Starbucks Sells Coffee. It Also Runs a Bank.
Starbucks carries $1.6 billion in unredeemed gift card balances.1 That is not a cute loyalty stat. It is a deposit base. Customers prepay, Starbucks sits on the cash for weeks or months, pays no interest, and settles the liability later in drinks with coffee-shop margins. On the balance sheet, Starbucks looks a lot like a mid-sized bank that happens to pour espresso.2
VCs and fintech operators have treated that as a success story. Angela Strange gave the pitch its slogan in 2019: every company will become a fintech company.3 A16z argued embedded finance expands addressable revenue. Shopify's merchant services growth became the standard exhibit. Apple and Uber joined the parade. The case sounds great. Add financial products to a captive audience, collect fatter margins, keep users inside the app.
That story leaves out the part that matters.
Most companies do not drift toward banking because they have some grand plan. They do it because software pricing stops carrying the growth story.
This piece makes a simple claim. The move into banking is usually defensive. SaaS companies hit a ceiling on subscription revenue, then reach for payments, lending, stored balances, or float to keep the numbers moving. Financial revenue turns into a fintech margin crutch. It props up a slowing core business, changes the company's risk profile, and eventually changes how the market prices the stock.
The Crutch Replaces the Legs
Look at Shopify. Payments and lending did not sit neatly beside the software product. They overtook it. Shopify's filings show Merchant Solutions grew with gross merchandise volume and deeper Shopify Payments penetration, and now accounts for most platform revenue.4 Subscription software still matters. But it is no longer the engine. It is the top of the funnel for the financial product below it.
Ben Thompson described this as platform logic: the money is in taking a cut of the activity, not in charging for software.5 Fine. But that makes the shift sound cleaner than it is. For a lot of software companies, taxing the flow is not just attractive. It becomes necessary once subscription pricing runs into resistance and rates make customer balances more profitable. McKinsey says embedded finance gives distributors a new high-margin revenue stream.6 Read that another way: when your old margins start to sag, finance starts looking less like expansion and more like pain relief.
Healthy software companies do not need pain relief from payments.
The Float and the Middleware Mirage
Float looks like free money. For a while, it is. Starbucks gets a $1.6 billion interest-free loan from customers and pays it back in coffee. Apple's savings account with Goldman Sachs passed $10 billion in deposits.7 Those numbers tell you something real. Consumers trust big tech and retail brands with money at a level that now competes with banks. The Bank for International Settlements has flagged BigTech's data advantage and distribution as a structural edge in finance.8
But deposits are liabilities. They are promises. Once a retailer or software platform holds customer funds, it takes on work and exposure it was never built for: reconciliation, compliance, and the legal duty to return money on demand. The brand gets the trust. The brand also gets blamed when the machinery breaks, even if some partner bank or middleware vendor actually held the license.
BaaS vendors and fintech product teams often pitch APIs as a way to outsource banking operations. Stripe Treasury sells the idea directly: embed financial services through an API and let the stack do the hard part.9 That promise deserves a name: the Middleware Mirage. Write some code, launch an account, let somebody else deal with the banking mess. It does not work that way. Regulators go after legal entities, not software abstractions, and ledger reconciliation is not some back-office annoyance you can smooth over with better product design.10 A ledger decides who owns what money and where it went. If that ledger does not reconcile, users lose access to funds.11 Nice onboarding screens do not save you when the books are wrong.
Synapse blew up this illusion in public. The middleware provider filed for Chapter 11. Millions of users at apps including Yotta and Mercury lost access to their accounts, and reports said $100 million or more could not be properly accounted for.12 The Federal Reserve later hit sponsor bank Evolve with an enforcement action over risk management failures tied to the Open Banking Division that handled those deposits.13 The lesson was ugly and obvious. Compliance and reconciliation do not disappear because an API sits in the middle.
"Regulators regulate legal entities, not software. The idea that an API could shield a platform from compliance was always a zero-interest rate illusion."14
The stack was always unstable. The sponsor bank carries the license and liability; the fintech owns the customer; middleware becomes the weak link.
The Regulatory Fuse and the Bank Multiple Trap
The retention case for embedded finance has the same problem with a different wrapper. Gig platforms learned fast that instant access to wages keeps workers around. Uber launched Uber Money by promising real-time access to earnings inside the app.15 Pay drivers faster and churn drops. Let some of that money sit in a closed-loop wallet and the platform gets loyalty plus float. Shopify Balance came from the same instinct. It was built to sit inside the merchant's daily workflow, not as a standalone bank app.16
That setup starts to look uncomfortable once a platform is effectively warehousing wages or operating cash. A closed-loop wallet for worker pay can start to resemble company scrip. Regulators were slow to react, but not asleep. The FDIC has already warned that third-party arrangements reduce a bank's direct control and create fresh risk,17 while interagency guidance has lagged far behind product launches.18 Fintech Brainfood gave the catch-up phase a name in 2024: the BaaS Winter.10
| Date | Event |
|---|---|
| 2019 | Uber Money launched, promising real-time earnings access |
| 2023 | FDIC/Federal Reserve publish interagency third-party risk guidance |
| 2024 (Apr) | Synapse files for Chapter 11; millions of users locked out |
| 2024 (Jun) | Federal Reserve enforcement action against Evolve Bank & Trust |
Now assume the company gets through all of it. Clean ledgers. No compliance disaster. Payments and lending become the main source of revenue. That sounds like victory. It is also where the trap closes.
Public markets reward software companies for predictable revenue and low incremental cost. Banks get priced for risk and regulation, not for software-style expansion. So when a "software company" starts making most of its money from payments spreads, lending income, deposit economics, or float, investors stop giving it a SaaS multiple. The business gets re-rated toward a financial multiple instead. Same company. Same revenue dollars. Lower ceiling.
That is the part embedded-finance evangelists usually skip. Adding financial products can rescue growth for a few years. It can also teach the market to value you like a bank.
What To Actually Do About It
Angela Strange was right about one thing:
"In the not-too-distant future, nearly every company will derive a significant portion of its revenue from financial services." Angela Strange, a16z, "Every Company Will Be a Fintech Company" (2019)
She was less clear about the cost. The day customer funds become a profit center is the day regulatory and valuation risk stop being optional.
The next blowup will not come from bad UX. It will come from a platform that rented banking infrastructure, enjoyed the margin uplift, and learned too late that customers still blame the brand.
And if finance becomes your growth engine, the market will eventually stop calling you a software company.