There is a temptation in every central bank right now, and Nepal Rastra Bank is not immune to it. The temptation is to leap. Wholesale central bank digital currency, or CBDC, is the frontier that everyone is talking about: tokenized central bank money that settles between financial institutions instantly, atomically, without the chain of correspondent banks that makes a cross-border payment today feel like posting a letter in 1985. The pilots are dazzling. The conference slides are dazzling. And for a country like ours, where remittances equal roughly a quarter of national income and every extra percentage point of transfer cost is money taken directly from a household in Jhapa or Jumla, the appeal is not academic. It is personal.

I want to make an argument that is less exciting than the frontier but, I think, more useful. We should finish the plumbing first.

I recently completed a study comparing 40 economies over eleven years, looking at whether moving along a wholesale CBDC path is associated with faster, cheaper, safer cross-border settlement once you account for everything else a country is doing at the same time. I want to be honest about the data, because honesty is the whole point: harmonized settlement figures across countries simply do not exist in the public domain, so the panel I built is illustrative rather than observed. The magnitudes are not measurements and should not be read as forecasts. What the exercise does well is show the shape of the relationship under a stated structure, and that shape is worth taking seriously.

Here is what it shows. Wholesale CBDC readiness is associated with improvement on every dimension I looked at: settlement time, cost, settlement risk, and the amount of liquidity a bank must tie up during the day. That part confirms the enthusiasm. But CBDC is not the largest factor. Two humbler things matter more. The first is messaging standardization, the migration to the ISO 20022 format that lets a payment instruction carry structured, machine-readable data instead of a free-text note a human has to investigate by hand. The second is renewing the real-time gross settlement system, the domestic engine on which everything else runs. In my estimates, both of these carry larger coefficients than the tokenized settlement asset itself.

More telling still is how these things interact. The benefit of a CBDC gets bigger when the messaging layer beneath it is more advanced, and smaller when it is not. This is not a surprise once you think about the mechanics. A tokenized settlement asset lets you attach conditions to a payment, settle both legs of a currency exchange at once, or cut out an intermediary bank. But every one of those tricks assumes the message can carry the instruction and the counterparty's system can read it. If the pipe cannot carry structured data, the clever settlement asset has nothing to act on. The frontier depends on the foundation.

There is one finding I want to press on, because it cuts against a story that is popular in developing economies. The story goes: countries with less legacy infrastructure can leapfrog, adopting the newest technology directly and skipping the boring intermediate steps that rich countries had to build. It is a seductive idea. My estimates give it no support. The returns to CBDC development were not systematically larger for emerging economies than for advanced ones. If anything the logic runs the other way. Less existing infrastructure means fewer of the enabling conditions a CBDC needs to deliver value. You cannot leapfrog to a destination that only works once the road is built.

Why does this matter for Nepal specifically? Because we face a constrained infrastructure budget that will not be replenished on any convenient timetable, and because the choice is real and live right now. We are in mid-transition. Our settlement platform renewal and our messaging migration are not finished. So the sequencing question is not a thought experiment for us. It is a budget decision this year and next.

The case evidence points the same way as the numbers. Every multilateral CBDC experiment that reported meaningful gains, from Project Jura to Project mBridge, ran between jurisdictions that had already completed their domestic modernization. None demonstrated gains where the messaging layer was still fragmented. And the constraints these projects kept hitting were not technical but legal: whether the law recognizes a tokenized transfer as final, whether cross-border disputes have a governing law, whether a digital identity credential is valid. Those are problems no amount of hardware spending can solve. They require legislation, and legislation takes years, so it should start before the technology does, not after.

So what would I actually recommend? A sequence, not a race.

In the near term, over the next two years, complete the ISO 20022 migration properly, with genuinely structured data rather than a minimum-compliance mapping that ticks a box. Extend our settlement operating hours to overlap with the jurisdictions our remittances actually come from. Commission a serious legal gap assessment on settlement finality and digital identity. None of these depend on anyone else's agreement. We can do all of them on our own schedule, and they carry the largest measurable returns.

In the medium term, two to four years out, run a narrow wholesale CBDC proof of concept, ideally a single corridor with a trading or remittance partner that matters to us. The goal is not deployment. The goal is to build in-house capability and discover where our own legal and operational walls are, before the stakes are high.

In the long term, the real decision is whether to join a regional multi-currency platform, and that decision will be about governance rather than technology. Who sets the rules? How are small participants treated when their volumes are tiny next to the giants? A small country negotiates those terms far better if it arrives with demonstrated technical competence. Which is the honest argument for doing the medium-term pilot at all: not because the pilot pays for itself, but because capability buys a seat at the table.

I am aware this is not a fashionable position. It is more thrilling to announce a digital currency than to announce a messaging-format upgrade. But the two agendas are not competing claims on the same rupees. They are stages of one program. The foundational work delivers benefit to Nepali households immediately, and it simultaneously lays the conditions under which the frontier becomes worth reaching.

Sequencing, on the evidence I have seen, matters more than selection. Let us build the road before we buy the faster car.

The author is Director of the Payment Systems Department at Nepal Rastra Bank. The views expressed here are personal. The study referenced is illustrative in its data and is disclosed as such