Inside the Indian Fintech Stack: UPI, RuPay, AePS, BBPS in 2026

The numbers are by now familiar but still hard to absorb. UPI — the Unified Payments Interface — crossed 18 billion transactions in a single month for the first time in November 2025 and has stayed above that since. RuPay debit cards are issued by more than 1,100 banks. Aadhaar-enabled Payment System handles cash withdrawal for rural India’s last-mile economy. Bharat BillPay quietly clears interoperable utility bill payments across 22,000 billers. Put together, India runs the largest live payments network humanity has built, and it is operated as digital public infrastructure rather than as a private oligopoly.

For builders outside India, the stack is sometimes treated as a curiosity — interesting policy, irrelevant engineering. That view ages badly. Brazil’s PIX, Nigeria’s eNaira, Singapore’s PayNow, the UK’s Open Banking refresh, and the European Payments Initiative all borrow ideas from the Indian playbook. If you build payments at scale anywhere, the Indian stack is worth understanding properly. This post is a practitioner’s tour.

The four primary rails#

There are more, but four cover most of what builders interact with.

UPI (Unified Payments Interface) is a real-time, account-to-account, peer-to-peer (and peer-to-merchant) payment rail operated by NPCI — the National Payments Corporation of India. Its standout property is that it is a layer on top of existing bank accounts, not a separate wallet. The user authorizes a payment via a PSP (payment service provider) app — PhonePe, Google Pay, Paytm, Cred, NaviPay, plus increasingly the banks’ own apps — and money moves bank-to-bank in under three seconds. There is no per-transaction fee for either side. NPCI runs the rail; banks settle through it.

RuPay is India’s domestic card network. It competes with Visa and Mastercard for debit, credit, and prepaid issuance. In 2024 NPCI permitted RuPay credit cards to be linked to UPI, which collapsed a previously separate use case — credit-on-UPI — into the existing UPI flow. As of 2026, roughly 40% of new credit cards issued are RuPay, driven by the UPI integration and lower interchange fees.

AePS (Aadhaar-enabled Payment System) lets people withdraw cash from a bank account using only their Aadhaar number and a biometric (fingerprint or iris). It is delivered through a network of business correspondents — small kirana shops with a fingerprint reader — and is the rail that keeps rural cash-out functioning. It is unglamorous, vital, and the part of the stack most outsiders ignore.

BBPS (Bharat Bill Payment System) is an interoperable bill payments network. Any registered biller — electricity, water, gas, telecom, insurance, mutual fund SIPs, school fees — is reachable from any registered customer-facing app. It runs on NPCI’s BBPCU central unit and a network of Bharat Bill Payment Operating Units (BBPOUs). It is the part of the stack a fintech founder underestimates and then ends up needing.

The architecture pattern#

What is distinctive about all four is that they are operated as public infrastructure with private participation. NPCI is a non-profit owned by Indian banks, regulated by the RBI, and chartered to run rails that no single bank would build alone. Apps compete fiercely on user experience; the rails do not compete at all. This is the inverse of the US model, where Visa, Mastercard, ACH, Zelle, Venmo, and the various closed-loop wallets are all separate networks with their own economics.

The technical consequence is that interoperability is not a feature — it is the design. Any UPI app can pay any other UPI handle. Any BBPS app can pay any registered biller. Any RuPay card works at any RuPay merchant. A new fintech does not have to negotiate with three card networks and four wallets to launch; it integrates once and reaches the whole market.

The economic consequence is that the apps make money on adjacencies — lending, insurance distribution, mutual fund sales, advertising — not on payment interchange. PhonePe and Paytm have never made money on UPI transactions. They use UPI as the loss-leader that brings users in.

What a builder integrates in 2026#

If you are building a fintech product targeting Indian users in 2026, the integrations look like this.

You become a “TPAP” (Third-Party Application Provider) by partnering with a sponsor bank that holds your PSP licence. The sponsor bank is your interface to NPCI for UPI and RuPay; you do not talk to NPCI directly unless you become a bank. Common sponsor banks for fintechs are Yes Bank, ICICI, Axis, and increasingly the small finance banks like Equitas and AU. Each has its own developer portal, latency profile, and tolerance for transaction limits.

For UPI, you implement the UPI 2.0+ specification — collect requests, intent flows, mandate (recurring) flows, AutoPay, and now UPI Lite for sub-₹500 transactions that bypass the central switch entirely. UPI Lite, launched in 2022 and refined since, is the rail that handles the truly high-volume small-ticket use case (groceries, snacks, transit) without the central switch having to authenticate every transaction. As a builder, your app maintains a small balance pre-funded from the user’s bank account, and Lite payments debit it locally with hourly sync.

For RuPay-on-UPI, you implement the credit-on-UPI flow which is a small variant of the standard UPI collect/intent flow — the difference is that the underlying instrument is a credit line rather than a bank account, and the merchant pays a small MDR (merchant discount rate) that the bank-account-UPI flow doesn’t carry.

For AePS, you do not build it yourself. You integrate with a BC (business correspondent) aggregator like Pay1, Spice Money, or PayNearby, who handles the biometric, the device certification, and the settlement.

For BBPS, you become a “BBPOU customer ou” by integrating with a sponsor bank or directly with one of the BBPOUs. Coverage is the lever — pick a sponsor with strong coverage in the biller categories you care about.

What is coming in 2026 and 2027#

Three large changes are visible from where we sit in early 2026.

Cross-border UPI has been quietly expanding for two years. Singapore (via PayNow), UAE, Bhutan, Nepal, Sri Lanka, France, and as of 2026 the early phases of the Brazil corridor with PIX are live. The implication for international fintechs is that an Indian tourist or expat increasingly pays via UPI abroad, and merchants in those corridors are starting to accept UPI QRs. If you build a merchant POS product, this is a checkbox you will need.

UPI for credit lines is broadening beyond credit cards. From late 2025, banks have been permitted to extend pre-approved credit lines that are drawable via UPI directly — no card, no separate journey. This collapses the distinction between debit and credit at the moment of payment. It is also the most significant disruption to the credit card business model in fifteen years.

ONDC convergence is the wildcard. ONDC — the Open Network for Digital Commerce — is not a payments rail but it heavily depends on UPI for the payment leg. The ONDC architecture post covers this in detail. The short version: if ONDC takes off the way UPI did, the next wave of fintech volume will come from embedded commerce flows, not from standalone payment apps.

What outsiders should learn#

If your team is building payments anywhere — Brazil, the EU, ASEAN, the Gulf — three lessons from the Indian stack are worth carrying:

1. Rails as utility, apps as products. The competitive layer should not be the network. If you can lobby for the network to be a regulated utility, do it; if you can build on one that already is, do it. Competing on rails is expensive and produces worse user experience for everyone.
2. Last-mile design matters more than the headline rail. AePS — the unglamorous biometric cash-out for rural India — is what made the rest work. Without it, the UPI numbers would not be the UPI numbers. If your country has a digital divide, the rail that bridges it is the one to design carefully.
3. Free at the point of use, monetize the adjacency. Zero-MDR payments did not destroy fintech in India. It moved the value capture to lending, insurance, asset management, and merchant services. The apps that figured this out — PhonePe, Paytm, Cred — are the largest. The apps that tried to charge per transaction did not survive.

Where pdpspectra fits#

We have engineering teams in Boston, London, Sydney, and Kathmandu, and the Kathmandu team has been integrating with NPCI rails for cross-border Nepal-India use cases since 2023. If you are a fintech building into the Indian market from outside, or an Indian fintech trying to scale internationally, our team does the platform engineering — including the regulatory architecture and the operational rails work.

Related reading: the ONDC architecture deep-dive, the Brazil PIX vs India UPI comparison, and the digital public infrastructure post.

---

The Indian fintech stack is the best-documented payments experiment in the world. Talk to our team about what you can borrow from it.