Does the Best Case for Free UPI Hold
| AUTHOR | Anupam Manur |
| DATE | August 27, 2026 |
| CATEGORIES | Economic Policy |
Ajit Ranade has written the most serious defence of zero-MDR I have read. Most zero-MDR advocacy amounts to “UPI is free and free is good.” His piece builds an actual cost-benefit ledger, engages with tax incidence, and proposes a concrete funding mechanism rather than assuming the money appears.
What’s the best form of the argument?
UPI, he argues, is not a payments product but economic infrastructure, and infrastructure with large positive externalities should not be tolled for every use. The benefits spill well beyond the transacting parties: digital literacy, formalisation, tax trails, and even pressure on telecom operators to improve coverage, since a weak signal that blocks a shopkeeper’s payment becomes an economic bottleneck. Zero MDR was central to adoption precisely because it made digital payments resemble cash, which charges nothing at the point of use. More than 85% of person-to-merchant transactions are small-ticket: ₹20 for tea, ₹150 to an auto driver, made without either party incurring a transaction cost. Cash is not free either, he argues. It has to be printed, transported, sorted, stored, guarded and replenished, and banks maintain ATMs, branches and staff to handle it.
Finally, he does propose a funding mechanism for keeping UPI free. The industry’s annual cost, he puts it at ₹20,000 crore, is roughly 7% of the RBI’s ₹2.86 trillion dividend to the Union government, and he proposes a transparent reimbursement mechanism tied to audited costs and efficiency standards. That beats the current arrangement, where the allocation is a discretionary line item that fell from ₹3,500 crore to ₹437 crore before recovering.
Distinguish between the Protocol and the Transaction
This is the central move I want to contest. UPI’s public-good characteristics belong to the standard: the interoperability specification, the virtual payment address scheme, the mandate that every bank must join. That layer is non-rival, non-excludable, and should stay free forever, budget-funded, no conditions.
A transaction is neither. It consumes server capacity, fraud-detection compute, reconciliation effort and dispute-resolution labour at the margin, and it is trivially excludable. TCP/IP is a public good; bandwidth is not, and nobody argues bandwidth should be free because the protocol is. So, yes, we can subsidise the standard, but must charge for the transaction.
A positive externality justifies a subsidy, not a zero price
UPI has positive externalities. The efficient response to a positive externality is a subsidy equal to marginal external benefit, which almost never equals the full cost of provision. For zero to be optimal, marginal external benefit must exceed marginal social cost at every volume. That maybe was plausible at 5 billion transactions a year, when the marginal user was a first-time adopter being pulled out of cash. It loses explanatory value at 240 billion transactions. The externality is concentrated at the extensive margin: bringing new users and merchants in. The subsidy is paid on the intensive margin, on every repeat transaction by users long since converted. A tiered structure that exempts small merchants and small transactions while pricing large ones is the correction.
The obvious objection is that the intensive margin matters too: even a converted user might shift a marginal transaction back to cash if UPI is priced. This could potentially be true, but we have to note where the price lands. MDR falls on the merchant, and surcharging is prohibited, so the consumer sees no price at the till. Substitution has to run through merchants steering customers to cash, and above the exemption threshold, those merchants already accept cards at roughly 2%. If a payment fee were enough to push them back to cash, they would have steered away from cards long ago. Brazil’s Pix has charged merchants from the start, at around 0.22%, and volumes still grew 26% last year.
The elasticity is not zero; some transactions will be lost. But underfunding destroys transactions at the intensive margin too. A five-hour outage on a peak trading day loses transactions. Rising technical declines lose transactions. Fraud losses above ₹1,000 crore a year erode the trust that sustains repeat usage. The case for MDR is that the second cost is now larger than the first, and growing.
On the RBI Dividend
Three problems. The surplus is not free money. It is seigniorage and returns on reserves, already budgeted, and there is an opportunity cost. Plus, it is volatile, driven by forex valuation gains and the rate cycle, which makes it a poor match for recurring operating expenditure.
Competition and Innovation!
Under zero MDR, the only firms that can sustain UPI at national scale are those with balance sheets deep enough to run payments as a loss-leader for something else. Alphabet and Walmart can absorb that. An Indian fintech challenger will find it difficult to do so. Zero MDR is an entry barrier that only global capital can clear. The duopoly is not despite the policy, but it is downstream of it.
The same logic extends upward. If nobody earns on the marginal transaction, nobody builds sophisticated fraud detection, dispute resolution, merchant analytics or credit products on top of UPI; they build them somewhere monetisable instead. The cost of zero MDR is not only underinvestment in the pipes, but it is the innovation that doesn’t happen on top of them.
Where this leaves us
And whatever the rate, the right to charge should carry obligations: published monthly decline rates by PSP, uptime commitments with financial penalties, mandated dispute-resolution turnaround.
Finally, with an MDR, we should allow New Umbrella Entities to come in as competition at the top level to UPI.
Other articles on UPI: Zero MDR for UPI no More (Maybe?); India must hedge against Weaponised Payments and UPI doesn’t cut it; Introduce MDR and Make UPI Sustainable