The rails of development | What Pix and UPI reveal about the next generation of public infrastructure

Key reasons to read this article

• How could digital public infrastructure change the way development programs are designed and delivered?
• Can systems built to reach millions also leave the people they fail to recognize even further behind?
• Who should control the digital frameworks that connect people to money and public services?
• If digital infrastructure is public infrastructure, who should pay for it and who should be responsible to keep it running?

A government would rarely build a new road every time it launched a public program. Yet in the digital world, governments and development organizations routinely do exactly the opposite: separate beneficiary databases, payment systems, and identification platforms are built for individual programs, with limited ability to communicate beyond these.

Digital public infrastructure (DPI) offers another model: build shared digital systems once and allow multiple programs and providers to use these.

In this respect, Brazil’s PIX and India’s Unified Payments Interface (UPI) are more than examples of instant payment systems. They represent a test of whether digital public infrastructure can become as fundamental to delivering development as roads, electricity or telecommunications, and what governments, donors, development banks, NGOs and contractors will need to do differently if it does.

What made Brazil’s Pix work?

When Brazil’s Central Bank launched Pix in November 2020, it addressed a basic weakness in the country’s payment system: although people could have bank accounts, they still faced costly, fragmented ways of transferring money.

The bank therefore built and now operates a core infrastructure, having established the rules of use and invited large financial institutions to participate. Banks, fintechs and other payment providers are able to connect their own services to the same interoperable national network.

That combination helped to resolve the network-effect problem that can prevent new payment systems from gaining traction. A payment platform is of little use if the person expecting payment paid cannot receive money through it.

Cost was another factor. A BIS analysis found that the average cost to merchants of a Pix transaction was around 0.22%, compared with approximately 2.2% for credit card transactions at the time of the study. Meanwhile, person-to-person transfers are generally free.

The result was extraordinarily rapid adoption. Within a little more than a year, 67% of Brazilian adults had used Pix, while around 50 million people had made a digital payment for the first time through the system during its early expansion, according to BIS.

Moreover, a recent study into the introduction of Pix found that the system was associated with an approximately 2.6% increase in companies being established in underbanked Brazilian municipalities, with this involving a higher number of smaller firms and areas with greater Pix adoption.

Renato Gomes, Director of Financial System Organization and Resolution at Brazil’s Central Bank, says Pix “enabled new business models and allowed people to become entrepreneurs in ways that were not previously on the radar”.

India: infrastructure around the infrastructure

India’s UPI offers a different but complementary lesson. Launched in 2016 and operated by the National Payments Corporation of India (NPCI), UPI now processes more than 23 billion transactions per month through an ecosystem that involves hundreds of banks.

But its significance lies partly in what sits around it.

UPI was developed alongside Aadhaar, India’s digital identification system, the expansion of bank accounts under the Pradhan Mantri Jan Dhan Yojana, mobile connectivity, and systems through which governments can send payments directly to beneficiaries.

An IMF study shows that India’s experience suggests DPI is most powerful when several layers of infrastructure work together: digital identity, financial accounts, interoperable payments, data exchange, and public-service delivery.

The whole infrastructure matters because an interoperable payment system has limited value for an unbanked recipient if there is no account into which money can be transferred. An account has limited practical value without connectivity, a usable phone, a nearby agent, or the digital literacy required to use it.

World Bank data shows that account ownership among Indian adults increased from 35% in 2011 to approximately 89% by 2024. However, 16% of those accounts were still inactive in 2024, twice more than the global average. This points to the fact that building digital access is not the same as ensuring that people can actually use it.

Can DPI function as development infrastructure?

Since the implications extend well beyond everyday transactions, there is no one straightforward answer to this.

Governments distribute enormous amounts of money through pensions, salaries, subsidies, social protection and emergency assistance. Humanitarian organizations distribute billions more through cash-based assistance. Every transfer requires a system to identify the recipients, determine eligibility and move the money.

The potential of DPI lies in replacing individual program-level systems with a shared infrastructure. Evidence suggests that an interoperable payment and identity infrastructure can reduce transaction costs, speed up transfers, improve reconciliation and decrease duplication.

But treating DPI as a shared infrastructure also raises questions about how it is financed and maintained. Digital infrastructure requires continuous investment in cybersecurity, software maintenance, fraud detection, governance, technical staff and upgrades. Donor-funded projects launched without credible plans for long-term financing risk becoming another layer of a fragmented infrastructure.

For development banks, donors and humanitarian organizations, this raises another question: should funding move from isolated digital projects toward a shared infrastructure, long-term institutional capacity and systems that can be used across diverse programs?

But the risks of doing so are equally important. What happens when the government that controls this infrastructure is a party to a conflict, when refugees lack national identification, or when data protection is weak? These are not technical details. They will determine whether DPI can function as development infrastructure.

A new role for the development industry

This shift could also change what development organizations procure.

When each development program builds its own system, contractors compete to deliver standalone databases, platforms and applications. But if a service already exists, there would be no need to procure a different one. So, if governments invest in shared infrastructure, procurement demand may move toward interoperability, APIs, cybersecurity, data governance, open-source components, cloud infrastructure and systems integration.

The procurement question would change from “Can you build this system?” to “Can what you build become part of a system?”

That means procurement may increasingly have to examine whether a platform uses open standards, whether APIs allow other services to connect, whether governments retain control over data, whether another provider could maintain the system, whether technical documentation is transferred, and whether domestic institutions have the skills required to operate the infrastructure after a contract ends.

This also brings an old development issue into the digital age: vendor dependency. A government can formally own a platform while remaining dependent on the private contractor that possesses the technical knowledge required to operate it.

When infrastructure excludes

The strongest argument for DPI is scale, but its greatest risk may also stem from scale. A malfunctioning database for one social program may prevent someone from accessing one benefit. A malfunctioning digital identity that is used across governments can potentially prevent someone from accessing many payments.

India’s experience with Aadhaar illustrates the risk. Researchers have documented cases in which biometric authentication failures, incomplete linking between databases and insufficient fallback mechanisms have been associated with eligible households being unable to access welfare benefits.

This problem illustrates what might be referred to as an exclusion paradox. Digital infrastructure can lower the average cost of reaching millions of people while substantially increasing the cost of access for the minority whom the system fails to recognize.

That means DPI should be designed around inclusion, not simply efficiency, and introduce alternative authentication methods, offline options, grievance mechanisms, data protection and ways to access essential services without being entirely dependent on digital systems.

Who controls the infrastructure?

There is one final question that development institutions will increasingly have to confront. If digital infrastructure becomes essential to accessing the economy and public services, who should control it?

Who controls the data? Who determines the technical standards? Who can audit the system? Can governments change technology providers? What happens when an identity is incorrectly rejected? Can citizens access essential services without participating digitally? What safeguards exist against surveillance?

Pix provides one model. Brazil’s Central Bank operates the core infrastructure and sets its rules, while banks and fintechs compete to provide services on top of it. India offers another, with UPI operated by NPCI and an enormous ecosystem of banks and technology companies building services around it.

Neither model can simply be replicated elsewhere. Brazil and India had particular institutional, regulatory, and technological capacities that helped to make their systems work. Infrastructure may be transferable. Institutions are not necessarily so. That may be the most important factor for development finance.

The emerging challenge is not primarily to finance more apps. It is to decide which digital foundations should be shared, how they should be financed and maintained, which institutions should govern them, and how to ensure that people who cannot or do not use them are not excluded from the state itself.

As Coyle, Eaves and Vasconcellos put it, the shift required is from isolated projects toward shared infrastructure, from short-term savings toward long-term public value, and from fragmented ownership toward institutional stewardship.

The development challenge is to build this without creating new forms of exclusion, dependency, or control.