Stripe Annual Letter: A new level of cognitive density, especially the Level 5 model of "AI + Payments"

Original Title: Stripe Annual Letter: What Was Said? A New Level of Cognitive Density, Especially the “AI + Payments” 5-Level Model!

Original Author: Chen Tianyu Universe

Hello everyone, I am Chen Tianyu Universe.

Today, we’re going to analyze a letter—the Stripe 2025 Annual Letter. Why analyze it? Because through Stripe’s $1.9 trillion in transaction volume, we see many future trends that we haven’t noticed before!

If you’re not in the industry, you might not be very familiar with Stripe. Simply put, Stripe is a global payment aggregation service provider, similar to Adyen, PingPong, and Airwallex.

This letter, less than 3,000 words, is extremely dense with information. It’s not only a year-end summary for the company but also a deep analysis and forecast of global economic activity driven by payments.

In 2025, Stripe’s total transaction volume reached $1.9 trillion. What does that mean? A 34% year-over-year growth, accounting for about 1.6% of global GDP.

In other words, for every $100 of wealth produced on Earth, $1.60 flows through Stripe’s channels. In 2024, this number was $1.4 trillion. An increase of $500 billion in just one year.

After reading this letter, I only have one feeling: the old world we’re familiar with is collapsing and restructuring at an astonishing speed.

I’ve tried to break down the key insights from Stripe’s annual letter in my own words, though some parts are directly translated to preserve the original flavor. The amount of information is huge, but each point is worth pondering.

If you’re an entrepreneur or involved in payments, I recommend reading this multiple times. Every trend here is influencing everyone’s future survival.

A Sorting Machine

At the start of the letter, a very hardcore concept is introduced—the “sorting machine.”

They say that market economy is essentially a giant algorithmic machine. Its role isn’t to make everyone equally rich but to ruthlessly filter profits, capital, and talent, redistributing them to the most productive companies.

In the past, this machine turned slowly, and everyone had enough to eat. But now, AI has turbocharged this machine.

Data shows:

The top third of US profitable companies account for two-thirds of the stock market’s total value. This is the highest share since data began in 1963.

The top 10% of companies in the S&P 500 contribute 59% of profits. And this isn’t a valuation bubble—it’s real profit concentration.

This divergence isn’t just between big and small companies; it’s a life-and-death struggle within industries. The letter gives several examples:

Retail: Offline physical sales have grown only 5% after inflation over the past three years, while e-commerce grew 30% in the same period.

Aviation: The three major US airlines—American, Delta, United—have seen market share and profits rise over the past decade. But by 2025, Delta and United have nearly captured all profits in the US airline industry. Other airlines are basically just along for the ride.

Healthcare: Hospital and insurance profits have shrunk significantly since 2019, but the medtech sector is expected to reach over $110 billion EBITDA by 2029. Money is flowing from traditional players to new entrants.

The horror of this “sorting machine”: previously, big fish ate small fish; now, fast fish eat slow fish—especially AI-powered fast fish devouring sluggish, less responsive fish.

More macro data:

Demand for software, computers, and data centers drove nearly half of the US GDP growth in 2025. This proportion will soon become the majority.

What does this mean?

We used to say software is devouring the world; now it’s driven by computing power. Industries that don’t leverage computing and software are being ruthlessly left behind by this machine.

AI-Driven Entrepreneurial Boom

What has Stripe observed?

In 2025, the number of new companies joining Stripe hit a record high, with 57% from outside the US.

Moreover, these new companies are growing insanely fast—50% faster than those in 2024.

Even more astonishing, companies that reach $10 million annual revenue within three months have doubled.

But AI companies see this as routine.

Why?

Because GitHub code commits increased by 41%. Previously, annual growth was around 10-12%, but in 2025, it skyrocketed to 41%.

iOS app releases in December 2025 grew 60% year-over-year.

Previously, building an app required a team of ten and half a year of effort. Now, with a programmer and a few AI agents, you can code over the weekend, and by Monday, the product is live and collecting payments.

Stripe itself has also pushed this wave. They launched “claimable sandboxes,” allowing you to start a Stripe account directly from AI programming tools like Manus, Replit, Vercel.

Want to go live after testing? Just one click, and the sandbox becomes a full account—no configuration needed. Over 100,000 such sandboxes have been created so far.

Another product, Stripe Atlas—the world’s simplest company registration tool—grew 41% last year. And these new companies are monetizing faster: in 2025, 20% of Atlas startups received their first payment within 30 days, compared to only 8% in 2020.

Given these data, we have to ask: Is 2025 an anomaly or the start of a new normal?

Stripe’s judgment: The acceleration in 2025 marks the beginning of a major turning point driven by large language models, entrepreneurship, and creativity.

This “sorting machine” is speeding up. If you can’t make it into the top 10%, you might not just face slow decline but be instantly folded out of the system.

Globalization Is Dead, Localized Growth Is New

A particularly interesting point in the letter is the idea that “globalization has become the default.”

In the past, Chinese entrepreneurs expanding abroad followed a fixed path: first, they develop domestically, then test waters in Southeast Asia or the US, and finally consider expanding into Europe.

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.

After the internet arrived, free services launched globally simultaneously, but payment tools lagged behind.

When Facebook opened to the public in 2006, anyone could register, but advertising was difficult. It wasn’t until 2009 that Facebook supported international currencies—by then, the company had been around five years. Google launched global search in 1998, but it wasn’t until 2002 that it received its first GBP ad revenue in the UK.

This approach is dead.

What do AI products look like now? OpenAI’s ChatGPT, Anthropic’s Claude, Replit, Cursor, Midjourney—all launched globally at the same time, covering all markets instantly. Every AI product we’ve heard of is wildly popular in every country we know.

Stripe’s data shows a very counterintuitive phenomenon:

For companies earning most of their revenue overseas, 30% of their income isn’t from their home country or the top 10 economies like the US, China, Japan, Germany, but from lesser-known “transparent” countries rarely seen in the news.

What does this mean? Demand is dispersed, but tools are unified.

Previously, if you wanted to collect money from Kenyans, you had to open a local branch and connect with local banks via M-Pesa. Now? Stripe handles over 120 payment methods, just a click in the backend.

For example: Gamma, a California-based AI platform with 70 million users for presentations. When Gamma integrated Stripe to enable UPI payments in India, its Indian revenue shot up 22% that month.

This is the benefit of new infrastructure: your product can be based in San Francisco, but your money can instantly flow from a street vendor in Mumbai.

Ironically, the most geographically restricted are not traditional industries but fintech companies themselves:

Chime, after 12 years, still can’t go abroad. Want to open a Chime account outside the US? Not possible.

Nubank, founded in 2013, served only Brazil for the first six years, then added just two more countries in the next six.

Even Stripe itself, with its Issuing product, took 7 years to cover only 22 countries—faster than most, but still not fast enough.

But this is changing.

Next-generation fintech players—Sling Money, DolarApp, Felix, KAST—started building global financial apps from day one. Last year, Stripe launched its first inherently global product, Financial Accounts, allowing companies to hold, send, and receive funds. On launch day, it covered over 100 countries.

Why? Because of stablecoins. Their borderless nature enables fintech to build a universal global infrastructure.

Born global—finally making it possible.

Crypto Winter, Stablecoin Summer

2025 might be a crypto winter, but it’s definitely a summer for stablecoins.

Over the past decade, stablecoin trading volume has followed crypto asset prices. But in 2025, that correlation breaks.

This year, Bitcoin’s price plummeted 50% from its high.

One might think the crypto winter is here again. But strangely, stablecoin payments doubled, reaching $400 billion. About 60% of that is B2B payments.

This is the famous phrase from the letter: “Crypto Winter, but Stablecoin Summer.”

In the past, people traded crypto to get rich. Now, they use stablecoins to save money and make quick profits.

Stripe’s acquisition of stablecoin orchestration platform Bridge saw trading volume increase over fourfold.

Now, startup paths have changed:

A Y Combinator founder can raise funds using stablecoins.

They can earn interest in Stripe’s financial accounts.

When needed, they can pay global engineers—who could be anywhere in the world.

SaaS platforms can use smart contracts to collect stablecoin subscriptions, with wallet owners not needing to manually sign each transaction.

Companies expanding internationally with stablecoins now have better payment tools, embedding digital wallets directly into their core products.

With Privy, Ramp, and Deel, companies can manage both custodial and non-custodial wallets via a single API.

Interoperability between stablecoins and fiat currencies is also rapidly improving.

In April 2025, Bridge partnered with Visa to launch a card allowing businesses and consumers to spend their stablecoins like any credit card. When swiping, funds are deducted from the stablecoin balance and automatically converted to local currency; merchants see no difference from other payment methods, and users might not even realize they’re paying with stablecoins.

Phantom, one of the most popular crypto wallets with 20 million monthly active users, is issuing stablecoin cards via Bridge.

Even Visa has started issuing stablecoin cards. This means your USDC in your wallet can be used at corner convenience stores to buy cola.

Stripe Builds a Chain for Payments

To prepare for the era of large-scale stablecoin payments, Stripe has spent considerable time researching blockchain.

Current blockchains are designed for transactions and DeFi; throughput, reliability, cost predictability, and privacy for payments are not yet the focus.

Bitcoin can process fewer than 10 transactions per second.

In 2025, a wave of Meme coin trading on a major blockchain caused a Bridge user’s payment to be delayed over 12 hours, with transaction fees soaring 35 times.

These issues are critical, but will become more so in the future.

Stripe predicts that most internet transactions will be completed by agents—AI agents—and we need blockchains capable of supporting over a million, even a billion transactions per second.

In September 2025, Stripe and Paradigm incubated Tempo, a dedicated blockchain for payments.

Tempo’s logic is simple: no fancy DeFi, just focus on payments.

Dedicated payment channel: no competition with Meme coins.

Sub-second final confirmation: as fast as swiping a credit card.

Optional privacy: protecting business secrets while remaining compliant.

Interoperable with compliance and accounting systems: no money laundering, more anti-fraud measures.

These features may sound subtle, but they are crucial for supporting real economic activity.

Visa, Nubank, Shopify are already testing Tempo for various scenarios, including global payments, embedded finance, remittances, and more.

Klarna, whose CEO once called himself a crypto skeptic, is now the first bank to issue stablecoin—KlarnaUSD—on Tempo’s testnet, using Bridge’s Open Issuance to accelerate and reduce cross-border settlement costs.

Tempo’s architecture is especially suitable for agent payments and microtransactions.

Mainnet is launching soon. We look forward to ambitious companies creating new innovations with it.

Small Business Financing Dilemma

Since 2008, per capita GDP growth in OECD countries has been a meager 1.0%, compared to an average of 2.8% over the previous 46 years.

Aging populations in Japan, Brexit, Europe’s energy restructuring…

But strong evidence points to a key culprit: a sharp decline in global capital availability.

Post-financial crisis, banks in most OECD countries increased capital requirements, reducing small business access to capital.

Basel III raised global banking capital standards.

In Ireland, from 2011 to 2019, bank loans to small businesses fell over 66%.

In the UK, after 2012, small business lending remained weak.

In the US, slightly better—non-bank lenders provided some channels, with per capita GDP growing 1.7% over the past 15 years. But the overall trend post-Dodd-Frank is similar: tighter banking rules, less access for small businesses.

Since 2010, loans over $1 million increased 68%, while loans under $1 million decreased 5%. Last year, approval rate for small business loans in the US was only 41%, down from 50% in 2015.

Stripe Capital was created to address the global capital scarcity, especially for small businesses.

For companies using Stripe for payments, real-time revenue data simplifies lending decisions. They can repay with a small portion of future sales—essentially borrowing from their own growth.

From 2024 to 2025, financing volume grew 45%, serving over 80,000 businesses. Many of these companies obtained funds through vertical SaaS platforms—GlossGenius for salons, Tekmetric for auto repair, Pixieset for photographers, etc.

In the past two years, Stripe conducted a random study to understand capital’s impact: how fast can companies grow when we “lubricate the wheels”?

The results: much faster.

Companies that received Stripe Capital grew 27 percentage points faster in the following year than those that didn’t.

The fastest 10% of companies grew three times faster than their peers. The next tier was nearly 100 percentage points faster.

For example: Xirsys, a California-based server hosting company, used Stripe Capital to set up servers in China, India, and Japan, and then doubled its revenue.

Even companies with low credit scores saw growth accelerate by 11 to 18 percentage points after financing.

Stripe speculates that as AI improves investment returns, capital availability may become an even more critical variable in future economic outcomes.

“Low-Income Mode” OR “High-Income Mode”

For mature companies, obvious growth levers are scarce, and the most promising ones often don’t work.

Take advertising.

In 2019, Gillette’s ad “The Best Men Can Be” won a Silver Lion at Cannes. Later that year, Procter & Gamble wrote off $8 billion in that business because the award-winning ad failed to boost growth.

Recently, an Indian railway ad won the Grand Prix and six Gold Lions at Cannes 2025—a daily lottery ticket with a chance to win 10,000 rupees. But after eight weeks, it stopped because ticket sales didn’t increase.

That’s why Stripe has been passionate about payments for 15 years.

Compared to uncertain growth paths, optimizing payment setups is almost a guaranteed way to increase revenue and has one of the highest ROI among growth activities.

Microsoft evaluates payment providers monthly. From June to October 2025, Stripe used adaptive acceptance, card account updater, network tokens, and other techniques to significantly improve authorization rates. As a result, Microsoft shifted more payment traffic to Stripe.

Gatwick Airport’s previous payment provider had issues with failed payments, disputes, and customer complaints. After switching to Stripe, payment success rate increased by 2.5 percentage points.

FICO ran an A/B test comparing previous providers with Stripe, then fully switched, increasing authorization rate by 1 percentage point.

Telehealth company Ro, over the past 12 months, improved authorization rate by 2% and reduced dispute rate by 3%, adding millions of dollars annually.

Most companies are now living in Stripe’s “low-income mode”: payment infrastructure not optimized, losing money on conversion, authorization, fraud prevention.

What does “high-income mode” look like?

Checkout pages dynamically tailored per customer. Local currency pricing, and a handful of the most relevant local payment methods.

For Polish customers, BLIK increases checkout conversion by 46%; for Brazilian customers, Pix increases it by 31%.

Behind this is Stripe’s decade-long AI investment, including their “payment core model” and a suite of AI-driven services—Stripe Radar, Optimized Checkout Suite, Authorization Boost—that silently optimize billions of dollars in transactions daily.

But there’s still room for improvement!

Last year, Stripe tested a new authentication method: customers place their card on their phone. This NFC tap verifies the card’s chip, effectively proving the cardholder’s possession. DoorDash participated in testing and found that compared to previous fraud prevention methods, conversion rates improved significantly, and chargebacks decreased.

Stripe also built a new model for Radar to combat emerging AI-driven fraud—especially those abusing free trials to steal AI inference power, which is now common.

Here’s an excerpt:

CEOs: Flashy ads are fun, but don’t overlook the growth opportunities right under your nose. Your payment team deserves more recognition. Maybe host an award ceremony? In southern France, celebrating payment optimization. Or if the weather’s better, come to San Francisco in April.

Agentic Commerce

A recent hot topic in AI circles: models can not only think using training data but also go online, browse, code, and do things on the internet.

For Stripe’s world, the most relevant is agentic commerce: your AI will soon buy things for you.

Like other AI concepts, agentic commerce has been overhyped prematurely. Some paint a big picture: autonomous agents will know all your preferences, plan, and execute all your shopping.

Stripe’s approach is to break down this big picture into five levels:

Level 1: Eliminate web forms

You research and decide what to buy. But filling out forms is disliked. If an agent can be sent a link, fill in payment and shipping info, and confirm back, that’s much easier.

The system makes no decisions; it just types “buy” for you.

Level 2: Descriptive search

No need to search for products or attributes; just describe the scenario.

“I need back-to-school supplies for a third-grader in Chicago, including clothes (not too tight or itchy!), pencils, notebooks, lunchbox. My son likes KPOP Demon Hunters and tennis. School starts at the end of August.”

The system considers weather, materials, sizes, durability, tastes, reviews, delivery time. Niche and long-tail products become easy to find. Tedious keyword searches will be history.

Level 3: Persistent memory

No need to reintroduce yourself.

“Find back-to-school clothes options for Bobby.”

The system remembers your preferences, inferred from previous conversations and purchases. You still decide what to buy, but options are filtered by your taste and budget.

Level 4: Delegated authority

No more multiple-choice decisions.

“Handle back-to-school shopping within $400.”

The system searches, evaluates, and purchases on your behalf. You trust it to weigh options and pick what your kid likes. You only set the budget.

This is what most people mean by “agentic commerce” today.

Level 5: Proactive prediction

No prompts needed.

The system already knows the school calendar, your kid’s preferences, your usual budget. You just get a notification: here’s the completed shopping list.

This is the most sci-fi version: everything you need appears just before you realize you need it, without you saying a word.

Currently, the industry is on the cusp between Level 1 and Level 2.

Stripe looks back to the mid-1990s, when the internet structure we use today was solidified. Netscape created the graphical browser. HTTP and HTML became the shared application layer. URLs and DNS exploded in popularity. Back then, no one knew which protocol or player would win. Every Google faced an AltaVista.

Now, agentic commerce is at a similar critical moment—it has the potential to be a generational revolution.

Like early internet, its future depends on universal interoperability.

To that end, Stripe did several things last year:

• Co-developed the Agentic Commerce Protocol (ACP) with OpenAI, establishing a shared technical language between AI platforms and enterprises. Open design, usable across payment providers and AI platforms.

• Launched Shared Payment Tokens, a new primitive allowing agents to initiate payments without exposing credentials. Even companies not using Stripe for payments can forward these tokens to their vaults or other processors as secure credentials.

• Released the Agentic Commerce Suite, a toolkit enabling enterprises to sell across multiple AI interfaces and protocols (including ACP and Google’s Universal Commerce Protocol) with a single integration. Checkout, payments, fraud prevention—all working seamlessly underneath. Brands already onboard include Anthropologie, Urban Outfitters, Etsy, Coach, Kate Spade.

• Introduced machine payments, allowing developers to accept micro-payments from agents via stablecoins—through API calls, MCP, or HTTP requests. Autonomous agents are becoming a new customer type for internet companies.

• Collaborated with OpenAI to embed shopping experiences in ChatGPT, and with Microsoft to bring similar capabilities to Copilot.

No one can predict exactly where agentic commerce will be by the end of 2026, but it’s clear it has moved beyond hype into real development and experimentation. The pace of change will only accelerate.

If all goes well, those small agents won’t be confined within walled gardens but will run freely on open protocols, racing on broad, high-speed highways.

Licensing Republic and the Cost of Innovation

Today’s entrepreneurs and innovators have tools and reach unimaginable to previous generations of industrialists. If human ingenuity’s collective effect ultimately boosts productivity and living standards, that’s fantastic.

Last year, Joel Mokyr received the Nobel in Economics. Mokyr is known for emphasizing the importance of culture relative to capital, labor, and technology. 18th-century industrialists relied not just on coal or geography but on a new culture—a “mindset of improvement,” believing that the status quo was imperfect and improvable.

In “The Political Economy of Technological Change,” Mokyr also observed that new technologies often failed not because they weren’t advanced enough but because technology adoption involves a complex web of non-market aggregators—regulators, committees, courts—that influence what gets adopted.

As AI and the internet expand the boundaries of possibility, the costs of overcoming adoption and adaptation barriers grow. The ongoing economic polarization indicates that growth depends on the application of useful knowledge, not just the existence of abstract knowledge.

AI promises to revolutionize drug discovery… but only if regulatory processes (including clinical trials) become faster and cheaper.

European entrepreneurs can use new tools to boost a sluggish economy… but only if they cut burdensome regulations that do more harm than good, like the EU AI Act.

Next-generation nuclear tech could bring energy abundance… but only if outdated regulatory regimes are thoroughly reformed.

Autonomous logistics—from long-haul trucks to drones—can drastically reduce physical goods costs… if local regulations don’t harden into barriers.

Mokyr wrote about the “Intellectuals’ Republic” as a catalyst for the Industrial Revolution. Today, we live in a “Licensing Republic”: a filter composed of non-market aggregators.

Many regulations are well-intentioned, but it’s more important than ever to ensure they carefully weigh the benefits against the potential to block progress.

It’s fortunate that we can support resilient companies pushing the boundaries of possibility:

• Mistral AI and Bending Spoons demonstrate that top European talent can pierce through regulatory frozen layers.

• Zipline and Varda are gradually obtaining permits for complex new hardware.

• Spring Health and Maven Clinic are weaving new software layers into modern healthcare.

We still believe that ideas are crucial for economic progress, and many of the best ideas are undervalued. Stripe Press’s reading list includes many people, stories, and models we think will contribute to the next wave of improvements. Our team recently celebrated selling 1 million books. “Works in Progress,” our magazine about underappreciated ideas that can improve the world, has just started print editions.

The Event Horizon of a Black Hole

This letter devotes a large part to AI progress, sometimes feeling like it’s happening faster than we can follow. The gap between newly released products and the most advanced from last year is stark.

It reminds us of falling into a black hole. If you’re unlucky enough to experience this, crossing the event horizon doesn’t feel particularly different: the path is locally smooth, even though the future possibilities have irreversibly changed once you pass that boundary.

When we write this letter, we might be standing on the eve of a different, hopefully better singularity.

Though 2026 may seem similar to previous years in many ways, it’s also clear: the next decade will be completely different from the last.

We continue to believe that vibrant entrepreneurial spirit and wise culture can help build a more successful future society. We hope Stripe can play a small role.

If you’re also working to drive economic growth—whether as an entrepreneur, business leader, or infrastructure builder—we look forward to seeing you at Stripe Sessions in April. As always, there’s much to discuss.

That “sorting machine” won’t stop; it will only spin faster. Whether you become a winner selected by it or just data discarded as redundant depends on how you respond now.

That’s all for now. See you next time.