The UPI Playbook Doesn't Export. That's the Point.

There is a slide that has appeared, in some form, at every central-banking conference I have sat through for the last three years. It shows a line going up and to the right — monthly transaction volume on India's Unified Payments Interface — and beneath it, a single word: "Lessons." The speaker is usually from a regulator, occasionally from a development bank, and the conclusion is always the same. India built a public payments rail that the whole world should now copy.

I want to argue the opposite, and I want to be precise about it, because the precise version is more interesting than the contrarian one. UPI is not over-hyped. It is genuinely one of the most consequential pieces of public digital infrastructure built this century. The problem is not the praise. The problem is the lesson being drawn from it — and the lesson, as it is currently taught from that slide, is close to unexportable.

The export narrative has hardened into consensus faster than almost anyone noticed. Brazil's Pix, Thailand's PromptPay, Nigeria's recent NIBSS overhaul — all of them now cite UPI as the template. NPCI International, the body that takes the rail abroad, has live corridors in seven countries and signs new memoranda of understanding at a rate of roughly one a quarter. The word "stack" has entered the diplomatic vocabulary. And underneath all of it sits an assumption so quiet that it is almost never stated: that the thing India built was, fundamentally, a piece of software.

The panel that runs at every conference

Start with what is actually true, because the case for UPI is strong and you cannot understand the misreading without it. In a little under a decade, India went from a country where the overwhelming majority of consumer transactions were cash to one where a street vendor selling forty rupees of coriander expects you to scan a QR code. The numbers are genuinely without precedent.

• ₹0 merchant discount rate — the per-transaction fee that funds card networks everywhere else — set, deliberately, to zero.
• ~16B transactions a month — more monthly transactions than the rest of the world's instant-payment systems combined.
• 7 live export corridors — countries with a working UPI link or acceptance deal, with more under negotiation.

So when a finance ministry in a middle-income country looks at that and says "we want this," the instinct is completely reasonable. The mistake is in what they reach for. They reach for the artefact — the API specifications, the QR standard, the switch architecture — because the artefact is the part you can put in a procurement document. You can license a switch. You can hire integrators. You can stand up a sandbox in eighteen months and cut a ribbon.

What you cannot put in a procurement document is the reason it worked.

"You cannot import a political settlement through an SDK."

— the recurring theme of this essay, stated once, plainly

What the rail actually sits on

Underneath UPI sit three preconditions, and not one of them is technical. The first is identity. Before there was a payment rail, there was Aadhaar — a biometric identity system the Indian state spent the better part of a decade and an enormous amount of political capital building, fighting in court, and forcing through. By the time UPI launched, the expensive, contentious problem of "how do you know who is at the other end of this transaction" had already been paid for. Most countries trying to copy UPI are quietly trying to skip this step, because the step took India ten years and a Supreme Court case.

The second precondition is demographic, and it is the one foreigners systematically underweight. India had, at the exact moment of launch, an unusually young population coming online for the first time through cheap smartphones and — crucially — through a mobile-data price collapse so severe it reset the entire market. A new entrant flooded the country with effectively free data in 2016. That was not a payments decision. It was not even a planned input to the payments system. But without sub-cent-per-megabyte data in the hands of four hundred million first-time internet users, the adoption curve on that conference slide is a gentle slope, not a wall.

The third precondition is the one that gets sanitised most thoroughly in the retelling, because it does not flatter anyone: the regulator was willing to let the banks lose. Zero MDR was not a clever pricing innovation. It was a decision to remove, by fiat, the revenue pool that funds payments infrastructure everywhere else on earth, and to make the banking system absorb the cost as a condition of participation. That is not a product call. That is a state with the leverage and the political will to tell its financial sector to eat a loss for a decade in service of a public good.

Put those three together and the UPI "stack" starts to look less like a technology and more like a snapshot of a particular country at a particular moment, with a particular state capable of particular things. The software is real and well-built. But the software is the part that was, comparatively, easy.

Pix didn't copy the API. It rebuilt the politics.

Here is where the standard narrative would have you predict failure for every imitator. But Pix, Brazil's instant-payment system, is a roaring success — arguably as complete a domestic victory as UPI itself. Doesn't that disprove the whole argument?

It does the opposite. Look at how Pix succeeded and you find that Brazil did not copy India's API at all. The Banco Central built its own system, on its own schedule, with its own technical choices. What Brazil actually copied — or rather, independently arrived at — was the political structure underneath. A central bank with the authority to mandate participation. A decision to make the rail cheap-to-free at the point of use. An existing, if imperfect, identity layer to anchor accounts to. The resemblance between Pix and UPI is real, but it lives in the settlement, not the software.

This is the chart I keep coming back to, because it shows the thing the export panels never quite say out loud: adoption like this does not happen incrementally. It happens when the preconditions are already in place and someone flips a switch.

When you frame it this way, the imitators sort cleanly into two groups. The ones who succeed are the ones who did the unglamorous, multi-year political work of assembling the preconditions, and used India's rail only as proof that the destination exists. The ones who stall are the ones who bought the switch and waited for the curve.

The corridor counterargument

I owe the other side a fair hearing, because the people running UPI's international expansion are not naïve, and they have a genuinely different theory of how this works. I put a version of this essay's argument to someone close to the corridor strategy. The response was sharp.

"You're reading it as a domestic-replication story, and that was never the plan. The corridors come first — remittances, tourist acceptance, cross-border settlement. You build the plumbing between countries, the volume follows the diaspora, and the domestic politics catch up afterward because the rail is already there and already useful. We're not exporting the revolution. We're exporting the pipe, and letting each country decide what to run through it."

— a senior figure close to NPCI's international strategy, paraphrased from a conversation on background

This is a serious argument and it might be right. The Singapore–India PayNow–UPI link is the proof of concept they point to: two systems, two sets of domestic politics, joined at the settlement layer to move remittances cheaply. On a fifteen-year horizon, the corridor-first theory could absolutely produce the thing the replication theory says is impossible — by sneaking in through cross-border utility rather than domestic mandate.

But notice what the counterargument concedes. It abandons the conference-slide promise entirely. It does not claim that a country can buy the stack and get India's domestic adoption curve. It claims something much more modest and much more credible: that you can connect to India's rail and capture a slice of cross-border flow. That is a real prize. It is also a completely different product from the one being sold on the panel.

What the five-year read misprices

I am, mostly, an investor's reader of these things, and the gap between the fifteen-year story and the five-year story is exactly where capital keeps getting set on fire. Here is what I think the next five years actually misprice, in order of how expensive the mistake is.

None of these is a reason to be bearish on UPI itself. They are reasons to be extremely careful about what you believe UPI proves. The rail is a triumph. The generalisation drawn from it on that conference slide is a category error, and category errors are expensive precisely because they feel like insight.

"The rail is the visible ten per cent. The preconditions are the ninety per cent nobody can screenshot."

— if you remember one line

A method, not a manual

So is there a lesson worth exporting? Yes — but it is a method, not a manual, and it is far less flattering to the people selling the stack abroad.

The exportable lesson is this: identify the preconditions in your own country that are currently doing the work that India's identity layer, data collapse, and regulatory will did in theirs. Maybe you already have a strong national ID and your missing piece is acceptance density. Maybe you have smartphone saturation but no political appetite to make the rail free. Maybe your banks are too weak to absorb a decade of zero-MDR and your real first move is to fix that, not to launch an app. The work is diagnostic, country-specific, and slow. It looks nothing like buying a switch.

This is, in the end, the whole thesis of how I try to read Asian technology from the inside out. The copyable artefact — the rail, the super-app, the QR sticker on the paan-shop counter — is almost never the thing that made it work. It is the visible residue of a set of preconditions that were political and demographic before they were ever technical. Read for the rail and you will keep buying switches and waiting for curves that never come. Read for the preconditions and you might actually build something.

The slide will keep appearing at the conferences. The line will keep going up and to the right. I would just change the single word underneath it. Not "Lessons." "Preconditions."