Things to know 👀
Circle got the final green light from the US Office of the Comptroller of the Currency to open First National Digital Currency Bank, which will operate as Circle National Trust. It's a bank in the narrow sense: no deposits, no loans, no FDIC insurance. What it does is put the infrastructure behind USDC, the largest regulated stablecoin, under direct federal supervision, starting with custody of digital assets for Circle and its affiliates. Circle applied in June 2025, got conditional approval in December, and now has the full charter, nicely before the GENIUS Act rules come into force. For a company that made its name as the credible onshore choice for stablecoins, this is the biggest notch yet.
🧠 Read the fine print though: management of the USDC Reserve itself is listed as a "future capability." The part of this charter everyone cares about, federal oversight of the actual dollars and Treasuries backing USDC, comes later.
🧠 Charters have stopped being rare. The OCC approved Circle, Ripple, Paxos, BitGo and Fidelity in December, then Bridge, Crypto.com, and even Sony followed. The long drought of new de novo charters made these look impossible to get. They weren't. Nobody was granting them.
🧠 Keeping a charter is the work, times ten, forever. Ask Anchorage, who got a trust charter in 2021 and an AML consent order by 2022. These approvals require conforming stablecoin activities to the GENIUS Act and every future implementing rule. Fall short, and the remedy runs from exam failures to consent orders to, at the extreme, a cease-and-desist.
🧠 And "conforming" now includes things Circle has historically resisted. Treasury's proposed rules require issuers to freeze, block and burn tokens on lawful orders. Circle took heavy criticism after the $285m Drift hack for waiting on a legal order while $232m in USDC moved across chains. This week, an ICIJ investigation revealed Wisconsin prosecutors filed a criminal complaint against Circle in a stolen-funds case. Circle became a federally approved bank the same week a state prosecutor accused it of obstructing a warrant.
🧠 What I wonder about is the Fed master account. That's what would get USDC much closer to honoring $1 for $1, because a dollar ultimately needs to become a dollar. It's a separate fight, and trust banks land in the Fed's Tier 3 review, the slowest and most scrutinized lane. The hope is Kraken, who snuck in with a limited "skinny" account in March before the Fed finished writing the framework, showing non-bank payment companies there may still be a path.
🧠 Revolut learned this under MiCA, delisting USDT to keep its European license. Full licensing changes what you have to say no to. Circle is about to find out what it has to say yes to.
JPMorgan, Bank of America, Wells Fargo and PNC held talks to buy Fiserv's STAR debit network (WSJ). No agreement has been reached; the talks remain preliminary, and the WSJ's sources say several of the banks have already decided they're unlikely to proceed. STAR is one of the networks that process payments when someone pays with a debit card, uses an ATM, or buys online. Per Fiserv's website, it connects more than 115 million debit cardholders and over 2,800 banks and credit unions. (Fiserv owns a second debit network, Accel; the reporting leaves open whether one or both are for sale.)
Why would a bank want to own a debit network? Because of a 15-year-old price cap. Every time you pay with a debit card, the shop pays a fee to your bank. Since 2010, the Durbin Amendment has capped that fee for big banks (over $10bn in assets) at 21 cents plus 0.05% of the sale. But the cap only applies when the fee is set by an outside card network like Visa or Mastercard. If the bank owns the network itself, the argument goes, there is no outside fee to cap. The bank can charge shops what it likes.
🧠 It worked for Capital One. Capital One's $35.3bn acquisition of Discover (announced February 2024, closed May 2025) came with the Discover Network, and the debit prize is that “capped transactions” average 23 cents of interchange, exempt ones average 51, and Discover-network debit averages 83 cents per transaction. That’s more revenue for Capital One, which understandably is moving all of its debit cards to that network now.
🧠Would this work? Durbin also requires every debit card to carry two unaffiliated networks, and the merchant picks the route. STAR has volume today because merchants choose it to save money. Reprice it above Visa's capped Interlink rate, and the traffic leaves the same afternoon. So banks might not get their revenue win.
🧠STAR is a back-of-card network with no default volume at all. Durbin also requires every debit card to carry two unaffiliated networks, and the merchant picks the route among them. That applies to Capital One's Discover cards too, which is why the exemption has a ceiling for everyone. The difference is default volume. Discover is the brand on the front of the card, so tapped, tokenized and e-commerce transactions ride it unless the merchant actively routes eligible ones away.
🧠Could it even support Apple Pay? Apple Pay tokenization runs on Visa and Mastercard plumbing inside EMVCo, where issuers aren't members, so a bank that owned STAR still couldn't provision STAR tokens into Apple Pay. Discover sits inside EMVCo with its own token service. That is a big part of what Capital One bought and precisely what STAR lacks.
🧠 Is this a game of commercial leverage or a test balloon? Possibly both, and neither requires a deal to close. For Fiserv (down roughly 70% from its 2025 highs, activists circling, a turnaround to fund), a WSJ story full of megabank buyers is a for-sale sign with excellent placement. For the banks, a credible threat of an owned debit network is negotiating in every incentive renewal with Visa and Mastercard. The headline is useful even if the deal never closes.
🧠 The regulatory timing is good. The political timing is terrible. The agencies that approved Capital One–Discover in 2025 are as deal-friendly as banks have seen in years. But as Tom Noyes has pointed out on X, Trump has been publicly attacking swipe fees, Senators Durbin and Marshall are still pushing the Credit Card Competition Act, and the banks need goodwill in Washington for other fights (like CLARITY). Acquiring the largest independent debit rail specifically to route around a consumer-protection law isn’t it.
🧠 How would merchants react? Badly, and with options. It depends how big they are. Noyes has been asking them directly. The biggest merchants can defend themselves: every debit card carries a second network, so if a bank-owned STAR raised prices, a Walmart-sized retailer simply sends its transactions down the cheaper rail and STAR's volume dries up.
🧠 Most small merchants don’t have the right tech. Most businesses route payments however their processor routes them. A bank pricing STAR at 125bps would have enough margin to pay processors to keep STAR switched on. That's buying the routing decision outright.
🧠 The biggest users of STAR would keep the lobbyists busy if this went through. STAR's cards are issued by 2,800 mostly small, Durbin-exempt banks. Selling their routing infrastructure to their four largest competitors is something that would likely make them angry at FiServ (a major supplier), and the community bank lobby is a strong voice in D.C.
3. Klarna wants to become a bank in the USA
Klarna filed applications on Monday with the Utah Department of Financial Institutions and the FDIC to create Klarna Bank USA, a Utah-chartered industrial bank. Utah is where America keeps its fintech charters, and Klarna is treating it that way: the proposed bank gets its own board, its own governance, FDIC insurance, and a CEO hired straight from the Utah banking bench. Gary Harding has run Milestone Bank and Prime Alliance Bank and done risk at TAB Bank and Rakuten Bank America, which is to say he has spent a career doing exactly this, for exactly this kind of company, in exactly this state.
Klarna has been a licensed bank in Europe since 2017, with a Swedish license that covers 23 other countries, and it funds much of its European lending with customer deposits. In the US it has always rented: partner banks, mainly Utah's WebBank, sit between Klarna and its 30 million American users. The US business has extended $91.3bn of credit to Americans since 2019, a debit card that passed 5 million users this year, and a savings account that just launched (held at WebBank, for now).
🧠 Now is the perfect time to apply. The current administration is handing out charters like conference lanyards. American Banker counts two dozen fintechs, crypto firms, lenders and payment companies that have applied for or conditionally received bank charters in the first half of 2026
🧠 The charter gets you better economics as a lender. Today, every Klarna loan in America involves paying WebBank to originate it and paying capital markets to fund it. With its own bank, Klarna keeps the origination economics, funds loans with FDIC-insured deposits (the cheapest, stickiest money there is), and launches products without negotiating with a sponsor.
🧠 Klarna's economics in Europe are stronger than in the US. In Europe, billions in Swedish and German savings deposits fund the loan book at deposit rates. In the US, Klarna pays partners and borrows at market rates, so every loan carries a thinner margin. The charter is how you copy the European model across the Atlantic.
🧠 The USA is becoming its biggest market. Thirty million of Klarna's 119 million users are American. The credit book, the card growth and the savings launch are all US stories now, and when Klarna finally listed, it chose the New York Stock Exchange over Stockholm.
🧠 This helps Klarna move towards its "super app" vision. The app already spans shopping search, price comparison, installment credit, a debit card, cashback, savings, an advertising business and even a US mobile phone plan. The missing piece helps them continue to grind towards profitability on all of that.
🧠 It's a very different kind of charter to the one the big high street banks have. It looks more like how Block did it. Square Financial Services, Block's Utah industrial bank, opened in 2021 and does the lending and deposits while the parent company gets on with Cash App and everything else.
🧠 This benefits tech businesses. The bank itself is supervised by the FDIC and Utah, but the parent escapes the Bank Holding Company Act, so the Federal Reserve never supervises the group. It can stay a shopping app, an ads business and an AI company that happens to own a bank, which is precisely why bank lobbyists have fought this charter for years.
🧠 But don’t expect them to offer classic checking accounts. Industrial banks generally can't offer classic checking accounts (that limitation is part of the legal carve-out), so expect Klarna Bank USA to look like a savings-and-lending bank first, that serves the parent’s super app and card ambitions.
Sony Bank said on July 7 it has received preliminary conditional approval from the US Office of the Comptroller of the Currency to establish a national trust bank in the United States. The planned subsidiary, Connectia Trust, would issue and manage US dollar stablecoins, hold the reserves behind them in cash and Treasuries, and custody digital assets. It would take no deposits, make no loans, and carry no FDIC insurance. The target opening is 2027, subject to approvals on both sides of the Pacific, and Sony was careful to say the preliminary approval guarantees neither the timing nor the stablecoin itself.
The application was filed last October, and it did not go unopposed. The big-bank lobby and community groups wrote to the OCC arguing a trust charter shouldn't stretch this far, and that letting a commerce company own a bank crosses a line US regulation has held for decades. Nine months later, the OCC granted preliminary approval anyway.
🧠 Sony is the first company from outside finance to get one of these. The OCC's stablecoin-era approvals so far have gone to crypto and finance natives: Circle, Ripple, Paxos, BitGo and Fidelity Digital Assets (Anchorage got a trust charter back in 2021). Sony makes PlayStations. The Bank Policy Institute objected on the basis of the separation of banking and commerce.
🧠 Which gives us a window into how the OCC will interpret the GENIUS Act. Chloé Dolsenhe pointed out on LinkedIn that in the letter confirmed some positions like “Payment stablecoins are NOT deposits under the Federal Deposit Insurance Act. They are not covered by FDIC insurance.
🧠 This isn’t a one-time approval; it’s oversight for life. As Chloé writes “The OCC requires Sony’s national trust bank to conform its stablecoin activities to the GENIUS Act and all future implementing regulations.” Failure to do so would result in a cease-and-desist order.
🧠 We saw with MiCA that once rules pass, they start to bite. Revolut recently delisted USDT, the world's largest stablecoin to comply with MiCA. Its choice was to continue to comply in its core, home markets without Tether, or close its crypto operations in Europe.
🧠 Sony's crypto ambitions are fascinating. It already runs Soneium, its own Ethereum layer-2 blockchain, launched in January 2025. It owns a Japanese crypto exchange, S.BLOX. Sony Bank has trialed a yen stablecoin at home. Now add a US-regulated dollar issuer and look at what it all plugs into: PlayStation, Sony Music, Sony Pictures, Crunchyroll.
🧠 What’s the use case? Creator payouts, in-game purchases, anime merch, cross-border fan payments, settling on rails Sony owns end to end. Plenty of stablecoins are coins in search of an economy. Sony is an economy in search of a coin.
Judge Analisa Torres of the Southern District of New York denied Kalshi's request for a preliminary injunction against the New York State Gaming Commission on July 7, ruling that New York's gambling laws, as applied to Kalshi's sports-event contracts, are not preempted by the Commodity Exchange Act. In plain English: New York can treat Kalshi's sports contracts as gambling and enforce accordingly while the case plays out. Kalshi has already appealed to the Second Circuit. (Yes, the same Judge Torres from SEC v. Ripple. No, that has nothing to do with anything.)
Kalshi's argument, in every state, is the same: it runs a federally licensed exchange, the CFTC has exclusive jurisdiction over it, and state gambling laws simply don't apply. That one theory is what lets Kalshi offer sports contracts in California and Texas, where sports betting is illegal, and in New York without the licenses and taxes that sportsbooks carry. Kalshi did $33bn in trading volume in June. Sports is a lot of what makes that volume possible.
🧠 You need a map for this one. More than a dozen states are now in open conflict with a federal agency over whether the same product is a trade or a bet. It has lost at the district level in Maryland, Nevada, Massachusetts, Michigan and now New York. Minnesota passed a law making trading sports contracts a felony from August 1. Nevada has a contempt hearing against Kalshi on July 16.
🧠 Of all the district courts, this was the one to lose. The Southern District of New York is the most influential markets courtroom in America, and Torres said explicitly that she wasn't bound by the Third Circuit's view. Gambling regulation is a traditional state power; courts presume federal law does not displace it, and Kalshi's claimed harms were mostly monetary, which doesn't get you emergency relief. The court's answer to the whole dispute is: go get licensed like everyone else.
🧠 The legal question is really a tax question. If Kalshi is a market, it charges trading fees and keeps exchange economics. If it's a bookmaker, New York taxes online sports betting at 51% of revenue, on top of licensing, consumer protections and age rules, and California and Texas are closed.
🧠 The Supreme Court is closer than people think. Justice Alito gave New Jersey until August 4 to file its petition asking the Court to review the Third Circuit's ruling. The Ninth Circuit heard consolidated arguments in April in cases involving Kalshi, Robinhood and Crypto.com. Appellate courts disagreeing on the same federal question is the classic trigger for the Supreme Court to step in, and after this week the disagreement runs deeper.
🧠 Watch the CFTC as closely as the courts. The CFTC has filed against Arizona, Connecticut, Illinois and Kentucky to stop enforcement against its licensed exchanges, while simultaneously running a rulemaking on event contracts that could redraw the rules for everyone.
🧠So the endgame arrives one of three ways: the Supreme Court settles preemption, the CFTC writes rules that define what an event contract can be, or Congress does it for both of them. Until one of those happens, the world's largest prediction market operates legally in some states, under injunction in others, and as a felony in Minnesota from next month.
6. Ramp launches Ramp for agents to give them bank accounts and cards
Ramp agents to give your AI agents cards, transfers and bank accounts, with controls governing every dollar they move. Each agent gets its own identity, its own budget, its own spend limits and its own audit trail. Credentials expire on a schedule, exceptions route to a human for approval, and every transaction lands in the Ramp dashboard attributed to the agent that made it, receipt attached. Setup is a connector inside Claude, ChatGPT, Cursor or Perplexity, or a command-line install. Setup takes 15 seconds according to marketing.
🧠 The mental model is agents as employees. An identity, a card, a budget, an expense policy, an approval chain, an audit trail. It's the most natural extension of Ramp's founding product: imaginable spend management. Except for an employee who works around the clock, spends in milliseconds, and occasionally hallucinates a purchase, is exactly the customer a spend-controls company was built for.
🧠 This is the corporate half of agentic commerce. Agents with cards buying things on behalf of businesses is much, much more likely to see traction before consumers. Offload repeat purchases, or let them buy their own compute / online API access.
🧠 Ramp built “playbooks” to get started quickly. For purchases on sites that don't speak agent protocols, which is most of the internet, "Browserbase drives the checkout; your agent card pays." This doesn’t need new protocols. It works today.
🧠 There's no integration project here. The buyer installs a connector in the AI assistant their team already uses, authenticates their existing Ramp account, and their agents can spend within policy.
4 Companies 💸
1. Karta - US credit card for global non-residents (no SSN)
Karta gives global non-residents a US-issued credit card without needing a Social Security Number or ITIN. A high-net-worth Brazilian with a US bank account but no US credit file applies online, gets a line of credit plus buy-now-pay-later, and spends with zero international transaction fees. The whole account is run through a WhatsApp concierge backed by AI agents, available 24/7.
🧠The promise is "build US credit," but the bureaus key everything off an SSN or ITIN — so which file is actually getting built? Nova Credit cracked the adjacent problem by porting your home-country history into the US; Karta is trying to originate a brand-new US one instead. How many wealthy immigrants also want an Amex-like card? I imagine quite a few.
2. Humanos - Permission rails for AI agents
Humanos lets a platform check whether an AI agent is actually allowed to act before it executes, then produce a receipt proving it afterwards. A user signs a mandate like "Book my Milan trip, spend up to €1,200, max €250/night, valid until Friday," and the agent calls humanos.verify() at runtime to confirm the action sits inside those limits, capture any missing approval, or revoke it. It speaks the agentic-payment standards like Visa Intelligent Commerce, Mastercard Verifiable Intent, Stripe's Agentic Commerce, Google's AP2 but works across any network, any protocol.
🧠"AI can act. Make sure it was allowed to" is competitive. The second an agent can spend your money, someone has to prove it stayed inside the mandate. But Visa, Mastercard, Stripe, and Google are each shipping their own agent-authorization standard, and Humanos's entire pitch is being the neutral layer that speaks all of them.
3. Coverd - A cash-back card that doubles as a sweepstakes casino
Coverd issues a no-annual-fee Visa card with spending limits and "up to 100% cash back," where the rate is set by a draw rather than fixed. They also run a separate sweepstakes app: you buy "Practice Coins," collect bonus "Coverd Cash," and play games for a chance to redeem real prizes. It's the dual-currency, "no purchase necessary," mail-in-entry structure used by social casinos like Chumba and Stake.us.
🧠So is this a fintech card, or a casino with a card stapled to the front? They have a "Responsible Social Gaming Policy," session reminders, daily loss limits, and self-exclusion forms. This is gambling plumbing, and the no-fee Visa is the respectable wrapper that gets it past an app store and onto a card rail. "Up to 100% cash back" tells me a lot too; a fixed-rate card competes on economics, a prize-draw card competes on the dopamine of the pull. I really want this to not be another brick in the wall of “everything is a casino now.” But I’m struggling to see the path to financial freedom for users here.
4. El Dorado - The dollar SuperApp for Latin America
El Dorado lets anyone in Latin America hold, send, and spend US dollars as stablecoins from their phone. You buy and sell USDT through a peer-to-peer market with 70+ local payment methods like Pix, bank transfer, cash, then spend the balance on a prepaid Visa or send it fee-free over Tron. It runs across 13+ countries with more than a million users.
🧠 These guys are winning in the bricks and mortar; every day, business flows in countries like Bolivia. They have a physical branch, and customers who don't care about stablecoins, but they do care about lower fees and faster payments. That's the whole story. Keep your eye on El Dorado, and send them to every banker or trade finance nerd.
Good Reads 📚
AI labs want to displace software, and software wants AI to be an input. Both have an imperative to own the user. Because if the labs own the user, the software has all of its IP and know-how commoditized. If the AI distributes via the software, they’re a commodity input. Ben also says, if the AI labs win, then you’ll see a hollowing of the economy like we saw with globalization. The headline GDP figure will be fine, but the displacement is real.
🧠 Pay attention to this if you care even a tiny bit about your own company’s IP or survival past the AI-pocalypse. And think carefully about lock-in.
🧠 Anthropic's pitch to businesses: "In the end, you'll pay the same labor costs as you do today… but to us instead of individual humans, and at 10x the bang for buck."
Tweets of the week 🕊
That's all, folks. 👋
Remember, if you're enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)
Want more? I also run the Tokenized podcast and newsletter.
(1) All content and views expressed here are the authors' personal opinions and do not reflect the views of any of their employers or employees.
(2) All companies or assets mentioned by the author in which the author has a personal and/or financial interest are denoted with a *. None of the above constitutes investment advice, and you should seek independent advice before making any investment decisions.
(3) Any companies mentioned are top of mind and used for illustrative purposes only.
(4) A team of researchers has not rigorously fact-checked this. Please don't take it as gospel—strong opinions weakly held
(5) Citations may be missing, and I’ve done my best to cite, but I will always aim to update and correct the live version where possible. If I cited you and got the referencing wrong, please reach out
