The purchase was a bank in Chico, California: one branch, roughly $300 million in deposits, and a price of about $50 million. In 2021, William Hockey and Annie Robertson Hockey bought Northern California National Bank. Plenty of buyers would have seen a modest banking business. The Hockeys saw a way to stop asking other people to fix the machinery.
- Buy the regulated bank. Build the software inside it.
- Give fintech developers direct controls over accounts, payments, cards, and lending.
- Fund the business with founders’ money and profits.
William had co-founded Plaid, which connects financial applications to bank accounts. He had watched companies build ambitious products on top of institutions whose technology set the pace. In Column’s launch announcement, he identified banks and middleware as the bottleneck. Improving the layer above them could only get a developer so far.
01 / A $50 million way around the queue
Column says work began in 2019; it launched publicly in April 2022. The bank acquisition sits between those dates. That sequence matters. Buying a chartered institution supplied a regulated foundation. Turning it into programmable infrastructure still required a separate engineering effort.
At launch, William led product development and sales; Annie led regulatory, legal, and human resources work. The division put banking obligations beside product decisions from the beginning. William is now CEO. Annie, the co-founder and former co-CEO, remained a director and major shareholder in February 2026 reporting.
A starting cost, not the total cost of building Column.
Here is the founding argument in Column’s own words: “We believe that a bank needs to be a software company.” The practical consequence was ownership of the core and ledger, with direct connections to payment networks. The company making the API could also change what happened behind it.
02 / Follow the money, then count the handoffs
In a conventional arrangement, a fintech might depend on a middleware provider, a sponsor bank, and a separate core processor. Each relationship introduces another system to reconcile and another team to consult. Column puts the bank and much of that software together. Fewer organizational boundaries can make a technical change easier to coordinate.
Bank · API · core · ledger↓Payment networks
Consider account numbers. Column lets developers create multiple numbers pointing to one account, with particular permissions and limits. A platform could use that arrangement to distinguish incoming payments without creating an entirely separate pool of money for every identifier. The interesting feature is the control over the underlying banking object.
The customer list makes the positioning concrete. Brex uses Column beneath its business spending platform. Mercury uses it for personal and business banking. Ramp uses it for cards in certain international markets. Best Egg works with Column on lending. These companies sell different experiences; Column supplies parts of the financial machinery.
Column has company in this market. Lead Bank and Cross River also offer banking infrastructure through APIs. Treasury Prime connects fintechs with a network of partner banks. Column’s pitch rests on how much it owns and builds internally; a buyer still has to compare coverage, controls, and the particular banking relationship.
Developers can combine accounts with ACH, wires, checks, and transfers between Column accounts. Lenders can add origination and capital: warehouse lines finance loans before sale, forward-flow agreements establish purchase terms, and retention lets Column keep originated loans. The product reaches into the balance sheet as well as the code.
03 / The weekend became part of the product
In May 2025, Column made its real-time payments product generally available across RTP and FedNow. These networks can move available dollars at night and on weekends between participating institutions. A beautiful payment interface becomes more useful when its underlying rail is awake.
Then, in September 2026, Column announced four additions: stablecoins, its own issuer processor, global banking, and multicurrency accounts with local payouts. They extend the same idea into places where different vendors and settlement schedules have traditionally met.
The stablecoin product connects USDC and USDT with Column’s banking ledger. Column describes receiving stablecoins on a Saturday and paying dollars out through RTP, avoiding a wait for Monday’s wire window. Its card processor brings authorization decisions into the same stack as issuing and capital. Customers can still bring outside processing or funding.
Global banking adds accounts and cards for verified international customers. Multicurrency accounts let businesses hold foreign currency rather than convert every receipt into dollars and back again. The attraction is practical: fewer reconciliations, fewer funding arrangements, and fewer calendars governing a customer’s money.
04 / The small failure worth stealing
The bank purchase is difficult to copy. A smaller engineering decision is much more portable.
Column’s early API documentation occupied one scrolling page. As the API grew, that arrangement became awkward. More seriously, handwritten documentation lived separately from the OpenAPI specification. Two descriptions of the same product could disagree.
In June 2026, Column described its repair: generate the reference from the specification, inside the same repository. An endpoint and its documentation now arrive in one pull request and ship together. Each page also has a clean Markdown version for developers using language models.
An endpoint and its instructions now leave the building together.
That is a lesson a reader can use without a bank charter: find the duplicated truth, then remove the duplication. The docs rebuild follows the same logic as the business. Put responsibility where someone can actually change the thing.
05 / Fewer owners. More responsibility.
Column earns transaction fees and banking income, including loan origination fees and interest. It says founders and employees own the company and that profits help fund it. Its employee program includes annual opportunities to sell vested appreciation-based awards back to the company.
That independence comes with conditions. The founders could finance an expensive acquisition; most founders cannot. Employee liquidity needs cash. A direct bank relationship still requires onboarding, compliance work, and operational planning. Instant payments depend on network reach and are final. Concentrating infrastructure also makes partner selection consequential.
Column’s useful question is simple enough to borrow: which dependency keeps deciding what your product can do? Its answer required buying a bank. Another company’s answer may fit inside one pull request.