What I Saw at Stripe Sessions 2026: The AI Economy
TechFlow Selected TechFlow Selected
What I Saw at Stripe Sessions 2026: The AI Economy
Stripe data reveals AI is reshaping the global economy.
Navigating Web3 tides with focused insightsBy: Gao Fei
Translated by: AididiaoJP, Foresight News
In 1987, economist Robert Solow famously remarked: “You can see the computer age everywhere except in the productivity statistics.” This observation perplexed economists for nearly a decade—until the mid-1990s, when computers’ contribution to productivity finally became visible in the data.
In 2026, the same puzzle is repeating itself with AI. Bubble theories come and go; scholars debate endlessly; enterprises hesitate; macroeconomic signals remain fuzzy. Yet there is one place where AI’s economic impact no longer requires debate.
Let’s look at Stripe.
Over the past few days, I attended Stripe Sessions in San Francisco. Stripe processes transactions equivalent to nearly 2% of global GDP, handles $1.9 trillion in annual payments, and serves over 5 million businesses. Eighty-six percent of companies on the Forbes AI 50 list use Stripe. If the AI economy is a newborn infant, Stripe is the heart-rate monitor in the delivery room—recording the infant’s heartbeat earlier and more precisely than almost anyone else.
A study released by the Federal Reserve Bank of St. Louis in early 2026 shows that AI-related investment has contributed nearly 40% of U.S. marginal GDP growth—surpassing the peak contribution of the tech sector during the dot-com bubble. And when those investments convert into revenue, most settlements happen on Stripe. More importantly, Stripe isn’t merely recording the AI economy’s heartbeat. At this year’s conference, it announced its ambition to catalyze an entirely new economic paradigm: Agentic Commerce—the idea that agents themselves become transacting entities. In a group media interview, co-founder and president John Collison stated he expects agents to become mainstream buyers in commercial transactions within 12 to 18 months.
Over two days, 288 product and feature announcements, and more than 10,000 attendees—the defining phrase was repeated throughout: Agentic Commerce. Below are my observations from Stripe Sessions 2026—and my personal reflections.
How Fast Is the AI Economy Really Running?
Before diving into agentic commerce, let’s first examine the broader contours of the AI economy. Solow said in 1987 that computers were invisible in the statistics—nearly four decades later, AI is already unmistakably visible in Stripe’s data.
On Day One’s morning keynote, CEO Patrick Collison presented a set of metrics. Since the pandemic, the number of new businesses launched monthly on Stripe has remained high—but relatively flat. Starting in early 2026, that curve shot nearly vertically upward. The direct cause? AI coding tools have dramatically lowered the barrier to entrepreneurship: many developers now build monetizable products in just days using “vibe coding.” Patrick described this as part of a larger phenomenon—the entire economy is undergoing AI-driven re-platforming. Maia Josebachvili, Stripe’s Chief Revenue Officer for AI Business, added an external benchmark: iOS App Store app releases had been declining until 2024. After AI coding tools emerged, release volume rose 24% month-on-month.
The change is not only quantitative but qualitative. Stripe Atlas—one of the easiest ways for founders to incorporate a U.S. company—just celebrated its 100,000th company. At the conference, I heard a startling statistic: companies incorporated via Atlas in 2025 generated twice the revenue at the same lifecycle stage as their 2024 counterparts. Companies founded in 2026—many only months old—already generate five times the revenue of companies founded in the same period last year.
During Day One’s afternoon AI Economy Report, Maia Josebachvili named several drivers behind the AI economy’s rise. Lovable hit $100 million in revenue in eight months—and then doubled to $400 million in the next eight. Cursor reached $1 billion in annualized revenue in under two years—and tripled to $2 billion just three months later. Leading AI-native companies on Stripe grew 120% in 2025—and 575% so far in 2026.
Consumption is also surging steeply. Top-tier users spend $371 per month on AI products—more than the average American spends monthly on internet access, streaming services, and mobile bills combined. I did a quick mental calculation of my own monthly token spend—and it already exceeds my phone bill.
Patrick also offered a comparison: businesses on Stripe are growing 17 times faster than the global economy.
On Day Two, John Collison directly referenced the Solow Paradox—and explained it through historical analogy. In 1882, Edison lit up the first customer electric lights in Manhattan. Yet over the following three decades of electrification, productivity barely improved. Not because electricity didn’t work—but because factories were still designed around steam engines. Only after entire factories were rebuilt did productivity gains materialize. John’s assessment is that AI is in a similar phase: transformation is underway, but legacy systems haven’t yet fully absorbed it. “That said,” he added, “I suspect AI won’t take thirty years.”
Stripe’s data appears to support his optimism. On its platform, the AI economy is already exploding. Almost every traditional enterprise I’ve spoken with has its top leadership driving AI deployment with extraordinary urgency.
Global from Day One
Beyond speed, another striking feature of these AI companies is their inherent globalization—from Day One. Stripe calls this “default globalization.”
Since becoming an AI blogger myself, I’ve often experienced how AI content creation operates beyond time zones: AI news from across the Pacific carries equal weight with local news. AI products operate similarly. Large language models (LLMs) blur the interface languages and interaction habits that traditional software relied upon. A single unified chat box enables global users to interact with products via natural language. In this sense, LLMs make a unified global software market possible—for the first time.
Conference data confirms this observation. During the previous SaaS wave, the fastest-growing companies covered ~25 countries in Year One and ~50 by Year Three. AI companies move at a different pace: 42 countries in Year One, 120 by Year Three. Maia noted Kazakhstan has already appeared on many AI companies’ market lists. In the “Indexing the Economy” breakout session on Day Two, Stripe shared a median figure: the top 100 AI startups sold into 55 countries in their first year.
Emergent Labs offers a concrete example. Founded in the U.S. in 2024, it already derives nearly 70% of revenue from overseas—with at least 16 countries each contributing ≥1% of revenue. Among leading AI companies, 48% of revenue comes from outside their home markets—up from just 33% three years ago. Global revenue is no longer supplementary—it’s foundational.
Speed and globalization are the AI economy’s two core features—and both are deeply tied to Stripe. AI companies need to rapidly establish payment capabilities—to collect payments in 40+ countries and regions within their first week. That’s precisely what Stripe has done since its founding day.
Here, a brief note on Stripe’s origins is warranted.
Stripe’s founders, Patrick Collison and his brother John, are Irish—and they themselves are cross-border entrepreneurs. At the conference, I met an Irish colleague who told me that, in the eyes of Irish AI founders, the brothers are heroes. When they arrived in the U.S., they discovered online payments were extremely difficult: integrating payment systems required signing contracts with banks, passing PCI compliance reviews, and connecting to multiple intermediaries—a process that could take weeks or even months.
So in 2010, two twentysomethings dropped out of school, moved to San Francisco, and wrote a solution enabling developers to accept payments in just seven lines of code. Those seven lines arrived just as mobile internet and SaaS were taking off. Shopify needed to help millions of merchants accept payments; Uber needed frictionless passenger payments; Salesforce needed to handle global subscriptions—all chose Stripe. Growing alongside these global customers, Stripe built localized capabilities in 46 countries, serving 195 markets and supporting 125 local payment methods.
To consumers, Stripe isn’t a company in the spotlight. It lives behind Shopify’s checkout page, OpenAI’s subscription confirmation emails, and Uber’s fare notifications. But this invisibility hasn’t prevented it from becoming the underlying financial infrastructure of the internet economy. In the AI era, this global financial infrastructure gives Stripe a first-mover advantage in serving AI companies expanding internationally.
At this year’s conference, I also met Abhi Tiwari, Stripe’s Global Head of Product. He’d taken on the role just three months prior—and relocated to Singapore. Stripe maintains engineering hubs in San Francisco, Dublin, and Singapore, plus a Latin America office in São Paulo. Abhi told me many AI companies approach Stripe saying, “We’re default global—the location of our users doesn’t matter.” The old model—building products at HQ and pushing them globally—is being replaced by a new one: local teams building directly in-market.
Reaching global users is one thing; getting them to pay is another—far more complex, given each market’s unique currencies and payment habits. Here, Stripe supports AI companies and other customers primarily through two mechanisms: local-currency pricing and integration with local payment methods. The former lets Brazilian users see prices in reais—not dollars—boosting cross-border revenue by 18%. The latter enables Indian users to pay via UPI and Brazilians via Pix—lifting conversion rates by over 7%. Gamma, an AI presentation tool, saw Indian revenue surge 22% after integrating UPI. At the exhibition booth, I also spotted Chinese company MiniMax. From what I understand, many Chinese companies going global use Stripe’s financial services via overseas entities.
These AI-native companies share another trait: minimal headcount—many are solo founders. One or two people, plus a team of agents, can sustain a globally operating company generating real revenue. On Day Two, Emily’s talk included a telling stat: the density of solo founders on Atlas has approached 5,000 per million Americans—and an increasing number earn over $100,000 annually.
Emily used the term “solopreneur.” That reminded me of China’s rapidly growing OPC (One Person Company) wave. John invoked Ronald Coase’s theory of the firm to explain this trend. Firms exist because internal coordination costs are lower than market coordination costs. But AI may be reversing this logic. When agents can help you discover services, integrate software, and process payments, external coordination costs plummet—you no longer need a room full of employees to do what once required an entire department.
From Human Economy to Agent Economy
The AI economy described above—however fast-growing or globally scaled—still has humans as its transacting entities. Humans buy AI products; humans launch startups using AI tools. But the strongest signal I sensed at this year’s Sessions was Stripe’s next major focus: a further shift—toward an economy where agents themselves become market participants. This is Agentic Commerce.
This shift is already quietly appearing in Stripe’s own data. Will Gaybrick, Stripe’s President of Products and Business, showed a set of numbers. For years, Stripe CLI (Command Line Interface) was used by only a small cohort of highly technical users—and usage remained flat. Starting in 2026, usage spiked sharply—because agents don’t need polished graphical interfaces; clean CLIs are often more useful. Maia’s data shows traffic from agents reading Stripe documentation surged roughly tenfold in 2025. If current trends hold, agent readership of Stripe docs will exceed human readership by year-end. Stripe’s API documentation—refined over more than a decade—has found its next most loyal audience.
If agents spending money still sounds unfamiliar, consider two already-existing scenarios.
First: shopping interfaces may be shifting toward model chat windows. Consumers now routinely use ChatGPT, Gemini, or Instagram to research products. The distance between research and transaction has collapsed into a single interface. Similar cases have emerged in China—for instance, buying bubble tea inside an AI app.
In the group media interview, John Collison used his own purchase of a travel power adapter to illustrate why this compression is irreversible. If an agent completes the full research-to-purchase flow—and the product arrives days later—he won’t return to another site to manually re-enter personal information—even if that site offers a marginally better product. Once a shopping agent completes the search workflow, checkout is the natural next step.
The second example is even more intriguing: OpenClaw. Anyone familiar with the “lobster” wave knows it’s among today’s hottest open-source autonomous agent frameworks. Users issue instructions to agents via messaging apps like Feishu, Telegram, or WhatsApp—and agents execute tasks autonomously. Crucially, OpenClaw can consume hundreds—or even hundreds of dollars—worth of tokens per day. It manages its own token consumption and usage. Though human authorization remains common, the agent ultimately consumes the tokens—and tokens can be directly converted into money.
From agents managing token consumption to agents spending money directly is just one step. At this year’s conference, Stripe crossed that step.
Demo: Agents Buying and Selling
On Day Two’s main stage, a demo earned multiple rounds of applause.
John Collison gave an agent a simple instruction onstage: “Research how AI demand affects energy markets.” The agent began searching—and discovered Alpha Vantage offered an energy market dataset it needed, priced at $0.04. Judging the price within budget, the agent autonomously purchased and downloaded it using Tempo CLI’s stablecoin wallet—since paying $0.04 with a credit card would be uneconomical. Then it generated a full analytical report. That alone was astonishing. But John then instructed the agent: “Publish and sell this report. Set a price you deem reasonable—and make it discoverable and purchasable by other agents.” The agent checked Alpha Vantage’s licensing terms, confirmed commercial use was permitted, built a website, published the report, and generated an instruction file allowing other agents to purchase the data with a single request.
In just minutes, an agent completed the full chain: research, procurement, production, compliance review, publishing, pricing, and sales. It acted as both buyer and seller. After the demo, John declared: “Agentic Commerce has arrived.”
Two other demos on Day One were equally impressive. Will Gaybrick built an API review application that lets agents procure review services for users—without him specifying any payment details. During execution, the agent automatically detected the app used the Machine Payments Protocol (MPP) and independently completed a $2 payment. A human only authorized it once, via fingerprint. This zero-configuration payment discovery capability is MPP’s core design principle: developers don’t need to write custom payment logic for agents—the agents find it themselves.
Next, Gaybrick demonstrated streaming payments—combining Metronome (a real-time metering engine), Tempo (a blockchain built for payments), and stablecoins. An application charges in real time based on AI token consumption—$3 per million tokens. Multiple agents run simultaneously. The left dashboard shows rising token consumption; the right shows synchronous inflows of stablecoin micropayments. Opening the Tempo blockchain explorer reveals the $3.30 total payment comprises thousands of sub-cent micropayments—each worth just one-thousandth of a cent. Credit cards can’t do this. ACH can’t. Neither can UPI or Pix. Gaybrick announced onstage: “This is the world’s first streaming payment business.”
The Return of Micropayments—and a New Consumption Logic
Shopping via chat windows and OpenClaw exemplify agents acting as human proxies for consumption. But in the group interview, Collison made a bolder claim: agents may create entirely new demand.
He argued agents could make feasible a long-discussed—but never truly realized—business model: micropayments. Humans struggle with ultra-fine-grained consumption decisions. Spotify replaced per-track payments with a $9.99/month subscription because no one wants to decide whether a song is worth $0.15 every time they hit play. Agents lack this cognitive burden. If this holds true, entire categories of business models—previously doomed by human cognitive friction—may suddenly become viable for agents. Maia echoed a similar view in our one-on-one conversation, noting she’d recently spoken with dozens of AI founders—and pricing was the most frequently mentioned topic when discussing agentic commerce.
Every transaction has a buyer and a seller. If the buyer becomes an agent—what should merchants do?
In one interview, I asked Stripe’s Head of Product Jeff Weinstein: “Humans have the saying ‘the customer is always right’—merchants must please consumers. So how do you please an agent?” Jeff replied: “Imagine the agent as your best programmer. It wants perfect information, structured formats, rapid readability, and all contextual data needed to make decisions. Human consumers love beautiful images and smooth animations; agents want raw structured data, precise logistics information, and the ability to complete transactions in minimal steps.”
In another conversation, Ginger Baker, Meta’s VP of Product, summarized this shift even more starkly: “Payments will evolve from ‘moments’ to ‘strategies.’ Human purchases are discrete: you walk to the register, pull out your wallet, swipe your card, and the transaction ends. Agent consumption is continuous: you set rules—e.g., ‘groceries this week ≤ $50,’ ‘always prioritize this card,’ ‘manual approval required for >$500’—and the agent continuously spends autonomously within your authorization framework.”
Security: Compute Is the New Cash
If agents truly become a new class of consumer, they’ll bring new risks—fundamentally distinct from traditional SaaS transaction risks or those faced by human consumers.
During Sessions, I focused especially on this topic—and discussed it with several Stripe executives.
Emily Glassberg Sands, Stripe’s Head of Data & AI, described three rapidly growing fraud patterns. First: multi-account abuse. One person repeatedly registers different accounts to claim free quotas. Per Stripe network data, one in six AI company registrations involves such abuse. Second: malicious consumption during free trials—especially lethal for AI companies, as each trial burns real inference costs. She cited an example: a partner company’s token cost per paying customer exceeded $500, because acquiring one customer required 25 free trials—19 of which were fraudulent. Third, she called it “dine-and-dash”: customers consume large volumes of tokens—and refuse to pay at month-end. Emily also quoted a pithy line: “Compute is the new cash.” When traditional SaaS is abused, marginal costs are near zero. But every AI inference call incurs real cost. Stolen tokens = stolen money.
Yet here lies a dilemma that particularly troubles me. Many AI founders combat abuse by shutting down free trials altogether.
Emily said she asked everyone claiming to “solve” this problem how they did it—and found their solution was simply disabling the free tier. But Jeff argues this creates another problem: agents are becoming the primary way new services are discovered. If agents can’t self-serve trials, they’ll jump straight to another URL. Emily added that if the call-to-action presented to agents is “join waitlist” or “contact sales,” agents will leave immediately. Shutting down self-service registration to prevent fraud may mean surrendering your most critical growth channel to competitors.
Stripe’s answer to this dilemma is its fraud prevention system, Radar. Radar’s logic is simple: every transaction completed on Stripe trains Radar once. Transaction data from 5 million businesses flows into a shared risk-detection network. If one company encounters a fraud pattern, all benefit. Last month, Radar blocked over 3.3 million high-risk free-trial registrations across eight high-growth AI companies.
Jeff also offered a counterintuitive perspective: agent shopping may ultimately prove safer than human web shopping. Human trust verification relies on inference—how long users stay on a site, whether click paths look normal, etc. Agent transactions enable programmatic authentication. Stripe’s Shared Payment Tokens tokenize payment credentials—agents never touch raw credit card numbers. Users authorize via biometrics and can set transaction limits, time windows, and merchant whitelists. When trust shifts from inference to verification, the security baseline may actually improve.
Ecosystem, Protocols, and a Piece of History
By now, it should be clear: agentic commerce is impossible without a well-functioning ecosystem. At Stripe Sessions 2026, I met someone from the food industry. He attended to assess whether agentic commerce could become a new opportunity for his company—that’s the seller’s perspective.
So this cannot be accomplished by Stripe alone—it requires an ecosystem.
Over two days wandering the Sessions exhibition hall, I saw booths from numerous companies across the financial value chain. Stripe has also launched or joined a series of protocols with upstream and downstream partners—connecting different parts of the ecosystem: buyers and sellers, humans and machines, machines and machines. The Machine Payments Protocol (MPP) enables agents to discover and complete payments via HTTP. The Agentic Commerce Suite lets consumers purchase directly within AI apps from Google, Meta, OpenAI, and Microsoft. The Universal Commerce Protocol (UCP)—initiated by Shopify and joined by Meta, Amazon, Salesforce, and Microsoft—is a cross-platform commerce protocol. Stripe joined UCP’s Governing Council. A group of companies—both partners and competitors—agreed to jointly develop a shared protocol because fragmentation would hinder agents’ seamless cross-platform consumption—and hurt everyone.
Speaking of protocols, I noticed a distinctive Stripe partner in the exhibition hall: Visa. To me, Visa is fundamentally a protocol platform.
Spotting Visa immediately reminded me of a book I greatly admire: One from Many, by Visa founder Dee Hock. A central theme is how banks, money, and credit cards should be redefined in the electronic age. Money need not be coins or paper—it can be data, guaranteed by institutions, recorded on networks, and flowing globally. In the late 1960s, Bank of America’s BankAmericard expanded nationwide, flooding the system with interstate consumers—and the old infrastructure collapsed. Hock realized the problem was organizational: dozens of competing banks needed shared infrastructure, but existing structures couldn’t accommodate both cooperation and competition. Using decentralized design principles, he made all banks equal members of the new organization—and Bank of America relinquished sole control. This organization was later renamed Visa.
So do two companies from different eras, doing similar things, share some lineage?
Any agent can easily find the answer. Patrick Collison has publicly honored Hock. After Hock’s death in 2022, Patrick called him “a severely underappreciated innovator”—profoundly influential on himself and his brother. A clearer signal is the hiring decision: David Stearns, the authoritative academic historian of Visa, later joined Stripe.
There’s another detail that would make anyone familiar with payment history smile knowingly. Onstage, Georgios Konstantopoulos, CTO of Tempo Blockchain, showcased the validator lineup—and one name stood out: Visa. Hock’s Visa is now a participating node in Stripe’s incubated blockchain network. Students built a new network—and teachers became nodes within it.
When Patrick traced Stripe’s intellectual origins in his opening keynote, he said he began as a Lisp programmer. Lisp’s core idea is “code is data.” He translated that idea into Stripe’s own language: “Stripe’s fundamental belief is that money is data. When we launched Stripe in 2011, this wasn’t yet industry orthodoxy.” Hock approached money’s essence from organizational theory—concluding money is merely “a guarantee of value exchange.” Its medium can be anything. Collison approached it from programming languages—directly equating money with data: programmable, API-callable, agent-operable data. Both spoke the same truth—in different languages. That day onstage, Ginger Baker put it even more plainly: “Isn’t money just another form of digital content?”
If money is data, then data’s consumers naturally become money’s consumers.
Side Plot: Stripe’s Content DNA
At this point, the AI economy story nears its end. But let’s take a slight detour—Stripe is almost a peer to content creators.
This company excels not only at financial services but also at content products. Its publishing imprint, Stripe Press, boasts exceptional taste—many know Stripe because it published Poor Charlie’s Almanack. Its podcast, A Cheeky Pint, is distinctive and widely listened to—featuring guests like Google CEO Sundar Pichai, Anthropic CEO Dario Amodei, and a16z co-founder Marc Andreessen.
During Sessions, I met Tammy Winter, Senior Editor at Stripe Press, and designer Pablo Delcan. Tammy joked, “Stripe is a publishing house with a $10-billion company attached.” Pablo Delcan spoke about his understanding of taste: it’s accumulated over time, requiring patience. In design trends, he believes the new challenge—without abandoning simplicity and clarity—is adding a degree of complexity through detail and precision.
When discussing books, Tammy told me that Stripe Press’s series for founders and builders is called “Turpentine.” These books focus on practical knowledge—tools, techniques, maintenance, and the nitty-gritty operational tasks that keep things running. They’re not abstract theory—they aim to help readers solve concrete operational problems.
The name draws from an apocryphal Picasso story: art critics gather to discuss form, structure, and meaning; artists gather to ask where to buy cheap turpentine. This series aims to be founders’ cheap turpentine. Think about it: for AI companies going global, Stripe’s financial services are another kind of turpentine—you don’t worry about payments, compliance, or FX—you focus on building your product.
This side plot may seem unrelated to the main thread—but there’s a deep connection. Stripe also publishes a magazine called Works in Progress, centered on how economies grow. Its podcast interviews AI economy leaders. Sessions itself, in many ways, resembles an economics lecture. On Day Two’s morning keynote, John Collison devoted an entire talk to economic data, Coase’s theory of the firm, and the Solow Paradox. I suspect a financial services company cares so deeply about economics precisely because understanding structural economic shifts is how it discovers its next product opportunity.
As a podcast enthusiast, my first question for John Collison on Day One wasn’t financial—it was about podcasts. I asked: after interviewing so many diverse people, is there a single underlying question threading through all conversations? He paused—and said what genuinely interests him is how those people’s companies actually operate: what competitive equilibrium they occupy, and how they understand their own businesses.
Coincidentally, Day One ended with a small twist. The scheduled final fireside chat was Patrick interviewing OpenAI co-founder Greg Brockman—but moments before going onstage, the guest changed to Sam Altman. Patrick explained, “After all, AI is a rapidly evolving field.”
So surprise turned to delight—the crowd erupted in cheers.
They’ve known each other for nearly 19 years. Altman was one of Stripe’s earliest angel investors—when the Collison brothers were under 20. That’s why Altman looked so relaxed throughout the conversation.
Near the end, Patrick asked a personal question: Why invest in two teenagers back then? Altman recalled that they were building something to solve a problem they’d personally experienced—and he saw the opportunity could scale, because many others needed the same solution.
I think his answer about podcasts—and his answer about investing—point to the same thing: finding real needs, solving real problems. In the conversation, Altman divided OpenAI’s evolution into three phases: from research lab, to product company, to a “token factory” supplying intelligence to the world. Each phase carried a distinct mission. Stripe follows a similar arc. In 2010, two Irish young men solved the problem: “Online payments are too hard.” Along the way, they solved that same problem for 5 million users. In 2026, they’ve identified a new problem: their customers’ customers may soon no longer be human.
Holding a podcast in one hand and a publishing house in the other—discussing Coase’s theory and the Solow Paradox onstage, rolling out protocols and APIs in the exhibition hall—Stripe isn’t just creating the AI economy. It’s documenting it. At the conference, I had a seemingly crazy thought: Stripe holds transaction data representing nearly 2% of global GDP. It sees where every dollar of AI revenue comes from, where it goes, and how fast it grows. If Solow had had such a heart-rate monitor back then, perhaps he wouldn’t have waited a decade to spot computers in the statistics.
Maybe one day, Stripe will deliver a model for the AI economy—not a large language model, but a Nobel-caliber economic model. Who says it’s impossible? Just a few years before DeepMind founder Demis Hassabis won the Nobel Prize, who could have imagined it?
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@feigaobox
