The Sovereign Utility Pivot: How India Payments Transformed Zero-Fee Velocity into a $3.75 Trillion Financialization Engine

1. The Kirana Counter Fracture: Zero-MDR and the Upstream Infrastructure Bet

Play the scene out in any of several million Indian shopfronts. It is seven in the evening in a lane off Chandni Chowk in Old Delhi. The kirana store — the neighbourhood grocery, ten feet wide, stocked floor to ceiling — is three customers deep. Someone buys fifty rupees of loose tea and holds up a phone. The shopkeeper does not look at it. He does not have a hand free, and in any case he learned two years ago that a screenshot of a payment confirmation can be faked in about four seconds. Instead, a small plastic box wired above the counter speaks: "Paytm par 50 rupaye praapt hue." Fifty rupees received. The customer leaves. The next one steps up.

That box is the most interesting object in Indian finance, and the reason is arithmetic. The fifty-rupee payment that just cleared travelled across the Unified Payments Interface, the national real-time payment switch operated by the National Payments Corporation of India. In the fiscal year ending March 2026, that switch carried 24,161.69 crore transactions — 241.6 billion of them — worth ₹314.23 lakh crore, roughly $3.75 trillion.¹ The merchant discount rate charged on that fifty-rupee payment, and on essentially every one of those 241.6 billion transactions, was zero. Not low. Not competitive. Zero, by sovereign policy.²

So the payment itself generated no revenue for anyone. The box speaking above the counter generates ₹90 to ₹125 a month, every month, for as long as the shopkeeper keeps it plugged in. India's payment industry spent six years discovering that it could not sell the water flowing through the pipe, and would have to sell the plumbing, the meter, and eventually the mortgage.

Why we looked here

The proposition that made this industry worth a month of work is a single falsifiable claim: that in India, digital payment velocity functions as the foundational data-generation infrastructure for financialising a 1.4-billion-person consumer economy, so that the economic value of a payments franchise is determined not by the fees on payments but by the credit, subscriptions and cross-border flows that verified transaction history unlocks. If that belief is right, a company processing a sixth of the volume can earn several times the revenue of the volume leader, and the industry's profit pool sits almost entirely outside the payment itself. If it is wrong, the whole sector is a state-subsidised utility with venture-capital cost structures bolted on, and the correct expected return on its equity is unpleasant.

Three independent evidence streams pointed the same direction. The first is sovereign infrastructure data. UPI volume rose from 18,586.60 crore transactions in FY25 to 24,161.69 crore in FY26, a 29.9% increase, with value up 27.8% to ₹314.23 lakh crore; digital instruments now account for the clear majority of retail transaction count in India.¹ Growth is decelerating from triple digits, which is what a maturing utility does — and utilities with 240 billion annual touchpoints are interesting for reasons other than their growth rate.

The second stream is merchant hardware. Roughly 10.5 million active soundbox subscriptions were deployed across India by mid-2026, up from around 8.5 million a year earlier, with shopkeepers paying ₹90–125 a month for a device whose only function is to say a number out loud.³ These are people who resisted a 1% card fee for two decades. They pay a subscription because the box solves a labour problem — verification without looking — rather than a payment problem.

The third stream sits outside the industry entirely. Fast-moving consumer goods distributors report that a large majority of kirana inventory orders now settle digitally rather than in cash; quick-commerce platforms are digital by construction; and non-bank lenders including Aditya Birla Capital ($ABCAPITAL) and Poonawalla Fincorp ($POONAWALLA) attribute a meaningful and growing share of new merchant loan origination to transaction telemetry sourced from payment platforms rather than to conventional financial statements. These third-stream figures come from industry channel estimates compiled for this study rather than from audited disclosure, and we treat them as directional corroboration, not measurement.

The transmission mechanism is the part worth holding on to. Payment volume operates as a customer acquisition vehicle priced at zero, which is a strange but real advantage: it means acquisition cost is subsidised by the state and by the interoperability mandate, and the acquired asset is not the payment but the record. A shopkeeper with eighteen months of daily settlement history has, for the first time, a cash-flow statement — one that cannot be dressed up before a loan application, because it is generated by the same rail that moves the money. That record is what a lender buys, what a subscription is sold against, and what a cross-border processor underwrites. The same belief, followed one step further, implicates Indian retail banking, the non-bank lending complex, quick-commerce, and the electronics assembly base that builds the terminals — each of which appears here only where it changes the payment industry's economics.

Here is the shape of the utility everyone is trying to monetise around.

Exhibit 1 — UPI transaction volume and value, India, FY2020-21 to FY2025-26 Units: volume in crore transactions (1 crore = 10 million); value in lakh crore rupees (1 lakh crore = ₹1 trillion). Fiscal years end 31 March. Source: NPCI ecosystem statistics. Evidence status: reported operator data.¹

Read that table aloud and two things happen. First, the absolute numbers are absurd: volume grew nearly eleven-fold in five years, from 22 billion transactions to 242 billion, and India now runs more real-time retail payments than the rest of the world combined. Second, the growth rate has halved and halved again — 106%, 82%, 57%, 42%, 30%. The average ticket size has also drifted down, from about ₹1,840 in FY21 to roughly ₹1,300 in FY26, because the marginal new transaction is a ten-rupee chai rather than a rent payment. An industry whose revenue scaled with transaction count would be watching that deceleration nervously. An industry whose revenue scales with merchants monetised barely cares, because the count of paying merchants is on a completely different, much earlier curve.

That gap — between an adoption curve in its late innings and a monetisation curve in its first — is the entire investment question. To understand why the fee is zero and why it is likely to stay near zero, you have to understand who built the rail, and what they built it to prevent.

Subscribe on LinkedIn

↑ Back to Top

2. India Stack Architecture: Sovereign Rails and Regulatory Switches

Most countries let payment rails emerge from the private sector and then regulate the resulting rent. India did the opposite. It built the rail as public infrastructure and regulated the rent to nearly nothing before it could form.

The architecture is a stack, and each layer solved a specific problem in sequence. Aadhaar gave more than a billion residents a verifiable digital identity. Electronic know-your-customer turned a multi-week, paper-based account-opening process into a minutes-long authentication. DigiLocker made government documents portable. And UPI, launched in 2016, put an interoperable real-time payment protocol on top, so that any account at any participating bank could pay any other account instantly, addressed by a human-readable handle rather than an account number.¹ Collectively this is India Stack, and its defining feature is that the identity, consent and payment layers are public goods rather than private property.

Compare that to the alternatives. In the United States and Europe, retail payment rails run through card networks — Visa ($V) and Mastercard ($MA) — that sit as private toll operators between issuing and acquiring banks, and take a percentage of every transaction for the privilege. In China, the dominant rails grew inside two closed consumer ecosystems, where the wallet and the merchant network belonged to the same company and interoperability arrived late and under regulatory pressure. India's designers looked at both and built something that resembled neither: a switch owned by a not-for-profit consortium of banks, operating under the Reserve Bank of India's authority, with mandatory interoperability written into the protocol.

Mechanically, UPI is a two-tier system. At the centre sits NPCI, running the switch that routes and clears every transaction. Connected to it are sponsor banks — payment service providers such as HDFC Bank ($HDFCBANK) ($HDFCBANK), ICICI Bank ($ICICIBANK) ($ICICIBANK), Axis Bank ($AXISBANK) ($AXISBANK) and State Bank of India ($SBIN) ($SBIN) — which hold the actual accounts and bear the settlement obligation. Sitting on top of the sponsor banks are third-party application providers, or TPAPs: PhonePe, Google Pay, Paytm, WhatsApp Pay, Navi, CRED, Amazon Pay. The TPAP owns the interface, the user relationship and the fraud-detection layer. It does not own the account, the money, or the rail.

A useful analogy: UPI is a municipal water pipeline owned by the state; the apps are the faucets installed in homes and shops. The faucet manufacturer cannot meter the water. It can only compete on how good the faucet is, and sell you things that attach to it. The analogy has a limit worth stating, because it misleads in one important way: unlike a water utility, this pipeline generates a stream of information about every household's consumption, and the faucet manufacturer is allowed — within consent rules — to see and use its own customers' portion of that stream. The value is in the telemetry, not the flow.

Interoperability is the sharpest edge of the design and the most underrated fact about the industry's competitive structure. A QR code pasted on a shop counter by one company can be paid by any other company's app. There is no closed loop to defend. A merchant who dislikes his provider does not need to change anything about how he accepts money; the QR sticker keeps working. This single regulatory choice destroyed the network-effect moat that payment companies enjoy in almost every other large market, and it explains why so much of the strategic energy in India has gone into physical hardware and lending relationships. Those are the only things a competitor cannot route around with a software update.

Who actually enforces this

Three institutions determine outcomes here, and none of them is a company.

The Reserve Bank of India regulates under the Payment and Settlement Systems Act of 2007, and it regulates with unusual granularity. Its Master Directions on Payment Aggregators, updated in September 2025, require any entity aggregating merchant payments to maintain a minimum net worth of ₹25 crore, to conduct documented background checks on the merchants it onboards, to tokenise rather than store card credentials, and — critically — to hold all merchant funds in escrow accounts at scheduled commercial banks rather than on its own balance sheet.² That last requirement is the structural reason Indian payment companies cannot become quasi-banks by accident. The float belongs to the escrow, and the escrow lives at a bank.

NPCI, owned by a consortium of banks and operating under RBI's guidance, runs the switch, the RuPay card network and the IMPS rail. It also holds a power that no private participant has: it decides who may be a TPAP, and on what terms. Its most consequential and most repeatedly deferred rule is the 30% market-share cap on any single third-party app's UPI volume, first announced in November 2020 to prevent an entrenched duopoly, and pushed back most recently to a compliance deadline of 31 December 2026.¹ More on that rule and its unresolved mechanics later; for now, note that a company's right to keep growing its core business is a discretionary grant.

The Ministry of Finance provides the money that makes zero-MDR politically survivable. It allocates an annual budgetary subsidy to compensate acquiring banks and payment providers for processing UPI and RuPay debit volume for free — an allocation that has run in a band of roughly ₹1,500 crore to ₹3,500 crore a year. Set that against the traffic it is meant to subsidise: on ₹314 lakh crore of UPI value, ₹2,500 crore of subsidy is under one basis point. Industry participants have consistently argued it covers well under a quarter of the real infrastructure and fraud-management cost of running the rails. That gap is not an accounting curiosity. It is a permanent, state-imposed operating loss on the payment function itself, and every business model in this article exists to fund it from somewhere else.

One company in this story sits closest to that squeeze and shows what happens when you cannot fund it: Fino Payments Bank ($FINOPB), a payments bank running an asset-light network of banking agents and micro-ATMs across rural and semi-urban India. Its business was built on charging fees for cash-in, cash-out and domestic remittance — precisely the services that a free, instant, interoperable transfer rail makes unnecessary. It is the industry's clearest displaced incumbent, and we return to what its numbers show in section nine.

The architecture, then, prevents rent extraction at the switch and at the interface. Innovation and profit were pushed to the endpoints. But that pressure did not arrive gradually. It arrived on a specific morning, by government order.

↑ Back to Top

3. The Zero-MDR Trap and the Falsifiable Profit-Migration Thesis

On 1 January 2020, the merchant discount rate on UPI and RuPay debit card transactions in India went to zero by law. Not by competition. Not by a price war. By statute, applied to every merchant and every acquirer simultaneously, with no opt-out.²

Understand what that did to a business plan. Every payment company in India had modelled the same S-curve: acquire merchants at a loss, subsidise consumers with cashback, wait for volume to compound, then harvest a fee of somewhere between 0.3% and 1.8% on the resulting flow. Overnight the harvest line was deleted from the model while the acquisition line stayed exactly where it was. Companies that had spent thousands of crores acquiring merchants discovered they had built distribution to a product with no price.

This is the single most important distinction for anyone underwriting the sector: zero-MDR is structural, not cyclical. A cyclical price collapse mean-reverts when capacity exits. A sovereign choice about whether a piece of public infrastructure should carry a toll reverts only when the sovereign changes its mind, and Indian governments of both major political persuasions have treated free digital payments as a consumer-welfare achievement worth defending. Any thesis that requires MDR restoration as its base case is a thesis about politics wearing the costume of a thesis about business.

So the industry had to migrate its profit pool. That migration is the falsifiable claim this article tests, and it runs as a chain:

Zero-margin payment volume creates merchant reach at state-subsidised acquisition cost. Merchant reach is converted into physical lock-in through hardware — soundboxes at the small end, Android terminals at the organised-retail end — which carries a monthly subscription price the merchant will actually pay because it solves an operational problem rather than a payment problem. Hardware and gateway presence generate continuous, verified transaction telemetry. Telemetry underwrites credit distribution — merchant working capital, consumer credit lines routed over UPI — at origination commissions that dwarf any plausible payment fee. And the resulting mix converts chronic operating burn into positive EBITDA and free cash flow.

Each link is observable and each can fail. If merchants churn off hardware when a rival offers a cheaper box, link two breaks. If lenders decide payment telemetry is a worse underwriting signal than they hoped, link four breaks and takes the margin with it. If the state caps subscription fees or distribution commissions, the chain breaks at whichever link the rule touches. The clock is roughly three years: by fiscal 2029 the base-case version of this chain requires monetisation metrics that are one to two orders of magnitude above where they sit in mid-2026, which we quantify in section ten.

The capital cycle that forced the pivot

The migration was not a strategy chosen from a menu. It was chosen at gunpoint by the capital markets, in three distinct phases.

The first phase, roughly 2016 to 2021, was a capital super-cycle. SoftBank, Tiger Global, Ant Group and Sequoia's India arm funded an all-out land grab. The money went into consumer cashback, merchant acquisition subsidies and free processing. Return on invested capital across the sector was comprehensively negative, and everyone knew it; the argument was that share bought now would be monetisable later, an argument that zero-MDR had already quietly falsified in January 2020 while the funding taps were still open.

The second phase, 2022 to 2024, was the cleanout. Global rates rose, growth capital repriced, and Indian fintech valuations in the private market compressed hard. Simultaneously the RBI tightened. Digital lending guidelines constrained how fintechs could originate loans; card tokenisation rules forced expensive re-plumbing; the payment aggregator licensing regime put minimum net worth and merchant due-diligence obligations on entities that had been operating on growth-first assumptions. Companies cut headcount, rationalised employee stock plans, killed cashback and started reporting contribution margin. Several business models simply did not survive the transition from funded growth to self-funded operations.

The third phase, from 2025 into 2026, is the harvest — and it is the phase that makes this an investable theme rather than an interesting one. Operating leverage finally asserted itself. Cloud and engineering costs are largely fixed; subscription and commission revenue is not. One97 Communications, the listed parent of Paytm ($PAYTM), swung from a consolidated net loss of ₹663 crore in FY25 to a net profit of ₹552 crore in FY26 — a ₹1,215 crore swing — with EBITDA turning from ₹1,506 crore negative to ₹502 crore positive on operating revenue of ₹8,437 crore.³ Razorpay completed its corporate re-domiciliation from Delaware to India in May 2025 and by mid-2026 had pre-filed confidentially with the Securities and Exchange Board of India for a domestic listing.⁵ Pine Labs executed the same reverse migration to prepare for an Indian listing.

Three phases, one lesson. The industry now has a burden of proof it did not have in 2021: public markets in Mumbai are asking for audited profit and cash conversion, not gross merchandise value. That is a healthier discipline and a harder game, and it split the field into two monetisation battlegrounds — one fought at the physical counter, one fought inside the checkout page.

↑ Back to Top

4. The Hardware SaaS War: Soundbox Monetization and Kirana Counter Lock-in

The insight that rescued Indian merchant acquiring was behavioural, not technical, and it is worth stating precisely because it is so easily misread.

Small Indian merchants did not refuse card machines because they hated technology. They refused because a 1% fee on a ₹20 cup of tea is a rounding error to the processor and a visible, itemised, recurring tax to a shopkeeper working on a 6% gross margin. Fees on transactions are salient and feel extractive. But those same merchants had a genuine and expensive operational problem: with a phone-based QR code, confirming that money had actually arrived required stopping work, picking up a phone, unlocking it, opening an app and reading a screen — during the busiest hour of the day, while a queue formed, while an employee who might or might not be trustworthy handled the counter. And a customer could show a screenshot of a payment that never happened.

Paytm's answer, introduced in 2019, was a cellular-connected speaker. It has a SIM card, a battery, a speaker, and effectively no other capability. It announces the amount received, in the local language, loudly enough to hear over street noise. It costs the merchant ₹90 to ₹125 a month.³

Merchants who would not pay a 1% transaction fee pay the subscription without much argument, because the subscription is priced against staff time and shrinkage rather than against sales. On a shop turning over ₹4 lakh a month, ₹100 is 0.025% of throughput. The same merchant refusing 1% was refusing ₹4,000. That is the whole trick: the industry re-priced its product from a percentage of the merchant's revenue to a fixed fee against the merchant's labour cost, and the fixed fee turned out to be far more defensible.

The unit economics, and where they are contested

A soundbox costs roughly $12–15 to build, assembled by contract electronics manufacturers. Industry participants have identified Foxconn and Dixon Technologies ($DIXON) ($DIXON) among the assemblers supplying Indian payment companies with soundbox and Android terminal hardware, alongside specialist terminal vendors such as Pax; the specific allocation of volume between them is not publicly disclosed, and we treat named supplier-customer pairings as reported rather than audited. Semiconductor and cellular-module lead times set the deployment ceiling — a company can hire a thousand salespeople in a quarter, but it cannot conjure two million radios — which makes hardware sourcing a genuine throughput constraint on merchant acquisition, not merely a cost line.

Against a device cost of roughly ₹1,100–1,300, a ₹100 monthly subscription with better than 60% gross margin after cellular data costs returns about ₹60–75 a month of contribution. That implies a hardware payback closer to sixteen to twenty months than the eight-to-ten months the industry commonly cites. The reconciliation matters: the shorter payback figure is achievable only if devices are placed with an upfront rental or refundable deposit collected at installation, if device cost is materially below the quoted range at scale, or if the payback calculation is being credited with downstream lending and settlement revenue from the same merchant. All three are plausible; none is separately disclosed. An investor underwriting soundbox economics should treat the payback period as a contested input and the churn rate as the observable that actually matters, because a device that stays plugged in for four years pays back on any of these assumptions and a device that comes back in nine months pays back on none of them.

Reported monthly churn runs below 1.2%, which annualises to roughly 13% — high retention for a hardware subscription sold to micro-merchants with no contractual lock-in.³ The reason is workflow integration rather than switching cost in the classical sense. The QR code is interoperable and can be replaced in an afternoon. The box, however, has become the mechanism by which the owner supervises the counter when he is not standing at it. Removing it does not change how customers pay; it changes how the owner runs his shop. That is a switching cost measured in operational habit, and it is the closest thing to a durable moat anyone in Indian payments has constructed.

Who leads, on what, and by how much

Paytm leads the soundbox category on installed base, and the size of that lead is genuinely contested. The internally consistent reading of the available evidence is an industry active base of roughly 10.5 million devices in mid-2026, with Paytm holding around 60% — call it 6 to 7 million units.³ A separate line of reporting credits Paytm alone with about 10.5 million devices and PhonePe with 4.2 million, which cannot be reconciled with the industry total unless the denominators differ. They almost certainly do: "deployed" (cumulative shipped) and "active" (currently subscribing and transmitting) are different metrics, and companies are not consistent about which they quote. BharatPe, the third significant competitor, is credited with roughly 1.5 million audio units. The honest summary is that Paytm is the clear leader on active subscribing devices, PhonePe is a substantial and fast-growing second, and any precise share figure below the level of "Paytm roughly 55–65%" should be treated as a vendor estimate rather than a measurement.

Why Paytm rather than PhonePe, given that PhonePe has three times the consumer relationships? Because the soundbox is sold, installed, explained, serviced and collected on by human beings. Paytm built a field sales organisation running to over ten thousand people, structured around dense route-based coverage of individual market clusters — the same logistical shape as an FMCG distribution network, applied to a payment device. That organisation is slow to build, expensive to run, and it is the actual asset. Software companies can copy a plastic speaker in a quarter. Copying a functioning ground force that can install two hundred thousand devices a month across four hundred cities takes years and tolerates a level of operational grind that most engineering-led organisations find culturally intolerable. PhonePe has been closing the gap by leaning on its consumer brand for lead generation, which lowers acquisition cost but does not remove the servicing requirement.

At the other end of the merchant spectrum, Pine Labs ($PINELABS.NS) leads the organised-retail terminal category with roughly 550,000 Android point-of-sale units. Its lead rests on something different again: exclusive brand-EMI arrangements with major consumer electronics manufacturers, so that a shopper buying a phone or a television at an organised retailer gets the no-cost instalment offer processed through a Pine Labs terminal. That is a two-sided distribution asset — the brand needs the terminal footprint to deliver its financing promotion, and the retailer needs the terminal to close the sale — and it is much harder to dislodge than a QR sticker. It is also a fundamentally different customer: a few hundred thousand organised-retail lanes with high average ticket sizes, rather than several million kirana counters with ticket sizes under ₹200.

What the hardware pivot did to the P&L

Exhibit 2 — One97 Communications (Paytm), consolidated results, FY2024-25 vs FY2025-26 Units: ₹ crore. Fiscal years end 31 March. Consolidated, audited. Source: One97 Communications FY26 results and investor presentation. Evidence status: audited company disclosure.³

Say those numbers out loud and the shape is unmistakable: revenue grew 22%, and EBITDA improved by ₹2,008 crore. That is not a revenue story. Costs came down by more than revenue went up. The FY25 base included the aftershocks of the regulatory action on Paytm Payments Bank and the associated merchant and user attrition, so part of the swing is a recovery from a self-inflicted trough rather than pure operating leverage. But the direction is real and the cash position — over ₹13,000 crore, more than an eighth of the company's market value — means the turnaround does not depend on capital markets staying friendly.

Set against that, the counter-example in the same category. PhonePe, majority owned by Walmart ($WMT), reported revenue of ₹7,920 crore in FY26 on 11.5% growth, with a GAAP net loss of ₹2,792 crore driven substantially by non-cash employee stock compensation and goodwill amortisation.⁷ Two companies, near-identical revenue bases, opposite bottom lines — and the volume leader is the one losing money. Understanding why requires leaving the physical counter and going online, where the fees still exist.

↑ Back to Top

5. Online Gateways, PA-CB, and Developer APIs: The Monetization of Digital Checkout

Before 2015, integrating a payment gateway into an Indian website was a procurement exercise. A startup founder submitted physical documents to a bank, waited four to six weeks for a merchant account approval, received integration documentation of variable quality, and then discovered that a meaningful fraction of transactions failed at the bank's end with no useful error message. The gateway was a bank product sold on bank timelines to companies the bank did not particularly want.

Razorpay's founders built the opposite: a self-serve signup, digital know-your-customer, an API a developer could integrate in an afternoon, and — the part that actually mattered commercially — obsessive engineering on transaction success rates. In Indian online payments, the difference between an 82% and a 91% success rate is the difference between a merchant's revenue and a merchant's complaint queue, because failures are dominated by flaky bank endpoints rather than by customer behaviour. A gateway that intelligently retries and routes around a struggling issuer bank is worth paying a premium for, and it is a capability that compounds: more volume produces better real-time data on which bank endpoint is degrading, which produces better routing, which produces higher success rates.

That combination gave Razorpay a reported share above 50% of India's startup and small-business online payment gateway volume, against roughly 25% for PayU India.⁵ Both figures deserve a caveat that applies across this entire subsector: unlisted payment aggregators do not publish audited transaction volume broken out by customer segment, so segment share claims are vendor-defined and self-reported. Razorpay's claim of SMB leadership and PayU's claim of parity in enterprise processing are both defensible on their own definitions and are not directly comparable. Treat them as credibility-tested analytical estimates, dated mid-2026, rather than as measurement.

Where the fees still live

The reason the online layer matters at all is that zero-MDR applies to UPI and RuPay debit, and not to everything else. Credit cards, most debit routing, net banking, corporate cards and buy-now-pay-later instruments all still carry a merchant fee. The aggregator's net take on the non-UPI portion of the mix runs roughly 15 to 35 basis points after interchange and network costs are passed through. On cross-border flows the economics improve dramatically, because foreign-exchange spread, compliance handling and settlement complexity command 100 to 150 basis points.

Exhibit 3 — Where economic value sits in the Indian payments value chain, mid-2026 Take rates are per-transaction economics to the named layer; gross margin is the layer's typical contribution after directly attributable cost of service. India only. Source: industry take-rate and margin ranges compiled for this study from company disclosure and regulated fee schedules. Evidence status: analytical estimate, not audited; ranges rather than point values.

The exhibit tells the story in one glance: the layer with all the volume has none of the margin, and the layer with all the margin has almost none of the volume. Move down the table and per-transaction economics improve by two to three orders of magnitude, from zero on a UPI payment to several hundred basis points on a loan origination. Every strategic decision in this industry over the past five years has been an attempt to move down that table while keeping the customer acquired at the top of it. The catch, which section six deals with, is that the bottom row is where the credit risk lives — even when it sits on someone else's balance sheet.

Enterprise, government, and the gross-versus-net trap

While Razorpay was winning the startups, a different company was quietly processing the country's airline tickets, hotel bookings and municipal tax payments. CCAvenue, operated by the listed AvenuesAI Limited ($INFIBEAM) — formerly Infibeam Avenues ($INFIBEAM) — has more than two decades of enterprise banking integrations and a dominant position in exactly the flows that are tedious to win and hard to lose: government portals, utilities, hospitality chains, airlines. Enterprise customers negotiate take rates down aggressively, but they transact in high tickets and in volume, and they change providers roughly never, because the integration touches their reconciliation systems.

AvenuesAI reported FY26 revenue of ₹8,115.8 crore, up 103.3%, with EBITDA of ₹387.3 crore and net profit of ₹332 crore, on a debt-free balance sheet.⁴ Put that next to Paytm's ₹8,437 crore and the naive conclusion is that two companies of similar size sit at opposite ends of profitability. The conclusion is wrong, and the reason is the single most important accounting trap in payments analysis. Aggregators that book payment processing on a gross basis run the merchant's full transaction settlement value, or a large pass-through component of it, through the revenue line; the economic revenue is the spread retained. AvenuesAI's 4.77% EBITDA margin on gross-basis revenue and Paytm's 5.95% on a substantially net-basis revenue base are not comparable numbers, and neither is their growth: revenue that doubles because a large enterprise mandate was onboarded is a different event from revenue that grows 22% on subscription and commission expansion. Compare gross profit, not revenue, whenever the disclosure permits it — and in this subsector it frequently does not.

BillDesk occupies the adjacent enterprise bill-payment niche, processing recurring utility, insurance and loan repayment collections at institutional scale. Its relevance to the theme is as a reminder that a large share of India's payment value moves through unglamorous recurring-billing plumbing rather than through consumer apps, and that this plumbing is priced like infrastructure rather than like software.

Cashfree Payments is the specialist at the opposite corner: outbound. Where most aggregators are built to collect money, Cashfree built its position on automated bulk payouts — vendor settlements, refunds, marketplace seller disbursements, insurance claim payments — through APIs, plus cross-border rails for global platforms paying Indian sellers. It has raised over $95 million and remains private with no audited public disclosure, so its economics cannot be compared with the listed names. Its strategic significance is that payouts are stickier than collections: a marketplace that has wired its seller settlement into one provider's API is not switching over a five-basis-point quote.

The cross-border inflection

The most interesting near-term margin development in the online layer is regulatory. The RBI created an explicit licensing category for cross-border payment aggregation — PA-CB — covering inward collections for exporters and outward payments for importers, and issued authorisations to major aggregators including Razorpay and Pine Labs in early 2026.² Before this, cross-border flows for Indian software exporters, e-commerce sellers and freelancers ran through a patchwork of bank arrangements and offshore intermediaries with poor documentation and slow settlement.

The commercial logic is straightforward. India exports services and, increasingly, direct-to-consumer goods; millions of small businesses and independent professionals receive foreign currency. Those flows carry 100 to 150 basis points against 15 to 35 domestically, and the customer is far less price-sensitive because the alternative is a bank wire with a worse rate and a three-day delay. If a licensed aggregator can move even a modest share of its existing merchant base onto cross-border rails, the blended take rate on its book improves without winning a single new customer. This is the strongest available counter to the consensus fear that online gateway pricing trends inexorably toward zero.

Razorpay's own financials show why the fear exists and why it is incomplete. Its FY25 revenue was ₹3,783 crore, up 65%, with a reported net loss of ₹1,209 crore driven substantially by one-time tax and restructuring costs from the Delaware-to-India re-domiciliation, and with its core payments operation described as EBITDA-positive.⁵ Note the fiscal mismatch: this is FY25 against FY26 numbers elsewhere in this article, because Razorpay is private and its FY26 audited accounts are not in the public domain. Note also that a ₹1,209 crore charge to move a company's legal home is a real cash cost to shareholders even when it is genuinely non-recurring — the "reverse flip" that Razorpay and Pine Labs both executed carried a tax bill on the appreciation in the offshore holding structure, and it was paid to make an Indian listing possible.

Behind all of these companies stands the same set of counterparties, and it is worth naming the dependency plainly. Every payment aggregator in India, without exception, is required by RBI directions to settle merchant funds through escrow accounts held at scheduled commercial banks; HDFC Bank, Axis Bank and ICICI Bank are among the principal escrow and nodal sponsors for Razorpay, Paytm, PayU and Cashfree.² The aggregator has customers and technology; the bank has the licence that lets money legally sit anywhere overnight. If a sponsor bank's own regulatory standing deteriorates, or if it decides an aggregator's merchant book carries unacceptable compliance risk, the aggregator's operations are affected within days. This is a real, documented single-point dependency, and it has already been demonstrated in practice, as section eight describes.

On the other side of the same ledger, aggregators are themselves suppliers — to the card networks and to lenders. Visa, Mastercard and RuPay supply the card rails on which non-UPI take rates depend, and receive routed volume in return; when interchange schedules change, aggregator gross margin changes with them, and the aggregator has no negotiating position whatsoever over a network's published rate. And the aggregator supplies merchant transaction data to lenders. Which brings us to where the actual money is.

↑ Back to Top

6. Credit on UPI and the Lending Super-Rail: Monetizing Telemetry

For its first several years, UPI moved only money a person already had. The account had to be funded; the transaction was a debit. That single design fact capped the industry's economics, because a debit rail carrying no fee produces no revenue for anyone, ever, regardless of scale.

The RBI's decision to permit credit on UPI changed the instrument, and it changed it in two distinct ways that are frequently conflated. The first is card-on-UPI: linking a RuPay credit card to a UPI handle, so that a scan at any QR code draws on the card rather than on a bank balance. The second is credit-line-on-UPI: a bank pre-sanctions a credit line to a customer, and the line becomes a payment source inside the UPI app with no card involved at all. Both routes accomplish the same commercially critical thing. They reintroduce interchange.¹

Credit-instrument transactions on UPI above the small-ticket exemption threshold carry an interchange of roughly 1.50% to 2.00%, split among issuer, network and acquirer. Against a zero-MDR debit transaction, that is the difference between a rail that costs money to run and a rail that pays for itself several times over. Suddenly the 60 million merchant QR codes that took years and billions of rupees to deploy become credit acceptance points — without a single new terminal, sticker or merchant conversation.

The scale of the opportunity is best understood by how far it currently is from being realised. Credit routed over UPI was running at roughly ₹2,500 crore a month in mid-2026, growing at better than 200% year on year.¹ Annualise that and you get about ₹30,000 crore — against total UPI value of ₹314.23 lakh crore. Credit is therefore running at approximately one-tenth of one percent of the rail it sits on. The industry's base case for fiscal 2029 has credit at 10–12% of UPI value. That is not an extrapolation of a growth rate; it is a requirement for roughly a hundred-fold increase in absolute credit volume in three years. It may happen — 200% annual growth compounds to about 27x over three years, which gets you to 3% penetration, not 10% — but the base case as commonly stated demands more than the current trajectory delivers. This is, in our judgement, the single most demanding assumption embedded in the sector's optimistic scenarios, and section ten treats it as the first crux KPI for exactly that reason.

The other credit rail: merchant working capital

The second and, today, larger financialisation channel does not run over UPI at all. It runs over the data.

A kirana merchant with a soundbox generates a continuous, timestamped, tamper-resistant record of daily receipts. So does an online merchant on a payment gateway. Traditional lenders could never underwrite these businesses, because the businesses have no audited accounts, frequently no formal registration, and no collateral a lender wants. What they now have is verified cash flow, observed by the same party that could originate the loan.

The economics of that origination are the highest-margin activity in the industry. The payment platform does not lend. It acts, in regulatory terms, as a distributor — sourcing the borrower, presenting the offer inside an app the merchant already opens every day, and handing the file to a lending partner. Non-bank lenders including Aditya Birla Capital ($ABCAPITAL), Poonawalla Fincorp ($POONAWALLA) and KrazyBee supply the balance-sheet capital and absorb the credit risk; the platform earns an upfront origination commission in the range of 2.5% to 4.0% of the disbursed amount, sometimes plus a collection-linked share. Gross margin on that commission is 70% or better, because the marginal cost of showing an offer to a merchant already inside your app is close to nothing.

Walk the dependency in both directions, because it defines the theme's fragility. The lender needs the platform for two things it cannot replicate: distribution reach into merchants no branch network profitably serves, and the transaction telemetry that makes the underwriting possible at all. The platform needs the lender for the one thing it is structurally forbidden from providing: risk capital. Under RBI's digital lending framework, arrangements in which a distributor guarantees a portion of the lender's losses — first loss default guarantees — are capped at 5% of the loan portfolio. That cap is the industry's most consequential and least discussed regulation. It protects payment companies from taking large credit losses onto balance sheets that were never capitalised for them. It also caps how much of the credit spread they can ever capture, because a party bearing 5% of the risk does not get to keep 100% of the return.

The consequence is a business model with fee-like economics and lender-like exposure to someone else's risk appetite. If merchant loan performance deteriorates, the platform's revenue does not fall because it took losses. It falls because the lender stops funding originations. The revenue line is therefore hostage to a credit metric the payment company does not control and does not report.

That metric, as of mid-2026, reads about 3.1% gross non-performing assets on payments-led merchant origination — respectable for uncollateralised lending to micro-enterprises, and derived from partner lender disclosure rather than from the payment platforms themselves. It has not yet been tested through a genuine downturn in Indian small-business activity. Uncollateralised small-ticket credit in India has historically behaved well for several years and then not, and the reasons for the "not" are usually correlated across the whole industry: a monsoon, a policy shock, a liquidity squeeze at the non-bank lenders. Anyone underwriting the financialisation thesis is, whether they price it or not, underwriting a non-bank credit cycle.

There is a second-order concern that a sharp analyst would raise here. Payment telemetry is an excellent measure of a merchant's revenue. It is a poor measure of the merchant's obligations — the informal supplier credit, the family borrowing, the gold loan against the same shop. As payment-led lending scales, multiple platforms observe the same merchant's receipts, each concludes independently that the merchant can service a loan, and none of them sees the others. This is the classic mechanism by which good underwriting signals produce bad portfolios at scale, and it is exactly the pattern India's credit bureaus were built to catch — for formal loans, with formal identifiers. Whether the coverage extends cleanly to this cohort is not established by available evidence.

For the moment, though, the arithmetic works, and it explains why the sector's profit inflected in FY26 rather than FY24. Consider Paytm's mix: an active soundbox base of six to seven million devices at roughly ₹100 a month generates on the order of ₹750–850 crore of annual subscription revenue, which is around a tenth of its ₹8,437 crore revenue base but a much larger share of gross profit given 55–65% margins. Layer distribution commissions on merchant and consumer credit at 70–85% margins, and a company can reach positive EBITDA on revenue growth of 22% while its transaction volume grows faster and earns nothing. That is the thesis working. It is also, note carefully, a thesis in which the payment business is a cost centre that buys distribution — which means the payment leader and the profit leader need not be, and currently are not, the same company.

↑ Back to Top

7. Parameter-Specific Leadership: How the Leaders Built Their Leads

There is no leader of Indian payments. There are five leaders of five different businesses that happen to share a rail, and confusing them is the most common analytical error in the sector. Let us take them one at a time, each with its parameter, its date, its closest rival, and the reason the lead exists.

Consumer UPI transaction volume. PhonePe led with roughly 46.5% of UPI volume in mid-2026 — about 112 billion transactions annually — against Google Pay at roughly 33.2% and Paytm at 7–8%.¹ The measurement is NPCI's monthly TPAP volume report, which is the one genuinely independent market-share series in this industry, and it counts transactions initiated, not value settled or revenue earned.

How PhonePe built that lead: it was acquired by Flipkart in 2016 and came under Walmart's ownership through the Flipkart transaction, which gave it patient capital and an e-commerce distribution channel at exactly the moment UPI was launching. While competitors were still defending wallet balances — stored-value products that UPI made obsolete — PhonePe went all-in on interoperable UPI and on the mundane transactions that create daily habit: mobile recharges, electricity bills, gas cylinder payments. It invested heavily in app reliability in low-bandwidth conditions and in transaction success rates, which matters disproportionately in tier-2 through tier-6 cities where a failed payment at a counter is socially embarrassing and permanently churns a user. Google Pay competed with a superior engineering organisation and an Android distribution advantage but has never matched PhonePe's rural depth.

Why has nobody copied it? Largely because it is not copyable — it is an accumulated habit across hundreds of millions of users, protected by nothing except inertia. Which is precisely the problem, and it brings us to the number that reframes this entire ranking.

PhonePe generated ₹7,920 crore of revenue in FY26 across approximately 112 billion UPI transactions. That is about seventy paise — under one US cent — of company revenue per UPI transaction processed. Paytm generated ₹8,437 crore across roughly 18 billion UPI transactions, or about ₹4.65 per transaction: more than six times as much revenue per transaction on one-sixth the volume.³⁷ The ratio is crude, and it should be labelled as such: neither company's revenue is purely UPI-derived, PhonePe's includes insurance and wealth distribution, Paytm's includes commerce and lending, and the two are on the same fiscal year but not on identical revenue recognition. Even after every one of those caveats, a six-fold gap in revenue per transaction is not a rounding difference. It is the measurement of what zero-MDR does to a company that leads on volume and has not yet built the layers underneath.

Merchant soundbox installed base. Paytm leads, as section four established, on roughly 55–65% of an approximately 10.5 million active-device industry base, with PhonePe second and BharatPe third.³ The lead came from inventing the category in 2019 and from a field organisation of over ten thousand people. What could erase it: the soundbox is not technically defensible, and PhonePe has both a larger consumer brand and a lower cost of merchant lead generation. The category leader here is being attacked by a company with a structurally cheaper funnel, which is why churn and average revenue per user, rather than installed base, are the metrics that will show damage first.

Small-business online payment gateway. Razorpay leads on reported SMB gateway volume share above 50%, with PayU India the closest rival at around 25%; both figures are vendor-defined and dated mid-2026.⁵ The lead was built on developer experience and transaction success rates, and has been deepened by product expansion into adjacent merchant workflows — business banking, payroll, vendor payouts — that raise the cost of leaving beyond the checkout integration itself. This is a real, if modest, moat: not a network effect, but a switching cost created by having quietly become the merchant's financial operating system.

PayU India, owned by Prosus ($PRX), is worth understanding as more than a runner-up. It has been the more aggressive builder of credit at checkout, through PayU Finance and the LazyPay buy-now-pay-later product, which places it earlier on the same financialisation path Paytm walked — with the important difference that PayU's Indian credit ambitions have run into the same digital lending tightening everyone else faced. Prosus reports fintech revenue above $1.1 billion globally, of which PayU India is a substantial but not separately quantified part.⁶

Enterprise and government gateway. CCAvenue, under AvenuesAI, leads on enterprise and government processing, with BillDesk the closest comparable in recurring bill payments.⁴ The lead is twenty years of integrations into systems that are painful to touch — airline reservation platforms, municipal revenue portals, hotel property management systems. Enterprise procurement in India rewards incumbency and audit trail over interface quality, which is exactly the opposite of the selection criterion that made Razorpay the SMB leader. Two companies, opposite advantages, same product category, different customers.

Organised-retail terminal footprint. Pine Labs leads with roughly 550,000 Android terminals and the brand-EMI relationships that make them valuable. Its exposure to the theme is different from everyone else's: it monetises high-ticket discretionary purchases and instalment financing, which makes it more cyclical and more sensitive to consumer credit availability than a soundbox business selling to grocers.

Exhibit 4 — Consumer UPI volume share by third-party app, India, mid-2026 Units: percent of monthly UPI transaction count (not value, not revenue). Source: NPCI monthly TPAP volume reports. Evidence status: reported operator data; shares fluctuate month to month.¹

Two facts jump out of that table. First, two companies control roughly 80% of India's national payment rail, and neither is Indian-owned — which is the entire reason NPCI's 30% cap rule exists. Second, and less obviously, the "others" bucket is where the interesting strategic action sits. Navi has been growing volume share from a small base under NPCI's explicit desire for diversification. CRED occupies a narrow, wealthy niche built on credit card bill payment, which gives it the most affluent user base in Indian payments and, correspondingly, the highest-value cohort for credit and wealth products despite trivial volume share. Amazon Pay serves its parent's commerce checkout more than it competes for open-market UPI volume. WhatsApp Pay has enormous latent distribution and has consistently failed to convert it, which is itself evidence that distribution alone does not win payment share in an interoperable market.

The composite picture is a market where competitive advantage is segmented by layer rather than concentrated in one firm, and where the segment with the most volume has the least pricing power. That structure is stable only as long as the regulator lets it be — and the regulator has already demonstrated, twice, that it will not.

↑ Back to Top

8. Regulatory Battles, NPCI's Duopoly Dilemma, and the Capital Cycle

At the end of January 2024, the Reserve Bank of India ($BANKINDIA) directed Paytm Payments Bank to stop accepting fresh deposits and top-ups in its accounts and wallets after a specified date, citing persistent non-compliance and supervisory concerns.² The action did not name a fine or a fraud. It simply switched off a bank.

The consequences were immediate and instructive. Paytm's listed parent lost roughly half its market value within days. Millions of merchant settlement accounts and consumer wallets had been hosted at the payments bank; those relationships had to be rebuilt at other institutions or lost. The company migrated its nodal and settlement arrangements to third-party scheduled commercial banks — Axis Bank, HDFC Bank and Yes Bank among them — a technical re-plumbing that in normal circumstances takes a year and was executed under existential pressure in weeks.

What the episode actually demonstrated is worth separating from the drama. The RBI's stated concern centred on know-your-customer documentation and on whether the payments bank was operating at genuine arm's length from the app company that shared its brand, its founder and much of its customer base. The regulator's implicit doctrine — that a payment application and a licensed bank may not be functionally the same organisation wearing two hats — is now understood by every participant in the industry. It is the reason no Indian payment company today argues for vertical integration into deposit-taking, and the reason the escrow-at-a-scheduled-bank requirement is treated as non-negotiable rather than as a cost to be optimised.

It also demonstrated the recovery path. In August 2025, the RBI granted a full payment aggregator licence to Paytm Payment Services, removing the restriction that had prevented the company from onboarding new online merchants during the overhang period.² PayU India's aggregator authorisation was similarly resolved in the same broad window. Only after that licence was restored could Paytm run a full year of unconstrained merchant acquisition — which is a material part of why FY26 and not FY25 was the turnaround year, and a caution against reading the entire ₹2,008 crore EBITDA swing as operating leverage on the soundbox model.

For an investor, the durable lesson is about the shape of the risk rather than the size of the drawdown. Regulatory risk in Indian payments is not a tail event that reduces earnings by a percentage. It is a binary switch applied to a specific licence, with a recovery period measured in quarters to years, and it arrives without a negotiation phase. It should be underwritten as an option that has been written, not as a volatility input.

The cap that nobody knows how to enforce

The second live regulatory battle is structural rather than punitive. NPCI's rule that no single third-party application may exceed 30% of UPI transaction volume was announced in November 2020 with a phased compliance path. PhonePe sits near 46.5% and Google Pay near 33.2%; between them they carry roughly four-fifths of India's national payment traffic.¹ The rule has never been enforced. The compliance deadline has been extended repeatedly, most recently to 31 December 2026.

The reason for the repeated deferral is that nobody has published a workable mechanism. There are only three ways to hold an app under a volume cap. You can stop it onboarding new users, which penalises growth but does nothing about the existing base and takes years to bind. You can throttle or decline transactions once a monthly threshold is crossed, which means telling a real person at a real shop counter that their payment failed for regulatory reasons — an outcome no regulator in a country that has made free instant payments a point of national pride wants to own. Or you can redirect volume, which requires the consumer to have and want an alternative app. NPCI has not published a technical roadmap resolving this, and the absence of one is the strongest available evidence about the probability of hard enforcement on the stated date.

That said, treating the cap as permanently unenforceable is its own error. The rule's mere existence has already shaped behaviour: it is why NPCI has encouraged new entrants, why Navi's share growth is welcomed rather than resisted, and why neither leader can plan on compounding share indefinitely. And the distributional consequence if it ever binds is large and asymmetric. Volume displaced from PhonePe and Google Pay has to land somewhere, and the beneficiaries would be the sub-scale apps — Paytm, Navi, CRED, Amazon Pay — none of which earns anything on the incremental UPI transaction. Which produces the sector's most elegant irony: strict enforcement of the cap would transfer enormous transaction volume to companies that make no money from transaction volume, while damaging the two companies whose entire strategic asset is the consumer relationship that volume represents. It is a redistribution of cost, dressed as a redistribution of market share, and its second-order benefit — access to more consumers for credit and subscription distribution — accrues only to whoever has already built those layers.

The capital cycle, repriced

Set that regulatory picture beside the funding picture, because the two have converged in a way that defines the current moment.

In 2021, an Indian payment company could raise at a revenue multiple in the high teens on the strength of processed volume. In 2026, the exit route runs through Indian public markets, and Indian public markets have priced payment companies with an unusual combination of scepticism and enthusiasm. Razorpay's confidential pre-filing targets a raise of $500–600 million at a valuation in the $5–6 billion range.⁵ Pine Labs pursued the same domestic path after its own re-domiciliation. PhonePe, which had been positioned for a listing at a valuation reported around $15 billion, paused its offering in mid-2026 citing market conditions.⁷

Note what the sequencing says. Two companies that generate identified, fee-bearing revenue — an aggregator and a terminal business — proceeded toward listing. The company with by far the largest consumer franchise and the weakest per-transaction monetisation paused. Public markets are not paying for volume in this cycle. They are paying for evidence of capture.

The capital cycle framework applies awkwardly to payments because there is no factory. But there is a real capacity analogue: field sales headcount and deployed hardware, both of which are capital-hungry and both of which were built out aggressively between 2023 and 2026. The industry is in a deployment phase rather than an overbuild — installed devices are still growing and average revenue per device has been stable rather than collapsing. The evidence to watch for a shift into shakeout is a soundbox price war: if PhonePe or BharatPe decides that ₹49 a month buys share worth having, the entire hardware SaaS profit pool compresses toward zero for everyone, including the leader, and Paytm's installed-base advantage becomes an installed-base liability with a depreciation schedule attached. Nothing in the current evidence shows that happening. Everything in the current market structure — three well-funded competitors, an undifferentiated product, no contractual lock-in — says it could.

↑ Back to Top

9. Expectations, Variant Views, Scenarios, and Portfolio Exposure

The decision context here is a general institutional public-equity mandate: global listed expressions, a multi-year thematic horizon, no position sizing and no recommendation. That framing matters more in this theme than in most, because the single most likely way to be right about India payments and lose money is to buy the correct thesis at the wrong layer of the capital structure or the wrong point of the expectations curve.

Start with the layers where the theme is real but the exposure is not.

Walmart owns the majority of PhonePe, India's largest payment application by volume, and PhonePe contributes an estimated 2–3% of Walmart's economics. At a mooted $15 billion standalone value against Walmart's roughly $580 billion market capitalisation, PhonePe is under 3% of the parent — genuine option value, entirely swamped by the operating performance of US retail. Prosus is the same problem in a different currency: PayU India is a strong asset, but Prosus's equity value has long been dominated by its stake in Tencent, and PayU India represents an estimated 8–10% of the relevant fintech exposure inside a holding company whose share price is driven by an entirely different Asian equity.⁶⁷ Alphabet's ownership of Google Pay, the second-largest UPI app in the world's largest real-time payment market, is not a measurable line item in Alphabet's disclosure at all. These three are the theme's canonical false positives — not because the businesses are weak, but because available disclosure does not establish measurable exposure at the level of the listed security. Owning them to express this theme is buying a lottery ticket stapled to an unrelated balance sheet.

Exhibit 5 — Listed expressions: exposure, purity and the expectations question, as of 2026-08-17 Market capitalisations as of mid-August 2026. "Payment revenue purity" is the share of company revenue attributable to payments and payment-adjacent activity. Source: company disclosure and market data compiled for this study. Evidence status: purity figures are analytical estimates from disclosed segment mix, not audited segment reporting.³⁴⁶⁷

Read down the purity column and the investable universe collapses fast. There is exactly one large, liquid, high-purity listed expression of this theme in the world, and it is Paytm. Everything else is either small, diluted, or displaced. That scarcity is itself a risk: a single-name theme concentrates all the sector's regulatory, credit-cycle and multiple-compression risk into one security, and it means the pending listings of Razorpay, Pine Labs and eventually PhonePe are the most important supply event in the theme's future.

Three expectations gaps

Paytm. The market's evident concern is that the FY26 profit is a recovery bounce off a regulatory trough rather than a durable earnings base, and that soundbox competition from a better-capitalised PhonePe will compress the highest-margin recurring line. The variant case is that soundbox retention — under 1.2% monthly churn on an installed base with no contract — reflects workflow dependence that survives price competition, and that lending distribution commissions compound on the same merchant relationship at 70%-plus gross margins.³

But the security analysis has to be done separately from the business analysis, and here the arithmetic is sobering. At a market capitalisation of roughly ₹1.04 trillion against FY26 net profit of ₹552 crore, the shares trade at approximately 188 times trailing earnings; stripping the ₹13,000-plus crore cash balance still leaves enterprise value at roughly 180 times FY26 EBITDA.³ What does that require? Our own illustrative calculation — assumptions stated, not a forecast: if a holder requires a 12% annual return and the shares are worth 25 times earnings in five years, the company needs roughly ₹7,300 crore of net profit in fiscal 2031, against ₹552 crore today. That is about 68% compound annual profit growth for five consecutive years, which at a 20% net margin implies revenue somewhere above ₹36,000 crore, more than four times the FY26 base. Paytm may well grow into a fraction of that. The point is that the FY26 turnaround, impressive as it is, is not what the price is paying for. The price is paying for the credit-distribution and subscription flywheel scaling by an order of magnitude, and the observable evidence for that scaling is early. The first smart reason a senior investor rejects this security is not the business — it is that the expectations embedded in the price leave almost no room for a regulatory interruption, a credit cycle, or a soundbox price war, and this industry has produced all three within the last four years.

PhonePe. The consensus reading is that a 46.5% share of the world's largest real-time payment rail commands a double-digit-billion valuation more or less automatically. The variant view starts from the seventy paise of revenue per transaction established earlier, and from a FY26 GAAP net loss of ₹2,792 crore against 11.5% revenue growth in a year when the underlying rail grew 29.9%.¹⁷ Revenue growing at roughly a third of the pace of the volume it rides on is the defining fact about this company, and it says the monetisation layers are being built more slowly than the volume base is expanding.

Management has emphasised contribution-margin positivity and cash generation in core payments, and has attributed the GAAP loss substantially to non-cash employee stock compensation and goodwill amortisation. Both framings can be true simultaneously and the distinction is not cosmetic: stock compensation is a real transfer of value from existing shareholders even when it is non-cash, and a listing prospectus is where that dilution becomes explicit. The honest position as of August 2026 is that PhonePe's payment operations appear self-funding and its monetisation build-out is unfinished, and that the paused IPO removes the disclosure that would let anyone verify either claim. What would make it merit deeper work: audited disclosure of merchant subscription revenue and lending distribution revenue as separate lines, at scale comparable to Paytm's.

Razorpay. Consensus fears that online gateway pricing grinds toward zero as it has in several other markets. The variant case rests on mix rather than price: cross-border processing at 100–150 basis points, payout APIs with high switching costs, and business-banking products that make the gateway the merchant's default financial interface.⁵ The rejection test here is disclosure. Razorpay is private; its most recent public revenue figure is FY25's ₹3,783 crore, and its profitability claim — payments EBITDA-positive with the loss attributable to one-time re-domiciliation costs — has not been independently verifiable at the segment level. A prospectus will settle it. Until then, the sensible position is that the business is interesting and the security does not yet exist.

Three worlds, not three numbers

The FY27–FY29 scenarios below describe different causal paths, not different growth assumptions applied to the same path.

In the bear world (we assign it roughly 20% probability), zero-MDR becomes permanent and politically untouchable, and the state additionally scrutinises the substitutes: capping hardware subscription fees as a merchant-protection measure or restricting distribution commissions on loans. A soundbox price war breaks out because at least one well-funded competitor decides share matters more than margin, and average revenue per device falls below ₹40 a month. UPI volume growth decelerates toward 12% as saturation binds. Credit on UPI stalls below 3% of value because banks discover that pre-sanctioned lines to thin-file customers behave badly. Sector EBITDA margins compress below 2%, and the profit pool stays with the banks that own the balance sheets. Note what makes this world coherent: the losses are caused by the same force in each link — the state deciding that payments are infrastructure and infrastructure should not have owners who earn returns.

In the base world (roughly 60%), UPI volume compounds at about 22% and stops being the interesting variable. Soundbox pricing holds at ₹90–120 because three competitors reach an unspoken understanding that a price war destroys a pool none of them can rebuild. A narrow, selective MDR appears — something like 15 basis points on commercial UPI above a ₹2,000 threshold — as the fiscal cost of the subsidy becomes untenable. Credit on UPI reaches a meaningful but not transformational share of value. The top three operators reach 15–20% EBITDA margins, driven by mix rather than price. This world requires the state to accept that a permanently loss-making acceptance layer will eventually degrade, and to fix it with the smallest intervention that works.

In the bull world (roughly 20%), a broader MDR of around 25 basis points returns to commercial UPI, and the effect is immediate and enormous because the acceptance infrastructure is already built and fully depreciated. Credit exceeds 20% of UPI value as bank risk appetite catches up with the rail's convenience. Cross-border and hardware attach rates rise together. EBITDA margins reach 28–35% and the market re-rates payment operators onto software multiples. Sanity-check the MDR component: the claim that a tiered MDR unlocks $1.5 billion of gross profit implies roughly ₹13,000 crore of new fees, which at 15 basis points requires about ₹87 lakh crore of qualifying merchant volume — roughly 28% of total UPI value passing the threshold and the commercial test. That is a demanding but not absurd assumption, and it is worth stating explicitly rather than accepting the headline number.

Across all three worlds, one exposure is common and frequently unhedged: every listed and pre-listed expression of this theme is levered to the same non-bank credit cycle, the same regulator, and the same domestic equity flows. A basket of Indian payment names is one bet expressed several times.

↑ Back to Top

10. Crux KPIs, Thesis Kill Criteria, and Future Value Migration

If you tracked only four numbers in this industry, these would be the four. Each is upstream of revenue rather than a restatement of it, each discriminates between the bull and bear worlds described above, and each has a threshold at which the argument changes rather than merely weakens.

1. Credit routed over UPI, monthly run-rate. What it measures: the rupee volume of transactions on UPI funded by a credit instrument — RuPay credit cards linked to UPI and pre-sanctioned credit lines — reported monthly by NPCI and by partner banks.¹ Why it leads: interchange revenue is a mechanical function of credit volume, so this number moves one to two quarters before any payment company's interchange or distribution revenue does, and it moves before banks disclose changed risk appetite. What it discriminates: the bull case holds that UPI credit substitutes for traditional card spending at 60 million acceptance points that cards never reached; the bear case holds that bank risk committees will cap line sizes for thin-file borrowers and the product stays a niche for existing prime cardholders. Latest reading: approximately ₹2,500 crore a month in mid-2026, growing above 200% year on year, which annualises to roughly 0.1% of UPI value. Confirmation: a sustained run-rate above ₹5,000 crore a month during fiscal 2027. Break: growth stalling below ₹1,200 crore a month, which would indicate lenders pulling back rather than users declining to adopt. This is the KPI that tests the upstream belief most directly, because it measures whether payment velocity actually converts into credit.

2. Soundbox average revenue per device and monthly retention. What it measures: monthly subscription revenue per active device, in rupees, and the percentage of devices that stop subscribing each month; disclosed in quarterly filings and investor presentations by Paytm and, less consistently, by PhonePe.³ Why it leads: pricing moves before revenue and long before margin, because an installed base repriced downward takes two to three quarters to show up in reported subscription revenue. It also sits directly on the binding constraint, which is whether hardware lock-in is real or rented. What it discriminates: the bull case says workflow integration produces pricing power in a commodity device; the bear case says three funded competitors selling an identical box eventually compete on price. Latest reading: ₹90–125 a month with monthly retention above 98%. Confirmation: average revenue per device sustained above ₹100 with net quarterly additions above 400,000. Break: average revenue below ₹50 a month, or monthly churn above 3%, either of which would mean the physical moat is not a moat.

3. Cross-border processing as a share of aggregator gross profit. What it measures: the proportion of a payment aggregator's gross profit — not revenue — derived from PA-CB licensed cross-border flows, disclosed at investor days and, once listings complete, in prospectuses and segment reporting.² Why it leads: mix shift shows up in gross profit before it shows up in blended take rate or in net margin, and it is the earliest observable that an aggregator has found a defensible price floor. What it discriminates: the bull case says high-value international flows insulate aggregators from domestic price compression; the bear case says regulatory friction and compliance cost cap the addressable volume. Latest reading: growing quickly from a small base following the early-2026 authorisations, with no company having disclosed a quantified gross-profit contribution — which is itself a limitation worth stating rather than papering over. Confirmation: cross-border exceeding 15% of gateway gross profit at a disclosing operator. Break: a regulatory restriction on cross-border flows, or the failure of any listed operator to disclose the line at all after listing, which would be its own kind of answer.

4. Gross non-performing assets on payments-led merchant credit. What it measures: the percentage of loans originated through payment platform distribution that are non-performing, reported quarterly by the partner lenders — Aditya Birla Capital, Poonawalla Fincorp and peers — rather than by the payment companies themselves. Why it leads: origination volume is set by lender risk appetite, and lender risk appetite responds to portfolio performance one to two quarters before the payment platform's distribution commission revenue reflects it. A payment company's lending revenue does not decline because it took losses; it declines because someone else decided to stop lending. What it discriminates: the bull case says real-time transaction telemetry is a genuinely superior underwriting signal for businesses with no financial statements; the bear case says uncollateralised micro-enterprise credit behaves well until it does not, and correlates across the whole sector when it turns. Latest reading: approximately 3.1% in fiscal 2026, untested through a downturn. Confirmation: holding below 4.0% through a full macro cycle. Break: above 7.0%, at which point co-lending capital withdraws and the highest-margin layer of the value chain closes.

Two of these four are company-controlled and two are not, which is the correct proportion. Theme kill criteria and security kill criteria differ: a soundbox price war kills Paytm's earnings without killing the theme, whereas a policy decision to cap subscription fees or distribution commissions kills the theme regardless of who executes well.

Where the value could move next

A tiered merchant discount rate, plausibly fiscal 2027–2028. The mechanism is fiscal rather than commercial: the state subsidises free payments at a scale that covers well under a quarter of the real cost of running the acceptance layer, and that gap grows with volume. A narrow reintroduction — 15 to 20 basis points on commercial merchant transactions above a threshold, exempting small merchants and small tickets — would restore direct fee revenue to an acceptance network that is already built, staffed and depreciated. Almost the entire incremental fee would fall through to profit. The beneficiaries are the acquirers with the largest merchant bases: Paytm, PhonePe, the bank acquirers and the terminal operators. The adoption hurdle is not technical but political, and the observable milestone is a Ministry of Finance budget document or an RBI discussion paper putting a threshold and a rate on paper. Nothing of that sort has been published as of August 2026, and every previous rumour of it has been denied.

The retail e-rupee, fiscal 2027–2029. The RBI's retail central bank digital currency pilot has been running for several years and its consequential feature is offline capability — a payment instrument that clears without connectivity and without commercial bank intermediation.² If it scales, the winners and losers are non-obvious. Commercial banks lose deposit float and settlement revenue. Card networks lose the most, because a sovereign digital bearer instrument competes directly with the one rail that still charges interchange. Payment applications could gain or lose depending on whether the RBI distributes the e-rupee through existing apps — which it has done in the pilot — or through a state-operated wallet. The observable milestone is whether retail e-rupee circulation crosses from pilot scale into figures NPCI reports alongside UPI. It has not.

Biometric and palm-based authentication at the counter, fiscal 2026–2028. Aadhaar already provides a national biometric identity layer, so authenticating a payment by palm or fingerprint rather than by phone is an integration problem rather than an invention. The commercial mechanism is that it removes the customer's smartphone from the transaction entirely, which extends digital acceptance to the cohort that still cannot participate, and it deepens hardware lock-in because the merchant now needs a capable terminal rather than a printed sticker. Razorpay and Paytm have both worked on terminal-based biometric acceptance. The distinction to hold onto is between announced products and scaled deployment: nothing here has reached commercial scale, and the privacy politics of biometric payment authentication in India are unsettled.

Enforcement of the 30% cap, nominally 31 December 2026. Covered above; the milestone to watch is the publication of a technical enforcement mechanism, not the deadline itself. A deadline without a mechanism has been extended five times.

Does the belief still hold?

Return to the proposition we started with: that in India, digital payment velocity is the foundational data-generation infrastructure for financialising a 1.4-billion-person consumer economy, and that the value of a payments franchise is set by what the data unlocks rather than by what the payment earns.

The evidence as of August 2026 says the belief is holding, and holding in a narrower and more demanding form than its enthusiasts usually state. It is holding because a company with a seventh of the transaction volume of the market leader earns more revenue and is the only one of the two making money, and the difference is entirely explained by subscriptions and credit distribution built on top of transaction data. It is holding because a payments bank whose business was charging fees for moving cash is watching its revenue contract while the industry's revenue grows. It is holding because private capital voted with its feet: two companies paid substantial one-time tax bills to move their legal domicile to India specifically so that Indian public markets could price a financialisation story that Indian regulators created.

It is fraying at exactly one place, and it deserves to be named without euphemism. The monetisation pools are still small. Roughly 10.5 million soundboxes at ₹100 a month is on the order of ₹1,250 crore a year of industry subscription revenue — about $150 million, against $3.75 trillion of payment value flowing over the same rail. Credit on UPI is a tenth of one percent of that value. The bull cases require these pools to grow by one to two orders of magnitude within three years, and the current growth rates, while high, do not obviously get there. The belief that payment data financialises an economy is correct in direction and unproven in magnitude, and the gap between those two states is precisely where the money is made and lost.

That distinction — a right forecast about a society producing a wrong answer about a security — is the sharpest edge in this theme. India will keep digitising. UPI will keep growing. Whether the listed equity that trades at 188 times earnings on the strength of that fact delivers a return depends on four numbers reported by three regulators and two non-bank lenders, none of which the company controls. The kirana shopkeeper in Chandni Chowk will keep his soundbox plugged in for as long as it helps him watch the counter. Everything else is a claim on the layers that shopkeeper's data makes possible, and those layers are still, in August 2026, mostly under construction.

↑ Back to Top

Glossary

UPI (Unified Payments Interface). India's real-time retail payment protocol, operated by NPCI, which lets any bank account pay any other instantly using a human-readable handle. It matters because it is interoperable by mandate, which prevents any private participant from building a closed payment network in India.

TPAP (Third-Party Application Provider). The consumer-facing app layer — PhonePe, Google Pay, Paytm, Navi, CRED, Amazon Pay — that initiates UPI transactions through a sponsor bank. A TPAP owns the interface and the customer relationship but neither the account nor the rail, which is why app market share and revenue diverge so sharply in India.

Zero-MDR. The sovereign policy, effective January 2020, setting the merchant discount rate at 0% on standard UPI and RuPay debit transactions. It is the single most important economic fact in the industry: it makes the payment itself unmonetisable and forces every business model to earn elsewhere.

MDR (Merchant Discount Rate). The fee a merchant pays for accepting a digital payment, shared among the acquirer, processor and card network. It survives on credit cards and some other instruments in India, which is why the credit rail matters disproportionately.

Soundbox. A cellular-connected speaker at a merchant counter that announces payment amounts aloud. It converted an unmonetisable QR scan into a ₹90–125 monthly subscription by solving a labour and verification problem rather than a payment problem.

PA / PG (Payment Aggregator / Payment Gateway). RBI-regulated entities that let merchants accept multiple payment instruments, aggregating transactions and settling funds through escrow accounts at scheduled commercial banks. The escrow requirement is what stops aggregators from becoming quasi-banks.

PA-CB (Payment Aggregator — Cross Border). A specialised RBI licence for processing inward and outward international payments. It matters because cross-border take rates of 100–150 basis points are several times domestic rates, making it the strongest available offset to gateway price compression.

Credit on UPI. The set of arrangements linking RuPay credit cards or pre-sanctioned bank credit lines to a UPI handle, so that a scan at any existing QR code draws credit. It reintroduces interchange of roughly 1.5–2.0% into an otherwise fee-free rail.

Interchange. The portion of a card or credit transaction fee paid to the instrument's issuer. In this industry it is the mechanism by which credit-funded UPI transactions generate revenue that debit-funded ones do not.

NPCI (National Payments Corporation of India). The bank-owned, RBI-supervised body that operates UPI, IMPS and the RuPay network. It sets the rules of participation, including the unenforced 30% cap on any single app's share of UPI volume.

India Stack. The layered public digital infrastructure — Aadhaar identity, electronic KYC, DigiLocker document storage, UPI payments — that made low-cost onboarding and instant payment possible at national scale. Its public-good design is the reason India's payment economics differ so radically from the United States or China.

FLDG (First Loss Default Guarantee). An arrangement in which a loan distributor absorbs a defined first slice of a lender's losses, capped by RBI at 5% of the portfolio. The cap protects payment companies from credit losses and simultaneously limits how much of the credit spread they can ever earn.

Nodal and escrow accounts. Bank accounts, mandatorily held at scheduled commercial banks, in which merchant funds sit between collection and settlement. They are the reason every payment aggregator has a hard operational dependency on a bank sponsor.

Gross versus net revenue recognition. Whether a payment company books the full transaction settlement value or only the spread it retains as revenue. Ignoring this distinction produces false comparisons between companies of apparently similar size, which is the most common analytical error in the sector.

Take rate. The share of transaction value a participant retains, expressed in basis points. It runs from zero on UPI to 15–35 bps on domestic non-UPI gateway volume, 100–150 bps on cross-border, and 250–400 bps on loan origination — a range that describes the entire strategic logic of the industry.