# Platforms, embeddingembeddingAn embedding is a numerical vector that represents data (text, images, or items) in a way that captures meaning, so similar items sit close together in space.Voir la définition complète → and the fight for the customer interface
When you buy an iPhone on installments and pay through Apple, tap to pay with Apple Pay, and stash cash in a savings account, you may never see the name of the bank actually holding your money or lending it. That bank exists. It is just invisible. Apple owns the screen, the brand and the customer. The bank owns the risk and the regulatory paperwork.
This is the central battle of modern banking: not banks versus banks, but banks versus the platforms that sit between banks and customers.
Traditionally banking had a simple stack: a licensed bank did everything. It held the license, took deposits, made loans, and owned the customer relationship.
That single role has now split into distinct layers, each fought over by different players:
The fight is for the top layer. Whoever owns the interface owns the customer, the data and the pricing power. The layers below get commoditised.
Embedded finance means financial services delivered inside a non-financial company's product, at the point of need, so the customer does not go to a bank at all.
Examples in the wild:
None of these three is a bank. Each partners with a licensed bank behind the curtain.
For a platform to embed banking, it needs a licensed partner. Banking-as-a-Service (BaaS) is the model where a licensed bank rents out its license, balance sheet and regulatory permissions to a non-bank, via software connections called APIs (Application Programming Interfaces, the technical pipes that let two systems talk).
The typical division of labour:
| Layer | Who does it | Example |
|---|---|---|
| Customer relationship, brand, app | The platform | Apple, Shopify |
| BaaS coordination / middleware | Fintech enabler | (historically firms like Marqeta for card issuing) |
| License, deposits, lending, compliance | The licensed bank | Goldman Sachs (Apple's long-time card partner) |
The key power question: who captures the margin, and who carries the risk?
Often the platform captures the customer and a large share of the economics, while the bank carries the credit risk, the capital requirement and the regulatory liability. That is a bad trade for a bank if it becomes the default, and a good trade for the platform.
Andreessen Horowitz's essay "Every Company Will Be a Fintech Company" is a useful free primer on why non-financial firms rush to embed finance.
Platforms do not embed banking because they love regulation. They do it for three concrete reasons:
1. Retention. A Shopify merchant with a Shopify loan and a Shopify account is far less likely to leave. Finance is a lock-in.
2. Data. Amazon already knows a seller's exact sales history, so it can underwrite a loan faster and more accurately than a bank staring at a spreadsheet. That data advantage is structural.
3. Margin. Payments and lending are lucrative. Skimming a slice of every transaction inside your own ecosystem is pure incremental revenue.
Notice that Apple, Amazon and Shopify each embed finance where they are already strongest: Apple at the phone (payments), Amazon and Shopify at the merchant (working capitalworking capitalWorking capital is the difference between a company's current assets and current liabilities, measuring short-term liquidity and the funds available to run daily operations.Voir la définition complète → lending). They are not trying to become full banks. They are cherry-picking the profitable, high-frequency, data-rich moments.
Here is the constraint that stops platforms from simply becoming banks: you cannot take deposits or lend at scale without a banking license, and getting one is slow, capital-heavy and heavily supervised.
In the US, deposit-taking is regulated by bodies including the OCC (Office of the Comptroller of the Currency), the FDIC (Federal Deposit Insurance Corporation, which insures deposits), and the Federal Reserve. In Europe, the ECB (European Central Bank) supervises large banks, alongside national regulators.
So platforms rent, rather than buy, the license. This is why the licensed bank does not disappear entirely. It becomes a supplier.
Renting your license is not free money. US regulators have made this painfully clear. Several BaaS-focused banks have faced enforcement actions over weak oversight of their fintech partners, particularly around AML (Anti-Money Laundering) controls and knowing who the end customer actually is.
The 2024 collapse of Synapse, a BaaS middleware provider in the US, froze funds for many end users and exposed how messy the ledgers between platform, middleware and bank can be. It was a warning: when you rent your license, you still own the blame.
Think of it as a tug of war over the value chain.
The platform's leverage: it owns the customer, the data and the distribution. It can switch bank partners. Apple famously wound down aspects of its Goldman Sachs partnership, a reminder that the platform, not the bank, decides who supplies the balance sheet.
The bank's leverage: it holds the one thing that is legally scarce, the license, plus deep compliance capability the platform does not want to build. Regulation is the bank's moatmoatA lasting edge over competitors: a resource, capability or position they cannot easily replicate, letting a firm earn above-average returns over time.Voir la définition complète →.
The uncomfortable middle outcome for banks is becoming a commoditised balance-sheet supplier: interchangeable, competing on price, invisible to the end customer. If ten banks all offer the same rented-license service, the platform picks the cheapest. That is the fate incumbents fear.
Not all banks accept this. Some are fighting back:
Vérification des acquis
1. In the modern banking stack described in the lesson, which layer confers the greatest pricing power and control over customer data?
2. When a customer uses Apple Pay and a savings account without knowing which bank holds their money, what strategic dynamic does this illustrate?
3. Which scenario best fits the definition of embedded finance?
4. Select ALL correct answers about why the license and balance sheet layer tends to get commoditised in embedded finance arrangements.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers describing the traditional (pre-platform) banking model versus the redrawn one.
Sélectionnez toutes les réponses correctes.
The embeddingembeddingAn embedding is a numerical vector that represents data (text, images, or items) in a way that captures meaning, so similar items sit close together in space.Voir la définition complète → fight looks different across the Atlantic, largely because of one rule.
In Europe, PSD2 (the second Payment Services Directive, EU law) forced banks to open customer data to licensed third parties via APIs. This is open bankingopen bankingCadre réglementaire (PSD2 en Europe) obligeant les banques à partager les données clients via des API standardisées, avec consentement, transformant les données bancaires en actif compétitif.: the customer can authorise a third party to see their bank data or initiate payments. It lowered the barrier for platforms and fintechs to build on top of banks. A successor framework, often discussed as PSD3 and a Payment Services Regulation, has been progressing through the EU legislative process, aiming to tighten and extend these rules.
In the US, there is no single equivalent mandate, though the CFPB (Consumer Financial Protection Bureau) has worked on a personal financial data rights rule under Section 1033 of the Dodd-Frank Act to push open bankingopen bankingCadre réglementaire (PSD2 en Europe) obligeant les banques à partager les données clients via des API standardisées, avec consentement, transformant les données bancaires en actif compétitif.. Its final shape and enforcement have been contested. Note this is a fast-moving area as of early 2026, so treat specifics as subject to change.
The practical effect: Europe pushed embeddingembeddingAn embedding is a numerical vector that represents data (text, images, or items) in a way that captures meaning, so similar items sit close together in space.Voir la définition complète → forward by law, while in the US embeddingembeddingAn embedding is a numerical vector that represents data (text, images, or items) in a way that captures meaning, so similar items sit close together in space.Voir la définition complète → has been driven more by commercial deals and the sheer scale of platforms like Apple and Amazon.
Imagine a customer pays $100 at a checkout using an embedded pay-in-installments product.
Platform keeps roughly $4 and owns the customer. Bank keeps roughly $1 and owns the risk.
These numbers are purely illustrative, but they show the structural problem: the party that owns the interface captures the fat margin and the low risk. The party that owns the license captures the thin margin and the real risk.