Happy Wednesday, Fintech Listeners!

Here’s a fact that I did not know (but that my brilliant wife did know): The development of standardized time zones (or “clock lands” as Jason Mendoza calls them) started in England in the 1840s, and it started because the growth of steam-powered locomotives, which made local mean times (i.e., every city and town setting their own clocks by the position of the sun) confusing and increasingly dangerous.

By the 1850s, most towns and cities in England had converted to this standard “railway time,” which was, even then, being referred to as Greenwich Mean Time because the Royal Observatory in Greenwich would determine the time by marking when a star of known location would pass through the aimpoint of a telescope.

In 1884, 25 countries got together at the International Meridian Conference and drew up a global standardized time zone system, with each zone being 15 degrees of longitude wide, corresponding to roughly 1 hour of time. They decided — because being first sometimes really does matter — to make Greenwich the location which the prime meridian would run through, and against which all other time zones would forever be measured.

My apologies for that impromptu history lesson, which you absolutely did not ask for. I was a history nerd before I was a fintech nerd and sometimes I just can’t contain that OG nerd energy.

Anyway, switching into fintech nerd mode!

— Alex 

🏀 P.S. 🏀 — Fintech Takes The Court (my annual 3x3 basketball tournament in Las Vegas) returns! This year’s tournament will be held in the morning on Sunday, Oct. 18.

Whether you're a proficient amateur, a semi-proficient and out-of-shape amateur (like me!), or just want to watch the madness, I'd love to see you there!

Sign up here. All are welcome!

Was this email forwarded to you?

Subscribe Here

Payments People, This Is for You

At some point, as payment volume grows, a platform stops being a tech company and starts being treated like a financial institution.

Regulators start asking different questions.

Most teams don't notice the shift until they're already behind on it.

WEX has spent years running the regulated side of payments so platforms don't have to learn it the hard way.

On August 19th, I'm talking with Ron Griswold, WEX's Director of Market Development, about exactly where that line sits — and what it takes to get ahead of it instead of catching up to it.

If this sounds familiar, join us here.

Register today

3 BIG IDEAS FROM THE PODCAST

Simon Taylor is back from his vacation to Italy, tanned and armed with strong opinions about Lake Garda and traveling with children.

This also means that your cohosts are back with a new episode of Not Fintech Investment Advice, where Simon and I do what we always do: Find four companies that made us think harder than we expected to, and definitely don't give investment advice on any of them.

This week, we have an agentic payments company giving bots their own bank accounts, a startup financing AI data centers in people's homes as if they're rooftop solar, a home savings rewards platform that runs the inverse of Bilt's model, and a stablecoin platform solving an unglamorous B2B problem for importers and exporters in Bolivia, among other things.

Tune in for the full conversation here

And read below for my three big ideas ...

#1: Agentic Payments’ Jurisdiction Problem

Natural, the agentic payments startup, looks — at least on a quick read of its terms — like it's setting up to disclaim responsibility when an agent transacting on its rails does something unexpected.

For commercial accounts, that's a reasonable choice. Nobody signs up to underwrite that kind of risk before they have to. (Consumer accounts are a different story. Reg E applies no matter what the contract says, and every one of these disclaimers is qualified "subject to applicable law.")

What's striking isn't the disclaimer itself. That language is fairly standard. It’s that it was written for a world where the sender was a person typing an account number and the worst case was a typo. Natural is taking that inherited framework and applying it to a probabilistic technology.

Long term, it's worth considering how sustainable that position is.

Think about credit cards. The merchant discount rate hovers around 2-3% on average, but most of that goes from the merchants to the issuing and acquiring banks. The networks' slice of that merchant fee is about 13 basis points. The banks get paid for the risk they take (and the commerce behavior they encourage). Visa and Mastercard get paid for developing and enforcing the rulebook for how that risk is apportioned.

When you layer AI agents on top of existing payment rails, you aren’t just adding new infrastructure into the mix. You’re introducing new risks and new points of failure. That means two roles need filling, and they aren't the same role. Someone has to absorb the losses when an agent gets it wrong, and get paid for taking that risk. Someone else has to write and enforce the rules that decide who absorbs what.

To be honest, it’s not really clear to me what underlying payment rail(s) these new rules and adjudication processes will be layered on top of. 

Simon shared something interesting. A large payment service provider told him their fastest-growing rail right now isn't stablecoins. It's RTP, the real-time bank-to-bank payment rail from The Clearing House, finally taking off (alongside FedNow).

Here's the catch. Instant bank-to-bank payment rails make the accountability problem harder, not easier. Visa's dispute system works because card payments are reversible by design. RTP and FedNow are the opposite. They are final the second they clear. So, building a Visa-style accountability layer on instant payment rails (bank-to-bank, stablecoins, etc.) means manufacturing reversibility.

Interestingly, Natural is already reaching in this direction; outbound payments sit in a hold until the payee accepts, auto-cancel after 30 days, and the sender can pull them back before acceptance. That's essentially a private escrow service bolted onto a payment rail, which is a smart direction to build. But escrow solves the mechanical problem, and that’s not the hardest problem.

Look at what RTP already has. The network supports a Request for Return of Funds message for exactly this situation — a payment sent in error or obtained by fraud. The sending bank can ask. The receiving bank is under no obligation to say yes. The plumbing for reversal exists. The obligation doesn't.

That's the difference between a mechanism and a rulebook. Visa's rulebook works because Visa can make it stick. It binds every participant as a condition of access, it enforces them through settlement rather than the courts, and it can fine or expel anyone who ignores the outcome. Natural has a contract and a balance it happens to be holding. That's enough to adjudicate between two of its own customers. It isn't enough to bind anybody outside its own walls.

So the real gap in agentic payments — at scale — isn't reversibility. It's jurisdiction.

I don't know who fills that role. The card networks have the enforcement apparatus, but their authority runs over card rails. The Clearing House and the Fed have the membership structure, but they've treated finality as a feature rather than a bug. A startup can write whatever rules it likes but has no way to make anyone else follow them.

#2: The Evolving Market for Compute

One of Simon's picks this week was only loosely related to fintech, but it was fascinating: An AI data center … in someone's house.

Sky Fusion puts a liquid-cooled AI accelerator in your home alongside rooftop solar and battery storage, then recovers the waste heat for domestic hot water where the plumbing allows. The financing skips the bank loan. Investors fund the hardware through an SPV, and you get a lower energy bill plus recurring payments once Sky Fusion aggregates your node with the neighborhood’s and sells the inference capacity. The Series A target is about 300 homes and 660 accelerators, scaling to a full network of 1,000.

The optimistic case is easy. We can't build data centers fast enough, nobody wants a hyperscale campus next door, and plenty of people would take a panel, a small server, and a check.

Simon asked the question I didn't have a good answer for: How would we know if this is working?

Electricity is easy to answer because we spent a century building the answer. A kilowatt-hour is a kilowatt-hour everywhere. Utilities are obligated to interconnect. And there's a settlement process that runs whether or not the two parties trust each other.

Compute has none of that. There's no standard unit. A GPU-hour means something different across chip generations. There's no obligation to interconnect. And there's no way to verify the work was done. A meter spins whether or not you trust your utility; nothing comparable proves that a node in someone's basement actually ran your inference instead of returning a cached result. CME's compute futures, announced in May, price off datacenter GPU rental benchmarks. That's a reference price for hyperscale capacity, not a measuring stick for 300 basements.

Which reminds me of the solar version of this fight, which I watched from a distance.

Montana's net metering law dates to the late 1990s. It credits your excess electricity generation and rolls it forward month to month, which sounds fine. But it also provides that on your annual settle-up date, whatever credit you haven't used is forfeited to the utility, uncompensated. My cousin was one of the first people I knew with panels on his roof, and every year the surplus he'd banked and hadn't spent went to the power company for nothing.

That's the part I'd watch with residential compute. The rule that decides who captures the surplus gets written early, by whoever bothers to show up, before the surplus is worth anything.

#3: The Anti-Bilt

I brought Quarters to the podcast this week. It's a rewards program for Canadian consumers where the rewards can't buy you anything except a place to live.

Quarters has you link your existing credit and debit cards through open banking, then pays you rewards for qualified spending, including for the redemption of special offers provided by Quarters’ partners. The catch, and the whole design, is redemption. You can only spend what you've earned on a qualified housing activity — a lease deposit, a down payment, moving costs. And when you do, you transact through one of Quarters' partners, and that partner funds the reward. For them it's marketing spend. Quarters makes money on merchant fees for special offers and on fees from the real estate and landlord partners on the other end.

The most interesting part of this to me is the fact that Quarters' rewards aren't a general liability. They can't be cashed out, can't be transferred, and can't be redeemed against anything Quarters has to fund itself. The obligation only ever lands on a partner who's simultaneously acquiring a customer.

Compare that to Bilt, which is the model Quarters is running in reverse. Bilt — which requires customers to use its card rather than linking any card they want — uses housing spend to drive everyday spend and monetizes the interchange. Wells Fargo (Bilt’s original card issuer) projected around 65% of card volume would be non-rent; it came in around 35%. This resulted in massive losses for Wells and, eventually, the end of its co-brand issuing relationship with Bilt.

Quarters inverts the Bilt model. Everyday spend (on any linked card) drives housing spend, and the money comes (primarily) from lead generation rather than interchange.

From a financial nihilism perspective (which is a perspective my brain is permanently trapped in these days), I love this. Nihilism really sets in when consumers conclude that their long-term financial ambitions are permanently out of reach. Owning a home is one of the most common and meaningful long-term financial ambitions. We need to find ways to make it feel more achievable, and a service that helps tie everyday spend to future housing expenses (especially that first down payment) is an important step in that direction.

The challenge, of course, is that rewards on future purchases are not as motivating (for most people) as rewards that can be redeemed today. That’s the advantage that Bilt has over Quarters, and it’s one that I hope Quarters can overcome.

WHAT I'M LISTENING TO

#1: Is Gen Z Financially Nihilistic or Just 25? (Bank Nerd Corner) 🎧

Two of my favorite people talking about one of my favorite topics. Enough said!

#2: The Anthony Bourdain Special Episode (The Watch) 🎧

Nothing to do with fintech, but I am a huge fan of Bourdain (and The Watch) and this was a nice retrospective episode.

Thanks for the read! Let me know what you thought by replying back to this email. 

— Alex