Just now! The Stripe 2026 Conference exploded! The Solow Paradox was broken, and the AI economy's ECG skyrocketed! The retail investors' last chance to get on board?

Let me tell you something. I just returned from San Francisco, attending the conference of the payment giant that handles nearly 2% of the global GDP, Stripe Sessions 2026.

What I saw at this meeting gave me chills, not out of fear, but because I realized that all of us, including many large institutions, are all a half step behind in understanding the AI economy.

There is a term called “Solow Paradox,” coined by economist Robert Solow in 1987: you see computers everywhere, but productivity data shows nothing. It wasn’t until the mid-90s that the paradox was broken. Now, the same confusion is back, with fuzzy data signals and bubble theories flying everywhere.

But two Irish brothers in Stripe installed a “heart rate monitor.” They process $1.9 trillion annually, with 5 million businesses on the platform, and 86% of Forbes AI 50 companies are on it. The AI economy is just born, and Stripe is like an ECG in the maternity ward, hearing the heartbeat earlier and more accurately than any economist.

A study from the St. Louis Fed early 2026 states that AI investments have contributed nearly 40% to the US marginal GDP growth, surpassing the internet bubble period. And most of the income generated from these investments is settled through Stripe.

So, stop listening to scholars still debating whether AI is a bubble; their data is too outdated. Real-time data already shows a steep upward trend on this “ECG.”

First, speed—it’s terrifying. Stripe CEO Patrick Collison showed a graph: since the pandemic, the number of new companies each month has been quite steady. But early 2026, the curve is nearly vertical. The reason is simple: AI coding tools let ordinary people create billable products in days, called “vibe coding.”

This isn’t just quantitative change; it’s qualitative. Stripe Atlas, a tool for registering US companies, celebrated its 100kth company last week. Companies registered through it in 2025 have twice the revenue at the same lifecycle stage compared to 2024. Companies founded just a few months into 2026 are already five times more revenue than last year.

For example: Lovable made $100 million in eight months; Cursor reached $1 billion annualized revenue in less than two years, doubling to $2 billion in three months. Leading AI-native companies on Stripe grew 120% in 2025, and have already grown 575% in 2026.

On the consumer side? The highest-spending users spend $371 per month on AI products, more than the combined internet, streaming, and mobile bills of average Americans. I calculated my own Token expenses, which have long exceeded my mobile bill. The growth rate of enterprises on Stripe is 17 times that of the global economy.

Brother John Collison (co-founder and president) mentioned Solow’s paradox on stage, using a historical analogy: Edison lit the first electric lamp in 1882, but productivity hardly increased for the next thirty years because factory skeletons were steam-powered. Only after rebuilding entire factories was the magic of electricity unleashed.

John’s judgment: AI is at a similar stage. But he added, “I doubt AI will take thirty years.” Stripe’s data supports his optimism. Top executives of traditional companies are pushing AI with high urgency, I have seen it myself.

Now, about globalization—again, a paradigm shift. These AI companies have been global from day one. Stripe calls this “default globalization.”

Large language models blur interface language; a single chat window allows users worldwide to use products in natural language. This makes a unified global software market possible for the first time.

Data confirms this: during the previous SaaS wave, the fastest-growing companies covered 25 countries in their first year, 50 by the third. AI companies: 42 countries in the first year, 120 by the third. Kazakhstan is now on the market list of AI companies. The top 100 AI startups sold to 55 countries in their first year.

A US company called Emergent Labs earns nearly 70% of its revenue overseas, with 16 countries contributing at least 1% each. 48% of leading AI companies’ revenue comes from markets outside their home country, up from 33% three years ago. Global revenue isn’t supplementary; it’s fundamental.

Speed + globalization—both directly related to Stripe. AI companies need to receive payments in 40 countries and regions within the first week. And Stripe’s founders are cross-border entrepreneurs themselves; in 2010, they solved payment collection with just seven lines of code, riding the wave of mobile internet and SaaS.

Now, Shopify, Uber, Salesforce all choose Stripe. Stripe has established local capabilities in 46 countries, covering 195 markets, supporting 125 local payment methods. This global financial infrastructure is a huge first-mover advantage in the AI era.

Many AI companies’ first words when approaching Stripe: “We default to global, user location doesn’t matter.” Reaching users is one thing; getting paid is another. Stripe solves this with local currency pricing and local payment connections. Indian users use UPI, Brazilians use Pix, boosting conversion rates by over 7%.

Gamma, a game demo tool, saw a 22% revenue jump in India after integrating UPI. I also saw the Chinese company Minimax at the booth; many Chinese outbound companies are using Stripe through overseas entities.

These AI-native companies also share a common trait: very few personnel, many are solo founders. One or two people plus a few intelligent agents can run a profitable global company. The density of solo founders on Atlas is nearly 5,000 per million Americans, with more and more earning over $100k annually.

John explains with Ronald Coase’s theory of the firm: firms exist because internal coordination costs are lower than market costs. But AI might be reversing this logic. When you can use intelligent agents to discover services, integrate software, and handle payments, external coordination costs plummet, and you no longer need a whole team of employees.

All these, no matter how fast the growth or how high the globalization, the transaction subjects are still humans. But the strongest signal from this year’s conference is that Stripe’s next major focus is: Agentic Commerce, where intelligent agents become market participants.

This shift is already quietly reflected in Stripe data. Usage of Stripe CLI surged in 2026 because intelligent agents don’t need fancy GUIs. Traffic from agents reading Stripe documentation increased about tenfold last year. If the trend continues, by year’s end, the number of agents reading documentation will surpass humans.

They’ve refined their API documentation over a decade and found a new batch of the most loyal readers. If you think agents spending money is still unfamiliar, think of these two things:

First, shopping interfaces are shifting toward model chat windows. Consumers research products with ChatGPT or Gemini; research and transactions are compressed into one interface. John Collison explained why this is irreversible: if an agent completes the entire process from research to checkout and product delivery, it won’t go to another website to re-enter information. Once the shopping agent finishes searching, the next step is checkout.

Second, more directly: OpenClaw, one of the hottest open-source autonomous agent frameworks. Users give commands via Feishu or Telegram, and the agent autonomously executes tasks. It can spend hundreds or even thousands of dollars in tokens daily, managing its own consumption. Moving from token management to direct spending by agents is just one step away.

At this year’s conference, Stripe demonstrated this step. On stage, John Collison gave an instruction to an agent: “Research how AI demand affects the energy market.” The agent searched and found that Alpha Vantage has an energy data set it needs, priced at 4 cents. The agent decided within its budget to autonomously purchase and download using Tempo CLI’s stablecoin wallet.

Then it generated a complete analysis report. But it didn’t stop there. John said, “Publish and sell this report, set a reasonable price so other agents can find and buy it.” The agent checked the data set’s licensing (allowing commercial use), then built a website, published the report, and generated a request file so other agents could purchase with a single request.

In just a few minutes, one agent completed the entire chain: research, procurement, production, compliance, publishing, pricing, and sales. It was both buyer and seller. John said on stage: “Agentic Commerce is here.”

Another demo: an agent automatically discovered an API review app using Machine Payments Protocol, autonomously completed a $2 payment, with human only providing fingerprint authorization. Developers don’t need to write separate payment logic; the agent can find it itself.

Even more impressive is the streaming payment demo. An app charges in real-time based on AI Token consumption, at $3 per million tokens. Multiple agents run simultaneously. The left dashboard shows rising token consumption; the right shows stablecoin micro-payments flowing in. Opening the Tempo blockchain explorer, the total of $3.30 paid consists of thousands of micro-payments, each only a third of a cent.

Credit cards, ACH, UPI, and Pix can’t do this. This is the world’s first streaming payment system. John believes that intelligent agents could make a business model long discussed but never truly realized—micro-payments—feasible.

Humans are not good at making extremely granular consumption decisions; no one wants to decide if a song is worth 15 cents every time they press play. But agents don’t have this cognitive burden. If this judgment is correct, many business models that failed due to human cognitive friction could suddenly become viable in front of intelligent agents.

If agents become new consumers, new risks will emerge. Stripe’s data head Emily Glassberg Sands described three rapidly growing fraud patterns: multi-account abuse (one in six AI companies involves this); malicious free trial consumption (some partners incur over $500 in token costs per paying customer, with 19 out of 25 trials being fraud); “free-riding” (customers consume大量tokens and refuse to pay at the end of the month).

She quoted: “Compute power is the new cash.” Traditional SaaS abuse has near-zero marginal cost, but each inference call in AI incurs real costs; stolen tokens are stolen money.

But there’s a paradox: many AI founders respond by shutting down free trials. Emily asked everyone claiming to have “solved” this, and found the solution is simply to turn off free tiers. But product lead Jeff argued this creates another problem: agents are becoming the main way to discover new services. If they can’t try freely, they will just jump to another URL.

Emily added: if the call to action presented to agents is “Join the waitlist” or “Contact sales,” they will leave immediately. Shutting down self-service registration to prevent fraud hands over the most important growth channel.

Stripe’s answer is the fraud prevention system Radar. It learns from transaction data of 5 million businesses; if a company encounters certain fraud patterns, all benefit. Last month, Radar blocked over 3.3 million high-risk free trial registrations among eight high-growth AI companies.

Jeff also offered an counterintuitive view: autonomous shopping by agents might ultimately be safer than humans. Human trust relies on inference (how long a user stays on a site, click paths), while transactions by agents can be programmatically verified. Shared Payment Tokens tokenize payment credentials; agents never see the original credit card number. When trust shifts from inference to confirmation, the security baseline could actually improve.

To sustain this ecosystem, protocols are key. Machine Payments Protocol enables agents to discover and complete payments via HTTP. The Universal Commerce Protocol, initiated by Shopify, with Meta, Amazon, Salesforce, Microsoft, and Stripe on the steering committee, aims to create shared standards. These companies, both partners and competitors, agree to develop common protocols because fragmentation benefits no one.

Regarding protocols, I noticed a special partner: Visa. Visa is essentially a protocol platform itself. Its founder Dee Hock’s book “One from Many” describes how, in the late 1960s, decentralized design allowed dozens of competing banks to share infrastructure. Two different eras, two different companies doing similar things.

Patrick Collison has publicly paid tribute to Hock, calling him “a seriously underrated innovator.” A clearer signal: David Stearns, a renowned academic historian of Visa, later joined Stripe. On stage, among the validators of Tempo blockchain, there’s a name: Visa. Visa, founded by Hock, is now a node in the blockchain network incubated by Stripe. Students built a new network; the teacher became a node.

When Patrick traced the origins of this idea at the opening, he said he was originally a Lisp programmer. Lisp’s core is “code is data,” which he translated into Stripe’s language: “money is data.” Dee Hock approached the essence of currency from organizational theory, concluding that currency is just a “guarantee of value exchange”; Collison from a programming language perspective directly equates currency with data, programmable, callable via APIs, operable by agents. Both speak different languages but tell the same story.

Finally, an interesting detail: the conference’s originally scheduled closing fireside chat was Patrick interviewing OpenAI co-founder Greg Brockman, but just before going on stage, the guest was replaced by Sam Altman. The audience cheered. They’ve known each other for 19 years; Altman was one of Stripe’s earliest angel investors, back when the Collison brothers were under 20.

Patrick asked a personal question: why invest in two teenagers? Altman said they wanted to build products that solve problems they personally faced, and he saw the potential for scale.

Altman divides OpenAI’s transformation into three stages: from research lab to product company, and then to a “Token factory” supplying intelligence to the world. Stripe is very similar: in 2010, they solved “online payments are too hard”; by 2026, they found a new problem: their clients might soon no longer be humans.

This conference gave me an idea: Stripe controls transaction data close to nearly 2% of global GDP, capable of seeing where every dollar of AI income comes from, where it goes, and how fast it grows. If Solow had a heart rate monitor back then, perhaps he wouldn’t have to wait ten years to find computer productivity in the statistics.

Maybe one day, Stripe could provide a model for the AI economy—not a large language model, but a Nobel-level economic model. Who says it’s impossible? Just a few years before Demis Hassabis won the Nobel, who could have imagined it?

To you, I want to say simply: this wave of the AI economy has been clearly mapped out through Stripe’s data. Its heart is a distributed computing protocol; its blood is intelligent agent consumption commands; its skeleton is a new payment and trust protocol. Traditional financial intermediaries are being reconstructed, and the dawn of micro-payments is already on the horizon.

Stop viewing this market with old mindsets. The main players in this reshuffle are not humans—they are code.