What Link data tells us about AI spending

We surveyed 394 customers who use Link, a digital wallet built by Stripe, on their AI habits. Eighty percent had used a chat-based agent in the past month, and half had used AI for shopping research at least monthly.
Has that usage translated into spending? To find out, we analyzed spending patterns across the 250 million customers paying with Link. We found that Link customers are spending more on AI than they were three months prior, investing heavily in platforms that let them build with AI.
Here’s what the data shows.
Link customers are spending more on AI products each month
We grouped Link customers by their monthly spend on AI products. The top 10% of Link customers with the highest spend on AI products increased their monthly spending from $183 in December 2025 to $359 in March 2026—nearly doubling in a single quarter.
What’s notable is the timing: it took 22 months for that same cohort to grow from $84 to $183, but only three months to double from $183 to $359. Even at the 50th percentile, monthly spending on AI products increased from $60 to $72 over the same period.
Link customers are spending more on AI app-builder platforms
When we isolate spending on app-builder platforms like Replit, Lovable, and Bolt, the growth is even more pronounced. The top 10% of Link customers with the highest spend on AI products are now spending 5x more on app-builder platforms each month compared to January 2025.
Building the infrastructure to support agents
Link customers are spending more on AI and using agents for shopping research. As these agents become more capable, they’ll increasingly need the ability to transact with businesses and one another.
That’s why we built Link’s wallet for agents. It lets customers authorize agents to pay on their behalf with spending controls they set, gives agents broad purchasing access across any seller on Stripe, and gives businesses verified transactions without complex integrations.
Learn more about how you can give agents the ability to pay with Link’s wallet for agents.

Facts Only

* 394 customers were surveyed regarding their AI habits using the Link digital wallet.
* Eighty percent of surveyed customers had used a chat-based agent in the past month.
* Half of surveyed customers had used AI for shopping research at least monthly.
* Link customers are spending more on AI products than they were three months prior.
* The top 10% of Link customers with the highest spend on AI products increased their monthly spending from $183 in December 2025 to $359 in March 2026.
* This top cohort nearly doubled their spending in a single quarter.
* Monthly spending at the 50th percentile increased from $60 to $72 over the same period.
* Spending on app-builder platforms like Replit, Lovable, and Bolt is also increasing.
* The top 10% of Link customers are now spending five times more on these app-builder platforms monthly compared to January 2025.
* Link’s wallet was built for agents to authorize payments and enable broad purchasing access across sellers on Stripe.

Executive Summary

Link customers are demonstrating increased spending on AI products compared to three months prior. The top 10% of Link customers with the highest AI spending saw their monthly spend increase from $183 in December 2025 to $359 in March 2026, nearly doubling in a single quarter. This growth was also visible at the 50th percentile, where monthly spending on AI products increased from $60 to $72 over the same period. Furthermore, Link customers are showing accelerated investment in AI app-builder platforms such as Replit, Lovable, and Bolt. The top 10% of high-spending customers are now spending five times more on these platform types monthly compared to January 2025. This trend reflects a shift toward building infrastructure to support autonomous agents that need transactional capabilities, which is why Link was designed for agents to authorize payments across sellers.

Full Take

The data highlights an accelerating investment cycle where AI usage is rapidly translating into platform expenditure, suggesting a shift from consumption of AI tools to infrastructure development. The doubling of spending in the top cohort, while occurring over a compressed timeline (three months versus twenty-two months for initial growth), points toward hyper-acceleration driven by perceived utility and capability gains. This pattern suggests that the incentive structure is highly effective at driving immediate scaling of investment, effectively rewarding rapid adoption of building tools over simple usage. The focus on app-builder platforms implies that the next logical step for these users is not merely interacting with AI agents but controlling the environments those agents operate in, demanding a shift from end-user consumption to agent infrastructure. This introduces a systemic challenge: ensuring that the infrastructure built (like Link) serves genuine user agency and democratized commerce rather than simply optimizing platform monetization of increasingly complex AI workflows. The question becomes whether this spending reflects organic growth in capability or market forces steering users toward integrated, self-governing financial systems facilitated by agents.

Sentinel — Human

Confidence

The text presents a structured, data-driven analysis typical of investigative reporting. While highly polished, it lacks the uniform rhythm and excessive hedging often associated with purely synthetic content, suggesting human editorial oversight.

Signals Detected

Natural variance in sentence structure and rhythm; slightly complex temporal phrasing.

Clear, focused argument flow linking data points to a specific product vision (agents); exhibits appropriate analytical depth.

Data presentation is specific and tied directly to the narrative claims, suggesting reliance on primary source reporting rather than generalized talking points.

Human Indicators

The use of highly specific numerical shifts ($183 to $359) across different cohorts suggests reliance on detailed data analysis that is difficult for generic LLMs to fabricate without context.

The narrative flows logically from observed usage (AI agents) to the underlying infrastructure need, which demonstrates synthesized reasoning rather than simple information recitation.