Concentrated and unmonetised at once, which is why the share cap cannot bind: India FS Pulse

Generated 2026-08-22 by analysis/06_competitive_structure.py. Every figure below is interpolated from a computed value, so the prose cannot drift from the data.

The answer

India’s payments market is concentrated and unmonetised at the same time, and that combination is what makes the share cap unenforceable as written. In 2026-07, PhonePe held 45.9% of national UPI volume and Google Pay 32.3%, together 78.2% of every transaction in the country and 81.8% of the value. NPCI’s cap is 30% per app. Both leaders breach it, and neither earns a rupee of MDR on the volume that puts them in breach.

To comply by December 2026, 4.3 billion transactions a month would have to change app, 18.2% of the entire national market, relocated inside a policy window. That is the whole argument in one number.

Three supporting arguments

1. The gap is not closing at anything like the required rate. PhonePe’s share has moved from 46.9% in 2023-12 to 45.9% in 2026-07: a drift of -0.03 percentage points a month. Extrapolated, reaching 30% takes about 470 months at the observed rate. Market concentration overall tells the same story: the HHI has fallen from 3,702 to 3,256 over 31 months, which is real movement but still leaves the market highly concentrated by any competition-authority standard.

2. The challengers taking share are not substitutes for the leaders. Average ticket size separates them completely: CRED runs Rs 4,172 a transaction while FamApp runs Rs 137: a 30x spread across apps on the same rails. These are different businesses serving different customers, not rivals competing for the same payment. Share cannot simply be redistributed from a leader to a challenger, because the challengers are not built to absorb general-purpose everyday spend.

3. Nobody is fighting for share that pays. The leaders’ value share (81.8%) runs above their volume share (78.2%), so they carry the larger transactions too, and under zero MDR that additional value converts to no additional revenue. A cap is normally a remedy for market power being exploited. Here the market power produces no direct rent, which is why the deadline has moved before and why enforcement pressure is structurally weak.

So what

Do not underwrite a plan that assumes the cap binds on schedule. The arithmetic above (4.3bn transactions a month) is the reason. Treat the cap as a step-function risk with a low probability in any given quarter, not as a dated event.
If it does bind, the beneficiary is not the runner-up. Redistribution would have to go to apps whose average ticket says they serve a different customer. Model the transition cost, not just the share transfer.
This reframes the monetisation question. Concentration is not the reason the rails are unprofitable: price is. Splitting 78.2% of the market three ways changes who processes a free transaction, not whether it is free.

Method and its limits

NPCI publishes the top ten apps by volume and no more, so the tenth-place cutoff moves between months and a smaller app can drop out of view. Shares are computed against the NPCI national monthly total, not against the sum of the ten, so they are true market shares and the “All other apps” residual is genuine rather than a rounding artefact. The ten reconcile to between 94.3% and 99.6% of national volume across the period: an independent cross-check between two separately transcribed NPCI tables. The series starts 2023-12 because NPCI renders earlier months alphabetically rather than by size, capped at ten rows, which omits the leaders entirely.