A16Z Narrator | The a16z Show Summary

The a16z Show featuring John and Patrick Collison

1. Key Themes

The Real Economy is Quietly Entering a Phase Transition

While markets have been volatile and executive surveys claim AI isn't delivering value, Stripe's actual transaction data tells a starkly different story. The 2025 cohort of businesses is both larger in count and performing better on a per-business basis than any prior cohort — a simultaneous improvement in quantity and quality that is historically rare.

"There's been a phase transition in 2025, where there are both more of them, and on a per business basis, they are on average doing better. Which is really striking, because you might think, okay, well, there's this cavalcade of new lightweight vibe-coded applications or something. But, you know, there's not really a lot of substance there. We're actually seeing both numbers move together." — John Collison 00:08:18

"We're only a couple of weeks into 2026, but it looks tentatively like 2026 may plausibly be an acceleration even over that significant leap of 2025." — John Collison 00:08:47

Agentic Commerce Will Require Entirely New Financial Infrastructure

The current web and payment rails were built for humans, not agents. CAPTCHAs, form-filling, and low-throughput blockchains are bottlenecks. The Collisons believe that agentic commerce will require blockchains capable of billions of transactions per second — infrastructure that does not yet exist — and this is the thesis behind their incubation of Tempo.

"Where we think things will go is just there will be a huge amount of agentic commerce. And again, we're seeing a little bit of it today. We think there'll be a torrent of it. And that is what unites stablecoins and AI, because we think you're going to need blockchains and better blockchains. Honestly, I mean, this was our thinking behind incubating Tempo." — Patrick Collison 00:06:07

"The world is going to need platforms that support billions of transactions per second, which no payment rail or platform does today." — John Collison 00:00:00

Software is Shifting from Mass-Produced Product to Bespoke, On-Demand Creation

John Collison articulates a fundamental economic reframe: software has historically been a fixed-cost product infinitely monetized, creating winner-take-all dynamics. The introduction of inference costs and custom generation breaks this model entirely — software becomes more like a freshly made pizza than a frozen product.

"Software should be, like Pete's incentive, it should be cooked right then and there at the moment of use. And so it's this quite fundamental shift where you don't want mass-produced industrial-scale software. You want bespoke custom software made for you that moment." — John Collison 00:15:55

"Up until now, the economics of software have been conceived of as fixed costs and then infinitely monetized or monetized as much as possible. That has these kind of winner-take-all dynamics. But once there are inference costs and custom creation involved, it really shifts." — John Collison 00:16:24

2. Contrarian Perspectives

80% of Executives Saying AI Has No Value Is Simply Wrong — Follow the Behavior, Not the Surveys

Patrick Collison flatly dismisses the widely-cited executive survey claiming no AI value. His argument: no one is asking for refunds on tokens, no one is reverting to pre-AI workflows. Behavior is the ground truth, not stated opinions.

"Oh, come on, that's hogwash. Like, find me one executive who wants a refund on their tokens. Find me one executive who said, oh yeah, we started augmenting our customer service with AI so people are more productive, but we're just going to go back to doing it the old-fashioned way." — Patrick Collison 00:09:47

You Cannot MBA Your Way Into Building a Great New Product

Patrick Collison explicitly argues against TAM-first, spreadsheet-driven product reasoning. Atlas (Stripe's company incorporation product) would have failed any traditional MBA adjacency analysis — yet it became one of the most impactful fintech infrastructure products ever built. Founders should reason in "product specifics," not market size.

"I think any MBA would have told you that the adjacency of incorporation makes no sense. It's not related to what's our right to win... whereas you actually go talk to founders, they're like, guys, this is the single biggest issue I ran into starting my company." — Patrick Collison 00:13:29

"We actually never thought about Stripe in GDP terms until one day we realized, oh, hang on. Oh, wow." — John Collison 00:12:50

Stripe's Growth Data Reflects Selection Bias — And That's Actually the Investment Signal

John Collison acknowledges that Stripe's outperforming cohorts may partly reflect a composition effect: the most innovative companies choose Stripe. Rather than weakening the thesis, this is itself the signal — innovative activity is accelerating, and it concentrates on modern infrastructure.

"There is this qualitative sense that once a company decides to do something innovative, new, retool, what have you, they're more likely to come to Stripe." — John Collison 00:05:12

Stablecoins and AI Are Not Parallel Trends — They Are Convergent Infrastructure Requirements

The conventional view treats stablecoins and AI as separate narratives. The Collisons argue they are actually unified by a single forcing function: agentic commerce at scale requires programmable, high-throughput, machine-native financial rails that legacy systems and even current blockchains cannot provide.

"That is what unites stablecoins and AI, because we think you're going to need blockchains and better blockchains." — Patrick Collison 00:06:07

3. Companies Identified

Stripe Global payments infrastructure processing over $1 trillion annually. Mentioned as the primary lens through which real economic trends are being observed — growing 34% in 2024 with accelerating cohort performance in 2025.

"Stripe processed more than a trillion dollars in payments last year. It grew 34%." — A16Z Narrator 00:00:53

Tempo A Stripe-incubated high-throughput blockchain designed for agentic commerce. Mentioned as the infrastructure solution to blockchain congestion and throughput limitations that will become critical as AI agents conduct commerce at scale.

"This was our thinking behind incubating Tempo because you're going to need really high throughput blockchains for the agents." — Patrick Collison 00:06:07

Stripe Atlas Stripe's company incorporation product. Mentioned as the canonical example of a product that defied MBA logic, took 10 years of compounding, and became transformative infrastructure for startups globally.

"Atlas is now this great overnight success, but we launched Atlas, I think, in 2014, maybe 2015. And so 10 years of compounding, and yeah, now it's at some pretty meaningful scale." — John Collison 00:11:30

Bridge A stablecoin infrastructure company (acquired by Stripe). Mentioned as a real-world example of why higher-throughput blockchains are needed — Bridge experienced operational issues due to blockchain congestion from competing use cases like meme coin trading.

"We talked in the letter about Bridge having operational issues, not because of Bridge, but because of blockchain congestion. Where you have blockchains both used for meme coin trading and also serious real world payments." — Patrick Collison 00:13:52

Stripe Press Stripe's book publishing arm. Mentioned for having sold its millionth (now 1.1 millionth) book — a remarkable milestone for what began as a curated publishing side project.

"We sold our millionth book... we've actually now sold our 1.1 millionth book." — John Collison 00:17:33

4. People Identified

John Collison Co-founder and President of Stripe. Identified for his ability to synthesize macro economic signals from transaction data and reframe entire industries conceptually (e.g., the "pizza" model of software).

"I feel like there's at least a reasonable chance that 2026 Q1 will be looked back upon as the first quarter of the singularity." — John Collison 00:07:24

Patrick Collison Co-founder and CEO of Stripe. Identified for his product intuition, specifically his insistence on reasoning from specific user pain points rather than top-down market analysis, and his clear-eyed view on AI adoption.

"You have to reason in very specific product terms." — Patrick Collison 00:14:12

5. Operating Insights

Talk to Founders Before You Build — Not Analysts

The single biggest product insight in this conversation is operational: the reason Atlas succeeded was not spreadsheet reasoning but direct founder conversations. Stripe discovered that incorporation was the number one pain point for new founders — something no market map would have surfaced. The operating lesson: primary research with actual users beats TAM analysis every time when entering a new product area.

"You actually go talk to founders, they're like, guys, this is the single biggest issue I ran into starting my company." — Patrick Collison 00:13:29

Long Obedience in the Same Direction Wins — Even Decade-Long Bets

Atlas took roughly 10 years to reach meaningful scale. Tempo is expected to follow the same arc — 5 to 7 years before it becomes significant. The operating insight: resist the pressure to declare product failures early. Compounding infrastructure businesses require sustained commitment that most operators abandon prematurely.

"So much of these things is just a long obedience in the same direction. Atlas is now this great overnight success, but we launched Atlas in 2014, maybe 2015. And so 10 years of compounding." — John Collison 00:11:30

Build Product Catalogs Into AI Apps Now — The Infrastructure Is Being Laid Today

Stripe is already doing the boring API and protocol work to make major retailer product catalogs (Etsy, Shopify, Best Buy, Walmart) purchasable inside AI applications. Operators in commerce or retail should be asking: is our catalog accessible to AI agents? This is not a future problem — it is a current infrastructure decision.

"We're working with all the big retailers that you would expect — the Etsy's and Shopify's and Best Buy's and Walmarts — to make product catalogs buyable within the AI apps." — Patrick Collison 00:14:51

6. Overlooked Insights

The "Non-Walrasian Software Regime" Is the Most Underappreciated Economic Shift in This Conversation

John Collison briefly uses the phrase "non-Walrasian software regime" almost in passing — but this is actually a profound economic observation that the other participants did not engage with. Walrasian economics assumes market-clearing equilibrium pricing. Traditional software, with near-zero marginal costs, naturally tends toward monopoly and winner-take-all outcomes (think Microsoft Office, Google Search). The introduction of per-inference costs and bespoke generation fundamentally breaks this — it reintroduces marginal costs into software, which means the structural conditions that created software monopolies may no longer apply to AI-native software. This is an enormous investment implication: the next generation of software companies may not produce the same extreme concentration and monopolistic moats that defined the last generation.

"Up until now, the economics of software have been conceived of as fixed costs and then infinitely monetized or monetized as much as possible. That has these kind of winner-take-all dynamics. But once there are inference costs and custom creation involved, it really shifts. It's kind of the non-Walrasian software regime." — John Collison 00:16:24

The "2025 Cohort Acceleration" Is Being Measured Right Now and Most Investors Are Sleeping on It

John Collison mentions almost as an aside that 2026 is tracking to be an acceleration over 2025's already extraordinary cohort performance. This is not a forecast — it is an early data observation from the largest real-economy payment platform on the planet. Most macro investors are focused on tariffs, rate policy, and public equity volatility. The actual ground-truth signal from Stripe's data suggests something categorically different is happening at the startup formation and revenue layer of the economy — and it is accelerating into early 2026.

"We're only a couple of weeks into 2026, but it looks tentatively like 2026 may plausibly be an acceleration even over that significant leap of 2025." — John Collison 00:08:47