a16z Tracks Real AI Spending for the First Time: 4.5% of Americans Pay, the Top 1% Spend About $900

Andreessen Horowitz's latest Top 100 AI ranking is the first to rely on actual US consumer spending. The result is surprisingly down to earth: nearly half of Americans use AI, but only 4.5% have a paid subscription. Meanwhile, the top 1% of payers spend about $900 a month, and most of that money goes to professional tools for development and automation.

Money Instead of Traffic

Consumer app rankings are usually built on visits and installs. Those metrics show interest, but they don't reveal who is willing to pay, or how much. A chatbot with hundreds of millions of visits can earn less than a niche tool with a handful of professional customers. By adding real payment data to the list, a16z for the first time compares reach with revenue at the user level.

For the industry, this is an important correction. Claims about "AI adoption by the public" have long rested on counts of active accounts, and free tiers plus answers built into search engines and messaging apps inflated those numbers.

The Funnel: From Half to 4.5%

As a rough estimate, about one in ten people who use AI at all pays for it. The rest get by on free tiers, trial periods, or features built into products they already pay for, or don't pay for at all. For the mass consumer, AI remains a nice-to-have rather than a budget line.

Competition from free offerings works against subscriptions here. As long as a basic model is available at no cost and handles everyday tasks like translation, drafting an email, or explaining a term, most people have no reason to move to a paid plan.

A Thin Layer of Heavy Spenders

The top 1% of payers spend about 900 dollars a month, or roughly 10,800 a year. For comparison, standard consumer chatbot subscriptions cost about 20 dollars a month, so the gap is roughly 45x. The money goes to professional tools: development and automation. Most likely, these are agentic coding services, automation platforms, and usage-based plans where the bill grows with load. That is an inference from the category descriptions, not data from the report itself.

The available summary does not give the share of revenue that this top 1% generates. But the size of the bill shows where paying demand is concentrated: among people for whom AI directly saves working time or replaces part of their labor.

Key figures from the report:

• nearly half of American consumers use AI;
• 4.5% have a paid subscription;
• the top 1% of payers spend about $900 a month;
• the main spending categories are development and automation tools;
• for the first time, Top 100 AI accounts for actual US consumer spending.

What Product Builders and Investors Should Take Away

Betting on a mass-market $20 subscription requires a very wide funnel: even with good conversion, paying users end up as a low single-digit percentage. Products aimed at professionals earn from a smaller audience, but with a vastly larger bill per user.

Founders should count not the number of registered users but the share of people for whom the product has become a working tool, and test whether that segment is ready for pricing above the standard tiers. Investors should separate reach from the paying base: the metric "nearly half the country uses it" says nothing about revenue. For ordinary users, the data offers a reason to rethink their own spending. If a subscription doesn't save working hours, free tiers, judging by the statistics, satisfy most people.