Fintech Giants' Perspective on AI Payments: Five Levels, Stablecoin Infrastructure, Next-Generation Global Business

Writing: Sleepy.txt

On February 25th, Stripe released its annual open letter.

In 2025, the total transaction volume flowing through Stripe’s payment network reached $1.9 trillion, accounting for 1.6% of global GDP—more than Australia’s annual GDP. But Stripe’s founders, the Collison brothers, didn’t use this annual letter to boast about their performance; instead, they discussed the Industrial Revolution, Nobel laureates in economics, and black hole physics.

Why would a payments company talk about these topics? What are they really trying to say?

Stripe believes that a war over “who will build the next generation of global infrastructure for commerce” has quietly begun. And it aims to be the one setting the rules. This letter is its pre-battle mobilization, a call to CEOs and entrepreneurs worldwide.

This machine is spinning faster and faster

Stripe sees this as a critical moment because the “market”—the sorting machine—is operating at an unprecedented speed.

Its role isn’t to promote shared prosperity but to ruthlessly filter out profits, capital, and talent, redistributing them to the most productive companies. In the past, when this machine spun slowly, everyone could get by. But now, AI has powered this machine with a new engine.

Stripe cites data: the top third of profitable U.S. publicly traded companies account for two-thirds of the entire stock market’s value—this is the highest proportion since data collection began in 1963. J.P. Morgan’s early 2026 outlook also describes a winner-takes-all landscape, highlighting market over-concentration. The top 10% of companies in the S&P 500 contribute 59% of profits.

This divergence isn’t just between big and small companies; it’s a life-and-death struggle within industries. The letter provides some industry examples, and here’s some background:

Retail: Over the past three years, offline retail sales, adjusted for inflation, grew only 5%, while e-commerce grew 30%. This means that if you’re a purely brick-and-mortar retailer, you might feel your business is still barely holding on, but in reality, you’re already falling behind.

Aviation: Delta and United, the giants, nearly captured all profits in the U.S. airline industry in 2025. Other airlines are struggling just to survive.

Healthcare: Traditional hospitals and insurance companies have seen their profit shares shrink significantly, while the healthcare tech sector’s EBITDA (earnings before interest, taxes, depreciation, and amortization) is projected to surpass $110 billion by 2029. Money is flowing from old models to new ones.

More macro data shows that demand for software, computing, and data centers drove nearly half of the U.S. GDP growth in 2025. Once we said software was devouring the world; now, growth is driven by computing power. Industries that haven’t embraced computing and software are suffering more each day.

Let’s look at some startup-related data: code commits on GitHub surged 41% in 2025 (compared to an average growth of 10-12% in previous years), iOS app releases increased 60% year-over-year in December, and the number of companies generating $10 million annualized revenue within three months doubled.

AI is pushing startup speed to the limit.

Stripe’s own company registration service, Stripe Atlas, saw a 41% increase in registrations in 2025, with 20% of new Atlas companies receiving their first payment within 30 days—up from only 8% in 2020.

They also launched Claimable Sandboxes, allowing developers to deploy Stripe accounts directly from AI coding tools like Vercel and Replit with one click. Over 100,000 sandboxes have been created this way. This means a developer can go from idea to code to first payment in an unprecedentedly short time.

The sorting machine is accelerating, new species are emerging in large numbers, and they are inherently global. This raises the next question: these new species are born global, but can they truly collect payments worldwide?

Products are borderless, money is not

The answer is almost certainly no.

The internet has made information and products borderless, but the flow of money is still blocked by invisible walls. This is the biggest structural contradiction in today’s global commerce and the most important battlefield for Stripe.

What did globalization look like in the past? Coca-Cola took 20 years to bottle its first soda in Cuba; McDonald’s and Starbucks took 27 and 16 years respectively to open their first stores in Canada. In the internet age, Facebook took five years to support international currencies, and Google took four years to receive its first GBP ad revenue.

But now, the old approach—building a strong local base first, then expanding abroad—is no longer popular.

Today’s AI products, from day one, see the entire internet as their “domestic market,” launching globally and covering all markets instantly. Yet, while their customers are worldwide, their ability to receive payments is still blocked by national borders.

Behind capital flows are an ancient infrastructure built on nation-states—SWIFT, national clearing systems, local payment licenses, foreign exchange controls, AML compliance. This system was designed for money to flow within countries, not across the internet.

A developer wants to sell software online; they need to apply for a merchant account, which can take weeks. They need to integrate a payment gateway, which involves writing a lot of code. They need to handle different currencies, involving complex FX conversions. They must comply with local regulations, requiring a legal team. For a small team of two or three people, this is nearly impossible.

Stripe’s founders, the Collison brothers, have firsthand experience with this.

In 2007, as teenagers from Ireland, they founded their first company, Auctomatic, a management tool for eBay sellers. They quickly realized that the hardest part wasn’t coding or finding customers but receiving money from clients around the world.

At that time, they had only two options: use PayPal, which was unfriendly to developers and could freeze accounts at will; or deal with banks, which was even more complicated.

That’s why Stripe was born. They wanted to turn online payments from a complex, permissioned, friction-filled process into something as simple as calling an API.

Stripe’s success was precisely because it addressed this pain point. It handled all the backend work—dealing with banks, card networks, regulators—and provided developers with a simple interface. Developers no longer needed to worry about the nitty-gritty; they could focus on their products.

But even Stripe can’t completely break down this wall. The open letter mentions that Stripe’s card issuing product, launched seven years ago, covers only 22 countries. Fintech companies are the slowest to globalize; for example, U.S.-based Chime has been around for 12 years and is still only in the U.S., while Brazil’s Nubank took six years to expand outside Brazil.

Yet, demand exists. For example, Gamma, an AI-powered presentation tool born in California, saw its Indian income jump 22% in the month after integrating Stripe’s UPI payments in India. This case shows that once infrastructure is in place, suppressed demand can explode instantly. Stripe’s data confirms this: for companies earning most of their revenue abroad, 30% of their income comes from small, lesser-known countries rarely seen in news, not from their local markets or the top ten economies like the U.S., China, Japan, or Germany.

If the old financial infrastructure was designed for the old world, how can we break down this wall?

Stablecoins: a narrative independent of cryptocurrencies

Stablecoins may no longer be considered cryptocurrencies. They are a new global payment infrastructure, making money flow on the internet as naturally as data.

In 2025, during the crypto winter when Bitcoin prices plummeted 50%, stablecoin payments doubled to $400 billion, with 60% being B2B payments. Stripe calls this the “Summer of Stablecoins.” People are no longer just trading coins; they’re doing business with them.

Stripe’s acquisition of the stablecoin platform Bridge saw trading volume increase more than fourfold. A YC founder can now use stablecoins for financing, earning interest in Stripe’s financial accounts, and paying engineers worldwide. This was unimaginable before.

More dramatically, Klarna’s CEO, once a well-known crypto skeptic, now the bank behind Stripe’s Tempo testnet, aims to reduce cross-border settlement costs by issuing stablecoins.

Stripe predicts that future commerce will be driven by AI agents capable of handling billions of transactions per second on blockchain. But current blockchain infrastructure can’t support this future. So, Stripe has built its own chain—Tempo.

Focused on payments, it offers near-instant confirmation, optional privacy, and interoperability with compliance systems. Visa, Nubank, and Shopify are already testing various scenarios on it. Stripe also launched Financial Accounts, which on its first day covered over 100 countries—its first truly global financial product.

Stripe’s ambition is to become the TCP/IP protocol of this new infrastructure. It aims not just to patch the old pipes but to build a new, internet-native global payment network.

Most companies are wasting money

Stripe’s letter mentions that most companies operate in a “low-income mode,” wasting a lot of money daily on payments.

What is low-income mode? It’s when payment infrastructure isn’t optimized, leading to losses in conversion, authorization, and fraud prevention. High-income mode, according to Stripe, includes real-world examples:

• Microsoft evaluates payment providers monthly, continuously optimizing authorization rates, significantly increasing revenue.

• Gatwick Airport switched to Stripe, improving payment success rate by 2.5 percentage points. While small, multiplied by millions of transactions annually, it’s a substantial cost saving.

• Credit scoring firm FICO switched fully to Stripe via A/B testing, increasing authorization rates by 1 percentage point.

• Telehealth company Ro, after adopting Stripe, improved authorization by 2%, reduced disputes by 3%, earning tens of millions of dollars more annually.

These cases show that payment optimization is essential.

Another challenge for companies is financing. Since the 2008 financial crisis, small business credit has been tightening globally. Loans to small businesses in Ireland have fallen 66%, U.S. microloans under $1 million declined 5%, and OECD GDP growth slowed from 2.8% to 1.0%. Traditional banks are reluctant to lend to small businesses because they lack sufficient data to assess risk, and approval costs are high.

Stripe Capital’s logic: I have all your transaction data; I understand your business better than any bank. Using real-time transaction data, it bypasses traditional approval processes. Companies financed by Stripe Capital grow 27 percentage points faster than those without, with some even tripling their growth rate.

Stripe is transforming from a payment tool into a business operating system. It’s not just helping you collect money; it’s helping you finance, issue cards, manage finances, and prevent fraud. It aims to be the financial brain of enterprises, not just a payment terminal.

But these are human decisions and human shopping. If decisions and purchases are made by AI agents, what kind of infrastructure will that require?

AI agents are coming—who will provide their wallets?

When AI agents become new consumers, the entire payment infrastructure must be redesigned. The control over this design will determine the rules of the next-generation business.

What is Agentic Commerce? Simply put, when AI becomes sufficiently intelligent, it’s no longer just a search tool but an authorized agent capable of completing tasks for you. You tell it, “Book me a window seat on a flight to Shanghai next Tuesday,” and it will compare prices, book, and pay—all without your intervention.

We are on the eve of this new world’s explosion. Like the internet in the mid-1990s, foundational protocols like HTTP, HTML, and DNS are still competing, with no clear winner—AltaVista gave way to Google. Now, it’s the same; no one knows who will become the “HTTP” of agentic commerce.

Stripe divides the evolution of agentic commerce into five levels:

• L1: Eliminating web forms—AI helps automatically fill out registration, login, and payment forms.

• L2: Descriptive search—using natural language, you tell AI what you want, and it finds and presents results.

• L3: Persistent memory—AI remembers all your preferences and history.

• L4: Authorization delegation—you can authorize AI to make purchases within certain limits.

• L5: Proactive prediction—AI can anticipate your needs and arrange everything before you realize you need it.

Stripe believes we are currently at the cusp of L1 and L2. Crossing into L3 and L4 will completely transform commerce. When thousands of AI agents trade on behalf of humans online, they will need their own wallets and payment protocols.

This is the future Stripe is fighting for. They’ve partnered with OpenAI to develop agentic commerce protocols, collaborated with Microsoft to enable payments for Copilot, and launched the Agentic Commerce Suite, allowing brands like Etsy and Coach to sell through multiple AI platforms with a single integration. They even introduced a “machine payment” feature, enabling AI agents themselves to be charged as new customer types—meaning AI can pay AI.

When AI agents become new consumers, Stripe aims to be the provider of their wallets and payment protocols—a much larger battlefield than payment processing itself.

Technological barriers, whether borders for payments or wallets for AI agents, Stripe is addressing one by one. But there’s an even older, more stubborn wall blocking all these possibilities.

The final enemy

“Permissioned Republics.” This is Stripe’s term, quoting Nobel laureate and economic historian Joel Mokyr at the end of the letter. It’s a rare public statement from a business that the obstacle isn’t technology but a system of regulation—composed of regulators, committees, and courts—that, under the guise of “preventing bad things,” systematically stifles “good things.”

Mokyr’s core idea is that the Industrial Revolution in 18th-century Britain succeeded not just because of coal and steam but because the political and social environment fostered a mindset of improvement, encouraging innovation and entrepreneurship. Many technological failures in history weren’t due to poor technology but because they were suppressed by non-market actors—governments, guilds, churches.

Stripe believes we live today in a vast “Permissioned Republic.” It lists a critique:

• In AI drug discovery, despite AI predicting protein folding in weeks—something that used to take years—the pace of new drug approvals remains slow, averaging over ten years.

• European entrepreneurs are hamstrung by the EU AI Act’s complex regulations, requiring significant time and money to comply instead of focusing on product development.

• New safer, more efficient nuclear technologies are hindered by rigid, veto-style regulation, despite urgent climate pressures.

• Waymo’s autonomous vehicles face regulatory hurdles in San Francisco, despite data showing they are safer than human drivers.

But Stripe isn’t entirely pessimistic. It highlights counterexamples:

• Mistral AI in France and Bending Spoons in Italy have grown into world-class AI companies despite strict European regulation.

• Zipline in Rwanda and Varda in the U.S. are carving out new business models in drone delivery and space manufacturing, areas with tight controls.

• Spring Health and Maven Clinic in the U.S. are transforming mental health and women’s health services through software and data, despite being in conservative industries.

This is Stripe’s deepest concern and the underlying tone of the letter. It compares the current AI revolution to falling into a black hole: crossing the event horizon, you feel nothing, but your future is irreversibly changed. Stripe believes we are on the brink of a “different, hopefully better singularity.”

The letter ends without offering optimism or despair. It simply states that the market’s sorting machine won’t stop; it will only spin faster. Whether you become a winner filtered out by the machine or discarded as irrelevant data depends on how you respond now.

From a small Irish village of just 102 people in Dromineer, Stripe took fifteen years to turn seven lines of code into a global business empire accounting for 1.6% of GDP. Its next step is to define the rules of the next generation of global commerce.