How BaaS Providers Actually Work

You want to launch a product that holds customer deposits. What do you actually need?

A bank charter, an examination cycle with federal regulators, a BSA/AML compliance program, capital reserves, and a connection to the Federal Reserve. That adds up to three to five years and tens of millions in legal costs before you process a single dollar.

The shortcut is Banking-as-a-Service. Rent a charter through an API and go live in weeks instead of years. For early-stage fintechs, the pitch is accurate.

But "BaaS provider" covers at least four distinct architectures, from banks that expose their own APIs directly to middleware platforms routing your requests through a sponsor bank you'll never interact with. The number of layers in that stack determines whose ledger tracks the money and what happens to your customers when one of those layers goes bankrupt.

The three-layer stack

A typical BaaS arrangement puts three layers between a customer's deposit and the Federal Reserve.

At the bottom sits a sponsor bank. This is the entity with the charter, the FDIC insurance, the legal obligation to regulators, and the examination relationship with banking supervisors. Evolve Bank & Trust, Cross River, Choice Financial, and Celtic Bank have all served as sponsor banks for fintech programs.

These banks are typically under $10 billion in assets. That size matters because banks below that threshold are exempt from the Durbin Amendment's interchange caps, which lets the fintech on top earn higher revenue per card swipe.

In the middle sits the middleware platform. This is what most people mean when they say "BaaS provider." It translates the bank's legacy core system into a modern REST API and handles KYC verification, ACH origination, card issuance, and compliance reporting.

Unit and Synctera are the major middleware providers today. Synapse was too, until its collapse in 2024.

On top sits the fintech brand, which owns the customer experience. Your users see your name, while the sponsor bank and middleware stay invisible.

Where the money sits

When a customer deposits $500 into your fintech product, that money goes into a pooled For-Benefit-Of (FBO) account at the sponsor bank, not one with the customer's name on it. A single bank account holds funds for potentially thousands of individual customers.

The middleware maintains a sub-ledger tracking each customer's share of the pool, while the bank's core system sees only one large balance.

This works until the middleware's ledger and the bank's records disagree. A small drift in reconciliation is a bug you can fix. A large one, where the middleware company goes bankrupt and nobody can reconcile the two ledgers, is what happened to Synapse's users in 2024.

Four ways to build this

Not every BaaS provider uses the three-layer model, and the differences in architecture determine where risk concentrates.

Column is a nationally chartered bank (Column National Association, supervised by the OCC) that exposes its own banking infrastructure through APIs. The founders bought Northern California National Bank for roughly $50 million in 2021 and rebuilt it as an API-first institution.

There is no middleware layer. When you integrate with Column, you integrate directly with the bank that holds your deposits, giving you one ledger instead of two. Mercury runs on Column.

Increase connects to FedACH, Fedwire, FedNow, RTP, and Visa. It started as a non-bank, accessing payment networks through partner banks.

In 2025, Increase filed to acquire Twin City Bancorp through the Federal Reserve and began operating as a chartered bank with a core built from scratch. Where the middleware model abstracts payment networks into a simplified API, Increase exposes the raw protocols. If you want to understand what a FedACH file actually looks like, Increase publishes the documentation.

Unit operates the full-stack middleware model, sitting between FDIC-member sponsor banks and fintechs to provide accounts, cards, payments, and lending through a single API. Unit partners with at least four banks, which gives fintechs options for deposit insurance coverage and risk appetite matching. The bank holds the deposits while Unit maintains the sub-ledger.

Synctera emphasizes compliance as its core product rather than treating it as a feature. It matches fintechs with sponsor banks and runs the ledger and reconciliation while providing real-time transaction monitoring through its acquisition of compliance fintech Cable. Synctera works with eight sponsor banks and gives each bank direct visibility into the fintech programs running on its charter.

There was a fifth player. Treasury Prime tried connecting fintechs to a multi-bank network, but cut half its staff in February 2024 and pivoted to selling software directly to banks.

What Synapse showed

Synapse Financial Technologies filed Chapter 11 in April 2024. It was the middleware layer for fintechs like Yotta and Juno, routing customer funds to partner banks including Evolve and Lineage Bank.

The court-appointed trustee, Jelena McWilliams (a former FDIC chair), found an $85 million gap between what partner banks held and what customers were owed. Some estimates put the shortfall between $65 million and $95 million across roughly $265 million in customer deposits.

Synapse maintained the sub-ledger that mapped individual customer balances to the pooled FBO accounts at partner banks. When Synapse collapsed, that mapping collapsed with it. The banks held the money, but they didn't hold the records showing which customers owned which portion of the pool.

Millions of users had accounts frozen. Many received partial recovery or nothing.

FDIC insurance didn't cover the gap. The guarantee applies to individually identifiable deposits at a failed bank, and the banks in this case stayed open. The middleware company between them went bankrupt, and with it went the records linking customers to their money.

The regulatory response

The Federal Reserve and the Arkansas State Bank Department issued a consent order against Evolve Bank & Trust in June 2024, requiring board approval for any new or expanded fintech partnerships. Blue Ridge Bank received an OCC consent order in January 2024 for BSA/AML failures tied to its fintech programs and wound down its BaaS business. Lineage Bank received an FDIC consent order in early 2024.

The FDIC proposed a rule in September 2024 (RIN 3064-AG07) requiring banks to maintain individual beneficial ownership records for any FBO account with transactional features. The proposed requirements include daily reconciliation of individual balances against the FBO total and independent validation when a third party maintains the records. The comment period closed in December 2024.

Outsourcing the customer-facing product to a fintech and the ledger to a middleware vendor leaves the regulatory obligation on the charter holder.

When renting stops making sense

At what scale does renting a charter stop making sense?

If you're processing under a million accounts, BaaS is still the faster and cheaper path to market. You can go live in weeks rather than years.

Your loaded cost per account through a BaaS provider runs roughly $3 to $5 per month, including platform fees and per-transaction costs. The interchange revenue on a debit card program often covers that and more.

But BaaS fees scale with volume while the cost of owning a charter is mostly fixed. At Chime's scale of 22 million accounts, the interchange split alone represented a significant fraction of its $1.5 billion in annual revenue. Chime paid $590 million to buy Stride Bank in September 2026 and expects to recoup that through eliminated BaaS fees and direct control of its deposit base.

When your sponsor bank receives a consent order, your product development stalls until the bank satisfies the regulator. Mercury migrated to Column after its relationship with Evolve deteriorated. That migration was possible because Mercury had the engineering resources and an alternative provider ready.

A smaller fintech on the same sponsor bank might not have either.

BaaS isn't going away. For most fintechs at most stages, renting a charter through an API beats spending years and tens of millions to get your own. What changed after 2024 is that the cost of choosing carelessly became visible. Your choice of middleware provider and sponsor bank, along with the architecture connecting customer records to actual deposits, now determines whether money survives a failure in the stack.