Push Rails Around the World: The Same Three Layers, Grown Differently

Part III · Bank Account Rails (Chapters 9–12) Builds on: Chapter 2 (clearing and settlement, RTGS and DNS, finality), Chapter 9 (push and pull), Chapter 10 (the four US rails) New concepts in this chapter: T2 and TIPS, China's HVPS and IBPS, NetsUnion, NEFT, TED and STR

This chapter takes the US recipe from the previous chapter to four other currency zones and tests it in each. Conclusion up front: the recipe is the same everywhere; what grows out of it differs everywhere — and none of it interconnects.

1. The Question the Previous Chapter Left Open¶

The previous chapter ended on this: ACH, Fedwire, RTP, and FedNow share one premise — every participant holds an account on the same central bank's ledger, and the last step of every payment lands on the Fed's books.

Switch countries and the premise swaps in a different central bank's ledger, but the same needs remain: batch must be cheap, large-value must be final, retail must arrive in seconds.

So every currency zone has grown the same three layers:

2. Five Currency Zones, One Master Table¶

Three things worth noticing up front.

One: these are public goods, not some company's product. Except for the US's RTP, every system in the table was built and is run by a central bank or an institution under one. A completely different species from the privately owned card rails.

Two: the names differ, but the recipes line up. The systems in each column solve the same problem with the same settlement mode. Think of NEFT as India's ACH and T2 as the eurozone's Fedwire, and on a first pass you won't go far wrong.

Three: Brazil's empty cell was eaten. Brazil's batch pipe, DOC, was formally shut down four years after Pix launched — once the instant layer is cheap and good enough, the batch layer has no reason to exist. Three layers is a recipe, not destiny; a later layer can eat an earlier one.

3. A Walk Through Each Zone¶

The Eurozone: First, Manufacture a "Domestic"

What makes the eurozone special is that its "domestic" is man-made. Twenty-plus countries share the euro, and SEPA (the Single Euro Payments Area) unifies their credit transfer and direct debit rules into one set of messages and deadlines — wherever the rules are unified, that is how far "domestic" extends, in the payment sense.

Push and pull said direct debit only works inside a region with unified rules; SEPA is that region made concrete. Sending euros from Helsinki to Lisbon goes over SEPA SCT, at the same price and speed as a transfer across town.

The three layers in place: SEPA SCT is the batch-net layer — the clearing houses compute net positions and settle them into T2. T2 is the euro's central bank large-value system, where the euro positions of national central banks and commercial banks are finally wiped clean. (T2 went live on 20 March 2023, replacing the TARGET2 system that had run for twenty years — the two names are not the same system, and the chapter on euro-area clearing rails takes them apart.) SCT Inst is the instant layer — arrival within ten seconds, settled payment by payment in central bank money on TIPS.

China: Two Pipes, an Instant Layer, and a Convergence Hub

China's interbank system is two pipes run by the People's Bank of China: HVPS, the high-value system — payment by payment, real time, final on arrival — maps to Fedwire; BEPS, the bulk system — batch net — maps to ACH. IBPS (the Internet Banking Payment System), launched in 2010, added an instant layer for personal transfers — but it settles unlike the others, and the next section takes it up separately.

What China alone has is a fourth thing: NetsUnion. Before 2017, nonbank payment companies like WeChat Pay and Alipay each built private connections to every bank — a replay of the N² problem from clearing and settlement, connection count growing as the square of the number of institutions. NetsUnion collapsed that into one hub: the leg where a wallet debits a bank card now clears through NetsUnion, all of it.

So China's answer is two systems in parallel: banks clear with each other through the PBOC's pipes; nonbank institutions go through NetsUnion. No other country has this layer, because no other country has two nonbank wallets with a billion users each.

India: Another Way to Do Batch

India's NEFT shows batch netting in another form: not one settlement at end of day, but a batch every half hour. The tighter the batches, the closer to instant — push and pull said Same Day ACH works by adding windows, and NEFT pushes that idea further.

Large value goes over the RBI's RTGS. The instant retail layer is UPI, launched in 2016 and now the bulk of India's electronic payments — but its interbank settlement is deferred too, the same family as IBPS; the next section measures them together.

Brazil: The Instant Layer Eats the Batch Layer

Brazil's large-value system is called STR; the product bank customers use to send money over it is called TED. The instant retail layer is Pix, launched in 2020: payment-by-payment real-time transfers on the central bank's ledger, final on arrival, free for individuals.

How good is Pix? Four years after launch, Brazil's central bank shut down DOC, the batch pipe, outright. That is where the empty cell in the master table came from.

4. The Measuring Stick Again: Seconds to Arrive — but Final?¶

Clearing and settlement handed you a measuring stick: for every new rail, ask when clearing happens, when settlement happens, how long the gap is, and who carries the risk in between.

Hold that stick up to these countries' instant layers and you measure something counterintuitive: the user experience is identical, but the interbank books split into two kinds.

The divide is not on the user's side. It is whether the money for the interbank leg is already there. The first four rows move money that is already in place: Pix, SCT Inst, and FedNow transfer gross, payment by payment, on a central bank ledger; RTP never touches the central bank's ledger, but the banks parked money into the pool beforehand. The last two rows book first, net, and settle later — at the moment the user sees the money arrive, nothing has yet moved between the institutions.

To judge whether an instant rail is final on arrival, look at one thing: for the interbank leg, is the money already there?

The RTP row makes the point best: it doesn't run on the central bank's ledger and is final anyway — so the key variable is whether the money is in place, not whether a central bank ledger is involved.

Walk down the UPI row and you can see who this difference lands on.

A user in India sends money over UPI, and it shows up in the other person's account within seconds. But at that moment, the receiving bank has fronted the money out of its own pocket — its account with the sending bank won't truly be squared until some settlement window later that day.

This is the opening scene of clearing and settlement, replayed word for word at national scale: Bank B credits Alice 100 first, out of its own money, while its claim on Bank A still hangs on the books.

"Real-time arrival" describes the layer the user sees. "Real-time settlement" describes the layer on the central bank's ledger. The two can happen at different times — and often do.

For anyone building products, the difference is concrete: on an instant rail that settles net, the receiving institution carries an intraday settlement exposure. When the system is well designed you never feel it; under stress it surfaces as volume caps, per-payment limits, or entry bars for certain kinds of institutions.

5. Why Adoption Diverged So Far¶

All central-bank-built instant rails — yet Brazil's Pix covered most of the country within two years of launch, India's UPI carries the overwhelming share of its electronic payments, and America's FedNow, three years in, is still climbing.

This is not a technology gap. Use the method from interchange — when you see markets differ, look for the institutional difference first:

Whether a new pipe spreads comes down to two things: whether anyone can force adoption, and how good the thing it has to replace already is. India and Brazil have both going for them. The US has neither.

When the previous chapter explained why RTP and FedNow both exist, the answer was governance, not technology. Seen globally, the conclusion sharpens: the shape of a country's clearing pipes is a direct projection of its financial governance structure.

6. The Question This Chapter Leaves Open¶

Five currency zones later, the recipe is the same everywhere and the implementations connect nowhere: each stack ends at its own central bank's ledger, each takes only its own currency, each reaches only its own territory.

That fact grows consequences in two directions:

• Upward: with the pipes disconnected, the gateways, payment methods, and merchant playbooks above them grow apart too — a global company faces not one market but dozens of mutually incompatible ones.
• Outward: if every currency zone is a world unto itself, what do countries run on between them?

The next chapter takes the upward half first — and once that's done, you will see a more surprising answer: step outside the border, and the choices collapse to just two.

7. Self-check questions¶

1. To the user, India's UPI and Brazil's Pix are both "money in seconds." Using the two words from clearing and settlement, state their difference precisely, and name who carries what risk on the UPI path.
2. Brazil shut down DOC, its batch pipe, yet America's ACH shows no sign of shrinking. Give at least one explanation using this chapter's framework.
3. Why does SEPA turn "domestic" into a question of rules rather than geography?

8. Answers¶

Answer for yourself before reading on.

1. Clearing happens in seconds on both: the instruction lands, and the numbers on both sides' ledgers change on the spot. The difference is in settlement.

   Pix transfers payment by payment, in real time, across accounts at Brazil's central bank — clearing and settlement complete together, final on arrival. UPI has NPCI compute net positions first, then settle them in batches through the RBI's RTGS, so a settlement window sits between clearing and settlement.

   During that window, the receiving bank has already credited the payee but hasn't yet been paid by the sending bank — it is carrying exactly the settlement risk described in clearing and settlement. The user never senses any of this, because the experience on the two rails is identical.

2. For the instant layer to eat the batch layer, it has to hold its own even in the scenarios batch is best at. Pix is free for individuals, arrives in seconds, runs 24/7 — everything DOC could do, Pix does better, so DOC had no reason to exist.

   The US is different: ACH carries the mega-batch scenarios — payroll, bills — and its few-cents-per-payment cost still has no rival. RTP and FedNow charge more per payment, and the originator has to process each one in real time, which for a company sending five thousand paychecks at once is a burden, not a feature. The batch layer still has scenarios it wins in America, so it stays.

3. Because SEPA's boundary is not a national border — it is the reach of a rulebook. Once twenty-plus countries adopted the same credit transfer and direct debit rules, a Helsinki-to-Lisbon transfer became indistinguishable from a crosstown one in cost, speed, and return rules — in the payment sense, they are "domestic."

   The reverse holds too: where the rules are not unified, even geographic next-door neighbors are "cross-border." This is also why direct debit can't leave its rule zone: the mandate standards and return rules it depends on are precisely what a regional rulebook like SEPA provides.

Previous: Chapter 10 · The Four US Rails, Side by Side Next: Chapter 12 · Domestic and Global: Fragmented at Home, Two Roads Across Borders