Who Actually Pays for AI Products

Most AI product strategy is built on vibes. The narrative comes from earnings call transcripts, VC demo day decks, and hot takes on social media. But there is harder data available, and it paints a picture that the hype cycle mostly ignores.

Two independent sources published recently tell a more grounded story. Stripe's Link team analyzed spending patterns across 250 million customers. Noam Segal and Lenny Rachitsky ran their second annual tech worker sentiment survey with thousands of responses. The two data sets converge on three findings that every AI product builder should internalize.

The Superuser Cohort Is Accelerating Faster Than Anyone Expected

Stripe tracked Link customers by their monthly spend on AI products. The top 10% went from $84 per month to $183 over 22 months, steady gradual growth. Then something shifted. In the next three months, that same cohort jumped from $183 to $359. Nearly double, in a single quarter.

That timing is the story. It took almost two years to build the first $99 of growth. It took three months to add the next $176. The curve did not slope upward gradually. It bent.

For product builders, this has a concrete implication. The people who use AI products most heavily are spending dramatically more, and they got there fast. If your product targets this cohort, your revenue trajectory could look nothing like a typical SaaS ramp. But if you are building for the median user, the numbers look different. At the 50th percentile, monthly AI spend went from $60 to $72 over the same period. Modest.

The question is not whether AI has product-market fit. It is which market. The superuser segment is already spending like enterprise software customers. The middle is still cautious.

The People Spending Are the People Building

Stripe isolated spending on app-builder platforms like Replit, Lovable, and Bolt. The top 10% of AI spenders now spend 5x more on these platforms than they did in January 2025.

That is a striking multiplier. The AI economy, in its current form, is dominated by people building things. The gold miners are buying shovels from each other. It is a market where the customers are also the suppliers.

This maps directly onto the tech worker survey. The workforce is bifurcating along exactly this line. 49% of respondents said AI has amplified them, letting them do more and better work. 14% said it has destabilized them. 5% said diminished. The AI identity question was the single strongest predictor of career optimism, stronger than role, company size, or seniority. The gap between the amplified and the diminished groups on optimism was roughly three times the size of the famously strong founder effect.

The people spending money on AI products are the amplified cohort. They are the ones for whom the tools are working. The rest of the workforce is either watching from the sidelines or actively struggling.

If you are building an AI product, this tells you something about your addressable market. The amplified cohort is roughly half of the tech workforce. They are your early adopters. The other half is not yet ready to buy. They need different products, different positioning, and a different value proposition than just go faster.

The Real Fear Is Not Job Loss. It Is the Squeeze.

The survey asked tech workers what they actually worry about regarding AI's impact on their careers. Only 22% said losing my job to AI. The top fear (51%) was being expected to do more for the same pay. Followed by getting trapped in an unsustainable pace (46%) and the quality of their work going down (41%).

This is the ambivalent center. 77% of respondents picked at least one positive and one negative emotion about AI. The average person selected more than five emotions. People find AI genuinely useful and also quietly dread what it demands of them.

The survey identified four emotional archetypes. The Energized (41%) lead with excitement and curiosity. The Conflicted (35%) hold positive and negative feelings at once. The Disoriented (12%) feel their role shifting beneath them faster than they can find footing. The Resentful (12%) are burned out and checked out, feeling pressured to use AI. The Conflicted are the largest group after the Energized. They are not opponents of AI. They are exhausted by the work of keeping up with it while holding two feelings at the same time.

This has implications for product strategy. Most AI products are positioned around speed and output. Ship 10x faster. Do the work of a whole team. For the amplified half, that message lands. For the Conflicted and Resentful, it sounds like a threat. They do not want to do more work faster. They want to do their current work with less anxiety.

There is a product opportunity here that most teams are missing. Positioning around reducing overwhelm, protecting quality, or giving people back time instead of expecting more output. The market for make my job sustainable is as large as the market for make me faster.

Three Decisions for Product Builders

First, calibrate for your market. If you target the superuser cohort, you are selling to people whose spending is accelerating fast. Usage-based pricing makes sense here. The superusers are demonstrating willingness to pay that scales with usage. If you target the broader market, you need a different pitch. Speed alone will not close them. Reducing friction and anxiety will.

Second, reconsider your default positioning. Most AI products default to move faster. That message only works on the amplified half. Consider a value proposition that addresses the actual top fear: doing more work for the same pay. Products that promise quality over speed, or that give people back control over their pace, address a market that is not being served.

Third, the builders market is real and growing. The 5x spending multiplier on app-builder platforms is a signal. If you are building tools for other builders (developer tools, design tools, content tools), this segment is willing to spend aggressively. They are also the most likely to churn if your product does not keep pace with their expectations. Speed of iteration matters more than polish.

None of this means the hype is wrong. It means the hype is incomplete. The real AI economy is not a single market. It is two markets with different customers, different fears, and different willingness to pay. Building for both requires more than one product strategy.