UPI: How India Built Its Advanced Payment System

India pulled off something that shouldn't have worked

Picture a chaiwalla in a small Rajasthani town taking payment through a printed QR code taped to his cart. No card machine, no POS terminal, only a smartphone and a piece of paper. The customer scans, enters the amount and taps a button, and the chaiwalla's phone pings with the confirmation seconds later. Scenes like that became ordinary across India within a few years.

It's remarkable when you think about it. A country where, a decade earlier, the vast majority of transactions happened in cash had leapfrogged the entire developed world in digital payments. It did that without credit cards or NFC terminals, using a homegrown system built by a non-profit entity most people outside fintech had never heard of.

From a technical standpoint, that's what makes the story remarkable: India didn't copy anyone else's homework. UPI, the Unified Payments Interface, wasn't a localized version of some Western payment rail. It was designed from scratch for Indian conditions, meaning low-cost smartphones, unreliable internet, millions of tiny merchants who'd never touch a card reader, and a population that overwhelmingly preferred cash. Building for that meant rethinking assumptions the rest of the world's financial infrastructure was built on.

By early 2026, UPI processes over 20 billion transactions per month, according to NPCI's published UPI statistics. That's more real-time payment transactions than any other system on the planet. More than Visa, more than Mastercard, more than WeChat Pay. And a non-profit entity built the whole thing on a relatively modest budget.

How did this happen? Why are countries from Singapore to France trying to replicate or integrate with it? I've been digging into this for a while now. The answers surprised me.

---

Before UPI: a fragmented mess

You can't understand UPI without the problem it solved. Before 2016, India's digital payment scene was a mess. Several systems existed. None talked to each other well. Ordinary people found most of them baffling.

NEFT and RTGS, the old guard

India had NEFT (National Electronic Funds Transfer) since 2005 and RTGS (Real Time Gross Settlement) since 2004. The Reserve Bank of India ran both, and they worked reasonably well for large bank-to-bank transfers. They also had serious limitations:

• NEFT processed transactions in batches rather than in real time, so a transfer could take hours to show up. Sometimes longer.
• RTGS was real-time but had a minimum transaction amount of Rs 2 lakh, which made it irrelevant for everyday payments. Nobody's buying groceries through RTGS.
• Both needed the sender to know the recipient's bank account number and IFSC code. Try asking your vegetable vendor for that. Good luck.

IMPS: getting warmer

In 2010, the National Payments Corporation of India (NPCI) launched IMPS (Immediate Payment Service), the first real-time, 24/7 interbank transfer system in India. You could send money instantly using a mobile number linked to your bank account, at any hour. IMPS was a real breakthrough and laid the technical groundwork for what came next.

Friction, though. Registering your mobile number with an MMID (Mobile Money Identifier) was confusing, and plenty of people who tried gave up halfway. The banking app interfaces were clunky and transaction limits were low. There was no ecosystem for merchant payments either; IMPS was purely a person-to-person transfer tool.

Three pieces of infrastructure

Meanwhile, the Indian government was laying the foundation for a digital economy through what became known as the JAM Trinity:

• Jan Dhan Yojana (2014) opened over 500 million bank accounts for previously unbanked citizens
• Aadhaar gave every citizen a unique 12-digit identity number linked to biometrics
• Mobile: smartphone penetration was exploding, driven by affordable devices from companies like Xiaomi and Jio's free 4G data revolution

All the pieces were there. What was missing was a system to tie them together, and that's where NPCI's ambition comes in, along with the part of the story I find most interesting as an engineer.

---

The birth of UPI

NPCI launched UPI on April 11, 2016, with 21 member banks. The vision was audacious: one platform where any bank account holder could send money to any other, instantly, 24/7, using just a virtual payment address (VPA) like anurag@upi.

No account numbers. No IFSC codes. No waiting, no minimum amount. Most important of all, no transaction fee for end users, which is probably the single biggest design decision in the whole system.

Under the hood

UPI's architecture is simple on paper, even though building it at scale was anything but. It works like this.

There are four pillars:

1. NPCI Switch. The central routing hub that processes every UPI transaction, a bit like the air traffic controller of Indian digital payments. Every single transaction passes through it.

2. Payment Service Providers (PSPs). The apps you use, like Google Pay, PhonePe, Paytm and CRED. They provide the interface and handle the customer experience, but they don't hold your money. That distinction matters.

3. Issuer banks. Your own bank (SBI, HDFC, ICICI, etc.), where your money sits. The issuer bank debits your account when you send money.

4. Beneficiary banks. The recipient's bank, which credits their account when they receive money.

How a transaction flows

When you send Rs 500 to a friend through Google Pay:

1. You initiate the payment on Google Pay (the PSP app)
2. Google Pay sends the request to the NPCI UPI switch
3. NPCI identifies your bank (issuer) from your VPA and routes the debit request
4. Your bank authenticates you (via UPI PIN) and checks your balance
5. If approved, your bank debits Rs 500 and confirms to NPCI
6. NPCI routes the credit instruction to your friend's bank (beneficiary)
7. Beneficiary bank credits Rs 500 to your friend's account
8. NPCI sends confirmation back through the chain
9. Both you and your friend get instant notifications

On average all of this takes under 3 seconds. Authentication, debit, credit and confirmation happen in roughly the time it takes you to blink twice. When I first learned how many systems are involved in a single transaction, I was surprised it could happen that fast.

Built on IMPS

Most people don't realize that UPI runs on top of IMPS infrastructure. Real-time settlement between banks uses the same IMPS rails NPCI built back in 2010, so UPI is essentially a clean, frictionless interface layer over a proven interbank transfer system. Smart. They kept the plumbing and fitted a much better faucet. That's probably why it scaled so quickly, since the underlying payment rails were already tested and running.

---

Why UPI worked when others didn't

Plenty of countries have tried digital payment systems, with mixed results. The UK's Faster Payments, Australia's NPP and the Eurozone's SEPA Instant all exist, and none has come close to UPI's adoption. So what made India different? I think it comes down to five factors, some intentional and some lucky.

Zero cost to users

The big one. UPI transactions are free for consumers: no per-transaction fee, no monthly subscription, no hidden charges. When a digital transaction costs literally nothing and the alternative (cash) means finding an ATM, standing in line and carrying notes around, the math is obvious. People switch when you remove every reason not to.

Interoperability from day one

UPI was designed to be bank-agnostic and app-agnostic. Any UPI app can send money to anyone at any bank. Compare that with closed-loop systems like early PayPal or WeChat Pay, where sender and receiver needed the same app or platform. Because of interoperability, network effects kicked in fast, with each new UPI user making the system more valuable for everyone else. It's arguably the smartest architectural call NPCI made.

Demonetization, the accidental catalyst

On November 8, 2016, just seven months after UPI launched, Prime Minister Modi announced the demonetization of Rs 500 and Rs 1,000 notes. Overnight, 86% of India's currency by value became invalid. The cash crunch that followed pushed millions of people and merchants to try digital payments for the first time, many of them through UPI.

Was the timing intentional? NPCI says the development timelines were independent, but the effect was undeniable. Possibly the biggest unplanned marketing campaign in fintech history.

Smartphones got cheap

Reliance Jio's launch in September 2016 with free 4G data brought hundreds of millions of Indians online for the first time, and as we've explored in our India 5G coverage analysis, that connectivity revolution keeps accelerating. Suddenly the smartphone in a farmer's pocket was a payment terminal as well as a WhatsApp and YouTube machine. UPI apps were among the first things people installed after messaging apps.

QR codes need no hardware

With the UPI QR code system, merchants didn't need any hardware. Print a QR code, tape it to the wall, done. The credit card ecosystem, by contrast, makes merchants buy or rent POS terminals, pay processing fees, wait for settlements and deal with chargeback disputes. For a small shop owner a UPI QR code is far simpler and cheaper. My neighbourhood paan shop has one. So does the guy who fixes shoes on the sidewalk.

---

Google Pay vs PhonePe vs Paytm: who won?

The UPI app market has been a fascinating three-way (now arguably two-way) battle.

PhonePe, the early mover

PhonePe, backed by Walmart through Flipkart, launched in 2016 and was among the first to build a UPI app normal people found easy. Its aggressive merchant acquisition strategy, which meant sending representatives to shops across India to set up QR codes, gave it an early lead. As of early 2026, PhonePe holds approximately 47-48% of UPI transaction volume and is the clear market leader. It's one example of how the Indian startup ecosystem has produced world-class fintech products.

Google Pay, the tech giant

Google Pay (formerly Tez) launched in 2017 and quickly became a favourite thanks to its clean interface, its tie-in with Google's ecosystem, and clever touches like scratch-card cashback rewards. It holds roughly 34-35% market share. Integration with Google search results and Maps for merchant payments has been a big driver. Clean design matters. Google got it right.

Paytm, a cautionary tale

Paytm was the undisputed digital payments leader before UPI, and its wallet was everywhere during demonetization. The shift to UPI made wallets less relevant and wore down that advantage, and RBI's regulatory action against Paytm Payments Bank in 2024 did further damage. Paytm's UPI market share has dropped to around 7-8%, a dramatic fall from its earlier dominance. I think it's probably the starkest example of how fast a market leader can lose ground when the platform underneath changes.

The 30% cap nobody has solved

NPCI introduced a rule that no single app should process more than 30% of total UPI transactions in a given month, to prevent monopolistic dominance and systemic risk. PhonePe and Google Pay both exceed the cap, and NPCI has repeatedly pushed back the compliance deadline. How this plays out will shape the competitive picture a lot. I'm not sure enforcement will ever happen, since forcing users to switch apps looks like it could backfire badly.

---

Recent UPI upgrades

UPI hasn't stood still. Recent upgrades push it well beyond simple peer-to-peer transfers, and a few of them are interesting from a financial infrastructure point of view.

Credit line on UPI

Banks can now extend a pre-approved credit line through UPI. It works like an instant personal loan you spend straight from your UPI app. That's potentially a very big deal, because it brings credit to people without credit cards, using their banking history and UPI transaction data for risk assessment. In effect it's a lightweight credit card built into your UPI app. Whether banks will price the credit responsibly remains to be seen. I'm cautiously optimistic, but I wouldn't be surprised by some rocky early implementations.

UPI tap-to-pay (UPI Lite X)

NFC-based UPI payments through UPI Lite X let you tap your phone on a merchant's NFC terminal, much like Apple Pay or Google Pay's NFC works globally. The offline part is what interests me: small payments can go through without internet on the device. The RBI raised UPI Lite's per-transaction limit to Rs 1,000 in late 2024, so check your app for the current offline cap. The phone stores a small pre-loaded balance and settles with the server later, a clever workaround for India's spotty mobile data in some areas.

Feature phones too

You can now link UPI to feature phones through UPI 123PAY (an IVR-based system), and to smartwatches and other wearables. That takes UPI beyond smartphone owners, which is critical for rural India where feature phones are still common. Small addition? Not really. Millions of people in India don't own smartphones.

Recurring payments (UPI AutoPay)

UPI AutoPay supports recurring mandates, so you can authorize automatic debits for subscriptions, EMIs, insurance premiums and utility bills. Each mandate can carry a maximum amount and frequency, and you approve the initial setup with your UPI PIN. It's a major enabler for subscription businesses in India, and probably something that'll grow a lot over the next few years.

---

Going global

UPI isn't India-only anymore. NPCI International (NIPL) has been pushing UPI acceptance abroad aggressively, and progress has been faster than most people expected.

Where it works abroad

As of early 2026, Indian travelers can use UPI in:

• Singapore (via PayNow-UPI linkage)
• UAE (direct QR code acceptance)
• France (the Eiffel Tower was one of the first to accept UPI payments, mostly symbolic but now expanding)
• Sri Lanka, Bhutan, Nepal (bilateral agreements)
• Malaysia, Thailand (in pilot phases)

The Singapore-India corridor is the most advanced. You can send money from a Singapore PayNow account to an Indian UPI account and back, in near real time, which is cross-border interoperability the developed world is still struggling to achieve.

Why other countries want it

Several countries have approached NPCI for help building their own UPI-like systems. The appeal is easy to see:

• Low cost, since UPI's per-transaction processing cost is a fraction of what card networks charge
• Financial inclusion, by bringing digital payments to the unbanked
• Sovereignty, with less dependence on American card networks (Visa/Mastercard)
• Real-time settlement, in seconds instead of days

Peru, Namibia and several Southeast Asian nations are at various stages of consultation with NPCI International. Whether they can replicate UPI's success depends heavily on local conditions like bank account penetration, smartphone adoption and government support. I suspect some will succeed and others won't, depending on how many of those preconditions they already have.

---

The money problem nobody talks about

UPI doesn't make money for most of the players involved. Under the zero-MDR (Merchant Discount Rate) policy, merchants pay no fee for accepting UPI. The government compensates banks and payment processors through a subsidy from the budget, but many industry participants say it isn't enough.

Who's paying for all this?

• Banks spend money maintaining UPI infrastructure, handling transaction load and managing failed transactions
• PSP apps (Google Pay, PhonePe) spend heavily on user acquisition, cashback and server infrastructure
• NPCI maintains the central switch and security infrastructure

PhonePe and Google Pay are reportedly losing money on UPI transactions. They earn through adjacent services like lending, insurance, mutual fund distribution and merchant services, with the payment itself as a loss leader. Sounds unsustainable. Might be.

Can free last forever?

Should UPI charge a small fee? The Reserve Bank of India has gone back and forth on it, and I don't blame them, because both sides have a point that's hard to wave away when you're the one responsible for a system half the country now leans on every day. Consumer groups and the government want to keep it free so adoption keeps its momentum. Banks and payment companies argue that a small fee (even Rs 1-2 per transaction) would make the system financially sustainable without putting users off in any meaningful way.

It's still one of the most debated topics in Indian fintech. For now the government has chosen to subsidize the system, but whether that holds up at 20+ billion monthly transactions is an open question. My guess is that some fee structure eventually appears, probably not for P2P transfers but maybe for merchant transactions above a certain threshold. That's only a guess, though.

---

Engineering at scale: where things break

Processing over 20 billion transactions a month is no small feat, and this is where the technical challenges get real.

When transactions fail

UPI's technical decline rate, the share of transactions that fail because of system issues rather than user errors, has been a persistent concern. During peak hours (salary day, festival season) failure rates can spike. NPCI has been working with banks to improve server infrastructure and has brought in measures like transaction throttling during extreme peaks. Still infuriating when it's you. "Transaction failed, amount will be refunded within 5-7 business days" is probably India's most anxiety-inducing notification.

Fraud is growing

UPI fraud keeps rising. The common attack routes:

• Social engineering, convincing users to share their UPI PIN
• SIM swap attacks, taking over a user's mobile number
• Fake collect requests, debit requests disguised as credits
• Malware, including screen overlay attacks that capture UPI PINs

NPCI has responded with device binding (tying UPI to a specific device), AI-based fraud detection and mandatory cooling-off periods for new device registrations. This is probably where the most work is left to do. Social engineering is especially hard to stop because it exploits human psychology rather than technical holes.

Settlement behind the scenes

Every UPI transaction triggers an interbank settlement. NPCI runs a multilateral net settlement process several times a day through RBI, where banks settle their net positions. Resolving billions of transactions into net positions across hundreds of banks takes enormous computational and financial infrastructure. Unglamorous. Essential.

---

Where this goes next

UPI's roadmap is ambitious. A few developments I'm watching:

• Conversational payments: in-chat payments through WhatsApp and other messaging apps (WhatsApp Pay is already live but hasn't gained much traction yet, which might change if WhatsApp pushes harder)
• Offline payments: extending UPI Lite's offline capabilities to areas with poor connectivity
• Cross-border remittances: making UPI the default rail for sending money to and from India
• IoT payments: UPI on smart devices, connected cars and vending machines
• CBDC integration: the Digital Rupee (RBI's central bank digital currency) may eventually run on UPI rails

What this means for everyone else

India built something world-class with UPI. Not perfect. Failure rates need to come down, the revenue model needs sorting out, and fraud prevention needs constant vigilance. As a piece of public digital infrastructure, though, UPI is arguably India's most impressive technology achievement of the 21st century.

The implications reach well past India. UPI showed that a developing country can build financial infrastructure that's better than what wealthier nations have, and it knocked down the assumption that payment innovation has to come from Silicon Valley or private card networks. A non-profit entity in Mumbai built something that Visa and Mastercard, with their combined hundreds of billions in market cap, haven't matched on real-time capability and reach.

For other developing nations, UPI is a blueprint to learn from rather than copy exactly. The preconditions matter (bank accounts, identity systems, smartphone penetration), and so do government support and interoperability from day one. Get those right and you can skip the legacy infrastructure entirely.

That chaiwalla with his QR code is proof that the best technology fades into the background and just works. When people who have never used a computer can send money through a phone app in seconds, something big has shifted. UPI changed how India pays for things, and it probably changed what's possible for any economy willing to build public digital infrastructure with the same ambition.