Nearly half of Americans have now used AI, and about a quarter of adults open an AI chatbot daily. But a new look at the money behind the boom shows Nobody Pays for AI in any meaningful number. The latest edition of a16z's Top 100 consumer AI apps ranking is the first to track real US consumer credit-card spending, and the numbers are stark: less than five percent of consumers hold a paid personal subscription to a major chatbot, while a tiny slice of power users carries almost the entire market.

For six editions, the rankings ran on traffic: web visits and mobile app downloads. This seventh edition, published in October 2026 by Andreessen Horowitz partner Olivia Moore, layered in spending data from YipitData's US e-receipt panel, and the results describe a market where Nobody Pays for AI even as millions use it daily. ChatGPT still dominates usage by a wide margin, and just eleven products debuted on the combined traffic lists, the fewest ever, a sign the market is consolidating. But the spending list tells a different story: twenty-nine of the fifty vendors collecting the most consumer dollars do not appear in the traffic rankings at all, according to Crypto Briefing's coverage of the release.

Nobody Pays, so the top one percent carry the market

The concentration is extreme. The top one percent of AI spenders account for roughly one-fifth of all observed consumer AI spending — more than the bottom half of paying users put together. These super users average about nine hundred dollars a month on AI products, a figure that has climbed roughly eighty percent in eighteen months. The median paying user, meanwhile, spends about twenty-five dollars a month, a budget that has barely budged, as reported by RuntimeWire.

What do the heavy spenders actually buy? Productivity, not playthings. Their cards over-index on automation and workflow tools like n8n and Manus, plus creative production platforms such as fal, Higgsfield, Figma and HeyGen. The money flows to tools, not toys — because when Nobody Pays for AI subscriptions broadly, the dollars concentrate where the work happens. Moore summed up the gap between reach and revenue this way: "Consumer AI usage is broad, but shallow for almost everyone," she said, as summarized in market coverage of the report's release.

ChatGPT has the crowd, Claude has the expense accounts

On subscribers, ChatGPT is still in a league of its own, with roughly three times as many paid US subscribers as Gemini or Claude. But the runner-up race flipped: Claude has passed Gemini for paid subscribers in the US. And Claude's mix is unusually top-heavy — about seven and a half percent of its subscribers pay for plans of one hundred dollars a month or more, versus roughly one percent for ChatGPT and Gemini.

That split explains the two monetization playbooks. Nobody Pays at scale on ChatGPT, so OpenAI is leaning on volume: its advertising business hit a one billion dollar annualized revenue run rate in August 2026, with shopping and travel queries flagged as the next frontiers. Claude is betting the opposite way — fewer, richer subscribers paying pro-tier prices for deep work. The pressure behind these numbers showed earlier when OpenAI restated its revenue by twenty billion dollars.

Why the free tier is the real product

The paid base is growing fast — the share of consumers with a paid chatbot subscription more than doubled from roughly two percent a year ago to less than five percent this August, and a Pew survey in June found forty-nine percent of US adults had used an AI chatbot, with about one in four using it daily. But doubling from a tiny base still leaves a market where Nobody Pays for AI the way software usually needs. When the median user won't open their wallet, subscriptions alone can't fund the compute bills.

So the firm's take is that consumer AI needs older internet business models to reach everyone else: advertising and transaction fees, as Memeburn's analysis of the report put it. The report is blunt about the tradeoff, saying the pre-AI internet was one where consumers "were the product instead of paying for the product." If Nobody Pays with money, everyone pays with attention — and product decisions start bending toward whoever actually foots the bill.

There are caveats. The spending figures come from a card-spending panel, not a census: they capture US consumer cards, miss corporate purchases and web-direct billing, and describe observed dollars rather than total revenue. Moore herself is more optimistic than the headline suggests — in a recent interview she argued the current economics don't reflect what's coming, pointing to untouched consumer categories from dating to health as future revenue territory, with agents as the wedge. Google is already pushing that direction: Gemini agents recently got their own Gmail and Drive.

For now, the state of consumer AI is a paradox the data makes plain: usage is nearly universal, Nobody Pays, and a handful of power users fund the frontier. Whether the fix is ads, agent fees, or a subscription more people can stomach, somebody eventually has to — because the compute bill is very real. More of our AI coverage lives on the AI News topic page.