Here's this week's Brainfood in summary

👀 Things to Know:

💸 4 Companies:

Content Corner: Duffy’s last dance - The CME vs Kalshi

As I alluded to a few weeks ago, I’m experimenting with paid sponsors for Fintech Brainfood. And for fintech, this is the perfect start. When I think of payments nerds' Mt Rushmore, I think of the founders, Dimitri, Matt and Sam, almost immediately. And, having been the ledger of choice for so long, they’ve finally taken the plunge and become the PSP that works with any rail, RTP, Wire, and even stablecoins seamlessly. So here’s their ad spot:

Stablecoins are here.

Checks, ACH, and wires aren’t going away.

RTP, FedNow, and new payment rails keep coming.

The challenge isn’t predicting which one wins. It’s supporting all of them without slowing down.

Build faster. Launch sooner. Scale with confidence.

Trusted by Procore, Navan, Morse, and companies of all sizes. Backed by $600B+ in payments.

Start building today: moderntreasury.com

1. Stripe and Advent have bid more than $53 billion to take PayPal private

Stripe and private equity firm Advent International have offered $60.50 per share for PayPal, valuing the company at more than $53bn, per a Reuters exclusive. The bid was submitted earlier this month, is backed by roughly $50bn in committed financing from banks, and lands about 28% above the July 14 closing price.

Under the proposal, Stripe and Advent would own PayPal in equal stakes and keep the company whole rather than break it up. At the same time, CNBC reports Stripe, Advent and Block are together contributing $17bn of equity for the offer. Same reporting says PayPal's board meets as soon as July 20, the day after this edition lands.

The context is that PayPal peaked at roughly $360bn in July 2021 and now trades at about an eighth of that due to slow growth. Branded checkout growth slowed to 1% in Q4 2025 (from 6% a year earlier), BNPL keeps losing share to Klarna and Affirm, and the board replaced CEO Alex Chriss with HP's Enrique Lores on March 1 after a 2026 forecast that dropped the stock 19% in a day.

🧠 The difference between Stripe and PayPal is growth. Stripe achieved $1.9 trillion of 2025 volume in 2025, up 34% YoY. PayPal's $1.79 trillion in 2025 was on an anemic 7% YoY growth. That’s what markets want and reward.

🧠 The combined entity would be the single largest PSP in the USA. With a combined TPV of nearly $3.7 trillion, they'd be enormous. The current largest incumbent, Chase, processed an estimated $2.8 trillion of payments volume in 2025.

🧠 But considering the possible Block involvement, perhaps this is more of a feast over assets. Would block take the POS plus BNPL business?

🧠 Braintree processing is a major prize. PayPal outsources actual processing to Fiserv and Global Payments. Stripe could insource it as a 3rd party provider and reap many of the benefits. But PayPal has been trying (half-heartedly) to do this for a decade or more and hasn’t pulled it off.

🧠 Stripe has the infrastructure but not the consumer brand. PayPal brings 439 million active accounts across PayPal and Venmo for consumers. (Link has 300m consumers but is not a brand in the same way)

🧠 This is a story that has been rumbling. Reuters dates the first approach to early April, with the first stories as early as February. What changed since previous reporting is a formal number and $50bn of committed bank financing. There’s a serious effort here at least.

🧠 At more than $53bn this would rank among the largest leveraged buyouts in history, ahead of Twitter's $44bn in 2022. That banks committed $50bn tells you as much about credit markets in 2026 as it does about PayPal. At roughly eight times PayPal's $6.4bn adjusted free cash flow, the price is private equity maths.

🧠 The problem with LBO’s is that debt needs servicing. And doing that debt while funding a multi-year replatforming could pull Advent's payback clock and Stripe's rebuild ambitions in different directions.

🧠 Advent has run this exact playbook before. With Bain, it bought Worldpay from RBS in 2010 and floated it in 2015; it built Nets and merged it into Nexi. Buy a payments asset out of an unloved corner, fix it in private, return it to market larger. But why do they need Stripe so close to it if that’s the play?

🧠 Venmo is a take-private argument. Roughly 70% of PayPal's transactions and about 5% of its revenue. Fixing that gap means years of lower margins that the public markets won’t tolerate. Private owners can sell off assets, strip costs and make the company much leaner to make room to invest in Venmo.

🧠 The cautionary tale is FIS-Worldpay. A $43bn acquisition that looked great on paper but ultimately produced $17bn+ of writedowns and an eventual divestiture.

🧠Payments mega-mergers die in integration. Look at Nexi, Worldline, Fiserv or any of the big incumbent “roll ups” where decades of M&A that gave a short-term boost to stock prices ultimately left a legacy of unintegrated tech and teams.

🧠I worry this is a big distraction for Stripe. They’re crushing in all directions. Benefitting from a massive AI tailwind, as new companies form with Stripe as their first choice for payments. They’re well placed in billing, stablecoins, and even cards.

🧠 It’s hard to tell what’s real here from just the reports, but here’s my guess. I can imagine some mega deal where the PayPal assets are being put up for bid, with Block and Stripe taking bits that are sensible but not a heavy lift (Braintree processing for Stripe, BNPL for Block). There’s also ancillary bits like Honey and Xoom (the remittance business) that could be interesting to others.

2. BlackRock, JPMorgan, Goldman and the NYSE live traded tokenized stocks and US Treasuries through DTCC

Almost 40 firms just live traded tokenized stocks and US Treasuries through DTCC, per the WSJ. JPMorgan turned shares of the QQQ ETF into tokens, then turned them back, proving you can tokenize and untokenize in production. Vanguard and Invesco, the giant funds that make those ETFs, took part too.

If you own a US share, DTCC is the plumbing that records it. The Depository Trust & Clearing Corporation custodies $114 trillion in US securities and processed $4.7 quadrillion in transactions last year. This is the biggest change to that plumbing since the 1970s, when Wall Street swapped paper share certificates for electronic records.

The opening basket was Microsoft shares, Circle stock, QQQ, SPY, SGOV, plus Treasuries across multiple maturities. Trades settled on Hyperledger Besu and Canton, under a three-year SEC no-action letter, with the full launch in October.

🧠 This is live in October! That’s insanely fast. The market is going to shift in a way we can’t yet imagine in time.

🧠 I cannot understate how massive this is. Every US stock and government bond would be a token. Fully redeemable for the original ownership obligation at the DTCC. Tokens are 24/7 by default, and global. This opens new markets, functionality and economics for anyone who touches those asset classes.

🧠 Not all tokens are created equal. There are three types of tokenized security.

(1) Wrappers track the price of a share held in a fund (SPV) under someone else's name but promised to you.

(2) Digital twins mirror a real share onchain while a transfer agent keeps the legal register (what the DTCC did). (3) Natively issued securities, where the onchain record is the security itself (what Figure has done with its own stock).

The DTCC twin matters because the DTCC is the register itself. When the record keeper issues the token, the dividends, the votes and the legal ownership travel with it.

🧠 The SpaceX IPO already showed what can happen when the token is a wrapper. Binance Wallet, Bybit and Bitget Wallet refunded customers after xStocks, Kraken's tokenized equities business, gathered more than $1bn in orders and couldn't fill most of them at allocation. Nearly every wrapper also routes through one broker-dealer, Alpaca, which turned out to be a single point of failure for the whole market. We broke this down on Tokenized.

🧠The fact that there could be many tokens for the same stock is ripe for scams and rug pulls. As tokens go onto public chains, the linkage back to the real ownership is going to become more and more important. The industry will need to develop protections against this.

🧠 This is about market plumbing efficiency for the DTCC. DTCC's CEO told the Wall Street Journal the focus is on ways to "free up trapped liquidity", and the day-one workflows are collateral transfers, repo, and margin. JPMorgan already posted tokenized assets to meet a margin call at CME. Moving collateral that currently sits stuck in settlement windows is gloriously boring, and exactly where the money is.

🧠 I'm so here for crypto's boring era. It seems every crypto fund is now LARPing at TradFi and trying to figure it out. It is unquestionably where the action and volume are (and always was). What’s interesting is that’s increasingly going on public blockchains. And because they’re public, what starts as small and closed-loop can in time become more open-loop. And that’s a whole other order of magnitude of potential.

🧠 You buy stocks with cash. Nobody has figured out the cash onchain yet. The DTCC is launching with Besu and Canton. Besu is used by some banks for tokenized deposits, and SWIFT for its ledger. But it’s an aging, poorly supported private chain. Canton is ideal for capital markets assets because it offers privacy, but less so for cash, as it is slow and expensive compared to payment systems. I’m spending a lot of time on this topic in my day job (so I’m probably a little biased). But I really think it’s the bottleneck we have to solve.

3. Kalshi launches GPU compute forward curves derived from prediction markets

Kalshi went live on July 14 with forward curves for the hourly rental prices of Nvidia's B200, H200, and A100 GPUs. A forward curve is a chart of what the market expects something to cost next week, next month, and further out. Fixing your energy tariff or your mortgage rate works the same way. You pay a known price now instead of an unknown price later.

Why would anyone want that for GPUs? A data center wants to lock in what its chips will earn next quarter. An AI company wants to cap what training will cost. Both sides need an agreed number for the future price, and compute has never had one.

Kalshi's curves come from its prediction markets, where people trade contracts on what chip rental prices will be each week and month. The curve isn’t tradeable. It is the reference number two firms write into a private contract, and anyone who wants to lock in an actual price trades the underlying markets on Kalshi's exchange. As CEO Tarek Mansour puts it, "Compute is the new oil. Like every commodity before it, it needs a real derivatives market."

Kalshi is not alone; CME Group said in May it would launch compute futures cash-settled against Silicon Data's daily GPU rental indices, pending regulatory review, and ICE is working on GPU capacity futures with Ornn. Kalshi got to market first because it needs no external index. Event contracts on chip prices already trade, and the curve is derived from them.

🧠 Prediction Markets are both controversial and fascinating, and we’re seeing this in real time. Venues where genuinely interesting markets and derivatives are being generated, whilst also fighting a dozen states over whether their biggest offering by volume, sports contracts, is gambling.

🧠 Ai’s cost explosion (still) needs a fintech solution. Compute is the next great financial infrastructure buildout, and that price discovery comes first. Larry Fink told Milken a new asset class will be buying futures of compute. Five days later, CME and Silicon Data announced their own compute futures market.

🧠 Self-certification is the speed advantage. The CFTC, the US regulator for derivatives, lets a licensed exchange 'self-certify' a new contract, meaning list it first without waiting for approval. CME chose a formal review for its compute futures, due later this year. Kalshi self-certified, the same mechanism that let it launch perpetual futures after Chairman Michael Selig loosened the delivery-date definition (see Good Reads below).

🧠 Read the mechanics closely. Kalshi publishes the price for free. Anyone who wants to lock that price in has to trade on Kalshi's exchange, and Kalshi earns fees on every trade. Owning a reference price is one of the best businesses in financial markets. Oil is priced against Brent, loans were priced against Libor for decades, and whoever owns the reference price earns fees on the trading around it. CME's 66% operating margins, covered in the Good Read below, are what benchmark ownership pays.

🧠 GPU prices already move like a commodity. H100 rentals went from above $8 an hour in 2023 to around $3 by late 2025, and by early 2026 Silicon Data showed zero on-demand availability across 90% of providers. A market that swings from glut to scarcity inside a year is the textbook case for futures contracts.

🧠 The customer that matters most is the lender. Meta, Microsoft, Alphabet and Amazon plan $700bn in AI capex and are increasingly funding it with debt, like Meta's $27bn Blue Owl data center deal. Nobody can underwrite GPU-backed loans without a reference rate for what a GPU earns. Power plants borrow cheaply because electricity has forward curves. A credible compute curve does the same job for data centers.

🧠 Liquidity decides whether this becomes a benchmark. Kalshi did $33bn in volume in June, roughly 87% of it sports. A benchmark requires thick markets with ample liquidity. Kalshi projects compute futures will one day out-trade oil's 800 million contracts a year. That only happens if data centers and labs hedge on it routinely.

🧠 This offering helps the legal story too. Kalshi spent last week losing a preliminary injunction in the SDNY over whether its sports contracts are gambling. Forward curves for data center capacity look more like the CME than a sportsbook, and every capital markets adjacent product Kalshi ships strengthens its argument that this is an exchange.

4. Anthropic, Blackstone and Goldman launch Ode, a $1.5bn AI services firm

Anthropic, Blackstone, Hellman & Friedman and Goldman Sachs formally launched Ode this week, a $1.5bn joint venture that sells AI implementation. Forward-deployed engineers (FDEs) embed inside a company, map how it runs, and build new AI systems on top. The wider consortium includes Apollo, General Atlantic, Leonard Green, GIC and Sequoia.

Ode is launching with 100 engineers, over half of them former founders, from Fractional AI, the applied AI startup the venture acquired. The pipeline starts with the sponsors' own portfolio companies, and the firm is Claude-first but “will use rival models where a customer needs them.” A services firm is an odd thing for a frontier lab to build on paper. But every enterprise wants AI but most cannot implement it. And the data is telling us that the businesses that accelerate are leveraging AI.

🧠 The bet is that implementation is a big business in AI. Blackstone and Goldman ran the numbers before writing $1.5bn of cheques. As Ode's chief technologist Eddie Siegel puts it, "Model selection matters, but it's not where the majority of calories are spent."

🧠 This is the Palantir playbook bolted onto private equity distribution. Palantir normalised the forward deployed engineer. Ode adds a captive pipeline, because the sponsors can point it at hundreds of portfolio companies that need margin improvement before exit. Distribution is usually the hard part of consulting.

🧠 Anthropic bought OpenAI's services partner. Fractional AI had spent 11 months partnered with OpenAI before Ode acquired it. OpenAI runs its own version, The Deployment Company, and Accenture and Deloitte have forward-deployed teams too.

🧠 The mid-market for FDEs is wide open. Ode's executives admit elite applied AI engineers are scarce, and a firm this size can only ever serve a few hundred clients. There are hundreds of thousands of US companies doing $2m to $50m a year, drowning in manual work across 20 disconnected tools, with full access to the best AI ever built and no idea how to apply it. Ode will never call them. Somebody will.

🧠 But will CEO’s be concerned regardless? Owning the implementation company is a helpful hedge against lock-in, but when that company is part-owned by a lab, have you not just locked yourself in again?

🧠 Regulated companies now have a fourth option. I argued in the Enterprise AI Operating Model Playbook that the harness, the control plane and your ability to build evals separate the AI haves from the have-nots, and that most companies cannot build that alone. Until now the choices were going all in with one lab (BBVA with OpenAI), building your own harness (JP Morgan, Ramp), or working with startups like Primitive. Ode is the buy-the-transformation option. The priciest engagements in regulated industries, where implementation is hardest.

5. Visa will now let banks mint and burn digital dollars

Visa announced the Visa Stablecoin Platform (VSP) this week, a single Visa-managed environment where FIs, fintechs, and crypto natives can mint, burn, hold, and transfer stablecoins, starting with Open USD (OUSD). It ships with a new "Wallet-as-a-Service" offering, dual approvals (the bank "four eyes" principle), allowlists to control where coins can be transferred, and full audit logging. A bank links a bank account, sets its approval policies, and can mint or redeem OUSD as part of treasury, settlement, and liquidity workflows. It's in beta with select clients now. Open USD, announced two weeks ago, is the Open Standard consortium stablecoin with 140+ partners including Visa and Mastercard, slated to go live later this year.

🧠 Visa is selling better settlement. Today, an issuer or acquirer settles its net position with Visa over bank wires, and those wires come with everything wires come with: cutoff times (closed on weekends and out of hours) and slow correspondent banks on the cross-border legs. Risk builds up over a weekend, and acquiring banks, merchants, and issuers can all be out of pocket waiting for those delays to catch up.

🧠 Mint OUSD and the wire problem goes away. A bank mints OUSD from a linked bank account inside its existing Visa relationship, and that same settlement becomes an onchain transfer. 24/7/365, no correspondent chain in the middle.

🧠 There's a yield benefit for banks too. With the incumbent stablecoins, the issuer keeps the reserve yield on your float. Open USD hands most of it back to the partners who mint and distribute it. The settlement balances a bank holds start earning for the bank.

🧠 Visa is commoditizing the coin. It had already enabled this via USDC, running stablecoin settlement with crypto-native acquirers Worldpay and Nuvei since 2023. VSP packages the coin and the minting process for any FI on the network, with the controls a bank needs to get it past risk and compliance.

🧠 Scheme settlement is a closed loop. Known counterparties, defined flows, net positions. That's where I expect stablecoins to hit real volume inside banks first, and the tooling now ships inside a contract every issuer and acquirer already holds. The open question is how banks will treat OUSD balances on the balance sheet (cash equivalent? something else?), because the accounting and capital treatment will decide how big the float gets.

1. Ground - Helping Fintech companies access onchain yield.

Ground provides a simple, clean set of APIs to allow fintech companies, wealth managers or banks offer onchain yield sources like Aave or Morpho vaults. The company aims to solve the complexity of integrating directly onchain, and helps with risk management, revenue generation and allows companies to offer their clients yield on idle cash (without becoming a bank for example).

🧠 Every fintech company that added stablecoins is now asking about yield. There are clean APIs to add stablecoin capability (almost too many), but yield is still a complex beast. Partly because you have to understand the products, the tradeoffs, and think about how to present those to customers. And partly because integration is hard. Ground aims to streamline all of it.

2. Daya - Mobile Treasury Management for African Businesses

Daya helps companies track cashflow, manage pay ins and payouts from a single mobile interface. They also have a payments infrastructure for cross-border payments via stablecoins or fiat rails. Users can add virtual accounts, multiple currencies and integrate with any POS, banking or other internal infrastructure with simple clean APIs.

🧠 Even corporate treasury can be mobile native. Much of Africa runs on infra-Africa and global import/export. Handling multiple currencies and cross border payments can be complex and slow with just fiat rails and bank accounts. Making this mobile-first, and low-lift is a natural fit for a generation of entrepreneurs who grew up mobile-first.

3. Kalipso - The compliance policy updater AI for Europe

Kailpso scans for new EU regulations, identifies what applies to your organization, gaps in your documentation and proposes ready to implement fixes automatically. This includes the EU AI Act, GDPR, Mifid, MiCAR, DORA, and then all the individual state level regulations.

🧠Updating policies is a cottage industry and a very expensive one. There’s a hidden iceberg of compliance, where whole armies of people fight to keep policies up to date, update training, and keep track of it all. This desperately needed AI assistance, especially in Europe.

4. Onyx Predictions - The sports prediction market

Onyx provides users with access to trade sports outcomes with exchange traded instruments. This more open market way to participate in sports outcomes.

🧠 I don’t think Onyx knows what it wants to be. It’s leaning into sports and parlay like bets, and even has a “responsible gaming” page on its site. But is a registered broker with the CFTC partnering with NinjaTrader for clearing. It turns out Onyx Odds launched using a sweepstakes and Daily Fantasy Sports (DFS) model to operate across more than 35 US states. It leverages virtual tokens "Onyx Coins" for free social play and "Onyx Cash" for premium, redeemable prize play. Then later, has pivoted to prediction markets. Kraken who provides its clearing, also invested and aims to use crypto as the backdrop. So. Go for sports. Build your own polymarket. Do it on Kraken infrastructure.

Duffy’s last dance - The CME vs Kalshi

Terry Duffy took over the CME in 2002 leading a transformation that turned it into the powerhouse it is today. As Marc puts it “it transacts 28.1 million contracts a day, up from 2.2 million contracts in 2002, earning a fee of around $0.65 on each contract. Margins, which have expanded from 34% to 66% under Duffy, rank among the highest in the S&P 500. Its market cap – $115 billion at its peak a few months ago – has risen 8,000% since the IPO.”

“Under US law, a futures contract is defined as an agreement for the purchase or sale of a commodity for delivery at a fixed future date” However the CFTC Chairman, Michael Selig recently changed that, with Kalshi then self certifying a dozen or more perpetual futures contracts. Derivatives with no fixed delivery date.

🧠Hyperliquid is reshaping markets. Oil futures on the offshore Hyperliquid platform spiked from $21m per day, to $3.7bn during the early phases of the strikes on Iran.

🧠The precedent is now set. It’s going to be much harder to unwind perps now they’re live than prevent them in the first place.

🧠This pressures the business model. CME makes money on fees on a delivery date. Being able to settle whenever puts a downward pressure on those fees.

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