Data and Technology

Gen AI Has Many Fans, but Few Consumers Pay For It

An analysis of billions of transactions by Mark L. Egan illustrates the fierce competition generative AI companies will likely face, especially as products become more interchangeable.

Generative AI seems to be everywhere, but fewer US consumers are paying to access the technology than the hype suggests, and the battle to win them over will likely be fierce.

Research by Harvard Business School Professor Mark Egan and colleagues finds that only 2.9% of consumers pay for at least one major gen AI service. Their analysis estimates that 6 million individuals paid for subscriptions as of April—about two-thirds the number of YouTube TV customers.

As Anthropic prepares to go public and OpenAI intensifies efforts to secure investments before a potential 2027 IPO, Egan’s research offers a sobering view of AI companies’ ability to turn household demand into sustained profits.

The team focused on paying customers, who accounted for about 60% of OpenAI’s revenue at the beginning of 2026. Enterprise clients, not included in the analysis, make up a larger share of Anthropic’s revenue, but individual subscribers still historically accounted for 20%, the authors detail in the October working paper “Competition in Generative AI: Evidence from ConsumerDemand.”

“Unlike free subscriptions, paid AI subscriptions reflect providers’ ability to monetize consumer demand and are a core source of their revenue,” write Egan and HBS doctoral students Dominic Russel and Claire Shi, who conducted the analysis with Vincent Yao, a professor at Georgia State University.

Their findings reveal several potential challenges to growth for some of the largest generative AI providers: OpenAI, Anthropic, Google, and xAI.

The research team analyzed 114 billion transactions collected by fintech firm Fiserv from early 2021 through April 2026. Gen AI subscribers tend to earn more than $200,000 per year, more than twice the median income in the US. People younger than 55 and those who live in coastal or major metropolitan areas are more likely to open their wallets.

While Americans routinely subscribe to multiple media companies, such as Netflix and Disney+, retention rates one year after an initial signup for the four major AI companies analyzed ranged from 20% to 45%, lower than those for popular subscription services.

As the major tools become more “general purpose”—offering coding and writing in one tool, for example—people will likely be less compelled to subscribe to more than one, the researchers say. The team created a “multihoming index” to measure consumers’ willingness to subscribe to multiple services. They found that the index was high but declining across all companies.

As part of their analysis, the research team used the data to model consumer preferences for paid AI services and how they might shape the industry. Among their findings:

Subscribers tend to be AI enthusiasts, but not platform loyal. Over time, Anthropic’s and OpenAI’s platforms have become more substitutable, fueling competition. The model points to “strong competitive pressure,” as companies know that setting prices too high will drive consumers to rival providers.

Some customers might be willing to pay a little more. “Consumers in the top quartile of the preference distribution for ChatGPT are willing to pay $37 more per month for ChatGPT than consumers in the bottom quartile.” However, companies would need to tread carefully: a 10% increase in price lowers demand by 46%.

Consumers have the upper hand for now. Researchers estimated AI subscriptions generate about $35 per quarter in extra value for paid subscribers. “Our results suggest that, absent a substantial increase in AI model quality differentiation, growing substitutability among services will constrain providers’ pricing power and allow consumers to capture a large share of the surplus generated by these services.”

Disclosures: Egan and Yao are unpaid members of Fiserv’s Small Business Index advisory council. Yao also serves as a paid consultant.

Competition in Generative AI: Evidence from Consumer Demand

Egan, Mark, Dominic Russel, Claire Shi, and Vincent Yao. "Competition in Generative AI: Evidence from Consumer Demand." Harvard Business School Working Paper, No. 27-037, October 2026.


Latest from HBS faculty experts

Expertly curated insights, precisely tailored to address the challenges you are tackling today.

Strategy and Innovation

Social Responsibility

Data and Technology