Editor’s Note

U.S. crypto regulation is entering a very different phase.

Last week, the SEC released its proposed Regulation Crypto Assets, an attempt to establish a clearer framework for certain crypto-asset fundraising activities.

After years in which much of the industry operated in regulatory gray zones—and companies often relied on legal opinion letters to determine whether a token might be treated as a security—the more important shift may not simply be that regulation is becoming more permissive.

It is that regulators are beginning to answer a much more fundamental question:

What are the actual rules?

For this episode of the Fintechnize Founder Series, I sat down with my longtime friend Victor Fang, founder and CEO of AnChain.AI.

Founded in Silicon Valley in 2018, AnChain.AI focuses on crypto investigations, AML, and blockchain forensics and has worked with U.S. regulators and law-enforcement agencies including the SEC, IRS, and FinCEN.

Victor’s background sits at an unusual intersection of cybersecurity, artificial intelligence, and crypto. That made this conversation particularly timely as three major shifts are happening at once:

crypto regulation is becoming clearer, financial crime is becoming more sophisticated, and AI agents are beginning to operate on both sides of the security battlefield.

We discussed what Regulation Crypto Assets could mean for the industry, how the SEC’s technical understanding of crypto has evolved, why AnChain.AI bet on AI-driven blockchain investigations as early as 2018, and what happens when hackers themselves begin deploying autonomous AI agents.

This transcript has been edited for clarity and length. Repetitive phrasing and verbal fillers have been removed, and technical terminology has been standardized without changing the substance of the conversation. Specific regulatory provisions, figures, products, and external cases mentioned during the discussion should be independently reconfirmed before being cited as authoritative facts.

What Actually Matters About the SEC’s Regulation Crypto Assets Proposal?

Charlie:

We are seeing a series of important developments in U.S. crypto regulation—from the SEC’s Crypto Task Force to the CLARITY Act and now last week’s Regulation Crypto Assets proposal.

It feels as though the United States is gradually moving away from regulatory ambiguity toward a more explicit framework.

How do you see this shift?

Read more

On August 25, Google Cloud launched Gemini Enterprise for Financial Services, pushing AI deeper into KYC, credit analysis, portfolio monitoring, bond issuance and financial research — workflows that sit much closer to the operating core of a financial institution than the customer-service chatbots and internal copilots we have become used to.

Two days later, MoonPay connected PayBox with Kamino on Solana, allowing users to initiate lending, borrowing and yield-generating activities through interfaces such as Claude and ChatGPT.

Around the same time, the Dallas Fed published a much less flashy piece of research asking what might happen to bank liquidity and maturity transformation if deposits could move around the clock and AI agents could continuously search for better yields on behalf of companies and individuals.

Then BIS entered the picture. In its latest discussion of stablecoins and tokenized deposits, it highlighted Project Agorá, the experiment bringing tokenized commercial-bank deposits and central-bank reserves onto a shared programmable infrastructure.

Google is teaching AI how finance works. MoonPay is giving agents ways to act on financial assets. BIS is testing what happens when bank money itself becomes programmable. And the Dallas Fed is already asking what all of this could mean for bank balance sheets.

Put together, they point to a more important question than whether AI will “transform finance.”

The real dividing line is whether AI is allowed to make financial decisions on our behalf — and whether the financial system underneath it is capable of executing those decisions.

That difference may prove larger than the jump from search engines to ChatGPT.

AI has spent the past few years standing next to the money

When I first started looking seriously at generative AI in fintech a few years ago, I was relatively conservative about its near-term impact.

The most credible applications were customer support, reporting, fraud detection, expense management, internal knowledge retrieval and other productivity workflows. They were useful, sometimes very useful, but they largely shared one characteristic: AI stood next to the money without actually touching it.

There was a good reason for that.

If a model writes a bad paragraph of marketing copy, someone can edit it. If a model sends $5 million to the wrong counterparty, there is no convenient Undo button.

Financial institutions have always had a lower tolerance for black boxes than ordinary software companies. Accuracy matters, but so do authorization, auditability, liability, regulatory obligations and the ability to explain why an action was taken. Until those questions are solved, even the smartest model remains an adviser.

What has changed over the past few months is not that the industry suddenly believes fully autonomous wealth managers or CFOs are about to arrive. The more meaningful development is that companies are breaking the act of “letting AI move money” into smaller, more controllable pieces of delegated authority.

Read more

If you haven’t been staring at the crypto market over the past week, opening the Bitcoin chart today feels a little like coming back to a TV series after missing several episodes.

Those who know me personally also know what our family has been through over the past two weeks. I barely had time to look at markets for a few days, so even I was genuinely surprised when I finally caught up.

Bitcoin was struggling around $64K early last week. After it briefly slipped below that level, there was still plenty of discussion about whether $60K would be tested again. Then $67K went. $70K went. $75K went. By Friday, BTC had traded close to $79.5K. After a modest pullback over the weekend, it pushed higher again today, trading above $79K several times and reaching as high as $79,888.

That is a gain of more than 20% in a matter of days.

Naturally, everyone wants to identify the news that caused it.

There are plenty of candidates. The U.S. Treasury unexpectedly increased the size of its long-dated Treasury buybacks. The Trump administration delivered several crypto-friendly regulatory signals in rapid succession. Spot Bitcoin ETF flows suddenly turned positive again. And traders, perhaps most straightforwardly, point to a massive short squeeze.

All of these explanations are directionally right. None of them, on its own, is enough.

A more useful way to understand the rally is that a market already compressed for nearly two months suddenly encountered several catalysts at exactly the right time. ETF money started coming back first. Treasury and regulatory news then broke the equilibrium. Finally, a large amount of bearish positioning was forced to cover, turning what might otherwise have been a normal 5% or 10% rally into a classic reflexive move.

To understand why Bitcoin could suddenly jump 20%, therefore, we have to start with what happened in the two months before the rally.

The prerequisite for the rally: sellers had largely run out

Bitcoin peaked around $126K last October and bottomed near $58K by late June, losing more than half its value in roughly eight months. At first glance, this looks like another familiar crypto winter. But I think the structure of this downturn was quite different from 2018 or 2022.

Read more

Disclaimer: This article is for general informational purposes and does not constitute legal, investment, tax, or compliance advice.

On August 18, 2026, the U.S. Securities and Exchange Commission (SEC) proposed Regulation Crypto Assets, a framework designed specifically for certain investment contracts involving crypto assets. [SEC link]

The proposal could give cryptocurrency token issuers clearer paths to raise capital in the United States. More importantly, it introduces a lifecycle model: disclose what the issuer promises to build, report while that work continues, and provide evidence when the promised managerial work is complete.

This is a significant shift from years of regulatory uncertainty and case-by-case enforcement. It is not broad deregulation. Antifraud and antimanipulation rules would remain in force, and the proposal would not replace AML, sanctions, commodities, custody, tax, cybersecurity, or other legal obligations.

What the SEC Is Proposing

The startup exemption would allow an issuer to raise up to $5 million over a period of as many as four years. Issuers would make public filings when entering and leaving the exemption and provide principles-based disclosures while relying on it.

A separate fundraising exemption would have two tiers.

Tier 1 would permit offerings of up to $20 million in a 12-month period.

Tier 2 would permit up to $75 million. Both tiers would require public offering materials, financial information, and ongoing reporting. Tier 2 would require audited financial statements.

The proposed investment-contract safe harbor addresses a more fundamental question: when can a crypto asset separate from the investment contract through which it was originally sold?

An issuer would need to show that it completed or permanently stopped all essential managerial efforts previously promised to purchasers, that it is not making new promises of that kind, and that it filed a public certification supported by legal and factual analysis.

The proposal is not yet effective law. Its requirements may change through public comments and the final rulemaking process.

What AnChain.AI Raised With the SEC

On September 9, 2025, AnChain.AI met with the SEC Crypto Task Force to discuss the interpretation of rules and forms in digital-asset compliance. [SEC memo link]

AnChain.AI’s submission focused on 5 practical issues:

1. How existing requirements—including Form D, Form 144, Form 13F, and Suspicious Activity Reports—should apply to digital-asset transactions.

2. Whether issuers and intermediaries should disclose wallet addresses, smart-contract identifiers, and transaction-traceability information.

3. How securities filings should interact with FinCEN and OFAC requirements for AML and sanctions compliance.

4. Whether regulatory forms should include crypto-native fields such as wallet identifiers, token-issuance details, and smart-contract audit attestations.

5. How AI-driven automation and cross-agency coordination could improve accuracy while reducing unnecessary compliance burdens.

The SEC meeting record documents the discussion. It does not constitute an agency endorsement of AnChain.AI or establish that the submission caused the later proposal. However, the comparison highlights where legal policy and technical implementation converge.

What the Proposal Resolves—and What It Does Not

Regulation Crypto Assets directly addresses two major issues: compliant token fundraising and the lifecycle of an investment contract. It also advances principles-based disclosure and creates a stronger foundation for SEC-CFTC coordination.

Important implementation questions remain unresolved.

The proposal does not appear to provide a complete answer on whether filings must identify controlling wallets, deployed contracts, upgrade keys, vesting contracts, bridge deployments, or related-party addresses. It does not fully modernize every existing securities form. It also does not consolidate separate SEC, FinCEN, OFAC, CFTC, banking, and state obligations into one compliance process.

These gaps matter because a disclosure is only useful when it can be connected to the real asset, contract, transaction, and controlling entity.

Six Actions Crypto Firms Should Take Now

1. Classify the token and analyze each fundraising or distribution transaction separately.

2. Create a register of issuer promises from white papers, websites, pitch decks, social media, governance proposals, and sale agreements.

3. Map treasury, foundation, vesting, liquidity, bridge, governance, market-maker, and insider-controlled wallets and contracts.

4. Reconcile financial statements and public disclosures with minting, burns, unlocks, transfers, upgrades, governance votes, and treasury movements.

5. Maintain AML, sanctions, and market-integrity controls independently of any securities-law exemption.

6. Preserve evidence of product milestones, audits, code releases, governance transitions, key rotations, and completion of managerial commitments.

From Legal Opinion to Verifiable Evidence

The emerging framework will require more than a legal conclusion. Effective compliance must connect four evidence layers: regulatory claims, corporate records, smart-contract controls, and blockchain transactions.

AnChain.AI’s blockchain intelligence can support that evidence layer through wallet screening, entity attribution, cross-chain tracing, smart-contract risk analysis, transaction monitoring, and reproducible investigation reports. These tools do not replace lawyers, auditors, or compliance officers. They help make professional conclusions testable.

The central opportunity in Regulation Crypto Assets is clarity. The corresponding obligation is accountability.

Token issuers that build continuous, evidence-based compliance systems will be better positioned to use the proposed exemptions, support future safe-harbor certifications, and earn the confidence of investors, intermediaries, and regulators.

Preparing for crypto fundraising, disclosure, or AML readiness? Schedule a free consultation with AnChain.AI’s experts: https://anchain.ai/demo

Primary sources:

[1] U.S. Securities and Exchange Commission, “SEC proposes new Regulation Crypto Assets,” Press Release No. 2026-76, Aug. 18, 2026. [Online]. Available: https://www.sec.gov/newsroom/press-releases/2026-76-sec-proposes-new-regulation-crypto-assets.

[2] U.S. Securities and Exchange Commission, Crypto Task Force Staff, “Meeting with representatives of AnChain.AI, Inc.,” memorandum, Sept. 9, 2025. [Online]. Available: https://www.sec.gov/files/ctf-memo-anchain-ai-090925.pdf.

Disclaimer: This AnChain.Ai article is for general informational purposes only and does not constitute legal or financial advice. Consult qualified legal, financial, or compliance professionals regarding your specific circumstances.

Stablecoin payments company Rain announced the launch of the Agentic Payments Alliance, or APA, with 26 founding members packed into one logo wall: Visa, Mastercard, Fiserv, Circle, Solana, Avalanche, Fireblocks, Chainalysis, Uniswap Labs, Remitly, Shift4, and many more.

At first glance, it looked like a fairly typical tech industry alliance announcement: get a bunch of logos onto one slide, declare that everyone is collectively building the future, and then everyone goes back to work on their own thing.

But after reading through a few posts and reposts from the CEOs of the member companies, I started to think there might be more to it.

For one, I had just written about Rain in January, when the company raised a $250 million Series C at a $1.95 billion valuation. What interested me most at the time was a very payment-native story: how do you combine the settlement efficiency of stablecoins in the backend with the enormous real-world acceptance of traditional card networks such as Visa?

Rain disclosed then that its active card base had grown 30x over the previous year, while annualized payment volume had grown 38x.

Then, just six days ago, Rain acquired stored-value payments startup Ansa. That news had a more personal connection for me: Ansa founder Sophia Goldberg is a former Adyen colleague and friend. Following the acquisition, she joined Rain as Head of Payments.

Fundraise. Acquisition. Agentic Payments Alliance.

Put those three events next to one another, and Rain’s strategic direction over the past year suddenly starts to look much clearer.

More importantly, they point to a problem in agentic commerce that is often buried underneath the bigger AI and crypto narratives.

The question may no longer be:

Can AI spend money for us?

It may be:

How much authority should we actually give AI to spend our money?

This Is Not Just Another x402

Rain describes APA as a “working coalition” run collectively by its founding members, rather than an industry association controlled by Rain itself.

Even the alliance’s charter and specific mission are expected to be shaped by members after launch.

Its first phase is supposed to focus on shared research and frameworks, testing emerging standards around agent identity and authorization, and participating in regulatory discussions.

That distinction matters.

Because over the past year, agentic payments have already become something of an alphabet soup created by every large tech and payment company with a seat at the table.

Coinbase has x402.

OpenAI and Stripe have ACP. Stripe and Tempo also have MPP.

Google has AP2 and UCP.

Visa has its Trusted Agent Protocol.

Mastercard has Agent Pay, and more recently Agent Pay for Machines.

So at least today, APA is not another payment protocol to add to this growing pile of acronyms.

It feels more like the industry suddenly realized that plenty of protocols have already been created—and perhaps everyone should sit down and figure out how these pieces are actually supposed to work together.

The Payment Stack Behind the Logo Wall

If you look at the logos alone, the alliance feels a little random.

Once you map them onto the payments value chain, it starts to make much more sense.

Read more

Editor’s Note

As consumers increasingly ask AI systems what to buy, the nature of online commerce is beginning to change.

Today, an AI assistant can understand a user’s problem, research possible solutions, and recommend a product. But the transaction usually breaks at the final step: the user clicks out of the conversation, lands on a website, compares prices again, evaluates the merchant, and completes the purchase through a traditional checkout flow.

Pivota is building infrastructure for that missing layer.

In the second episode of the Fintechnize Founder Series, I spoke with Jack, founder of Pivota, about his journey from investing to Uber China, Ant Group, consumer finance, and eventually Agentic Commerce.

We discussed what Uber taught him about marketplace growth, why interest-free installments are often a marketing product rather than a financial product, why entering a market after consensus has formed is usually too late, and how commerce may evolve from “search to shop” into “chat to shop.”

We also explored Pivota’s core thesis: that the most important customer in the next generation of commerce infrastructure may not be the consumer or the merchant, but the AI agent making decisions on their behalf.

This transcript has been edited for clarity and length. Repetitive phrasing and verbal fillers have been removed, and terminology has been standardized without changing the substance of the conversation. Specific figures, protocol names, and external references should be reconfirmed before publication.

From Investing to Uber China and Ant Group

Charlie:

Welcome to the second episode of the Fintechnize Founder Series.

Today’s guest, Jack, is someone I have known for more than a decade. We first met when we were both studying in the United States, and recently reconnected through our shared interest in Agentic Commerce and the intersection of AI and fintech.

Jack has had a particularly interesting career. He worked in investing, joined Uber China as an early employee, spent several years at Ant Group working on payments and installment products, and later raised capital from Sequoia, SIG, and other investors for his first startup.

Jack, welcome. Could you start by introducing yourself?

Jack:

Thank you for having me, Charlie.

After graduation, I started in investing. I worked first at a hedge fund and then in venture capital. I spent around three years in that world, but eventually realized that I wanted to be closer to company building and entrepreneurship.

That led me to Uber, where I became one of the earlier employees in China. I stayed until Uber China merged with Didi.

After that, I joined Ant Group and spent close to four years there. Most of my work was related to payments, including offline payment growth and later the growth of Huabei Installments.

After leaving Ant, I started my first company. We raised money from Sequoia, SIG, and several other investors to build something similar to Klarna for the Chinese market—a Buy Now, Pay Later product.

Starting last year, I began paying much closer attention to Agentic Payment and Agentic Commerce. One of the main triggers was seeing OpenAI, Google, and other major players begin to release protocols related to payments and transactions initiated by AI agents.

After studying the market more deeply, I started Pivota earlier this year. We are focused on the missing infrastructure layer in Agentic Commerce, particularly outside closed platform ecosystems.

Charlie:

You were responsible for Qingdao and some surrounding cities at Uber China, correct?

Jack:

Yes. I started as an operations manager and eventually took responsibility for Qingdao and several nearby markets.

The scope was broad. I worked on operations, growth, marketing, government relations, and almost everything else required to build a city from the ground up.

What Uber Taught Me About Focus

Charlie:

I know several people who were part of Uber China in its early days. The culture was very different from that of most Chinese internet companies at the time.

Looking back, what did you learn from Uber that later shaped your work at Ant and your approach to entrepreneurship?

Jack:

Uber was one of the biggest turning points in my career.

Read more

Editor’s note
What does it take to build a financial platform for founders and companies that operate globally from day one, while the banking system they rely on remains fragmented by country?

In this episode of the Fintechnize Founder Series, I spoke with Christian, founder of Infini Money, about the company’s evolution from a consumer stablecoin card product into a broader financial operating system for individuals and borderless businesses. We discussed why Infini is combining cards, payments, treasury, on- and off-ramps, and team financial management within a single platform, as well as why Christian believes the next generation of fintech companies will compete less on whether they are “Web2” or “Web3” and more on product execution, customer trust, and workflow design.

Christian also reflected on Infini’s major security incident, the decisions the company made to protect users, and how the experience reshaped his understanding of security, governance, and responsibility in financial services. We ended with Infini’s plans for AI-powered financial agents, its expansion across APAC, and the company’s next stage of growth.

This transcript has been edited for clarity and length. Repetitive phrasing and verbal fillers have been removed, and technical terminology has been standardized without changing the substance of the conversation. Specific figures, product claims, regulatory references, and comments relating to the security incident or fundraising should be reconfirmed with the guest before publication.

Try Infini via this link: https://join.infini.money/business/uGptk98

Opening

Charlie:

Welcome to Fintechnize, a podcast about the people building at the intersection of AI and fintech. In each episode, I speak with founders, investors, and industry leaders about the ideas shaping the future of financial services—from intelligent agents and new payment rails to digital-asset infrastructure and the next generation of financial products. I’m Charlie Liu. Let’s get started.

Hello, everyone, and welcome to Episode 3 of the Fintechnize Founder Series. Today, it is our pleasure to have Christian, founder of Infini.money. Christian, please say hello to our audience.

Christian:

Hi, Charlie, and hi, everyone. Thanks for having me. I’m Christian, founder of Infini.

1. From NFTs and DeFi investing to building Infini

Charlie:

Let’s start with your background. What led you to build Infini?

Christian:

My story is a little unusual. I’m a relatively young founder, and I did not begin my career in traditional banking or at a fintech startup.

I became interested in crypto while I was in college. During the previous cycle, I first entered the industry through NFTs. Curiosity gradually led me into DeFi, where, for the first time, I felt that there was something much more fundamental underneath the speculation.

I later co-founded an investment fund called NextGen Digital Venture and spent several years investing across crypto, DeFi, and public equities. The more I studied the industry, the more convinced I became that the largest opportunity was not another token or another trading product. It was using this infrastructure to build financial products that real people and real businesses could use every day.

That was the reason we started Infini in 2024.

Charlie:

What was special about the timing in 2024? What made you decide to pull the trigger and start the company then?

Christian:

DeFi was still very popular, yields were high, and strong products such as Ethena and Morpho were emerging. Our first thought was that we might build a yield-bearing stablecoin or another yield product.

But as we studied the market, we realized that the larger opportunity was in consumer finance. At the time, there were not many neobanks with a genuinely good crypto experience, and crypto cards were still at a very early stage. In APAC especially, we did not think there was a product delivering the experience users wanted.

We did not have deep payments experience at that point, but the demand was obvious. Users were already waiting for a product like this. We became one of the earlier teams to combine stablecoin yield and a crypto card in a single consumer product. That was the opening we saw.

2. What Infini is today—and who it is built for

Charlie:

After almost two years of building, how would you describe Infini today in one or two sentences? Who is it primarily built for?

Christian:

Today, Infini is a stablecoin-powered, AI-native financial operating system for global consumers and borderless companies.

A neobank normally begins with an account, and a payments company normally begins with a transaction. We are trying to begin with the customer’s entire financial workflow: receiving money, holding it, converting it, spending it, earning on it, managing a team, and eventually allowing an agent to help operate all of those activities.

We are trying to build the financial operating system for the next generation. And when I say the next generation, I am thinking first about the next generation of founders.

Charlie:

You mentioned several customer groups. For a scalable product, there may be multiple segments, but there is usually one bullseye customer. Who is that for Infini?

Christian:

If I had to name one group, it would be the next generation of founders.

Everything is changing very quickly. In the past, building a product might have required fundraising first, followed by months or even years of development. Today, AI coding tools and agents can compress that process into weeks. If you have a strong idea, you can test it and go to market much faster.

These founders may incorporate one company in one jurisdiction, but they are global from day one. They sell to customers in different countries, pay vendors and remote teams in different currencies, and increasingly receive or move money in stablecoins. That is our most important early customer group.

3. Global capital flows still run on fragmented local banking systems

Charlie:

When you meet these founders and companies, what are the number-one or number-two problems they are trying to solve?

Read more

On July 29, Robinhood reported a quarter that looked almost impossible to fault.

Revenue reached a record $1.31 billion, up 32% year over year. Diluted EPS came in at $0.62, well ahead of expectations. Net deposits approached $22 billion, another record. Options, equities, and prediction markets all grew, while event contracts generated $156 million of revenue in a single quarter—more than crypto and more than equities.

And yet, the stock did not celebrate.

It fell after the release, recovered part of the loss, then remained caught in a broader downward trend.

The market was not saying Robinhood had failed to grow. It was asking a more difficult question: how much of this revenue comes from financial relationships that can last for ten years, and how much comes from the World Cup, market volatility, and a temporary burst of retail risk appetite?

That is what makes Robinhood so interesting today.

The company has already proven that it is exceptionally good at capturing attention. Its next test is whether the money remains in the account after the attention moves somewhere else.

From crypto beta to financial-attention beta

For a long time, discussing Robinhood was almost inseparable from discussing the crypto market.

Crypto prices went up, retail traders rushed in, and Robinhood’s revenue followed. Crypto prices fell, risk appetite disappeared, and both revenue and valuation suffered. The Robinhood of 2021 was, to a large extent, retail trading beta wrapped in a brokerage interface. Even the stock’s previous peak roughly ten months ago lined up closely with last year’s crypto rally.

That relationship is now weaker.

In Q2 2026, Robinhood’s crypto revenue fell 38% year over year to $100 million. Total company revenue still grew 32%. Options revenue reached $342 million, equities revenue reached $129 million, and event contracts generated $156 million.

Robinhood has clearly reduced its dependence on crypto as a single asset class.

But we should not jump too quickly to the opposite conclusion—that Robinhood has already become a low-volatility wealth-management platform.

Transaction revenue represented roughly 59% of total revenue in Q2, up from about 55% a year ago. Options, crypto, and event contracts generated close to $600 million combined, or approximately 46% of company revenue.

Robinhood has achieved high-beta diversification. It has not yet achieved low-beta diversification.

The old Robinhood looked like a small casino with one crypto table. When crypto went quiet, nobody lined up outside. The new Robinhood looks more like a financial entertainment complex: equities downstairs, options upstairs, with side rooms for the World Cup, elections, interest rates, and crypto.

That is a much more resilient business than a single-product casino. But if overall risk appetite falls, users may reduce options, crypto, and event-contract activity at the same time. Several tables can still go cold together.

Robinhood has not left the casino floor. It has started building a bank, a retirement platform, and a wealth-management office behind it.

Whether those additions become the core business—or merely sit behind the trading floor—is what will determine whether Robinhood ultimately deserves a platform multiple.

The World Cup is not the business. The event calendar is.

Prediction markets are the easiest part of this quarter to underestimate—and the easiest to overestimate.

Read more

“不战而屈人之兵，善之善者也。(To subdue the enemy without fighting is the supreme excellence.)” — Sun Tzu

A note on this letter. Our last letter, tracing where profit settles in a value chain, ended at the company that sits above all the doorways. We previewed a letter on Visa; here it is — our first public study of the company, written in an exploratory spirit rather than a judging one. As the series prefers: one idea, tested against history, with the reader left to attach the numbers. Visa reported its June quarter the day this letter went to press; a postscript reads the print through the letter’s framework. This letter is analysis, not advice: it makes no recommendation, sets no price target, and expresses no view on where any security should trade. It is based entirely on public information, and BitStone and related accounts may hold positions, long or short, in companies discussed — including Visa. The legal disclaimer at the end is part of this letter; please read it.

A few plain words to know first

The four-party model — a cardholder, the cardholder’s bank (the issuer), the merchant, and the merchant’s bank (the acquirer). Visa is the fifth thing: the network in the middle that issues no cards, lends no money, and holds no deposits.

Interchange — the biggest fee in a card transaction — roughly $1.80 of a $100 U.S. credit purchase. Visa sets it and never keeps a cent; it flows to the issuer, where it funds your cash-back. When politicians attack “swipe fees,” they are mostly attacking money Visa prices but does not collect.

The network fee — Visa’s own slice: a few cents per transaction for the switch, a sliver of volume for the brand and rules, and a premium toll — several times the domestic rate — when money crosses a border.

Client incentives — the rebates Visa pays issuers and big merchants to stay on its rails, reported as negative revenue.

Take rate — what the network nets per dollar carried. For Visa, under three-tenths of one percent of payments volume — thinner than almost any fee in finance, on more volume than any fee in finance.

A2A — account-to-account payments — money moving bank-to-bank over public rails (Pix, UPI, FedNow), skipping cards entirely.

The 20-Second Read

Where this fits. Two letters ago we showed that busy money and held money are different products. Last letter, that profit migrates to whoever owns the moment of decision. Visa owns no float and no doorway — yet it has out-earned the doorways and the mints for decades. This letter asks how, and for how long.

The finding. Visa is neither mint nor doorway. It is the treaty between all of them — a rulebook with a P&L. Its killer move: it lets everyone else make money. It takes the smallest slice of each transaction, keeps risk controlled, and touches nearly a billion transactions a day.

The framework. Value the business with five questions: how many cars; the toll per car; the cost of the booth; how much must be reinvested; and how long the road stays the road. The first four have knowable answers. The fifth is the debate.

The open question. The companies once expected to kill Visa are on its payroll. What remains on the risk list is what cannot be paid off: statutes, states, and — possibly — blockchains and machines.

The posture. This is a first public study. We publish the framework, not our numbers. Forecasts are opinions, and nothing here is a recommendation.

I. The company our letters kept meeting

Every letter in this series has bumped into Visa on its way somewhere else. The Busy Dollar, i.e. USDC, used Visa’s own on-chain analytics to measure stablecoin volume. The Open USD consortium announced in June counts Visa among its founders. And when the last letter mapped a century of doorways, every doorway had one thing in common: it charged the manufacturers for access to the customer. Visa is the anomaly that map could not place. It has no customer relationship at all — no app you open, no balance you hold, no button that says Visa. By the last letter’s logic it should be a mint, paying rent to the doorways. Instead the doorways carry its product everywhere on earth, and its revenue has compounded so steadily that three straight years of 10–11% growth — through inflation, rate shocks, and a banking scare — read as boring.

A year-plus of mapping this industry has left us with an observation we can state plainly: very few companies have ever beaten Visa, and Visa is a profoundly un-product company. It does not make the card (the issuer does), the checkout (the acquirer does), the tap (the phone does), or the credit (the bank does). Its killer move is that it is willing to let everyone else make money: it prices well, controls risk well, keeps almost none of the risk itself, and simply earns.

What Visa makes is the agreement. It is best understood not as a technology company but as a private treaty with a P&L: the rulebook under which some fifteen thousand institutions that do not trust each other exchange seventeen trillion dollars a year and almost always get paid. The treaty says a card issued by a bank you have never heard of, in a country you have never visited, will be honored in two seconds at a shop that has never seen it, with known rules for who eats the loss when something goes wrong. Nobody in the chain loves the treaty. Nearly everybody in the chain is richer inside it than outside it. That — not the data centers — is the asset.

II. How Visa makes money, and a framework for valuing it

Read more

Editor’s note
What happens when AI agents stop merely answering questions and begin executing tasks, purchasing services, and participating in the economy on their own? What kind of money should they use, and what payment infrastructure will they require?

In the first episode of the Fintechnize Founder Series, I spoke with MK, co-founder and CEO of BitDance Labs, about his path from Tencent and mobile gaming to a third startup at the intersection of Bitcoin and agentic commerce. We also discussed why he believes the Lightning Network—an infrastructure that has struggled to find a mainstream human use case—may ultimately prove far more useful for machines.

This transcript has been edited for clarity and length. Repetitive phrasing and verbal fillers have been removed, and technical terminology has been standardized without changing the substance of the conversation. Specific figures, product names, and references to external events should be reconfirmed with the guest before publication.

1. From Honor of Kings to Sequoia’s investment to a 2nd Startup

Charlie:

Welcome to the first episode of the Fintechnize Founder Series.

Today’s guest is MK, a highly experienced founder. We will talk about his latest journey in agentic commerce, but let’s begin with your background. What did you do before BitDance Labs, and how did those experiences lead you to this particular direction?

MK:

Thanks, Charlie. I’m very glad to be here.

I’m MK, co-founder and CEO of BitDance Labs. We are building a neo-fintech company based on AI and blockchain, with payment and wealth-management products designed for both humans and agents.

My career can roughly be divided into three chapters.

The first began in 2009, when I graduated from university and joined a large internet company. I worked on social media, social networking, and mobile gaming products with hundreds of millions of daily active users. Looking back, 2009 was right at the tail end of the PC internet era. Once the iPhone 4 arrived in 2010, the industry entered a mobile internet expansion cycle that lasted more than a decade.

The second chapter began during the mobile internet boom. In 2016, I left the large-company environment and founded a gaming platform. It was still a favorable period for mobile startups: traffic was growing, device adoption was accelerating, and the policy environment was relatively supportive. The company eventually raised capital from leading dollar-denominated funds including HSG (formerly Sequoia China), Qiming Venture Partners, ZhenFund, and Morningside.

During the pandemic, a combination of regulation and other market factors forced us to wind down that business. That transition led to the third chapter—my current startup, which began in 2023.

The main catalyst this time was the maturation of large language models and blockchain-native technologies. Those are the three broad stages of my career.

Charlie:

MK makes that sound much simpler than it was. At Tencent, the mobile gaming business you worked on included what later became Honor of Kings, one of the most widely recognized games in China. Your previous startup also raised tens of millions of dollars from some of the country’s top investors.

When you decided to start again, you were drawing not only on the rise of AI and digital assets, but also on your operating experience at Tencent and your earlier startup. How did those experiences help you identify this market?

MK:

When ChatGPT emerged at the end of 2022 and we started looking for a new direction in 2023, there were broadly two choices: build in AI or explore Web3.

At the time, Web3 was beginning to move from a bear market toward the next cycle, so the environment was actually quite attractive. But our team did not yet understand AI deeply enough, so we initially focused on the Bitcoin-native ecosystem.

The most important lesson I carried from both the large-company environment and my previous startup was how to distinguish a good idea from a good business.

A good idea is often something a founder uses to persuade himself. You can create many narratives and identify many local pain points. A good business is more objective.

During the pandemic, I took one or two years to step back and think carefully about what I would choose if I started again. My conclusion was that the next venture had to be a fundamentally good business.

At a minimum, a good business should satisfy two conditions. First, it must be scalable. Second, it must compound. On top of those conditions, the business model does not need to be unprecedented. What matters more is whether you can identify a new “100x factor.”

A 100x factor might come from expanding the potential user base by ten times while making the solution ten times more efficient. Or it might come from a tenfold improvement in efficiency combined with a tenfold reduction in unit cost. Multiply two tenfold changes and you may have a 100x opportunity.

Honor of Kings is a useful example. It was one of the best mobile multiplayer online battle arena games of its time, but the early data showed that relatively few PC League of Legends players migrated directly to it. The real source of growth was the enormous number of people who had never played a MOBA game before.

That was the first tenfold expansion: the user base became much larger.

The second was monetization. Buying skins and other virtual items on a phone was a new kind of experience for many users who had grown up with mobile gaming. Those two factors combined into what I would call a 100x opportunity.

2. Why Agentic Commerce Could Be as Important as the Smartphone

Read more

It acquired stablecoin infrastructure company Bridge and crypto wallet developer platform Privy. At the beginning of 2026, it completed the acquisition of usage-based billing company Metronome. Then came the much larger moves: Stripe and Advent International submitted an offer of more than $53 billion for PayPal, followed only days later by reports that Stripe was in talks to acquire OpenRouter, the AI model aggregation platform reportedly valued at as much as $10 billion.

Viewed separately, these companies seem to belong to entirely different markets.

Metronome is billing infrastructure. Bridge is stablecoin infrastructure. Privy provides wallet technology. PayPal is a consumer payments and identity network. OpenRouter sits inside the AI infrastructure stack.

Put them together, however, and a different picture emerges.

Stripe is not buying five companies. It is buying five different entry points into economic activity.

How does a company calculate revenue? How does a developer select and consume an AI model? How does an agent purchase data or software on behalf of a user? How does a consumer prove identity and authorize a transaction? Through which asset and network is the payment ultimately settled?

Stripe is connecting these previously separate layers, one acquisition at a time.

Stripe’s original ambition was to “increase the GDP of the internet.” But processing a payment after that GDP has already been created may no longer be enough. Stripe now appears determined to move further upstream, into the formation of the transaction itself—and perhaps into defining what a transaction will mean on the agentic internet.

From Metronome to Orb: Payment Companies Are Fighting for the Right to Determine the Price

Stripe’s acquisition of Metronome and Adyen’s acquisition of Orb can be understood as two moves in the same war.

Stripe completed its acquisition of Metronome in January 2026. Less than six months later, Adyen announced that it would acquire Orb for approximately $335 million in cash.

Both targets provide usage-based billing infrastructure: systems that allow companies to charge customers according to actual consumption, API calls, data usage, computational resources, or business outcomes.

Not long ago, this looked like a narrow component within the SaaS finance stack.

Traditional SaaS pricing was relatively straightforward. A company purchased 100 seats, each seat cost a fixed amount per month, and Stripe Billing, Zuora, or another subscription-management platform handled plans, renewals, discounts, and invoices.

AI has broken that simplicity.

The cost of serving an enterprise AI customer may depend on input tokens, output tokens, model selection, inference time, GPU consumption, cache-hit rates, tool calls, and whether the task was successfully completed. The cost structure can change by feature, by customer, by model, and by day.

Billing therefore stops being a system that generates an invoice at the end of the month. It becomes a real-time data-processing layer.

It must ingest hundreds of millions—or billions—of usage events and convert them into billable items according to complex commercial contracts. It must support tiered pricing, minimum commitments, prepaid credits, negotiated rates, overage fees, and customer-specific terms. And it must produce a number that sales, finance, product teams, and the customer all believe.

This is the complexity Metronome and Orb were built to handle.

The basic strategic logic behind both acquisitions is easy to understand. Payments answer the question of how money is collected. Billing answers the more fundamental question of how much should be collected.

Without payments, a business cannot complete a transaction. Without billing, it cannot accurately describe the transaction in the first place.

But Stripe and Adyen are entering this market from different positions.

For Stripe, acquiring Metronome strengthens an already extensive billing product suite. Stripe Billing was originally designed around the subscription economy. As AI companies became one of Stripe’s most important customer segments, however, traditional subscription architecture became less suitable for real-time, high-frequency, multidimensional usage pricing.

Rather than continuing to bolt increasingly complicated metering capabilities onto a system built for subscriptions, Stripe acquired an architecture designed from the usage event upward.

Adyen’s acquisition of Orb carries a somewhat different strategic weight.

As I wrote when the deal was announced, this was a rare external acquisition for a company that has historically been almost stubbornly committed to organic development. Adyen built its core payments platform, acquiring capabilities, risk infrastructure, and global localization network largely in-house. It generally preferred building over buying.

Its willingness to pay $335 million for Orb tells us that usage-based billing is no longer a secondary feature that can wait on a product roadmap. It has become a critical entry point into the AI economy.

Stripe and Adyen are not merely competing for billing revenue. They are competing for what might be called pricing formation rights.

When a payments company sees only the final payment amount, it remains an executor at the end of the transaction chain. When it can also see usage events, pricing rules, customer contracts, and payment outcomes, it begins to understand how a company actually produces revenue.

Control that layer, and it becomes easier to move upward into revenue analytics, contract management, and finance automation—and downward into payments, tax, treasury, and credit.

That is why Metronome and Orb matter more than they initially appear to. They are not simply two similar billing tools. They are forward positions in Stripe and Adyen’s attempt to become the operating system for enterprise business models.

OpenRouter: Why Is an AI Aggregator Suddenly Worth $10 Billion?

Among Stripe’s reported acquisition targets, OpenRouter is probably the easiest to underestimate—and the most important to unpack.

Read more

How value finds its place in a value chain: what “distribution” actually means, tested against a century of middlemen — with Robinhood, Coinbase, and Circle as today’s examples

BitStone · Fintechnize

July 24, 2026 · An Ongoing Conversation on the Stablecoin Landscape

“There is only one boss. The customer.…

Read more

Looking out over the open waters of the Greater Bay Area, we kicked off the third summer gathering of 沙丘路的毛圈狮子. After previous stops, we had arrived in Shenzhen—the frontline of Chinese companies going global, and one of the most active testing grounds for financial innovation—to continue our discussion on the signal and noise surrounding AI, fintech, and agentic commerce.

I moderated both panels that day.

The first was an investor fireside chat with Herbie from Hack VC, focused on where investors are finding genuine opportunities across AI and fintech. The second brought together Zuo from Barker Money, Gene from pr402, and Melody from WIDTH to discuss a more specific question: once agents begin controlling money, how must the infrastructure for payments, yield, compliance, and risk be rebuilt?

The standard vision of agentic commerce is easy to understand.

An agent searches for a product, compares prices, calls the necessary services, executes an agreement, pays, and moves on. The internet finally shifts from a world in which humans click on software to one in which software calls other software.

Yet the number that stayed with me that day was not a trillion-dollar market forecast or the transaction speed of a new protocol.

It was 90%.

Barker Money had built its own dashboard to determine whether the activity described as “agent payments” represented genuine economic demand. According to its observations, more than 90% of the transactions it tracked may have involved project teams manufacturing volume, related addresses moving funds among themselves, or circular transactions designed to create the appearance of a thriving market.

Only a small fraction appeared to represent an agent independently paying for a real service.

The number should not be treated as precise. Different platforms define a “real transaction” differently, and identifying related wallets is an imperfect exercise.

But one week before the event, a study of x402 transactions on Base reached a remarkably similar conclusion. The researchers analyzed 136.7 million settlements and identified 21.2% as fabricated activity. Another 63.78% took place within clusters of related addresses.

The authors offered a line that the entire industry should remember:

Transaction count may be measuring manufacturability rather than adoption.

That captures both the most awkward and the most valuable reality of agentic commerce today.

On one hand, autonomous transactions between machines are almost inevitable. On the other, the industry is already writing three years of imagined demand into today’s transaction charts.

Agentic Commerce Is Repeating an Old AI and Crypto Story

During our conversation, Herbie said that investors need to be “de-mystified.”

Investors do not inherently understand technology better than founders. Most investors are generalists. They observe where capital is flowing, which markets are growing, and how industry structures are changing. They then try to identify the companies that can scale within a constantly shifting narrative.

This helps explain why today’s AI market produces such a strong flashback for anyone who lived through the 2021 crypto cycle.

Back then, attaching “ZK,” “cross-chain,” or “GameFi” to a project could attract immediate attention. Today, almost any product can connect to a foundation model, add an agent workflow, place a diagram about “redefining human work” in its pitch deck, and present itself as an entirely new species.

The problem is that fundraising may be driven by narrative, but revenue is ultimately driven by budgets.

Herbie’s view is that the easiest money to earn in AI over the past two years has come from three categories of relatively inflexible spending:

Read more

“Money is like an arm or a leg — use it or lose it.” — attributed to Henry Ford

A note on this letter. The latest in a series — after To Bank or Not to Bank and Everybody’s Dollar — and the first we write together. We began with a hypothesis we both liked and tested it the way you test an ide…

Read more

If you’ve been following the stablecoin industry over the past week, you’ve probably seen some version of that headline. After receiving final approval from the Office of the Comptroller of the Currency (OCC), Circle officially launched Circle National Trust Bank, becoming one of the few crypto-native companies to obtain a federally chartered trust bank in the United States. It’s an important milestone—but also one of the most misunderstood stories in fintech this year.

If your immediate reaction was that Circle has finally become the crypto equivalent of JPMorgan, almost every part of that conclusion is wrong. Circle’s National Trust Bank cannot accept retail deposits, make commercial loans, or offer FDIC-insured checking accounts. It isn’t a digital version of a traditional commercial bank, nor is it trying to compete with Bank of America or Wells Fargo. What the charter actually provides is a federally regulated framework for digital asset custody, institutional trust services, and the management of the reserves backing USDC. That is a meaningful expansion of Circle’s capabilities, but it is a very different business from banking as most people understand it.

The confusion doesn’t stop with Circle. Kraken has been called a “bank” for years through Wyoming’s Special Purpose Depository Institution (SPDI) framework, yet it looks nothing like Circle. Robinhood still doesn’t hold a banking charter of its own, but offers products most consumers would simply describe as banking. Coinbase is pursuing a National Trust strategy while simultaneously building wallets, custody, developer platforms, and Base. Stripe acquired Bridge—not because it wanted to become a bank, but because it wanted to make stablecoins disappear behind a single API. Even obtaining a banking charter doesn’t necessarily mean direct access to the Federal Reserve’s balance sheet. Some chartered institutions still rely on correspondent banks for settlement, while others spend years pursuing something arguably more valuable: a Federal Reserve master account.

Taken together, these developments raise an obvious question. Why does the United States suddenly seem to have so many different kinds of banks?

I think the answer is almost the opposite of what most people assume. The American financial system hasn’t become more complicated because regulators invented new types of banks. It has become more complicated because the very idea of a bank is being dismantled. Circle’s National Trust charter is simply the latest—and perhaps the clearest—signal that this transformation is accelerating.

America Isn’t Creating More Banks. It’s Unbundling Banking.

If I had to summarize the past two decades of American fintech in a single sentence, it would be this: the United States isn’t creating more banks; it’s quietly dismantling the traditional idea of what a bank is.

For most of modern financial history, banking was treated as an indivisible institution. A bank accepted deposits, processed payments, made loans, safeguarded assets, settled transactions, and ultimately connected to the Federal Reserve. Those functions naturally belonged together because they had always existed inside the same organization. Over time, we stopped thinking about those capabilities individually and simply called the entire bundle “a bank.”

The internet challenged that assumption. Payments no longer had to belong to banks. Custody no longer had to belong to banks. Customer relationships no longer had to belong to banks. Eventually, even dollars themselves no longer had to live inside banks.

Stripe demonstrated that payments could become software. Robinhood proved that customer distribution could be separated from balance sheets. Coinbase built an entirely new financial ecosystem around digital assets without replicating a commercial bank. Circle transformed dollars into programmable internet-native assets. Kraken focused on controlling the bridge between traditional finance and crypto instead of becoming another lender. Viewed independently, these look like separate fintech success stories. Viewed together, they reveal something much more fundamental: fintech didn’t invent an entirely new financial system. It gradually unbundled the traditional bank.

Once you look at banking through that lens, the seemingly chaotic collection of American bank charters begins to make much more sense. Money Transmitter Licenses, State Trust Companies, Wyoming SPDIs, National Trust Banks, Industrial Loan Companies, commercial bank charters, and Federal Reserve master accounts are often presented as if they form a hierarchy—as though every fintech company starts with a money transmitter license and eventually graduates into becoming a “real bank.” They don’t. These aren’t increasingly powerful versions of the same institution. They’re different regulatory frameworks built around different financial capabilities.

A Money Transmitter License answers one question: do you simply want to move money? A State Trust Charter answers another: do you want to custody assets or act as a fiduciary? Wyoming’s SPDI framework was designed for institutions that needed payment, custody, and deposit functionality without participating in the traditional fractional-reserve lending model. A National Trust charter allows firms like Circle to operate nationwide as federally supervised trust institutions, providing custody and reserve management without becoming commercial lenders. A full commercial bank charter remains the appropriate framework only if your business model depends on gathering deposits, building a balance sheet, making loans, and earning net interest margins.

At first glance, these all appear to be different versions of the same regulatory license. They’re not. They’re answers to a much more fundamental strategic question: which part of banking are you actually trying to own?

That, more than any individual charter, is what has changed over the past decade. People often describe American financial regulation as becoming more permissive. I think the opposite is closer to the truth. Regulators aren’t lowering the bar for becoming banks; they’re acknowledging that the future financial system won’t be built by institutions that all look like banks. Instead, different companies will specialize in different financial capabilities, each operating under a regulatory framework designed for the function it performs.

In that sense, banking regulation is becoming increasingly modular. The regulatory perimeter is beginning to follow capabilities rather than institutions. If you want to custody assets, there’s a trust framework. If you want to gather deposits and make loans, there’s a banking framework. If you want direct access to the core settlement infrastructure, that’s an entirely different conversation involving the Federal Reserve. The architecture increasingly resembles cloud computing more than twentieth-century banking: different layers perform different functions, some provide infrastructure, others provide applications or interfaces, and together they form a complete system without requiring any single company to own all of it.

That’s why the most valuable fintech companies today look less and less like traditional banks. Robinhood isn’t trying to become JPMorgan. Stripe isn’t trying to become Citi. Circle isn’t trying to become Bank of America. Each is trying to control a layer that once lived inside those institutions. Circle’s National Trust charter therefore isn’t simply another regulatory approval. It’s evidence that banking itself is no longer being organized around institutions. It’s being reorganized around capabilities.

And once you see that, the obvious next question isn’t about regulation at all. It’s about strategy. Why did Robinhood conclude it didn’t need a banking charter? Why did Kraken pursue Wyoming’s SPDI? Why did Circle choose a National Trust instead of a commercial bank? Why is Stripe moving deeper into stablecoin infrastructure without trying to become a bank itself?

The answers have surprisingly little to do with regulation.

They have everything to do with what each company ultimately wants to own.

Read more

Stripe, Visa, Mastercard, Coinbase, BlackRock, and more than 140 companies are backing a new stablecoin initiative called Open Standard, which plans to issue a U.S. dollar-pegged stablecoin called Open USD, or OUSD. According to Reuters, Open USD is designed to allow businesses to mint and redeem the token at no cost and without volume limits, while sharing reserve earnings with partners after a management fee.

The obvious market reaction was to frame this as a direct attack on Circle and USDC. That is understandable. Circle’s stock sold off sharply after the news, and the market naturally read the announcement as a competitive threat to one of the largest regulated dollar stablecoins. The Wall Street Journal reported that Open USD had backing from companies including BlackRock, Google, Coinbase, Visa, Stripe, and Mastercard, and noted that Circle shares dropped in response.

But I think that framing is still too narrow.

If we only see OpenUSD as “another stablecoin,” or as “a new competitor to USDC,” we are missing the more important point. OpenUSD’s real significance is not that someone is issuing one more digital dollar. It is that the most profitable and sensitive part of the stablecoin business is now being dragged into the open.

The question is simple: who should get the yield?

For the past few years, when people talked about stablecoins, they mostly talked about payment efficiency, on-chain settlement, cross-border transfers, exchange liquidity, or the larger geopolitical story of dollar expansion on blockchains.

Those are all use cases. But they are not the core economic engine.

The core economic engine is much more basic. A user gives dollars to a stablecoin issuer. The issuer gives the user stablecoins. The issuer then holds those dollars in cash, short-term Treasuries, money market funds, or bank deposits. The reserve assets generate interest. That interest becomes the issuer’s most important revenue source.

In a zero-rate world, this was easy to overlook. Nobody cared much about the yield on dollar reserves because there was not much yield to capture. Stablecoin stories were built around trading volume, compliance, crypto adoption, and developer ecosystems.

Then U.S. rates went up, and reserve income became a very good business.

Circle is the cleanest example. The larger USDC becomes, the more reserves Circle holds. The higher rates are, the more powerful the revenue engine becomes. Circle reported $733 million of reserve income in the fourth quarter of 2025, up 69% year over year, driven primarily by growth in average USDC circulation.

That is why OpenUSD matters.

The market is not mainly worried that USDC will disappear tomorrow. USDC’s liquidity, integrations, and institutional trust will not vanish overnight. What the market is really repricing is Circle’s margin structure.

If more platforms that control merchants, enterprises, developers, wallets, and payment entry points start saying, “Reserve yield should not belong only to the issuer; whoever brings transactions, users, capital balances, and distribution should participate in the economics,” then the challenge is no longer to USDC as a product.

The challenge is to the old stablecoin business model: I issue the coin, therefore I keep the yield.

This is also why OpenUSD is very different from something like PayPal’s PYUSD.

PYUSD is important. But structurally, it is a large platform issuing its own stablecoin. It is one more coin from one more powerful company.

OpenUSD is making a different argument. It is saying: stop letting every company issue its own coin and then go hunting for use cases. First gather the merchants, payment networks, exchanges, asset managers, developer platforms, and fintech infrastructure providers. Then make the stablecoin a shared value layer across that alliance.

That is not product versus product.

That is product versus ecosystem.

A product wins because it has better features, lower cost, stronger compliance, or a better user experience. An ecosystem wins because it puts enough players with different incentives onto the same economic map.

That is the clever part of OpenUSD.

It is not Stripe launching Stripe Coin. It is not Visa launching Visa Coin. The advantage of launching your own coin is control. The disadvantage is that everyone else immediately asks: why should I help you grow your balance sheet?

If Stripe launched its own stablecoin, Visa, Mastercard, Coinbase, BlackRock, Shopify, and everyone else would naturally be cautious. If the coin succeeds, are they partners — or are they just unpaid distribution for Stripe’s economics?

A consortium-style stablecoin at least solves the narrative problem.

It says: this is not one company’s coin. This is a neutral standard with shared governance, shared distribution, and shared economics. Instead of promoting someone else’s stablecoin, let’s coordinate around one and split the economics.

That is why I think OpenUSD looks more like Visa than Libra.

Read more

The field is still early. There is no real consensus yet. But the questions are getting sharper.

Only a few months ago, most discussions around Agentic Commerce still circled around the obvious consumer scenarios: Will AI agents shop for us? Will they compare products? Will they click the buy button? Will merchants need to optimize for agents instead of humans?

But in more serious conversations, especially with builders, fintech operators, and payment people, the discussion quickly moved one layer deeper. Where will real transaction volume come from? How will agent identity and authorization be established? Who owns the wallet, the account, the routing, the risk engine, the compliance layer, the settlement rail? How do Google, Coinbase, Circle, Stripe, Visa, and Mastercard differ in their approaches?

That shift matters. It means the debate is no longer about whether machines can technically initiate a payment. The real question is whether a machine has the right to move money on someone’s behalf, and who is responsible when that right is exercised.

Yet because this Agentic Commerce narrative is so natural, another equally important direction has been somewhat hidden in plain sight.

Most people still approach Agentic Commerce through a B2C lens. An AI agent helps a consumer discover products, compare prices, place an order, and make a payment. Merchants need better product data, pricing feeds, inventory accuracy, checkout readiness, and risk controls. Payment networks need to know whether an agent-initiated transaction was truly authorized by the user. Issuers and networks need new ways to think about fraud, chargebacks, disputes, and liability.

All of that is important. It is also the language the payments industry already understands.

But money does not start moving only when someone clicks a payment button.

In the broader economic system, money starts moving much earlier. It hides inside budgets, purchase requests, contract terms, vendor onboarding, invoice discrepancies, cash forecasts, tax treatment, and audit trails. In consumer Agentic Payments, the question is whether an AI agent can buy something for me. In corporate Agentic Finance, the question is whether an organization can make capital, contracts, budgets, procurement, cash, and risk judgment move faster and more accurately.

That is why the recent internal finance workflow disclosures from OpenAI and Google are worth looking at separately.

In May, OpenAI and PwC announced a collaboration to build AI agents around the core rhythms of the CFO office, including planning, forecasting, reporting, procurement, payments, treasury, tax, and accounting close.

The interesting part is not that PwC wants to advise CFOs on AI. That would be unsurprising. The interesting part is that the work starts inside OpenAI’s own finance organization, with a procurement agent, before extending the lessons into broader enterprise finance workflows.

OpenAI also said it has already been using ChatGPT and Codex internally to support investor relations, treasury, tax, reporting, corporate development, and contract review.

Google’s example feels more like a stress test inside a large and complex organization. According to recent reporting, Google’s finance team uses an AI agent to automatically compare vendor invoices against contract terms, increasing invoice review throughput by five times and potentially reducing $200 million in annual overpayments.

In treasury, Google is also using AI to manage cash across thousands of bank accounts, recommend investments for excess cash based on risk tolerance, and, after human review, allow another agent to execute transactions.

The most revealing detail is not just the efficiency gain. It is that Google’s finance team has stayed roughly the same size while producing more. People who used to work on invoice validation have moved toward higher-level audit and AI review work.

This is not a chatbot bolted onto the CFO Suite. It is not RPA with a large language model wrapper.

Over the past several years, the SaaS market has already gone through a full wave of CFO Suite innovation. Accounts payable automation, expense management, procurement, FP&A, close management, treasury dashboards — every category produced new companies, and almost every pitch deck sounded similar: reduce manual work, automate workflows, increase efficiency, shorten the close cycle, cut costs.

None of that was wrong. It was just incomplete.

Read more

The first event was at BAI Capital’s office. We talked about Kraken’s acquisition of Reap, the Hong Kong-based stablecoin payments and card infrastructure company. I initially thought it would be a relatively small discussion around an industry M&A case. Why would a global crypto exchange buy a Hong Kong payments infrastructure company? Was this simply about Kraken adding payment capabilities, or was it a sign that stablecoins are moving from being assets inside crypto trading venues to becoming settlement infrastructure for real commercial activity?

I was wrong about the “small” part. Once the event went out, sign-ups quickly crossed one hundred people. The room was not just crypto people. We had people from payments, cross-border commerce, compliance, enterprise software, fintech, and investment. What people cared about was not just Kraken and Reap as one deal. They cared about the bigger question behind it: if stablecoin infrastructure companies are now being acquired by exchanges, payment companies, and financial infrastructure players, are we about to redraw the boundary between traditional payments and crypto payments again?

That deal was worth discussing. Kraken’s parent company Payward agreed to acquire Reap in a transaction worth up to $600 million, with the acquisition expected to help Kraken accelerate its card issuance and stablecoin payment infrastructure, especially in Asia-Pacific.

The second event, focused on Agentic Commerce and Agentic Payments, gave me a similar feeling.

AI is obviously the hottest topic this year. But the hotter a topic gets, the easier it becomes to talk about it in a way that says almost nothing. Agentic Commerce is a perfect example. It is now a field where large platforms and vertical infrastructure companies are all announcing, experimenting, and positioning. Yet the actual product form is still far from consensus.

That was exactly why I wanted to host this event. The point was not to tell another grand story about the AI future. The point was to demystify it.

If agents really begin to find products, call services, place orders, and complete payments on behalf of users, what exactly are the large players building? What positions are payment companies, cloud platforms, search companies, and protocol organizations trying to occupy? More importantly, where can startups enter?

This event also attracted more than one hundred sign-ups. The room was unusually mixed: AI people, payment people, crypto people, cross-border e-commerce operators, security founders, DePIN builders, hardware people, and investors. Over the past few years, these groups usually appeared at different conferences, used different language, and cared about different KPIs. AI people talked about models and agents. Payment people talked about acquiring, fraud, and compliance. Crypto people talked about stablecoins and on-chain settlement. E-commerce people talked about traffic, conversion, and supply chain.

Agentic Commerce pulled them to the same table.

The reason is simple. Once agents stop merely answering questions and begin discovering products, comparing prices, calling services, placing orders, and making payments, several previously separate layers are suddenly wired together again. AI is no longer just the front-end interaction layer. Payments are no longer merely the last step. Merchant systems are no longer just back-office tooling. Crypto is no longer just about asset trading. It may become a settlement method among software, machines, APIs, and agents.

After sitting through the whole event, my biggest takeaway was this: Agentic Commerce felt less abstract.

That does not mean the market is mature. Quite the opposite. It is still early, messy, and full of protocols and demos that will probably disappear. But it is no longer just a vague statement that “AI will change commerce.” That statement is impossible to disagree with, but also difficult to act on.

Agentic Commerce is starting to land in specific questions: merchant product data, payment interfaces, agent authorization, security and fraud, stablecoin settlement, machine wallets, and growth for mid-tail and long-tail merchants.

Once a concept begins to make different parts of the value chain reposition themselves, it deserves attention.

Read more

Just two months earlier, Adyen had announced its acquisition of Talon.One, the German loyalty and incentives platform, for €750 million, roughly $867 million.

Put together, Adyen has now spent more than $1.2 billion on two acquisitions in just a few months.

For most companies of Adyen’s size, that would be notable. For Adyen, it is almost shocking. This is a company that had spent nearly twenty years building almost everything internally, with a strong cultural bias against M&A.

I had an immediate personal reaction.

Adyen is my alma mater. Orb’s founder, Alvaro Morales, is a friend from South Park Commons, another community I am proud to have been part of. When familiar people and familiar institutions suddenly meet inside the same strategic story, it is hard not to feel emotionally invested. So I wrote a quick note on LinkedIn. It ended up getting much more traction than I expected, including likes and reactions from several CEOs and senior executives close to the companies involved.

But the more I thought about it, the more I felt the real story was not that Adyen finally broke its twenty-year habit and started acquiring companies.

If you only read the headlines, the interpretation is easy: a payments company bought a billing company and a loyalty company. If you want to force it into the current Agentic Commerce narrative, you could say Adyen is buying its way upstream, trying to get closer to commerce data before AI agents intermediate the shopping experience.

That is not wrong. It is just too shallow.

Talon.One and Orb are not the same kind of “upstream.” Their relationship to payments is not simply that one is closer to the money and the other is further away from it. What they have in common is more interesting: both help answer a question traditional payment companies have historically struggled to answer.

Why did the payment happen?

Payments companies traditionally see the outcome.

A $100 transaction. An authorization. A refusal. A chargeback. A settlement. A reconciliation entry. For a platform like Adyen, those outcomes are tremendously valuable. They can be used to optimize routing, improve authorization rates, reduce fraud, support local payment methods, manage cross-border settlement, and help large merchants unify payment experiences across countries and channels.

But there is a natural blind spot here. A payments company knows whether money moved, how it moved, and where it failed. It does not always know why the money moved.

Why did the user buy this product? Why did this promotion work? Why did this SKU convert in one channel but not another? Why did the same discount protect margin in one segment and destroy it in another? Why should this enterprise customer pay $100,000 this month instead of $50,000? Why did usage grow without revenue growing proportionally?

Historically, those were not core payment questions. The job of payments was to move money safely, reliably, cheaply, and successfully.

But in an Agentic Commerce world, intent and causality become commercial infrastructure.

Talon.One sits before the payment.

Read more

On one side, Stripe, Visa, Mastercard, and potentially Coinbase are reportedly moving toward a new stablecoin platform. Stripe has already acquired Bridge. Mastercard has agreed to acquire BVNK. Coinbase remains the most important distribution partner for USDC, owns Base, and has been pushing further into stablecoin payments and business accounts.

On the other side, JPMorgan Chase, Bank of America, Citi, and Wells Fargo are reportedly planning a nationwide tokenized deposit network through The Clearing House, targeted for the first half of 2027. The idea is to represent commercial bank deposits in tokenized form for 24/7 settlement, corporate treasury, liquidity management, and cross-border payments.

At first glance, this looks like the long-awaited split: stablecoins versus tokenized deposits; crypto money versus bank money; Silicon Valley payments versus Wall Street balance sheets.

That framing is tempting. It is also too simple.

To borrow from Chinese history, it looks like a Chu-Han divide. In Chinese culture, “Chu River, Han Border” refers to a hard line between two opposing camps, originally from the struggle between Xiang Yu of Chu and Liu Bang of Han after the fall of the Qin dynasty. The phrase is still used today in Chinese chess and everyday speech to mean a clear boundary between two rival sides.

But the stablecoin market does not look like Chu versus Han.

It looks more like the Warring States period.

For readers less familiar with Chinese history, the Warring States period was the final stretch of the Zhou dynasty, when seven major states — Qin, Chu, Qi, Yan, Han, Zhao, and Wei — competed through war, diplomacy, reforms, and shifting alliances before Qin eventually unified China in 221 BC. One of the defining strategic concepts of that era was hezong lianheng, often translated as “vertical and horizontal alliances.” The weaker states sometimes tried to form a north-south coalition to contain Qin; other times, states cut east-west deals with Qin or each other to survive, expand, or buy time.

That is a much better analogy for what is happening in onchain dollars.

The surface-level story is two camps. The deeper story is seven powers, each with its own incentives, each forming alliances, and each quietly hedging against its allies.

The trigger is regulation, but not because the CLARITY Act “created” this market. Stablecoins already had product-market fit in crypto, cross-border dollars, exchange settlement, and increasingly payments. What regulation is doing now is forcing the industry to turn vague narratives into legal and economic boundaries.

For years, stablecoins could hide inside a broad story: faster dollars, cheaper cross-border payments, more open financial networks, better liquidity for crypto markets. But once stablecoins move into mainstream finance, the questions become painfully specific.

Can a stablecoin balance generate yield? If the issuer cannot pay interest, can a wallet, exchange, merchant, or membership program offer rewards? If a stablecoin cannot look like an interest-bearing bank account, can transaction-based rewards become a loophole?

That is where the Senate Banking Committee’s version of the CLARITY Act matters. The bill restricts passive, deposit-like interest or yield on payment stablecoins, while still allowing certain activity-based or transaction-based rewards. It also directs regulators including the SEC, CFTC, and Treasury to help clarify the boundary between prohibited interest-like returns and permitted rewards.

That line sounds technical. It is not. It decides where the profit pool goes.

If a stablecoin is only a payment instrument, banks can live with it. They may even participate in settlement themselves. But if stablecoins, through rewards, rebates, merchant incentives, membership benefits, or agentic finance accounts, become high-yield checking accounts by another name, banks are no longer competing over payment fees. They are defending the deposit franchise.

For banks, deposits are not just a product. Deposits are the mother of all products.

Credit cards, mortgages, loans, wealth management, corporate treasury, foreign exchange, cash management — all of them grow out of the account relationship. If that relationship migrates into Coinbase, Stripe, a merchant wallet, or an AI-agent-controlled financial account, banks lose more than basis points in payment margin. They lose the customer’s financial operating system.

That is why big banks have to respond.

But the way they respond is revealing. They are not simply saying, “Let’s launch a USDC competitor.” They are choosing tokenized deposits.

That choice is political language.

“Stablecoin” is crypto’s language. “Tokenized deposit” is banking’s language. The former says open dollars. The latter says regulated bank money. The former emphasizes programmability and global circulation. The latter emphasizes balance-sheet treatment, deposit status, and regulatory acceptability.

Banks do not misunderstand stablecoins. They are translating the stablecoin problem into a language they can control.

JPMorgan’s position here is especially delicate. JPMorgan did not discover onchain dollars last week. It has been building through JPM Coin and Kinexys for years. But if JPMorgan alone tried to define the future standard for tokenized bank deposits, Bank of America, Citi, and Wells Fargo would have little reason to enthusiastically support it. Large U.S. banks have shared interests, but shared interests do not mean they want one bank to become the de facto clearing standard for everyone else.

The Federal Reserve also has no reason to encourage a fragmented return to a banking world of private fiefdoms. The tone from regulators and central banks has increasingly favored interoperability, common standards, and technology-neutral treatment of tokenized assets rather than each institution building isolated rails. U.S. banking regulators, for example, have recently emphasized technology-neutral treatment for tokenized securities capital rules.

So putting this effort inside The Clearing House is politically more important than it is technically elegant.

The Clearing House is already a bank-owned clearing infrastructure. Putting a tokenized deposit network there turns “JPMorgan’s first-mover advantage” into “the banking industry’s common defense.” It may not be the technically optimal architecture. It may be the politically optimal one.

That is how alliances work.

In the Warring States period, vertical alliances were not formed because the six states loved one another. They were formed because nobody wanted to be the first one swallowed by Qin.

So who is Qin?

Read more