A founder sent me the Cari headline on Saturday, September 5, with three exclamation marks and a question about whether it validated his own payments thesis. I read the number first, the way everyone does. $32.5 million, decent size, nothing that moves a market.
Then I read the investor list, and I went back to my desk.
First Horizon. Huntington. KeyBank. M&T. Old National. SouthState. Glacier. Seven US banks, and not one venture fund. Cari is building a bank-governed network where chartered institutions issue and move tokenized commercial bank deposits, and the banks that will use it paid for it. Another thirty have joined the network already, with roughly forty more in conversation. That's not a funding round in the sense my industry uses the word. It's a consortium buying a utility, and it looks nothing like the six other deals announced that same week.
What I've Been Seeing
The week of August 31 to September 5 produced $292 million across ten crypto deals, and almost all of it pointed at the same seam: connecting digital dollars to the banking system that already holds everyone's money. Félix took $200 million for cross-border remittances, split between $87 million of a16z equity and a $113 million credit facility. At OpenReserve it was a $25 million seed, also led by a16z crypto. Cari took its money from the customers.
I've watched infrastructure categories mature for twenty years, and the investor mix is usually the earliest honest signal. When venture funds are the only buyers, the product is still a bet. When the eventual users start writing the checks, someone inside those institutions has already run the build-versus-buy analysis and lost the argument for build. Seven risk committees at seven mid-size banks reached the same conclusion in the same quarter. That's a lot of independent confirmation for a category most crypto people stopped tracking in 2024.
Look at which banks showed up, though, because that's the second signal. None of these are money-center institutions with innovation labs and a spare hundred engineers. First Horizon, Old National, Glacier — these are regional and community-adjacent lenders whose entire competitive position rests on deposits staying put. For most of the last four years the honest reading of a regional bank's crypto strategy was that it didn't have one and didn't need one. Something changed the arithmetic. My read is that once the largest banks demonstrated deposit tokens working in production, staying off the network stopped looking cautious and started looking like a slow deposit leak to whoever was on it.
It also isn't happening in isolation. JPMorgan's deposit token went into institutional rollout on Base in late 2025, and its Kinexys platform now clears more than $7 billion a day with over $3 trillion processed since launch. JPMorgan, Citi, Bank of America and Wells Fargo are separately working through The Clearing House on a shared network aimed at the first half of 2027. The largest banks in the country are building the top of this stack. Cari is building the part that lets a $40 billion regional bank in Tennessee participate without hiring a blockchain team.
The Bridge Thesis Applied
A stablecoin and a tokenized deposit look identical on a wallet screen and behave like different species underneath. A stablecoin is a claim on a reserve pool held by a non-bank issuer, transferable to anyone. A tokenized deposit is your existing bank balance with a blockchain wrapper on it, transferable between vetted clients, and — this is the part that matters to a bank CFO — still sitting on the balance sheet where it can fund lending. A February 2026 New York Fed staff report drew the line the same way: payment money on one side, bank money on the other.
So the banks aren't losing an argument to stablecoins. They're declining to have it. Cari's network is permissioned, anchored to Ethereum, and governed by the institutions on it. Every design choice says the same thing: we want the settlement speed, we don't want the deposit flight.
Which leaves the actual bridge unbuilt, and I think this is where the next several hundred million dollars of enterprise value ends up. There will be a permissioned layer holding regulated bank money and an open layer holding stablecoins and tokenized funds, and value will need to cross between them constantly — a treasury sweeping idle balances, a merchant paid in USDC settling into a deposit account, an asset manager funding a tokenized T-bill position before noon. Nobody owns that crossing yet. I wrote in Bridge Notes #15 that banning yield inside a stablecoin only relocates the yield, and the relocation traffic runs across exactly this seam. In our portfolio at PRIM3, Kima Network has spent years on settlement between systems that were never designed to talk, which is the unglamorous half of this problem and the half that decides who wins it.
If you remember one line from this piece: the rail is being financed by the people who will use it, which means the open question is no longer whether bank money comes on-chain but who gets paid when it has to leave.
Worth saying plainly what the bridge business actually consists of, because "interoperability" is a word founders use to avoid describing a product. Three things have to work at the boundary. Identity has to travel, so a permissioned network can accept that a counterparty on the open side has been screened to a standard its regulators recognise. On both sides, liquidity needs pre-positioning, because no bank accepts settlement risk on a rail it doesn't control. And someone has to carry the compliance record across, in a form an examiner will read without a translator. Whoever solves the third one has a moat, since the first two are engineering and the third is trust that takes years to earn.
Where the Analogy Breaks
I should be honest about the limits of my own framing here, because consortium infrastructure has a long and mostly disappointing record. R3, Utility Settlement Coin, half a dozen trade-finance networks — bank-owned rails have historically moved at the speed of the slowest legal department, and thirty signed banks is not thirty banks in production. Cari has raised a first tranche and onboarded nobody at scale yet.
The other thing I can't tell you is whether the two layers stay separate long enough for a bridge business to be worth building. If The Clearing House network arrives in 2027 with a native stablecoin on-ramp attached, the crossing gets absorbed into the incumbent stack and the independent bridge becomes a feature. I'd put that at maybe one chance in three. But two years ago I'd have said the regionals would never fund their own chain, and last week seven of them did.
What Founders Should Do With This
Stop pitching banks a replacement and start pitching them a connection. The single most useful reframe I can offer a payments founder right now is that the institutional decision has already been made in favour of bank-issued tokens, and the remaining commercial space is in translation, compliance, and liquidity between the permissioned and open worlds. That's a smaller-sounding story that closes far bigger contracts.
Then look hard at who funds you. A round from your customers prices lower and takes longer than a round from a crypto fund, and it comes with distribution attached that no term sheet from a fund can replicate. I've backed 60-plus projects and the ones that compounded through more than one cycle almost always had a strategic investor who was contractually and reputationally stuck with them. Cari's seven banks are stuck with Cari now. That's worth more than the $32.5 million.
And if you're building anything that touches settlement, assume that by 2028 the dollar leg of your transaction is at least as likely to be a bank deposit token as a stablecoin, and design for both.
One more thing, on timing. Founders keep asking me whether it's too early to sell into this, and my answer for the last few months has been that the window is open precisely because the networks aren't finished. A consortium that has signed thirty banks and integrated almost none of them has an eighteen-month gap between commitment and production, which is exactly when it buys rather than builds. Show up in 2028 with a beautiful bridge and you'll be quoting against an incumbent vendor already embedded in the compliance stack. Show up this quarter with something rough that solves one boundary problem for one bank, and you're the incumbent.
The bridges that matter in this cycle aren't between chains. They're between a system that has to stay regulated and a system that has to stay open, and I don't think either side gets to win outright.
If you're working on that crossing, I'd genuinely like to talk.