Things to know 👀
Robinhood’s Q2 earnings included data showing prediction markets overtook crypto and stock trading for the first time. Transaction revenue was $776m, up 44%. Options did $342m, event contracts did $156m, equities did $129m, and crypto did $100m, which fell 38% YoY. The company also posted record Q2 revenue of $1.31bn, up 32% YoY, with diluted EPS of $0.62 against a consensus nearer $0.42.
🧠 Prediction markets are now the #1 speculative product. Robinhood never misses a trick. If a speculative asset brings in users, they execute fast. They started by partnering with Kalshi, routing most of their orders there. Then they decided to build their own engine.
🧠 Robinhood has vertically integrated. In June, Robinhood turned on its own licensed exchange and clearinghouse (buying a 90% stake in a venue formerly called MIAXdx with Susquehanna). World Cup match outcomes now route straight through their own pipes.
🧠 They still route to multiple exchanges. Robinhood sends volume to its own exchange, but still routes some to Kalshi and Crypto.com. Not all exchanges (DCMs) carry the same events, so a broker (FCM) needs access to multiple venues to serve users.
🧠 Sports is the massive volume driver. Up to 90% of Kalshi’s volume comes from sports. It’s hard to look at this and not see plain old gambling (a debate where I can see both sides). But Sports is where Robinhood excels, and where Kalshi is strong.
🧠 Kalshi and Robinhood are the best of frenemies. Kalshi and Robinhood are the best of frenemies. Driven by consumer volume from Robinhood's app, Kalshi hit a $22bn valuation on $31.5bn in monthly volume. But Kalshi's CEO recently named Robinhood as his main rival, because they're both fighting to own the end-consumer app.
🧠 Do users eventually start investing longer term? There is some evidence that users who join for prediction markets eventually build long-term portfolios. Robinhood Gold hit 4.8m subscribers (up 39%), net deposits hit a record $21.7bn, and total assets reached $369bn. CEO Vlad Tenev now runs 13 separate business lines over $100m. But does a platform without speculative hooks grow that fast?
🧠 The monetization trap. My worry is that most users never transition to long-term investing, and Robinhood is simply a masterclass at monetizing speculation. Robinhood earns a lot more when a customer is actively trading.
🧠 Who is protecting the consumer? The consumer protection conversation is conspicuously absent. Institutions see a great hedging tool, brokerages see cheap customer acquisition, and states see a threat to their tax revenue. Who is actually looking out for user harm?
🧠 Expect prediction markets to continue to grow. Q3 brings the World Cup knockout rounds and the start of the NFL season, pushing their annual run rate well past $600m. This debate isn't going away.
New York sued KalshiEX in Manhattan state court on Friday, alleging it runs an unlicensed gambling business without a New York State Gaming Commission license. Hochul and James are seeking an injunction, restitution, forfeiture of all gains, and fines at three times the profit made in the state. Kalshi called it "political theater from the leadership in our own state" and moved to shift the case to federal court within eight hours. The filing followed a Manhattan appeals court refusing on Wednesday to pause a ruling that New York's gambling laws can apply despite federal regulation. Overnight, the CFTC asked the federal judge to bar New York from acting.
🧠 The jurisdictional fight. Kalshi and the CFTC hold that event contracts are swaps under exclusive federal oversight. States hold that sports contracts are sports betting, which they license and tax. Both readings have won somewhere, which is why this is going upwards.
🧠 The volume of state action is staggering. Kalshi and Polymarket face at least 20 actions from state regulators, tribes and individuals. 44 state attorneys general wrote to the CFTC on Monday saying it has no authority over sports event contracts. And, they’re losing potential revenue, so their motivation is obvious.
🧠 The American Gambling Association welcomed the lawsuit, which tells you who is funding the other side of this fight. For the lobbyists, this is about defending revenue, not consumer protection.
🧠 The problem is prediction markets are an incredible hedging product in a way that sportsbooks never have been. There’s a contract on Kalshi asking how many times the Fed would cut interest rates in 2026, in January, 0 cuts was at 6%, now its at 82%. As former CFTC comissioner Quintenz explained: “Imagine another kind of hedging opportunity where you can make 13x in 7 months”
🧠 Read the last part of the last bullet again.
🧠 Prediction markets are neither gambling nor are they free of consumer harm. They’re a speculative product, whose primary market is sports, competing with and selling to the same people who were sports betting, and at risk of addiction.
Increase Bank went live this week, an FDIC-member institution sitting underneath Increase's banking core, with direct connections to the Federal Reserve, The Clearing House, and Visa. It came out of the Twin City Bank acquisition rather than a new charter application.
🧠In the age of de novo charters, the acquisition route is less talked about. Twin City Bank in Longview, Washington has 16 staff, $114m of assets and one office on Vandercook Way. The software sitting on top of it at Increase now moves hundreds of billions of dollars a year for Gusto, Ramp and Stripe. Darragh Buckley was Stripe's first employee, and Stripe is now a customer of the bank he bought. So many subplots here.
🧠 The BaaS platform that bought the bank. “Banking as a Service” became a dirty word (and I doubt Increase really identifies as that). They’re deep ledgering and payments systems experts who care about details, engineering quality, and consistency. This is so different to the Synapse collapse where nobody knew where the deposits were. Increases’ whole schtick is they’re great at that, and reliable. Boring is sexy.
🧠 This new bank will be one of many. I asked Increase whether the new bank complements the existing sponsor partners. Their answer was that Increase Bank joins the network. Many users deliberately work with several banks for redundancy, capacity, and a spread of products and regulatory expertise.
🧠 There are now banks with incredible modern software. If you’re a sponsor bank and your competitive advantage was a charter, that’s no longer the case. Cross River, Increase, Column, and Lead all differentiate on their technology. I know some, like Coastal, are investing heavily to stay competitive here too.
X Money went live this week for US Premium and Premium+ subscribers, with a deposit account, free peer-to-peer transfers, and a Visa debit card. Cross River Bank holds the deposits. X Payments LLC has money transmitter licenses in 41 states and DC, and New York and Massachusetts are both still out. Users of Premium+ get 6% APY, and 3% cashback. Premium users get the 6% APY if they move their direct deposit.
🧠 X Money exists to sell X Premium. SpaceX's S-1 counted 4.4m X Premium and Premium+ subscribers on 31 March, worldwide, out of 550m monthly users. Musk's 2022 pitch deck had projected 69m by 2025. The filing lists increasing Premium subscriber conversion as a growth driver, and X Money sits behind the paywall. So the 6% is the cost of acquiring subscribers, not deposits.
🧠 This is the inverse of a neobank play. Neobanks start with the account and work to upsell the subscription. X sold the subscription first, so every X Money customer was already paying before they opened the account.
🧠 Two prices, one rate. Premium+ costs $395 a year and gets 6% straight away. Premium costs $84 and gets the same 6% once you link a direct deposit. On $10,000 at a 2% spread over a 4% savings account you earn $200, so the $395 tier needs about $20,000 parked before it pays for itself. The $84 tier breaks even nearer $4,200.
🧠 They’re giving 3% cash back which loses money, making me think it won’t last. Debit interchange pays nowhere near 3%, so every swipe costs X money. X has published no Truth in Savings disclosure, and it has not committed publicly to whether the 6% is permanent, whether it covers the whole balance, or what happens if rates move. Read that as an introductory offer.
🧠 Cross River holds the deposits, which raises the question of how the yield is shared. This was a prize every sponsor bank wanted and you have to imagine X negotiated hard, so I’m super curious what economics they got there.
🧠 X has gone state by state for money transmitter licenses, which is the slow and expensive route and still leaves it locked out of New York. In this regulatory climate, why not put a separate entity in for an OCC charter, or take the Utah ILC route that Block used for Square Financial Services? The answer decides whether the everything app plan means stablecoins, credit and stocks, or a bank.
Revolut is bundling ChatGPT Go into its plans at no extra cost, from three months on the free Standard tier up to twelve months on Ultra. Paid subscribers and Revolut Pro can redeem now, with Plus and Standard following over the coming weeks. Ultra gets 12 months plus a 12-month buddy pass to hand to a friend, Metal gets 12 months, Premium and Pro get 6 months, and Plus and Standard get 3 months as a promotion.
🧠 The neobank subscription is turning into the real loyalty product. I'm beginning to think it will be 10x more useful and adopted than Amex ever was. Amex bundled the physical world, lounges, concierge, and travel credit for a pre-digital affluent customer. Revolut is bundling the software you already live inside.
🧠 Go is the only paid OpenAI tier that carries ads, which makes me wonder why they didn't reach for Plus on the premium tiers. Ads went live for ChatGPT's UK Free and Go users on 6 June, the first European market to get them, so OpenAI earns from a customer who is paying nothing.
🧠 If you don’t opt out, you’ll auto-enroll into paid ChatGPT at the end of the term. Revolut's own terms then say Premium customers roll onto a paid ChatGPT plan when the benefit period ends unless they turn auto-renewal off (likely Paid Go). So this isn’t the benefit it seems like.
🧠 Subscriptions are Revolut's fastest-growing major line. £708m ($936m) in 2025, up 67%. Card payments did £1.0bn, up 45%. Interest income was £974m, up 23%. Compared to traditional peers that's a much larger proportion of the income base, and it shows Revolut's experimentation with modern partners is paying off.
🧠 This is an Amex alternative for people who care more about their software than their airport lounge. Ultra costs £540 a year and Revolut says the software inside it is worth over £4,000. Frankly, it's a compelling offer. 🧠 Imagine if your bank came with a proper ad-free ChatGPT, Netflix and Amazon Prime.
🧠 Subscriptions work. 63% of Revolut's new customers arrive by word of mouth, and perks are what people mention in that word of mouth. Paid plan adoption grew 42% last year.
🧠 Expect every neobank and brand to copy this within the next few years. The scarce asset here is a warm base of 75m people who open your app every day, and that is worth more to a software company than most banks realise they are sitting on. If you're writing a 2027 strategy this summer, you can do a lot worse than taking a closer look at Revolut.
Dakota has applied to the OCC for a national trust bank charter covering digital asset custody, stablecoin issuance, and related services. It started as business banking on stablecoins, scaled to around 500 business customers and billions in annualized volume, and is increasingly an infrastructure company supporting others to build neobanks. Today it runs as a registered money services business, holding state money transmitter licenses in place of a charter.
🧠 The GENIUS Act triggered all of this trust charter activity. A trust charter lets you custody stablecoins directly instead of paying someone else to do it, and it swaps fifty state regulators for one federal one. Both of those improve the economics for anyone building on top of you, and give you more control over risk appetite.
🧠 The queue keeps growing. The OCC conditionally approved Circle, Ripple, Paxos, BitGo and Fidelity in December. Circle went all the way through this month. Sony has conditional approval. Agora applied in April, Catena Labs in May, and Dakota now.
🧠 A trust charter is not a bank charter. No insured deposits and no lending. GENIUS is explicit that a payment stablecoin is not a deposit and cannot be sold as insured, so a bank still sits under anything that needs insurance.
🧠 I think we’ll see more apply and get approved. Circle took seven months from conditional to full approval, with more lawyers than most applicants have staff. Dakota filed this week, so they have time, though the clock in my mind runs out when this administration ends.
🧠 Two years ago this was a Circle-scale move. This year the queue holds Sony, Circle and a Series A startup that has raised $12.5m. Each one is a company deciding it would rather be regulated than rent someone else's license.
4 Companies 💸
1. SkyFusion - AI data centers in homes, financed like rooftop solar
SkyFusion puts liquid-cooled AI accelerators into homes alongside rooftop solar and battery storage, turns waste heat into domestic hot water, and aggregates AI inference capacity to sell to compute buyers. Homeowners get lower energy bills plus recurring participation payments; investors get the compute revenue wrapped in solar-style project finance (SPVs, power purchase agreements, first-loss equity). The first array targets 300 homes and 660 accelerators, scaling to 1,000 homes.
🧠 The finance stack is the product here. Solar and batteries are a proven, financeable asset; "contracted inference capacity" from accelerators in 300 living rooms is not, and compute buyers mostly pay for density, reliability and security. This model is unproven. NVIDIA and Span are already piloting mini AI data centers bolted onto homes, so the giants have noticed the same idea. If this works, SkyFusion is a structured finance company in an AI hat, and a smart way to monetize your roof.
2. Savi Security - Scam protection for your parents
Savi screens your family's calls, voicemails and texts for AI-powered scams: it answers unknown numbers and filters junk voicemail, joins a live call to flag social engineering in real time, and (on iOS) strips scam texts before they land. Anything that slips through, like your teen's "Your Snapchat will be deleted" text, can be pasted into its free Scamwise checker for a verdict. One subscription covers parents, teens and partners.
🧠 Savi's customer is the worried adult child. This is a real problem space but I don’t know that I buy the solution as a product. There’s a risk this just monetizes worry, and doesn’t actually protect anyone. I hope it does. But I remain to be convinced. A $7.99 app has to catch what Google and the bank apps miss, at the moment your mum is being coached by the scammer to ignore warnings. Detection is the easy half; getting a panicked human mid-scam to trust the warning is the hard one.
3. M1X - On-chain sovereign bonds backed by US Treasuries
M1X partners with governments to issue financial instruments natively on-chain. It’s first, USDM1 is a USD sovereign bond issued by the Republic of the Marshall Islands, collateralized 1:1 by short-duration US Treasuries. It pays a sovereign coupon, redeems at par under New York law, and funds what the company calls “the world's first nationwide UBI program,” paying Marshallese citizens quarterly.
🧠 A payment stablecoin can't pay you yield under the GENIUS Act; a sovereign bond can. Claude tells me that the Marshall Islands tried a national digital instrument once before (the SOV, 2018), the IMF hated it, and it never shipped, so this is attempt two but with tokens. So why this? BlackRock's BUIDL and Ondo already sell tokenized T-bill exposure without an exotic sovereign in the middle. But then. Why shouldn’t governments go onchain? And why shouldn’t that benefit citizens? Especially in the age of AI. There’s something here. I just don’t know if the Marshall Islands is it.
4. AIsa - Pay-per-call APIs for AI agents
AIsa gives an AI agent one API key for 1,000+ capabilities: Apollo contacts, X and Reddit data, Polymarket and Kalshi markets, stock fundamentals, web search, even its own email inbox. Every endpoint is metered per call and settled by the agent itself over Circle's USDC nanopayments or the HTTP-402-style Machine Payments Protocol; no subscriptions, no invoices. Onboarding is a prompt: "Read aisa.one/docs/agent-quickstart.md and help me safely connect."
🧠 I love this but my question is adoption and distribution. I hate having to sign up for a new service to get a bit of data. Surely this is the killer app for micro payments. But why is this a 3rd party service and why doesn’t it live inside Claude Code, Cursor, or Codex? If AIsa wins, it’s because the catalogue is the biggest and most used.
Good Reads 📚
1. Man Vs Token
For the first time in human history, humans are cheaper than some forms of AI token, and it looks like AI is creating more software jobs than it eliminates (although this is based on Ramp data). The piece draws a parallel to railroads, where the initial overbuild led to trains crashing and gave birth to modern management. Today’s AI has enabled anyone to do the same with software. Tokenmaxxing is throwing bodies at the problem, evals are the new OKRs, and wasted tokens are headcount bloat.
🧠 The meta skill of the AI workforce is management. Knowing what to build and how to measure the success of building it.
Tweets of the week 🕊
If you enjoy this kind of content, I can guarantee you’ll love being in a room of 1,500 other folks who love to go deeper into where finance meets AI. That’s a huge theme for us at this year’s Nerdcon in San Diego on the 19th and 20th November. I’m bringing my audience, the operators, the people who read this newsletter. And it’s the perfect place to find your next hire, client, or just get inspired. Let’s make events awesome again.
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
