ChatGPT subscribers who paid more after price increases responded by exploring more features and using the service more heavily — the opposite of what classical demand management predicts. That behavior, documented in reports cited by researchers at Texas A&M University, sits at the center of a new study arguing that price hikes can worsen rather than relieve congestion for AI and cloud services.

The paper, published in Production and Operations Management, was written by Rajiv Mukherjee of Texas A&M's Mays Business School together with Sreekreamer Bhaskaran of Southern Methodist University and Sanjiv Erat of UC San Diego. Its core finding: for prepaid and subscription services with high marginal costs, a higher price cuts the number of subscribers but pushes each remaining subscriber to consume more.

"These high prices can actually have a double-edged sword effect, where people say, 'We paid a lot, so let's just get our money's worth and use it more,'" Mukherjee said.

Two decisions, not one

Companies have long used price as a congestion valve. When demand strains capacity, raising the price thins the queue of willing buyers. The researchers argue this framework ignores a second decision: once customers pay, they choose how much to use the service.

The team built an analytical model that captures both margins — how many consumers buy access, and how much each one consumes afterward. The mechanism behind the second effect is mental accounting, a behavioral economics concept describing how people track spending against received value.

"When people pay for something, they expect certain value out of it," Mukherjee said. "As soon as you pay for a service, you create a mental account deficit."

The researchers call the resulting tendency to consume more "consumption bias." When that bias is strong, heavier usage among paying customers can outweigh the reduction in subscriber count and actually worsen congestion.

Why AI changes the economics

The problem matters more now because the marginal cost of digital service has flipped. Traditional digital goods cost essentially nothing to serve to an additional user. AI queries consume compute; cloud infrastructure carries real per-unit costs.

"Every query you make, that has a significant amount of cost that the firm has to bear," Mukherjee said of AI services. "Once the customer subscribes, they don't really care. They are just getting things done using the service and trying to get the money's worth in the process."

The paper cites two real-world cases that motivated the work. After price increases, ChatGPT subscribers reportedly expanded feature exploration and usage. At Amazon Web Services, customers responded by burning through more of their precommitted cloud spending before the billing period closed.

"Unlike traditional digital goods where the marginal cost was negligible, modern firms in the post-AI and cloud-computing era have a high marginal cost of service, and they haven't quite figured out how to incorporate that into a good pricing strategy," Mukherjee said.

When cheaper beats higher

The findings do not argue for universal price cuts. When consumption bias is low, raising prices to shed customers remains sound. But when customers are strongly motivated to recover their spend, lowering the upfront price can reduce how intensively they use the service — a counterintuitive result.

The model also shows that usage-based pricing, charging customers partly on consumption, becomes more attractive than a flat subscription fee when each unit of service is costly to deliver. That conclusion lands as hyperscalers and AI providers debate the mix of committed spend, metered inference, and flat-rate tiers.

For subscription businesses, the larger lesson is that the purchase decision is only the opening chapter. "The initial demand through the people who are coming into the system is not the end of the story," Mukherjee said. "That's pretty much the beginning of the story when the marginal cost of service is high."

The full paper, "Pay more, use more: Consumer bias and demand management for digital services," appears in Production and Operations Management (DOI: 10.1177/10591478261454768). As inference costs keep shaping how AI and cloud providers structure pricing, expect usage-linked models to draw heavier weight against flat subscriptions.