John and Patrick Collison on Stripe's Growth, Agent Commerce, and the Future of Software

What It Covers

John and Patrick Collison share Stripe's 2024 data revealing 34% growth and over $1 trillion processed, while outlining why agentic commerce will require blockchain infrastructure capable of billions of transactions per second, and reframing software itself as bespoke, on-demand creation rather than mass-produced product.

Key Questions Answered

• •Agentic Commerce Infrastructure: Agentic commerce will require blockchains handling billions of transactions per second — a throughput no existing payment rail achieves today. Stripe incubated Tempo specifically to build this capacity. Founders and operators building agent-facing products should architect payment infrastructure now for this scale, not retrofit legacy rails later.
• •2025 Cohort Signal: Stripe's 2025 business cohort is both larger in volume and performing better on a per-business basis than any prior year's cohort — simultaneously. This dual movement contradicts assumptions that new AI-era businesses lack substance, and suggests 2026 may accelerate further based on early data from the first weeks of the year.
• •Stablecoins + AI Convergence: Agents using today's payment rails must solve CAPTCHAs and work around infrastructure not designed for machine actors. The practical case for stablecoins is not ideological — it is throughput and latency. Builders designing agent payment flows should evaluate blockchain-native rails rather than defaulting to card-based or ACH infrastructure.
• •Software-as-Pizza Model: Software economics are shifting from fixed-cost-plus-infinite-monetization toward bespoke, inference-cost-per-use creation. Patrick Collison frames this as software moving from freeze-dried mass production to made-to-order, like pizza. Product teams should reconsider pricing models and build assumptions, as winner-take-all dynamics weaken when marginal creation costs are real.
• •Product Specificity Over TAM Reasoning: Stripe never modeled itself against global GDP percentages — it focused on solving a concrete developer pain point. The same logic applies to Tempo and Atlas, which grew from specific founder complaints, not market-size spreadsheets. Founders should validate by identifying one acute, named pain point rather than top-down addressable market calculations.