You Can Pool a Cost, But Can Pool a Business?

BankChain's release describes a settlement utility and a stablecoin issuer in one sentence...

Aug 30, 2026

The BankChain Alliance press release names 3,283 banks and $21.8 trillion of assets. So far, none of those banks has agreed to anything.

That ain’t suppose to be a gotcha, I think it’s more about the structure. So on Tuesday 25th August, 39 state bankers associations announced an industry owned blockchain network for their members. Kathy Kraninger, who ran the CFPB from 2018 to 2021 and now leads the Florida Bankers Association, is interim chair, and the target is 2027. The members of the alliance are trade associations. The $21.8 trillion is the combined balance sheet of those associations membership on FDIC call report data to 31st March, not a book of signed commitments. An association can convene its members, but cannot exactly commit their balance sheets, and no individual bank has been named as of yet.

Its obviously early days as no technology partner has been chosen either, and it sounds like the first phase of the RFP has only just closed. American Banker reports the group models its governance on the Federal Home Loan Banks, “without the controversies,” though the release itself doesn’t mention them, and neither the vote weighting nor the capital contributions have been published. Nine of the ten board seats belong to association chief executives. The tenth belongs to Kim Askwith, the founder of TekFactor, and I can’t find a source that explains what a technology executive’s seat is doing on the board of a body that is running a vendor selection. Eleven states are absent, and three of them, California, New York and Illinois, are some of the largest banking states in the country.

So the number in the headline is a market size, but what I think is interesting instead is the sentence describing what the network will do, because to me, it contains two businesses that end up owned by different people, and I wonder which one this ends up being.

What five consortia have actually committed so far

BankChain is the fourth bank led tokenized deposit network announced in the United States since April, and the fifth this year.

Cari, Eugene Ludwig’s network, surfaced first, in February. It picked ZKsync’s Prividium in March, shipped an MVP, and has a pilot running with design partners including Huntington, M&T, KeyCorp, First Horizon and Old National. It counts more than thirty committed banks across its network and pipeline, over $10 trillion in holding company assets, a custody partner in Fireblocks, and a route to the long tail through SouthState’s correspondent relationships with 1,300 institutions. In July it joined the American Bankers Association’s Premier Partner Network. Of all of them it feels like it has something, and it has done it with midcaps rather than money center banks.

FIS announced Project Keystone on 30th April with six institutions, including Citizens, Fifth Third, Huntington, KeyBank and M&T. Bank administered and built by the vendor, but no chain disclosed.

The Clearing House came on 5th June with seventeen named institutions and none of them hedging, including JPMorgan, Bank of America, Citi, Wells Fargo, US Bank, PNC, Truist and HSBC. TCH will operate the network itself and bridge it to CHIPS and RTP, which is to say it plugs the new thing into the rails it already runs. It has no chain, no announced vendor so far, and a first half 2027 target. The most committed institutions in the market, and the least actually built to date.

Hazel arrived on 18th June with Custodia Bank and Vantage Bank, two institutions, a white paper, and running code on Ethereum mainnet since March. The token is a tokenized deposit inside the consortium and a stablecoin outside it, which is a really interesting piece of design. Vantage is the San Antonio bank whose live deployment sits underneath the Texas Bankers Association’s technology program, and in February they opened pilot slots and early adopter pricing to the association’s roughly 400 member banks. That program is where press reporting says BankChain started. The bank at the root of BankChain’s origin story has been live with a network since March.

And then BankChain, 3,283 banks that have not agreed to anything yet, no chain, no vendor, no operator, and a date eighteen months out.

I tried lining each up with what is actually committed and I don’t see evidence of any ladder, infact it seems like the various signatures and code almost run in opposite directions across these five. TCH has some big signatures and the least code at the moment. Hazel probably has the most code and a few signatures. So at either end, seventeen named balance sheets have put themselves behind something that does not yet exist, while three thousand banks are represented by bodies that cannot commit them.

The middle is a bit of a scramble. Huntington and KeyBank are in all three of The Clearing House, Keystone and Cari. Citizens and Fifth Third are in two. Nobody in that tier is picking a side, which is the probably the right move when you don’t know which of these survives and the cost of joining is a name on a press release.

Why the associations had to do something

There were 4,278 FDIC insured institutions at the end of March. Of those, 4,124 hold less than $10 billion in assets, and between them they account for about 14% of the industry’s assets. Five hundred and fifty hold less than $100 million each. That smallest cohort earns a return on assets of 0.89% against 1.36% for banks in the $1 to $10 billion range, and 15.27% of them were unprofitable in the first quarter against 4.82% for the industry as a whole.

Now imagine if you were to ask an institution that size to evaluate a distributed ledger, select a custody provider, write the risk policy, satisfy an examiner who has never supervised one, and fund the integration! The people are not the problem here. The fixed cost of the evaluation on its own exceeds any plausible return at that asset size, and the integration is almost the harder bit, because a small bank doesn’t control the system it would have to integrate into. Fiserv serves roughly 42% of US banks, Jack Henry 21% and FIS 9%, together more than 70% of the institutions the Kansas City Fed surveyed. Whatever a community bank decides about tokenized deposits, the decision only becomes real if its core provider ships it.

Of course the volumes underneath are relatively tiny, and that’s the argument for pooling rather than against it. The Cleveland Fed looked at roughly 4,500 heartland institutions on Fedwire and found about 3,500 of them sent or received a cross border payment in a single month last year, most of them fewer than a hundred in the whole month. Three transfers a day are not exactly going to amortize a blockchain, a custody stack and a core integration at one bank. Across three thousand banks it might.

Pooling to buy what you can’t afford alone is one of the oldest moves in American banking. The Bank Service Company Act of 1962 exists for it. Post war check volume was rising faster than clerks could sort it, small and medium banks couldn’t afford computers individually, and Congress authorized them to pool and buy together, in exchange giving the banking agencies examination authority over whatever service companies the banks created. 39 associations forming a buying club in 2026 is pretty much that statute’s direct descendant, and on the constraint alone they’re doing the right thing.

Two businesses in one sentence

The thing about the press release that is interesting to me is that BankChain says the network will support

“smart payment tools, tokenized deposits, stablecoins, automated settlement, and other innovations.”

Those are not variations on a theme. A tokenized deposit is a better way to move a liability the bank already owes, on a rail the bank already needs. It replaces a correspondent leg and a cut off time. There is no new revenue inside it, which is exactly why a cooperative is the right owner. The members fund a shared cost because each of them saves more than their share of it, and nobody outside has a reason to want the thing.

A stablecoin is a different animal, because it has reserves and reserves earn. Under the GENIUS Act, which takes effect on 18 January 2027, the issuer is barred from paying that yield to holders. The float income stays with whoever holds the reserves, and at any scale worth building for that is a large and growing pool of money sitting inside the network. Who receives it, on what basis, and who can change that basis later are governance questions with real dollars attached, and they’re really not the same questions you’d ask about a settlement utility.

I think that distinction, rather than anything about bylaws or blockchain choice, is what determines who might own this network, or one like it, in ten years. A shared cost is stable because nobody else wants it. A revenue pool will attract a buyer, and a cooperative is structurally poor at defending one, because the members it would have to charge are the same members who own it.

I’m not sure any of that makes the stablecoin ambition wrong. It may be the only way the economics work at all, as somebody has to fund a network that is otherwise pure cost. But its pretty clear to me that the two paths need different governance, different capital, and different answers to the question of what happens when someone offers to buy it, and right now the announcement puts them in the same sentence without really distinguishing them.

What happened to the last networks banks owned?

Banks have owned networks before, and the dataset is illustrative, although the pattern in it is not particularly subtle.

Pulse was a Houston cooperative owned by its member institutions, 4,100 banks, credit unions and thrifts by the end, running PIN debit and ATM switching across the South and Midwest. Discover bought it for $311 million in January 2005, and the members voted the deal through. Star was the same shape, 3,500 member institutions across twenty two states, and Concord EFS bought it for stock in 2001, then First Data bought Concord for $6.7 billion in 2003, and the network its members had built now sits inside Fiserv. NYCE was owned by a group of New York banks, and First Data bought c60% of it in 2001 with Citi, Fleet, HSBC and JPMorgan keeping minority stakes, which lasted until the antitrust settlement made First Data sell it on. Cirrus was a bank consortium until MasterCard bought it in 1988. Plus was a bank consortium until Visa bought it in 1993. And Visa itself was a bank owned association for 38 years, until the banks took it public in March 2008 for $17.9 billion, the largest IPO the United States had seen.

It isn’t only an American pattern. Back in my old country, Vocalink ran the UK’s BACS and Faster Payments infrastructure and was owned by the British banks until they sold it to Mastercard for £700 million in 2016. 10 years later Mastercard is reported to be exploring the sale of a controlling stake back to a bank backed vehicle for around £400 million, which tells you the banks have concluded they should own it after all, and what it cost them to find out.

Now if you contrast with the ones that stayed bank owned. CHIPS has been bank owned since 1970. Zelle sits inside Early Warning, owned by seven banks, and nobody has bid for it. The difference isn’t that those owners were more principled. CHIPS moves a cost. Zelle moves a cost. Pulse, Star and NYCE earned switch fees on every transaction, Visa earned interchange, and Vocalink earned a processing fee on every salary in Britain. It feels like every bank owned network with a revenue pool inside it was eventually sold, and in each case the sellers were the members, one vote at a time, each holding a sliver of the thing and each offered cash for it. A cooperative with four thousand owners can find it very hard to resist a buyer. It counts the votes and takes the check!

That is what the 3,283 banks are being offered a share of, and it’s why I’m focus on the scope. If BankChain is a settlement utility it will look like CHIPS in 10 years. If it holds stablecoin reserves it will look more like Pulse, and the buyer won’t be Discover.

The alliance’s stated protection is to take an ownership stake in whichever technology partner it selects. That is a serious idea, which really needs scrutiny. It also gets negotiated after the members have joined, with a company that hasn’t been picked, by a board whose voting rules haven’t been published, and it isn’t obvious what a stake in the vendor protects against if the thing being fought over is float income rather than the vendor?

The Bank Service Company Act worked because what the banks pooled was a cost. Nobody really wanted to buy a check sorting cooperative, so the question of who owned it never needed a better answer than “the members.” Discover wanted to buy Pulse. A network that holds stablecoin reserves is a balance sheet earning a spread, and that is the structure the associations have described, whether they meant to or not. The 3,283 banks in the release are why it matters, because they are the ones who will be asked to join before anyone tells them what type of business they are joining.

Chances are that BankChain may end up as one of two things. A settlement utility its members run at cost, which is what the 1962 statute was written for and what the rhetoric describes. Or maybe a reserve pool with an association governance wrapper, which a buyer might want to pay for. The associations can build either. I’m not sure they can build both under one set of bylaws, and until they say which, the number to read in the release probably isn’t $21.8 trillion.