Fintech Stack: Rebuilding the Blueprint
Three years ago, the dominant narrative in fintech was about disruption: nimble startups eating the lunch of legacy banks. Today, that story looks almost quaint. The real disruption isn't a startup beating a bank β it's the underlying infrastructure of money itself being rewritten, layer by layer, in ways that most financial institutions are only beginning to understand.
The urgency is real. The Bank for International Settlements estimated in late 2023 that over 130 central banks β representing more than 98% of global GDP β were actively exploring or piloting digital currencies. Meanwhile, embedded finance revenues are projected to exceed $7 trillion globally by 2030, according to Lightyear Capital estimates. And AI is no longer a feature being bolted onto existing financial products; it's increasingly the product itself.
What's happening right now isn't an upgrade cycle. It's a platform shift.
The Three Layers Nobody Is Talking About Together
Most fintech coverage treats AI, embedded finance, and real-time payments as separate stories. They're not. They're converging into a single architectural question: who owns the financial operating system of the next decade?
Let me break down each layer and then explain why the intersection is what actually matters.
Layer One: Real-Time Payments Are Rewriting the Rules of Float
The United States β embarrassingly late to this party β finally launched FedNow in July 2023. India's UPI processed over 14 billion transactions in a single month (March 2024). Brazil's Pix, launched in 2020, now handles more transactions than credit cards in the country. Singapore's PayNow has been quietly setting the standard for cross-border interoperability.
The implications go far beyond convenience.
Float is dying. Traditional banks built entire revenue models around the lag between when money leaves one account and when it arrives in another. Real-time settlement collapses that window to near-zero. For context: Wells Fargo, JPMorgan, and Bank of America collectively earned an estimated $5β6 billion in float-related income in 2022 alone. That pool is shrinking.
But here's the less-discussed consequence: real-time payment rails are creating a new data layer. Every instant transaction generates a timestamped, structured data point about consumer and business behavior. Whoever controls that data infrastructure β whether it's a central bank, a private network like Visa or Mastercard, or a tech platform β gains an extraordinary intelligence advantage.
In Asia, this is already playing out. Ant Group's Alipay and Tencent's WeChat Pay don't just move money; they've built credit scoring, insurance underwriting, and wealth management products entirely on top of transaction data. The payment is the loss leader. The data is the business.
The lesson from Asia is that the payment rail is not the product β it's the distribution channel for every financial product that follows.
American and European banks are only beginning to grapple with this. FedNow's launch was a technical milestone; the strategic question of who builds the intelligence layer on top of it remains wide open.
Layer Two: Embedded Finance Is Eating the Bank Branch
The second major shift is structural. Financial services are migrating out of dedicated financial apps and into the workflows where people and businesses already operate.
Shopify Balance. Uber's driver earnings accounts. Amazon's small business lending. Stripe Treasury. These aren't fintech products in the traditional sense β they're financial services embedded so deeply into non-financial platforms that users often don't think of them as banking at all.
The numbers are striking. Bain & Company estimated that embedded finance could account for $3.6 trillion in US financial transactions by 2026. More importantly, embedded lending β where credit decisions happen at the point of purchase or invoice, not in a bank's underwriting queue β is growing at roughly 40% annually in Southeast Asia, according to various regional fintech reports.
The competitive dynamic this creates is genuinely new. A bank's traditional moat was distribution: the branch network, the brand, the customer relationship. Embedded finance dissolves that moat. If Grab or Gojek in Southeast Asia, or Mercado Libre in Latin America, can offer a credit product inside their super-app at the exact moment a user needs it β with underwriting powered by behavioral data the bank doesn't have β the bank's product becomes irrelevant regardless of its interest rate.
The actionable insight here is uncomfortable for incumbents: the question isn't whether to partner with platforms or compete with them. The question is whether your institution has the API infrastructure and the risk modeling capability to be the invisible bank powering someone else's customer experience. Most don't.
The banks that appear to be winning this transition β BBVA in Europe, DBS in Singapore, JPMorgan's embedded banking push in the US β share a common trait: they invested in modular, API-first core banking architecture years before it was fashionable. They're now licensing that infrastructure to fintechs and platforms rather than watching market share erode.
Layer Three: AI Is Not a Feature β It's the New Underwriting Engine
This is where I want to push back against the prevailing narrative.
Most financial institutions are deploying AI as a cost-reduction tool: chatbots to deflect customer service calls, document processing to reduce back-office headcount, fraud detection to cut losses. These are legitimate use cases. But they miss the more consequential application.
AI is fundamentally changing what can be underwritten, and at what cost.
Traditional credit underwriting is expensive and slow because it relies on structured data β credit scores, income verification, collateral assessment β that takes time and human judgment to gather and interpret. AI models trained on alternative data sources (transaction patterns, supply chain behavior, social signals, device metadata) can make credit decisions in milliseconds on borrowers who are entirely invisible to traditional scoring systems.
This matters enormously in markets like Southeast Asia, Sub-Saharan Africa, and South Asia, where hundreds of millions of people are "credit invisible" β they have economic activity but no formal credit history. Companies like Kredivo in Indonesia, Lendable in Africa, and Kissht in India are building lending books using AI underwriting that would have been impossible five years ago.
But it's not just about financial inclusion. In developed markets, AI underwriting is enabling entirely new product categories:
- Revenue-based financing for small businesses, where repayment is automatically calibrated to daily revenue rather than fixed monthly payments
- Dynamic credit limits that adjust in real time based on behavioral signals, not quarterly reviews
- Parametric insurance products that pay out automatically when measurable conditions are met, without claims processing
The risk, and it's a significant one, is that AI underwriting can encode and amplify existing biases at scale β and do so in ways that are far less transparent than human decision-making. The EU's AI Act, which took effect in 2024, classifies credit scoring AI as "high risk" and imposes explainability requirements. The US Consumer Financial Protection Bureau has been increasingly aggressive about algorithmic fairness in lending. Regulators are catching up, but the gap between what AI can do and what regulators have frameworks to govern remains wide.
The Convergence: Where These Three Layers Collide
Here's the thesis I keep coming back to: the most valuable position in financial services over the next decade is not being the best bank, the best payments company, or the best AI firm. It's being the entity that sits at the intersection of all three β controlling the real-time data rail, the embedded distribution channel, and the AI intelligence layer simultaneously.
In Asia, that entity increasingly looks like a super-app or a government-backed digital infrastructure project. In the West, the battle is more fragmented, and the outcome is genuinely uncertain.
Consider the competitive position of a company like Stripe. It processes payments (real-time rail access), offers embedded financial products to its merchant base (embedded finance), and has been quietly building AI-powered fraud detection and financial forecasting tools (intelligence layer). It's not a bank. It's not a pure fintech. It's arguably a financial operating system.
Or look at what's happening with JPMorgan's Chase. The bank's $12 billion annual technology budget isn't being spent on maintaining legacy systems β it's being used to build a parallel infrastructure that can compete with fintechs on their own terms. Chase's embedded banking partnerships, its developer APIs, and its AI-driven personalization engine suggest a deliberate strategy to become the infrastructure layer for financial services, not just a consumer-facing brand.
The smaller regional banks and credit unions face a genuinely difficult strategic choice. They can't build all three layers internally β the capital and talent requirements are prohibitive. Their options are:
- Partner deeply with a Banking-as-a-Service provider (Unit, Synctera, Treasury Prime) to access modern infrastructure
- Specialize aggressively in a niche where local knowledge and relationship banking still provide defensible advantages
- Consolidate β the current US banking landscape of ~4,500 FDIC-insured institutions will almost certainly compress significantly over the next decade
What This Means If You're Making Decisions Now
Whether you're a banker, a fintech founder, or an investor, here are the questions that actually matter in 2025:
For banks and credit unions:
- Does your core banking system support real-time API calls, or does it require batch processing? If it's the latter, you're already behind.
- Can you describe, specifically, what proprietary data asset you have that a tech platform doesn't? If the answer is "customer relationships," that's not sufficient anymore.
- What's your embedded finance strategy? Not "should we do it" β that debate is over. Where, specifically, will you be the invisible financial layer?
For fintech founders:
- The "unbundling the bank" thesis has largely played out. The next phase is rebundling β but at the infrastructure level, not the product level. Are you building a product or a platform?
- Regulatory risk is no longer a startup problem to defer. The EU AI Act, evolving CFPB guidance, and emerging data localization requirements in Asia are reshaping what's buildable. Factor this in at the architecture stage, not after Series B.
For investors:
- The highest-value fintech investments of the next five years likely appear to be in infrastructure and middleware β the picks-and-shovels of the embedded finance and AI underwriting buildout. Boring, but defensible.
- Geographic diversification matters more than it did. The real-time payment infrastructure and regulatory frameworks in Brazil, India, and Southeast Asia are years ahead of the US in some dimensions. Companies building cross-border interoperability between these systems appear to have significant structural advantages.
The Geopolitical Dimension Nobody Wants to Price In
One more layer that deserves more attention than it gets: the financial infrastructure being built right now is also geopolitical infrastructure.
The US dollar's dominance in global trade has been partly sustained by the fact that most cross-border transactions run through dollar-denominated correspondent banking networks. Real-time payment interoperability between non-dollar systems β India's UPI linking with Singapore's PayNow, Brazil's Pix exploring connections with other Latin American networks, China's CIPS expanding its reach β is quietly building an alternative plumbing system.
This isn't imminent dollar displacement. But it's the long-term infrastructure for a more multipolar financial system, being laid one API connection at a time. Fintech investors and strategists who ignore the geopolitical dimension of payment rail competition are likely missing a material risk factor.
I've covered this dynamic in my previous work on AI semiconductor supply chain control β the pattern is consistent: technology infrastructure decisions that appear purely commercial are increasingly inseparable from questions of national economic sovereignty.
The Stack Is Being Rebuilt β The Question Is Who Holds the Keys
The financial services industry is not experiencing a feature upgrade. It's experiencing a platform transition of the kind that happens once or twice a generation β comparable to the shift from physical branch networks to online banking in the late 1990s, but faster and with higher stakes.
The institutions that will emerge strongest from this transition share a common characteristic: they stopped asking "how do we defend what we have?" and started asking "what infrastructure position do we need to hold in 2030?" Those are fundamentally different questions, and the gap between them is where most strategic planning in financial services currently lives.
The real-time payment rail, the embedded finance distribution layer, and the AI intelligence engine are not three separate technology bets. They're three components of a single system. The organizations β banks, fintechs, platforms, or some hybrid that doesn't exist yet β that figure out how to control all three will define what financial services looks like for the next twenty years.
Everyone else will be a tenant in someone else's infrastructure.
Alex Kim
Former financial wire reporter covering Asia-Pacific tech and finance. Now an independent columnist bridging East and West perspectives.
Related Posts
λκΈ
μμ§ λκΈμ΄ μμ΅λλ€. 첫 λκΈμ λ¨κ²¨λ³΄μΈμ!