Moore’s argument. Ads already bring in serious money at ChatGPT’s scale, but per user they are still tiny next to a subscription, so ads only work for products with hundreds of millions of users. Medical search tool OpenEvidence, which a16z estimates is used by 50% to 60% of US physicians, now partly monetises through ads too.
Transaction fees: what Instinct’s volume implies
Personal agents open a third route for prosumer AI and consumer apps alike. According to a16z, Instinct founder Noah Shinn has reported that 40% of users connect a personal credit card within three weeks, that those users spend an average of $1,300 a month through Instinct once they buy something, and that more than $1 billion in annualised transaction volume already flows through the agent, half of it travel. Both Instinct and Meta’s Muse have described plans to take a fee on transactions, but neither does yet.
The table below is our illustration of what a fee could be worth on that volume. The take rates are assumptions, not company figures.
| Assumed take rate on $1 billion a year | Annual revenue | Equal to this many $96-a-year subscribers |
|---|---|---|
| 1% | $10 million | About 104,000 |
| 3% | $30 million | About 313,000 |
| 5% | $50 million | About 521,000 |
Instinct passed 100,000 users in its first three weeks, according to a16z. On those assumptions, even a 1% fee would earn as much as signing up every one of those users to an $8 plan, without asking anyone to subscribe. That is the business case Moore is hinting at. For the product details, see our hands-on comparison of Instinct and Muse.
The platform fights have started
Transactions bring a new risk: the stores have to let the agents in. a16z notes that Amazon shut down access for Muse in under two weeks, while Shopify, Instacart, OpenTable, Expedia, Ticketmaster and Plaid have signed official integrations. Muse itself reached a reported 250,000 daily active users in its first week and more than 5 million downloads in under a month, yet that is still about a third of Threads’ 15 million downloads in its first 22 days. As a16z puts it, “the primary battle is against apathy”. Our earlier piece on Meta’s Muse downloads tracks that launch.
The Whitespace: Categories Prosumer AI Has Not Reached
Moore’s list of empty categories is the most practical part of the interview for founders. Social, dating, marketplaces, retail, travel, finance and health have no entrant in the top 100. a16z’s report makes the same point from a different angle: “almost none of the top consumer AI products today have any real multi-player appeal”, even though many of the biggest consumer internet businesses were built on transactions and networks.
Why the gap exists
The list maps closely onto categories that need either a network of other people (social, dating, marketplaces) or trust with money and personal data (finance, health). Those are hard to build on the prosumer AI subscription model and hard to win with a single clever prompt. Health is a partial exception for professionals: OpenEvidence debuted at #47 on the web list serving physicians. But consumer health, like consumer finance, remains open. Travel is also moving, but through agents rather than apps, with half of Instinct’s transaction volume going on trips.
Where startups can still win
Incumbents are now prominent on the list. Canva and Notion are in the web top ten, Figma is in the top 15, Superhuman debuted at fourth on spending, and Google alone has five web entries. a16z says startups that are attracting lasting attention tend to own one of four things.
| What they own | Examples from the report |
|---|---|
| A differentiated model | Suno (#19 web, #7 spend), ElevenLabs (#25 web, #10 spend), Midjourney, HeyGen, Kling, Topaz Labs |
| A multi-model experience | Cursor (#41 to #35), OpenRouter, Lovable, Replit, Base44 (debut at #46) |
| An audience with specific needs | OpenEvidence (#47) for physicians, Venice (#43) for private AI |
| An experience a legacy interface cannot copy | Plaud (#16 on spending), which sells a physical notetaker plus a subscription |
What Prosumer AI Means for Businesses Buying AI
For a UK business, the prosumer AI pattern is not an abstract investor theme. It describes how AI arrives in the workplace: one employee at a time, on a personal or corporate card.
Expect tools to arrive through expense claims
If Gamma, ElevenLabs and Cursor can become majority-enterprise businesses within 18 months, many of their enterprise customers started with staff who signed up on their own. That makes expense reports a useful audit trail. Reviewing card spend on AI apps every quarter is a cheap way to find unapproved tools before they hold client data, and it belongs in any AI strategy alongside data-handling rules and basic cybersecurity checks on each vendor.
Budget for usage, not just seats
The a16z spending data shows prosumer AI bills that grow with use. A $25 prosumer AI subscription becomes a $900 bill when someone starts generating video or running agents all day. Budget owners should set credit caps and alerts per user, and compare the cost of a team plan with the sum of individual expense claims.
Watch the free, ad-supported tiers
If Moore is right that ads and transaction fees are coming, free tiers will become more capable and more commercial. That is good for access, but it changes the data trade. Before staff use a free AI tool for work, check what the provider does with prompts, whether ads are targeted on conversation content, and whether a paid tier with stronger data terms exists.
Prosumer AI and the Ugly Economics Debate
Brandom’s earlier column and Moore’s interview read the same facts in opposite ways. Our analysis of the ugly economics of consumer AI found that payer numbers are low and costs are high. Our report on why only about 2% of consumers pay for AI reached a similar conclusion from survey data.
Moore does not dispute those numbers. Her argument is that they describe the subscription model, not the size of the opportunity. Payer numbers more than doubled in a year (from 2.1% to 4.5%, a 2.1 times increase), the heaviest prosumer AI users keep spending more, and two business models that built the last internet barely exist yet in AI. The pessimists see a market that cannot charge enough; Moore sees one that has not yet learned how to charge differently.
The test over the next six months is visible in the data a16z will publish next. If Moore is right, the eighth edition should show more ad-supported products, the first transaction-fee revenue from agents, and at least one entrant in a category that is empty today.
Prosumer AI Questions Answered
What is prosumer AI?
It describes AI products bought by individuals for professional or semi-professional work, such as coding, design, video and meeting notes. Moore argues that almost all of what people call consumer AI today falls into this group.
How many people pay for AI in the US?
According to YipitData’s panel, as cited by a16z, 4.5% of US consumers had a paid personal subscription to ChatGPT, Gemini or Claude in August 2026, up from 2.1% a year earlier.
Which app leads a16z’s Top 100 list?
ChatGPT ranks first on web traffic, mobile users and consumer spending. Claude is now the clear third general assistant behind ChatGPT and Gemini.
Will more AI apps show ads?
Moore expects so. Today only 14% of the top 44 AI-native web products carry ads, but ChatGPT advertising has already reached a reported $1 billion annual run rate.
Which consumer categories have no AI app in the top 100?
Moore listed social, dating, marketplaces, retail, travel, finance and health.
References
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
