Things to know 👀
Sky News reports early talks valuing Monzo at £8bn to £10bn (up to $13bn), about double its £4.5bn valuation in 2024. Its board is also weighing a new round above £8bn ($10.6bn) to fund a push into mainland Europe.
🧠 Nu and Revolut are about to collide on three continents. Revolut launched a bank in Mexico in January, and Nu launched in the US two weeks ago, with both holding conditional US charters. Monzo would put Nu on Revolut's home turf.
🧠 Nu has escaped its home market; Monzo has not. Geo expansion is Monzo’s biggest weakness. For Nu, Monzo would be its shortcut into the UK and EU. Monzo has banking licenses in both, direct access to the rails, 15.2m customers and deep integrations that would take years to build. Revolut waited five years for its full UK license.
🧠 Nu and Revolut are making opposite bets on lending. Nu lends out about 87 cents of every dollar of deposits, and Revolut lends out about 6p of every £1. For thatNu makes a 22.9% net interest margin, while
🧠 Revolut has avoided building a large lending book in Europe for a reason. Lloyds, which lends out about 98p of every £1 of deposits, makes 3.19% NIM. Nik Storonsky told the FT he wants "effectively zero risk", so Revolut plans to cap lending at 10-20% of deposits and sell on the loans it makes.
🧠 Fee revenue offers more opportunity in Europe. Fees made up 76% of Revolut's revenue last year, and Storonsky puts its return on equity at 40-50%, double his best rivals. So would Nu adjust to this?
🧠 Monzo lends out less than 10p of every £1 of its £25.7bn in deposits. Nu could turn it into something very different, if its lending playbook works at UK margins.
🧠 The bigger question is whether Nu can change Monzo's culture, and whether it can afford the distraction. Nu is very big on culture and runs a single global platform. Both deliver feature velocity Monzo hasn’t had recently. But teplatforming Monzo would be a huge job, and a distraction while Nu is still focused on Brazil and Mexico.
🧠 I still prefer the Monzo product (dramatically so) as my daily driver to any other bank, including Revolut. But there's no question it needs a momentum shift.
Remember, these are early talks, and Monzo declined to comment. It might never happen. Still, it's about to get spicy between the big two neobanks. 🍿
2. Kalshi faces volume scrutiny (WSJ) and Files for margin trading with the CFTC
The WSJ found nearly 1 million trades of about $5,500 on Kalshi's ETH perpetual since August, worth more than $5B. It reports the CFTC is reviewing the data before deciding whether to open a case. Kalshi denies any wash trading and says the CFTC hasn't contacted it.
Kalshi Klear filed with the CFTC asking for approval of margin on event contracts. Regulated US prediction markets are fully collateralized today, so each trader posts their maximum loss up front. If the CFTC signs off, approved institutions could post less from as early as 9 November.
🧠 X user Beni accused Kalshi of faking its crypto volume last weekend. He pointed to $539M of ETH perp volume in a day against $3.1M of open interest, and found trades of about $5,500 made up roughly half of what he sampled.
🧠Kalshi says hundreds of different traders hit those $5,500 orders and mostly made money doing it. It pays market makers to rest orders of a fixed size, and faster traders keep picking them off. Barnard economist Rajiv Sethi finds that more plausible than wash trading, but says the rewards inflate volume without adding much liquidity.
🧠 Beni escalated with an unverified claim that 99% of “perp” (perpetual futures) volume was fake, said he needed a lawyer before sharing non-public information, and tagged Coffeezilla, but the information is unverified and hasn’t been responded to.
🧠 Parlays make the volume look bigger than the money. You pay 6¢ for a parlay that pays $1 if it wins, and Kalshi records $1 of volume. Gambling Insider found Kalshi's August parlays showed $17.8B of volume from $1.1B of bets.
🧠 Kalshi argues this is "notional." A $1bn oil futures position might sit on $100m of margin, but it's exposure to $1bn of oil worth $1bn today. A $1 parlay is worth $0.06 today, because the market gives it about a 6% chance of paying out. Kalshi will say, someone else put up the $0.94. But it’s counting outcome prizes as volume. That’s a generous use of the word notional.
Speaking of margin.
🧠 Kalsh wants its clearing agent to offer margin and requires CFTC approval before it can offer margin. The filing treats every eligible event contract as one. Klear filed under rule 40.5(a), which gives the CFTC 45 days to review. How will it work? We don’t yet know. The model's maths and parameters sit in a confidential annex.
🧠 Eligible contracts cover economic, financial, political, commercial and other objectively verifiable events. Sports are excluded. Kalshi told CNBC that culture and mention markets are out too. Klear judges margin risk on YES and NO sides separately, and new listings start fully collateralized.
🧠This starts with institutions. Margined contracts clear only through FCMs or big traders approved as self-clearing members, and Kalshi says it will begin with self-clearing members that meet capital thresholds it hasn't published. But an FCM can offer margin to its own customers, and Kalshi's affiliate Kinetic Markets registered as an FCM in March. So, expect prediction markets with margin soonish.
🧠 You'd post enough cash to cover one bad day's price move, which is a fraction of your full possible loss. Kalshi raises that amount before big scheduled events, and takes it back to the full amount as the result gets close. So margin mostly helps on contracts that settle months away, which Kalshi says is what institutions want. This is how margin works in most other contracts, but its untested here on events.
The OCC gave conditional national trust charters to three stablecoin companies on 18 September. Nick van Eck's Agora sells businesses their own branded dollar backed by its AUSD stablecoin, Bastion runs standalone stablecoins for brands like Sony Bank, and Circle co-founder Sean Neville's Catena gives AI agents stablecoin accounts with spending rules. The charters let them hold and manage customer assets ahead of the GENIUS Act taking effect in January 2027, while deposits and lending stay with full banks.
🧠 All three sit in the flow of funds, and each one makes its money in a different place.
🧠 Agora is the liquidity network + white-label stablecoin. Agora lets brands white-label its stablecoin. The stablecoin dollars are all AUSD underneath, so they move between each other with instant liquidity (e.g., MyUSD and YourUSD). Each brand controls its own token and earns a share of the yield based on its market cap.
🧠 Bastion gives each client a coin of its own, with its own reserves. That means more control over reserves if you have a sophisticated treasury, at the possible cost of speed of liquidity. (e.g., SonyUSD)
🧠 Cantena is using stablecoins to build wallets for businesses and AI agents. A business sets limits for its AI agents, and every payment gets checked against them before it moves.
🧠 The charter does a different job for each. Agora moves AUSD issuance from Bermuda into a US bank, Bastion's CEO says big banks need a federally regulated partner, and Catena can hold customer money itself where today it uses regulated partners.
🧠 These are conditional approvals, and the last batch shows how long the road is. Of the five crypto firms that got conditional trust charters last December, two were live banks by August.
🧠 This will take time, but the amount of innovation coming to how businesses manage their own treasury is fascinating.
Numeral raised a $100M Series C led by Insight Partners, with Salesforce Ventures joining, a year after its $35M Series B. It tracks where companies owe sales tax, registers them, calculates rates, files, pays the state, and manages exemption forms, with VAT and GST in 90+ countries. The company started in 2023 and has raised $157M in total.
🧠 Sales tax is a nightmare. I didn’t realize quite how much. Each state has its own idiosyncratic rule about what gets sales tax and how much. For example, California will tax SaaS and off-the-shelf software from 1 January 2027, joining 20+ states that tax SaaS in some form. I hated doing quarterly filings, but various rates, by state? Oof.
🧠 AI has solved sales tax. Numeral guarantees on-time filing and pays your penalties and interest if it misses (similar to Zamp). Customers buy a filed return, and the vendor pays for its own mistakes if AI can’t fix it..
🧠 It’s not all AI, and that’s a good thing. Numeral runs a rules-based engine for the rates, AI for that paperwork, and tax specialists for the hard cases. Calculating rates has been done for years; registering, filing, answering letters- that's where the cost and pain were, and the kind of thing AI is quite nifty at.
🧠 AI is about to disrupt all of taxes. The model fits tax jobs that are high volume, rule-based and deadline-driven.
SoFi Bank, N.A. is a nationally chartered, OCC-regulated bank and issues its own stablecoin, SoFiUSD. From this week, its debit and credit card transactions settle with Mastercard in SoFiUSD, on-chain, seven days a week. Merchants receive settlement into a SoFi Bank account and can withdraw to cash 24/7 at no cost.
🧠 If you issue cards to your customers, seven-day settlement solves a cash flow problem. Many banks and central settlement services are closed on weekends and bank holidays. If your customer is out spending, you have to hold a large buffer against that, to send when the banks do open.
🧠 This is by far the biggest card issuer to use stablecoin settlement. Both Visa and Mastercard let you settle 7 days a week with stablecoins. Visa's whole stablecoin settlement book runs at $20bn annualized across 160+ card programs. SoFi's program alone is $25bn.
🧠 SoFi has its own coin, meaning they keep the yield on the stablecoin float. Nearly every other issuer settling on-chain uses someone else's coin. Rain and Reap run roughly $3bn and $6bn on Visa in USDC, and Cross River, Lead Bank and Nuvei are lined up on Mastercard with USDC, PYUSD or RLUSD.
🧠 This is why the card networks are pushing OpenUSD (OUSD) quite heavily; that way, the card issuer wins. Why sacrifice that to someone else? I believe settling in a coin that shares the yield becomes the default once OpenUSD launches later this year.
🧠 Instant settlement for merchants is a huge bonus, but they’re still early. SoFi is looking to court more business customers, and offering them 24/7 settlement is a huge feature addition for merchants. But is it enough to make someone switch? Probably not. Although they don’t have to, they could just open an account for settlement. Interesting wedge.
4 Companies 💸
1. Apate.AI - AI agents that investigate scammers (SO COOL!)
Apate helps banks and telecom companies uncover scams by sending AI voice and text agents to pose as potential victims. It collects the bank accounts, crypto wallets, and websites scammers reveal, giving fraud teams information they can use to investigate and disrupt their operations. Commonwealth Bank is already using the technology.
🧠 I. Love. This. Scammers eventually have to tell you where to send the money. Getting a bot to that point gives a bank’s fraud team an account to investigate before more customers pay it. I’m curious how much of this intelligence is new to the bank, and how quickly it can act. There’s something very satisfying about getting scammers to supply the information needed to stop their own operation.
2. Rightcharge - Fleet spend management card for EVs (UK / EU)
Rightcharge helps employers pay for charging company vehicles at employees’ homes by crediting the cost directly to their electricity accounts. It also provides a card for public charging and brings both expense types into one bill for the fleet’s finance team.
🧠 Home charging your fleet vehicle only works if you can reclaim the cost on your energy bill as a driver. Those reimbursements are often manual, so giving the fleet to Rightcharge removes that admin. This makes the cheaper fleet option easier for companies and drivers to use, and as EVs grow, that’s smart. My worry is how big this niche can get: they’ve started expanding beyond the UK into France and Germany, but this seems like a great idea that should go global.
3. Diameter Pay - US dollar accounts for foreign banks and fintechs
Diameter Pay lets banks and fintechs outside the US offer customers dollar accounts and payments through its US banking partners. Customers can receive and send bank payments, with stablecoin conversion available too. The company says it has processed more than $10 billion in payments this year.
🧠 I love that they’re competing on better transaction data and compliance controls to support more demand. But why doesn’t Airwallex win this customer? The test is whether Diameter can get banks comfortable with legitimate customers they would otherwise turn away. And as it grows, can it keep local regulators comfortable without storing up a future AML problem?
4. Plenti - Dollar accounts for Colombians earning and investing abroad
Plenti lets Colombians receive international payments using US and European account details, hold dollars, and spend with a Visa card. They can also send money to brokerage accounts or access investments in the app. The company reports more than 150,000 active users and over $3.1 billion in annual transaction volume across its business and consumer divisions.
🧠 This space is becoming crowded. ARQ and Littio are chasing similar customers, so another dollar account needs more than a long feature list. And then there are longer-term questions for the whole category, like what local regulators allow and how well the AML controls hold up as volumes grow. I’m curious who wins this space, and what the enduring moat turns out to be.
Good Reads 📚
The term “AI Factory” gets a lot of hate, but the concept is simple: the factory buys energy (Megawatts) and turns it into compute (Tokens). On raw input vs output, in 2026 Anthropic achieved $50m per megawatt against a $10-15m cost. Anthropic booked its first profitable quarter, $10.9b of revenue & $559m of operating profit. But training, headcount, and buying up new data center capacity eat up their costs.
Where this gets interesting is how new models are innovating on this dimension. GLM 5.3-Flash scores similar to Opus 4.8 but at 97% lower cost, because it is so efficient on parameter use.
🧠 Ed Zitron makes some good points in isolation, but misses the big picture. Yes, a lot of the financing is circular; yes, the accounting is getting creative to say the least, and much of the debt is off-balance-sheet; the hyperscalers' revenue is mostly predicted on demand from OpenAI and Anthropic.
🧠But think about where OpenAI and Anthropic’s revenue and demand come from. If you’re not an engineer, or in a hyperscale growth company that’s using AI well, it sure does look dodgy. And you can reductively say “oh it writes code, great.” Turns out that’s INSANELY useful. What has AI ever done for us, other than make companies ship at least 3x faster?
🧠 It’s now fashionable to hate on AI. And there are plenty of things to dislike. But my word there’s also an issue of the benefits only being visible to a few. And that’s the thing we need to fix next: consumer-grade AI that makes people’s daily lives better. And not more dopamine-addled.
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
