Why Hardly Anyone Pays for AI Yet
Only 2.2% of consumers paid for AI services as of May 2026, paying about $31 a month on average. That small share of paying users and that small monthly amount describe the problem with selling AI to the public, according to data cited by TechCrunch.
Netflix had 325 million subscribers in the same reporting. Many people use AI every day. Very few pay for it directly.
Sales to businesses grew fast. Anthropic revenue rose 12-fold in 2025 to nearly $4.6 billion, as reported by Reuters. Earlier in that growth, the company reached $3 billion in annualized revenue, a figure reported in May 2025 by Reuters. OpenAI had projected at that time it would end 2025 with more than $12 billion in total revenue.
Costs grew faster. Anthropic spent $7.33 billion on computers and data centers last year. Its operating loss, the gap between spending and income, widened to $8.06 billion, according to Reuters. It also pays over $350 million in distribution fees to cloud platforms, as reported by Reuters. When AI is sold through app stores, cloud marketplaces and bundled assistants, store cuts and customer costs leave even less of that $31.
Cheap open models, systems anyone can copy and use, have caught up in ability with frontier models, the most advanced paid systems, according to Ars Technica. Paying for frontier models buys a 4-month head start at 5x the cost, in the same reporting. The wider economic payoff is still small. AI-related investment has lifted U.S. GDP, total economic output, by less than 0.4% since 2022, according to Reuters. Spending on equipment came first. Broad gains in productivity have not followed.
The broader context here is costs and income do not line up. Building and running large models needs steady spending on chips, power, networks and staff. Money from the public arrives in small monthly payments from very few people. Business contracts help fill the gap, and Anthropic revenue shows they are doing much of that work, but consumer sales remain high-volume with low return per person.
In my view, the PC era and the early commercial internet followed a similar payment pattern. Broad access lost money while paying uses grouped in workplaces, where saved time has a clear price. Consumers gained a great deal, just not through a direct subscription. Free, ad-supported, bundled and employer-paid access carried those earlier waves long before household spending caught up.
Looking at what this means for builders, three pressures stand out. Running answers must get cheaper, features must still matter after that four-month window, and selling must avoid heavy platform fees. If open models handle routine summarization, coding help, search and drafting at near-frontier quality, then paid plans need lasting strengths in memory, ability to act alone, connection to other apps, or trust. Not higher test scores alone. A brief lead rarely holds a 5x price once good-enough options exist. Otherwise $31 a month stays a ceiling paid only by heavy users, while most people stay on free plans paid for by someone else.
Worth flagging is this can still turn out well. Few people paying now does not fix how many will pay later. Once models sit inside daily routines and reliably finish multi-step tasks, payment can shift from novelty subscription to utility billing, like power or water. Infrastructure built at a loss today lowers delay and cost per answer for that future. Short-term math is hard. The longer direction still points to cheaper, capable help becoming widely available, even if businesses pay first.
