Skip to content
Velloreth
Financial Technology 10 min read

How UPI Changed Payments in South Asia

UPI turned India into the world's largest real-time payment system. A clear look at its design, scale, economics, and how its model spread across South Asia.

By

UPI did not win because it was clever engineering. It won because it solved an institutional problem: in a country with dozens of banks, hundreds of payment apps, and a vast population that had never used cards, the only way to make digital payments work was to make every bank talk to every other bank, and every app talk to every bank, through one shared rail. That decision — interoperability enforced at the infrastructure layer — is the thing worth understanding. Everything else follows from it.

What UPI actually is

Unified Payments Interface is a real-time payment system operated by the National Payments Corporation of India (NPCI), a not-for-profit company set up by banks and regulated by the Reserve Bank of India. It launched in 2016.

The mechanics are simple:

  • You link your bank account to any UPI-enabled app and create a Virtual Payment Address — something like name@bank — or use your phone number as the identifier.
  • To pay, you authorize a debit from your bank account and a credit to another account. The money moves bank-to-bank, typically within seconds, 24/7.
  • Payment requests work in both directions: you can push money to someone, or pull from them by sending a request they approve.

Three design choices did most of the work.

Interoperability. Any UPI app works with any bank. This sounds obvious and almost nobody does it this way. In most markets, Venmo talks to Venmo, Apple Pay works with certain banks, and each bank ships its own app with its own silo. UPI made the rail common infrastructure — the payment equivalent of roads — and let apps compete on user experience above it.

Push, not pull. UPI is debit-authorized by the payer in real time. There is no stored balance floating in a private wallet, no card number exposed to the merchant, and no credit line hidden in the transaction. The account debited is yours, the authorization is yours, and settlement is immediate. This kept the fraud surface smaller than card networks and eliminated the float that wallet operators used to earn on.

Phone numbers and VPAs as addresses. Before UPI, paying someone meant sharing account numbers and IFSC codes — bank-centric identifiers that normal people do not carry in their heads. UPI replaced them with handles you can say out loud, type into a chat, or scan from a QR sticker. The address system is what made person-to-person payments feel like messaging instead of banking.

The scale, in numbers

The growth curve is the most cited part of the UPI story, so it is worth stating plainly. NPCI’s annual figures, as reported by the government:

Financial yearTransactions (crore)Value (lakh crore)
2021-224,59684
2022-238,371139
2023-2413,113200
2024-2518,587261

Monthly data through early 2026 shows the same trajectory continuing: roughly 21.7 billion transactions in January 2026 and 20.4 billion in February, averaging around 700–730 million transactions a day. To put that in perspective, that is more daily retail transactions than any card network on earth, processed at a fraction of the per-transaction cost.

The user base has followed. Government figures put UPI users above 50 crore — half a billion — by early 2026, a remarkable share of a country of 1.4 billion where digital payments barely existed a decade ago.

Why it grew when other systems did not

Technology explains the rail. Adoption needed three more things.

Zero MDR for small merchants

A merchant discount rate is the fee a merchant pays to accept a digital payment — typically 1–3% on cards. India legislated zero MDR on UPI for small merchants, and the government subsidized UPI for larger ones for a period. The effect on acceptance economics was immediate: a street vendor could put up a QR sticker at no cost, and accepting a digital payment stopped being a margin decision.

This was the single most important policy choice. Card networks grew slowly in India precisely because acceptance economics never worked for small merchants — terminal costs, monthly rentals, and percentage fees made cash cheaper for anyone with thin margins. UPI removed the cost entirely.

The zero-MDR regime is not free — someone pays for the infrastructure, and the debate over who should keep paying for UPI as volumes mature is ongoing. But for a decade it did what it was designed to do: collapsed the cost of acceptance to zero and pushed digital payments into the informal economy.

QR codes as the merchant layer

Card acceptance required hardware. UPI acceptance required a printed QR code, or a dynamically generated one on a phone screen. India ended up with hundreds of millions of active QR codes — industry counts passed 700 million in 2025 — covering kirana stores, pharmacies, vegetable vendors, and transport hubs.

The economics deserve emphasis: a QR code costs nothing to print and nothing to accept. A POS terminal costs money on all three counts — hardware, connectivity, and a cut of every transaction. For a market where the median merchant is a single-person shop doing a few thousand rupees of daily revenue, the comparison is not close.

Government distribution

Direct benefit transfers, subsidies, and government payments moved onto UPI, giving hundreds of millions of people a reason to set up a UPI handle before they had any reason to shop online. The government became the largest acquirer of first-time digital payment users in the country’s history, not by marketing but by paying welfare, pensions, and subsidies into accounts that people then learned to spend from.

What actually changed in behavior

The most interesting UPI shift is not cash versus digital. It is the collapse of the distinction between person-to-person and person-to-merchant payments in daily life.

Before UPI, digital commerce meant e-commerce — buying things online, at a distance, with a card. UPI made digital payments local and immediate. Paying a friend for dinner, splitting a bill, paying a rickshaw driver, tipping a delivery worker: these micro-transactions never had a digital channel before. Cards were too expensive for merchants to accept at those ticket sizes, and cash was faster than any digital option.

Industry data through 2025 showed person-to-merchant transactions outpacing person-to-person, with average ticket sizes falling toward the low hundreds of rupees as UPI moved into groceries, transit, and food. A falling average ticket size is a good sign here — it means the system is being used for more of daily life, not just larger purchases.

UPI also quietly became the default for recurring payments: subscriptions, EMIs, utility bills. Auto-pay mandates on UPI now handle a large volume of recurring debits that previously required standing instructions or cards.

The economics: who pays for free payments

Zero MDR creates an obvious question: if merchants do not pay and consumers do not pay, who funds the system?

The answer, in practice, has three parts.

Volume subsidy from the apps. PhonePe and Google Pay — which between them handle the large majority of UPI volume — treat UPI as a customer-acquisition and engagement channel rather than a profit center. They monetize through financial products (loans, insurance, gold), advertising, and merchant services. The payment itself is a loss leader for the ecosystem around it.

Bank participation. Banks treat UPI as infrastructure they must be on, similar to ATM networks. NPCI’s rules and the RBI’s expectations keep participation open, and the competitive cost of not offering UPI outweighs the per-transaction cost of supporting it.

NPCI’s structure. NPCI runs the rail as a utility, not a venture-funded startup optimizing for revenue. That institutional structure — neither fully public nor fully private — is a large part of why nobody has managed to replicate the model exactly.

The model has a known weakness: app-level concentration. NPCI introduced a volume cap on third-party app providers to prevent any single app from dominating, with the deadline extended multiple times as the duopoly of PhonePe and Google Pay proved sticky. Competition policy around UPI remains unfinished business.

How the model spread across South Asia

UPI’s second act is as export infrastructure. NPCI’s international arm, NIPL, has spent the last few years linking UPI to other countries’ payment systems and enabling acceptance abroad.

The pattern is consistent: a domestic instant-payment system in another country connects to UPI at the infrastructure level, and the visible result is Indian travelers scanning local QR codes, plus remittance corridors where money moves bank-to-bank in near real time.

Nepal is a clear case. Nepali merchants on the Fonepay network accept UPI payments from Indian visitors directly at the point of sale — no card terminal, no currency exchange, no cash conversion. For a tourism economy that handles hundreds of thousands of Indian visitors a year, removing the cash friction at the merchant level is a meaningful improvement. Bhutan, Singapore, the UAE, France, Mauritius, Sri Lanka, and others have similar arrangements in various stages of coverage.

What makes this different from Visa or Mastercard expanding into a new market: the export is of a real-time, account-to-account, low-cost rail, not a card network with interchange economics. Countries adopting UPI linkages are getting something closer to public payment infrastructure than to a proprietary card scheme.

What is left unfinished

UPI is not a finished system. The open questions are worth knowing about.

Monetization stability. The zero-MDR regime works while apps subsidize participation and NPCI runs cheaply. Whether that holds at 800 million-plus transactions a day — and whether the economics of UPI Lite and offline variants work the same way — is unproven over the long term.

Fraud. Real-time push payments are hard to reverse. India has responded with risk-based transaction limits, cooling-off periods for new payees, and tighter rules on app-level security — NPCI’s comprehensive information security framework, for instance. Social engineering and mule accounts remain a growing problem as the user base expands downward and outward.

App concentration. Two apps still dominate. Interoperability at the rail level did not produce competition at the app level in the way early UPI advocates hoped, and policy tools to fix that are blunt.

Credit discipline. UPI is a payments rail, not a credit system — but lending against UPI transaction history is now a large and fast-growing business, with all the underwriting risk that implies. The rail’s success is pulling credit products into its orbit whether or not the infrastructure was designed for it.

The transferable lesson

If you are building or regulating a payment system anywhere else — and South Asia has several countries actively studying the model — the UPI playbook reduces to a short list.

  1. Make the rail interoperable by rule, not by negotiation. Every bank, every app, one network.
  2. Make authorization push-based and real-time. Kill stored-value float and card-like exposure.
  3. Make merchant acceptance costless at the small end. QR codes beat terminals for the bottom of the pyramid.
  4. Give people a reason to join before you give them a reason to shop. Government payments are the highest-leverage onboarding tool that exists.
  5. Accept that the payment layer will be nearly free, and decide in advance how the ecosystem around it will be funded.

None of that requires UPI’s specific technology. Most of it requires the political will to enforce interoperability over the objections of incumbents who would rather own their silos — which is, in the end, why the model has spread as infrastructure to places like Nepal but has not been easily rebuilt at home in larger, richer banking markets.

Frequently asked questions

What makes UPI different from a mobile wallet?
Wallets store your money with a private company and move it between closed accounts. UPI moves money directly between bank accounts in real time, and any app on the network can talk to any bank. You get wallet-like convenience with bank-account settlement.
How does UPI make money if MDR is zero?
Most consumer UPI payments carry no merchant discount rate, so the payment layer itself is nearly free. Apps monetize through advertising, financial products — loans, insurance, gold — and merchant services like working-capital advances, not through transaction fees.
Can other countries copy UPI?
Yes, but the hard parts are institutional, not technical. Interoperability across banks, a credible central operator, government willingness to forego MDR revenue, and a large base of small merchants with QR codes. Several countries have copied pieces of the design; few have copied all of it.
Is UPI safe for large payments?
UPI is designed for retail payments and applies risk-based transaction limits, which have tightened over time as fraud patterns emerged. Large-value transfers still typically go through IMPS or RTGS. UPI's sweet spot is everyday payments, roughly up to a few lakh rupees depending on bank and transaction type.