Is UPI free? For you, scanning a QR code at a tea stall, yes. No charge, no deduction, nothing. But a payment system that moved ₹314 lakh crore last year does not run on goodwill. Somebody pays for every one of those transactions and the bill is estimated at over ₹20,000 crore a year. The people paying are not who you think: taxpayers, banks, loss-making apps and, in ways most people never notice, merchants and you yourself, with your data. This post follows the money through India’s favourite payment system: who pays, who profits and who is quietly collecting the most valuable thing you hand over with every scan.
Note: I am not an economist or a lawyer. This is written from my common understanding as a founder who has spent years around payments, first integrating gateways for client websites and now building a booking platform. If you have a different perspective or if something here is not correct, do discuss it in the comments and correct me.
The scale of the thing
Numbers first, because the scale is what makes the economics matter. UPI processed about 23.2 billion transactions worth ₹29.9 lakh crore in May 2026 alone, roughly 74 crore payments every single day. Over the full financial year 2025-26 it moved around ₹314 lakh crore across more than 24,000 crore transactions. India now accounts for roughly 49% of all real-time payments on the planet, more than the next four countries combined. Over 50 crore people use it and around 6.5 crore merchants accept it.
Two details inside those numbers explain everything that follows. First, 63% of transactions are person to merchant but they carry only 29% of the value, because 86% of merchant payments are below ₹500. UPI is a machine for enormous volumes of tiny payments. Second, the average transaction is about ₹1,300. Small payments are precisely the ones where any percentage fee stings most, which is why the question of who pays was never going to be simple.
How the free machine was built
A quick detour on how we got here, because the architecture explains the economics. NPCI was set up in 2008 by RBI and the banks as an umbrella not-for-profit for retail payments. UPI launched quietly in 2016 as a way to move money between any two bank accounts instantly, using a phone number or a virtual address instead of account numbers. Demonetisation later that year threw fuel on it and the QR code did the rest. One credit is owed before going further. The QR code on the shop counter was not a UPI invention. Paytm had been signing up merchants with QR stickers and wallet payments well before UPI existed and when demonetisation hit, it was wallets that absorbed the first rush. What UPI did was take the habit Paytm proved, unbundle it from a single company’s wallet and put it on an open rail wired straight into bank accounts. Once that happened, the closed wallet era faded on its own. The lesson sits quietly under this whole post: the rail that won was the one nobody could charge for.
The design choice that changed everything was the third-party app model. UPI let technology companies build the customer-facing app while licensed banks did the regulated work underneath. That is why PhonePe, Google Pay and Paytm could reach hundreds of millions of users without becoming banks: each rides on partner banks that connect to NPCI. The split explains the strange economics you will see through this post. The companies that own the customer relationship and the data are not the companies legally responsible for the payment. The companies carrying the regulatory load and the cost cannot charge for it. And the not-for-profit in the middle collects a micro-toll on everything.
What a UPI payment actually costs
A UPI payment feels weightless but it runs on very physical infrastructure: your app, your bank’s UPI switch, NPCI’s central switch, the receiver’s bank, fraud monitoring, dispute systems, customer support and settlement machinery, all built and staffed by someone. The RBI’s own discussion paper on payment charges estimated the cost of processing an average person-to-merchant UPI transaction at about ₹2. Banks bear the biggest share of that, followed by the payment apps and NPCI itself.
Multiply ₹2-ish by 24,000 crore transactions and you see the problem. A parliamentary committee cited an industry estimate of about ₹20,700 crore a year as the cost of running the UPI ecosystem. The figure is debated, but even the conservative versions land in the thousands of crores. Free at the point of use has never meant free. It only means the bill is being paid somewhere you cannot see it.
The law that made it free
Until the end of 2019, digital payments worked on a simple model called MDR, the merchant discount rate: the shop paid a small percentage of each transaction and that fee was split between the banks and networks that made the payment happen. Then the Finance Act of 2019 inserted Section 10A into the Payment and Settlement Systems Act, which from 1 January 2020 prohibited any charge on transactions made through UPI and RuPay debit cards. Not reduced. Prohibited by statute.
The intent was genuine: get small merchants to accept digital payments by removing the one thing they hated about card machines. It worked spectacularly. The QR code conquered India precisely because accepting UPI cost a shopkeeper nothing. But the law did not make the ₹2 per transaction cost disappear. It only made charging for it illegal, which pushed the cost onto whoever could not refuse to carry it.
Who pays, part one: you, the taxpayer
The government partially compensates the ecosystem through an annual incentive scheme for low-value UPI merchant payments. The numbers tell their own story: ₹1,389 crore in FY22, ₹2,210 crore in FY23, ₹3,631 crore in FY24, then a sharp cut to ₹1,046 crore in FY25 as the scheme was narrowed to small merchants and transactions under ₹2,000. About ₹2,000 crore is allocated for FY27. Add it up and the government has paid out roughly ₹8,276 crore over four years.
| Financial year | Government UPI incentive paid |
|---|---|
| FY 2021-22 | ₹1,389 crore |
| FY 2022-23 | ₹2,210 crore |
| FY 2023-24 | ₹3,631 crore |
| FY 2024-25 | ₹1,046 crore |
| FY 2026-27 (allocated) | ~₹2,000 crore |
Hold that against the ₹20,700 crore annual cost estimate and the parliamentary committee’s observation lands hard: government incentives cover around 11% of what the system costs to run. The subsidy is real and it comes from tax money. It is also nowhere near enough. The gap is the interesting part, because someone else is eating it.
Who pays, part two: the banks
Banks are the someone else. Every UPI transaction touches at least two banks and their systems do the heavy lifting: authentication, debits, credits, reversals, disputes, fraud checks. They are barred by law from charging for any of it on UPI and the government incentive covers a fraction. So banks run one of their highest-volume services as a permanent cost centre.
The consequences are predictable and economists have called them out: when a service earns nothing, it gets funded reluctantly. Incentive amounts changed budget to budget, making them impossible for banks to plan around. There is a reasonable argument that banks underinvested in exactly the unglamorous things users need most, fraud monitoring and dispute resolution among them. Anyone who has chased a failed UPI transaction stuck between two banks has felt this underinvestment personally. Free had a price. It was paid in the quality of the plumbing.
Who pays, part three: merchants already pay, it is just not called MDR
Here is the part almost every article on free UPI skips. Zero MDR is a rule about one specific fee. It was never a rule that merchants pay nothing. In practice, large parts of merchant India already pay for UPI in other names.
Start with online businesses. If you sell through a website and your checkout runs on a payment gateway, you pay the gateway’s platform fee on every transaction, UPI included. I documented the actual rates in my payment gateways post: the popular gateways charge in the region of 2% plus GST as a platform or convenience fee and most of them price UPI inside that bundle exactly like cards. One or two price UPI lower, which only proves the point: UPI carries a charge at online checkouts, it is simply invoiced as a platform fee instead of an MDR. The statute banned a name, not the cost. An online seller doing ₹10 lakh a month through a gateway is paying real money on every zero MDR UPI payment, before GST on the fee.
Offline merchants pay through the side door of hardware and subscriptions. The soundbox that announces payments is typically a rented device with a monthly fee. QR stands, EDC machines and the merchant app’s premium tiers all carry charges. One large app alone reports over 90 lakh paid payment devices deployed. When the transaction itself cannot be charged, the ecosystem charges for everything around the transaction.
And the newer rails quietly reintroduce interchange. Pay through a RuPay credit card linked to UPI and normal credit card economics return for the merchant on larger payments. Pay from a wallet through UPI beyond ₹2,000 and an interchange of around 1.1% applies on the merchant side. The zero in zero MDR shrinks every year while the headline stays the same. So when someone tells you merchants pay nothing for UPI, the accurate version is: small offline merchants scanning a plain QR genuinely pay nothing per transaction. Nearly everyone else pays something, somewhere, under a different heading.
The August 2026 change and the ₹2,000 rumour
This year the ground moved. Legislative changes in August 2026 removed the blanket statutory bar that made charging for UPI illegal, which means the government can now design an MDR framework if it chooses. Reports suggest what is under consideration is charging only large merchants, with small merchants and consumers explicitly protected. As I write this, nothing is final: no rate, no threshold, no date.
Which brings us to the message forwarded in every WhatsApp group: “UPI transactions above ₹2,000 will now be charged.” That is not true. The ₹2,000 figure floats around because the government’s incentive scheme uses it as a cutoff and because wallet interchange applies above it, but there is no rule charging your UPI payments above ₹2,000. Person to person transfers are expected to stay free under every version of the proposals. If charges ever come, they are designed to land on large merchants first, not on you at the counter.
For what it is worth, the design question was mapped years ago. RBI’s 2022 discussion paper on payment charges laid out the honest menu: keep everything free and fund it from the budget, allow a small flat fee, allow a percentage MDR with caps, or price by merchant size and transaction value. The paper triggered such an outcry that the government declared UPI would stay free. The menu went back in the drawer. The August 2026 change effectively reopens the drawer. Whatever emerges, one design principle has survived every draft: the person paying should never see a charge, because the moment consumers are charged, cash starts looking attractive again and the whole two-decade project wobbles.
Who profits: the NPCI paradox
Now the question that puzzled me enough to dig into filings. If UPI loses money for everyone, how is NPCI, the organisation that runs it, sitting on a growing surplus? NPCI reported operating revenue of about ₹3,270 crore in FY25 with a surplus of ₹1,552 crore, up 42% in a year, debt free, with a net worth of ₹6,412 crore and a settlement guarantee fund of nearly ₹18,000 crore. For a not-for-profit, that is a remarkably healthy set of numbers.
The answer is position. NPCI sits at the exact centre of Indian payments and earns small processing and switching fees from banks across everything it operates: not just UPI but IMPS, the RuPay network, FASTag tolls, Aadhaar-enabled payments, NACH mandates and cheque clearing. The fees are tiny per transaction, paise not rupees. But NPCI is the only player whose economics improve with volume alone, because it carries none of the customer-facing costs that crush banks and apps: no cashbacks, no support centres, no merchant acquisition, no fraud losses on its own book. Micro-fees multiplied by twenty-four thousand crore transactions produce crores of surplus.
Is that wrong? Not exactly. NPCI is structured as a Section 8 not-for-profit, so the surplus is retained and recycled into infrastructure rather than paid out as dividends. A fat settlement guarantee fund genuinely protects the system. But it is worth being clear-eyed about the architecture: the one institution that profits from free UPI is the one that owns the rails, while the institutions required by law to serve you for free carry the losses. Volume is a business model only for the toll booth.
Who profits: the apps playing the long game
PhonePe and Google Pay together handle around 79% of all UPI transactions, with PhonePe alone at about 46%. Neither earns a rupee of MDR from your payments. So why have they spent years and thousands of crores competing for them? Because a free payment is the cheapest customer acquisition and engagement machine ever built. You open the app several times a day. Everything else follows from that.
PhonePe’s own numbers show the model. Around 86% of its revenue comes from payments-adjacent merchant services: device subscriptions like soundboxes, fees on bill payments and recharges, merchant platform charges. The faster-growing slice is financial services: distributing loans it does not underwrite (over ₹14,000 crore facilitated through more than 50 lending partners), broking insurance (over 1.8 crore policies sold), gold, mutual funds and a stock broking arm. Revenue crossed ₹7,000 crore in FY25. And it still ran an operating loss of about ₹1,444 crore in the first half of FY26. Google Pay and Paytm run versions of the same play: free payments in front, monetisation of attention, merchants and credit behind.
Read that model again and notice what powers the profitable parts. Which loan to offer you, which insurance nudge you see, which merchant gets which offer: these decisions run on knowing you. Which is the real answer to this post’s third question.
The two-app problem
Free payments produced one more thing worth naming: concentration. When nobody can compete on price, competition shifts to marketing budgets and staying power, which is a game only the deepest pockets win. Result: two foreign-owned apps, PhonePe under the Walmart group and Google Pay under Alphabet, together carry roughly four out of every five UPI transactions in India. NPCI saw the risk years ago and proposed a 30% market share cap per app, but enforcing it would mean deliberately degrading service for crores of users, so the deadline has been extended again and again. The cap exists on paper while the concentration exists in reality.
Why should you care? Because concentration turns an app outage into a national payments outage and an app’s business decisions into de facto policy for half the country’s transactions. India built a public payment rail and then let its front doors consolidate into a duopoly that funds itself by monetising users. The one encouraging sign in recent data: the combined share of the top two has finally slipped just below 80% as newer apps claw in. Watch that number. It matters more than most payment news.
Who gets your data: what the apps can see
Every UPI payment you make writes a line in a ledger somewhere: who you paid, how much, when, from which bank account, on which device, often with your location if you granted the permission. One transaction is trivia. Five years of transactions is a biography. Your salary date, your rent, your school fee, your chemist, your bar, your EMIs, the city you slept in last night. No survey, no credit bureau, no government census has ever held a picture of you this intimate, updated in real time. You built it yourself, one scan at a time.
Make it concrete with one ordinary week. Monday morning, a metro recharge and a coffee near an office: employer district known. Monday evening, groceries at the same store as every week: home locality known. Tuesday, a pharmacy payment and a diagnostic lab: a health event inferred. Friday night, a bar bill: lifestyle noted. Sunday, a temple donation and a transfer to a parent: family and faith sketched in. First of the month, rent to a landlord and an EMI to a lender: obligations mapped. No single payment said anything. The pattern said everything. Now extend that week across five years and 500 million people and you understand what the payment apps are actually holding.
The apps sitting on this are not neutral pipes. PhonePe belongs to the Walmart group. Google Pay belongs to the world’s largest advertising company. Paytm is a listed company hunting for profitability. Each of them runs lending, insurance and commerce businesses whose margins depend on knowing which user to pitch what. The payment is free. The knowledge of you is the asset.
Who gets your data: the business of knowing you
How is the asset used? The clearest case is credit. Transaction history is the new credit score: a lender looking at months of UPI inflows and outflows can estimate income, spot salary regularity and judge repayment capacity better than a bureau score for the crores of Indians who never had a credit card. Industry surveys say a large majority of lenders now combine traditional bureau data with this kind of alternative data for underwriting. When a payments app offers you a pre-approved loan within seconds, that speed is your own transaction graph being read back to you with interest attached.
Then there is cross-selling. The app that sees your motor insurance debit knows when your renewal is due. The app that sees your rent knows you might want a deposit loan. The app that sees your travel bookings knows when to show forex or travel insurance. None of this is hypothetical, it is the stated financial services strategy of every major UPI app and it is why the loss-making payments layer is worth funding. On top of that sits plain commercial analytics: aggregated spending patterns, merchant intelligence and offer targeting, businesses built from the exhaust of your daily payments.
Who gets your data: the rules, such as they are
There are rules and they have been tightening. RBI requires all payment data to be stored only in India. NPCI has repeatedly stepped in on conduct: its 2025 guidelines restrict how apps may call sensitive APIs, requiring that balance checks be customer-initiated and that account validation happen only with explicit consent. After apps were caught porting users’ UPI numbers without permission, NPCI made explicit consent mandatory for that too. The Digital Personal Data Protection Act, 2023 now sits over the whole ecosystem, making the apps and banks accountable as data fiduciaries and processors, with duties around consent, purpose limitation, breach notification and deletion. I wrote about the same law in the context of my APAAR ID petition: the pattern where convenience arrives first and consent is retrofitted later is not unique to payments.
There is also a glimpse of how consent-first design could look, from within Indian finance itself. The account aggregator framework, RBI’s system for sharing financial data, works on the opposite principle to app onboarding: data flows only when you approve a specific request, for a specific purpose, for a specific time, revocably. Nothing about UPI apps’ internal use of your transaction history works that way today. It runs on the one-time, everything, forever consent you gave at signup. The gap between those two models is exactly the distance between consent as a right and consent as a checkbox.
The DPDP Act rights are worth knowing in practical terms. You can ask a payment app what personal data it holds, demand correction, withdraw consent and seek erasure of data no longer required for the purpose you consented to. The apps must name a grievance officer and respond within defined timelines once the rules fully bite. None of this deletes your bank’s transaction records, which banking law requires them to keep. But it does give you, for the first time, a legal handle on the profile the app has built around those transactions.
The Google Pay episode is worth remembering as a case study. In 2020, media reports claimed the app’s terms allowed sharing customer data with third parties, Google publicly denied sharing transaction data for ads while a petition in the Delhi High Court pressed the question of whether the app was violating guidelines. What the episode surfaced was the uncomfortable official position: sharing with third parties is not banned outright, it is permitted within the rules with consent. And consent in an app is a checkbox most people never read. The honest summary of your protections: stronger on paper than three years ago, still resting almost entirely on consent screens designed to be accepted.
What you can actually do about it
Realistic steps, not paranoia. Read the permissions your payment app holds and remove what it does not need, location and contacts first. Say no when the app asks to become your default for SMS or notifications beyond payments. Treat the lending and insurance tabs as advertisements, not advice, because that is what they are. If an app has ported or seeded your UPI number somewhere you did not ask for, you now have the right to refuse and complain, so use it. Consider keeping one app for payments and doing your financial products research outside it, so the entity that knows your cash flow is not also the one selling you credit. And spread meaningful payments across two apps if the idea of a single company holding your complete transaction biography bothers you. It probably should.
The export nobody prices
One more wrinkle in the economics: UPI is now an export. Through NPCI’s international arm, UPI payments work in places like Singapore, the UAE, Sri Lanka, Mauritius, Nepal, Bhutan and at merchants in France, with more corridors in the pipeline. India pitches UPI abroad as digital public infrastructure, a soft-power story with real strategic value. The unresolved question travels with it: a system that runs on subsidies and cross-subsidies at home has to find a sustainable pricing model before it can run the world’s payments. Countries adopting the stack will ask who pays. India’s own answer is still “everyone, invisibly”. Getting the domestic economics honest is not just a local accounting exercise. It decides whether the export has a business model.
What this means if you run a small business
Bringing it home for the people who actually read this blog. If you sell offline on a plain QR, UPI remains genuinely free per transaction and no live proposal changes that for you, so ignore the scary forwards. Price the soundbox and app subscriptions as what they are, your real cost of accepting UPI. Negotiate them like any other vendor contract. If you sell online, read your gateway invoice and know your all-in UPI cost after platform fees and GST. My payment gateways post has the comparison table, including the one gateway that prices UPI meaningfully lower.
Watch the large-merchant MDR debate with mild interest rather than alarm: if a fee comes, it is aimed at large turnover businesses and every proposal so far protects small merchants. Keep your settlement account clean, one business account receiving all UPI settlements, which is the same discipline from my current account post. And remember that your business UPI data profiles you exactly the way personal data does: the app watching your settlements knows your revenue curve better than your CA and will price loan offers to you accordingly. Read those offers as negotiations, not favours.
So is UPI free? My take
Is UPI free? You now know the honest answer has three layers. At the counter, yes. For person to person transfers it should stay that way. Behind the counter, no: taxpayers fund a shrinking subsidy, banks eat losses that show up as underfunded fraud and dispute systems, online merchants already pay gateway fees on UPI under a different name and offline merchants pay through devices and subscriptions. And beneath all of it, you pay in kind, with a transaction graph that powers the lending and cross-selling machines of some of the largest companies on earth.
My view, for what it is worth: the zero MDR experiment achieved something genuinely historic and it is also an economic distortion that cannot run forever on hidden cross-subsidies. A transparent, capped fee on large merchants, with small merchants and P2P permanently free and the incentive money redirected to fraud and dispute infrastructure, would be more honest than the current arrangement where everyone pays and nobody sees the bill. And on data, the ask is simple: purpose limitation with teeth, so that the app processing my payment needs fresh, genuine consent before my transaction history becomes a lending file. Free should not quietly mean paid for in personal data.
FAQ on UPI charges and costs
Is UPI free for customers?
Yes. Paying by UPI costs you nothing per transaction and person to person transfers are free. No current proposal changes this, whatever forwarded messages claim.
Is UPI free for merchants?
Offline merchants accepting payments on a plain QR pay no per-transaction fee, thanks to the zero MDR mandate. But online merchants pay their payment gateway’s platform fee on UPI transactions, device users pay soundbox and equipment subscriptions and wallet or RuPay credit card payments via UPI carry interchange on larger amounts. Free applies to one slice of merchant India, not all of it.
Will UPI payments above ₹2,000 be charged?
No such rule exists. The ₹2,000 figure comes from the government incentive scheme’s cutoff and from wallet interchange thresholds, not from any charge on your payments. What is under discussion after the August 2026 legal change is a possible fee on large merchants, with consumers and small merchants protected. Nothing is final.
Who pays for UPI infrastructure?
Banks bear most of the running cost, the government contributes an incentive that covers roughly a tenth of the industry’s cost estimate, NPCI funds the central rails from its fees and the apps subsidise their share from investor money while monetising users through other products.
Why is NPCI profitable when UPI is free?
NPCI earns small switching and processing fees from banks across all its rails: UPI, IMPS, RuPay, FASTag, AePS and NACH. It carries none of the customer-facing costs. Tiny fees multiplied by twenty-four thousand crore transactions produced a surplus of about ₹1,552 crore in FY25. As a not-for-profit it retains the surplus rather than paying dividends.
How do PhonePe and Google Pay make money if UPI is free?
Payments are their customer acquisition engine. The money comes from merchant devices and subscriptions, bill payment fees, distributing loans and insurance, gold, broking and advertising. PhonePe crossed ₹7,000 crore in revenue in FY25 and still ran operating losses, which tells you how expensive the free payments layer is and how valuable they expect the attached businesses to become.
Is UPI free on RuPay credit cards?
For you as the payer, yes. But a RuPay credit card payment through UPI brings credit card interchange back on the merchant side for larger transactions, which is why some merchants hesitate to accept credit card UPI even while happily taking bank-account UPI.
Do UPI apps sell my data?
The apps deny selling personal data and the rules require consent and purpose limits. The more precise concern is use, not sale: your transaction history informs lending decisions, insurance nudges and offer targeting inside the app group’s own businesses, all resting on consents you clicked during onboarding. RBI’s localisation rules keep the data in India and the DPDP Act gives you rights over it, including deletion.
What is MDR in UPI?
MDR, the merchant discount rate, is the percentage fee a merchant pays on a digital payment, shared among the banks and network that process it. Cards still carry MDR. UPI and RuPay debit cards have had zero MDR by law since January 2020, which is what people mean by UPI being free for merchants.
Do banks get nothing at all from UPI?
Not quite nothing. Banks save on cash handling, keep deposits in accounts that UPI makes sticky and learn from transaction flows for their own lending decisions. But those indirect gains do not cover the direct processing costs at current volumes, which is why banks have lobbied consistently for MDR to return in some form.
Can the government make UPI chargeable now?
Legally, yes. The August 2026 legislative change removed the statutory prohibition, which is exactly why the design debate has begun. Politically, charging consumers or small merchants looks close to impossible, which is why every serious proposal targets large merchants only.
Is UPI used outside India?
Yes. UPI works in several countries through NPCI’s international arm, including Singapore, the UAE, Sri Lanka, Mauritius, Nepal and Bhutan, with acceptance at some merchants in France and more corridors being added. Indian travellers can pay by UPI at supported merchants abroad, though bank and app support varies by corridor.
Was zero MDR a mistake?
It bought India the fastest digital payment adoption in history, so calling it a mistake is too easy. It also created real damage: underfunded fraud and dispute systems, a two-app market and costs hidden in taxes, devices and data. The question now is not whether free was worth it, but whether the bill should finally be printed honestly.
Over to you
The next time someone says UPI is free, you can give them the full picture: free for you, subsidised by taxes, absorbed by banks, invoiced to merchants under other names and paid for by all of us in data. That is not a scandal. It is just how a free thing at this scale actually works and it deserves to be understood rather than forwarded.
Do you run a business that accepts UPI, online or offline? What do you actually pay, all in, once gateway fees and devices are counted? Do comment on the post with your numbers. And if any figure or rule above needs correcting, say so in the comments. I would rather fix the post than let a wrong number float around.
Table of Content
Join my newsletter
Stay up to date on recent posts, updates along with tips and hacks to grow your startup profitably.
