Digital Public Infrastructure Explained: India Stack and the G20 Model

A loan officer at a small finance company in Coimbatore can now approve a working-capital loan for a textile job-worker without collecting a single photocopy. The applicant's identity is confirmed through Aadhaar eKYC, eighteen months of bank statements arrive through an Account Aggregator after she taps consent on her phone, her GST and certificates come from DigiLocker, and the money lands by UPI or NEFT the same afternoon. None of those systems belongs to the lender, and none of them was built for lending. That is the practical meaning of digital public infrastructure, and it explains why so many governments now send delegations to study it.

It also explains the hard questions. Shared rails create shared points of failure, and a stack built by the state raises issues of privacy, exclusion and competition that a single app never would.

What "digital public infrastructure" actually means

The term entered diplomatic vocabulary in 2023. At their meeting in Bengaluru on 19 August 2023, G20 digital economy ministers agreed a G20 Framework for Systems of Digital Public Infrastructure, which the the G20 Leaders' Declaration of 9-10 September 2023 welcomed as a voluntary, suggested framework for how DPI is developed, deployed and governed.

In plain terms, DPI is a set of shared digital systems that any public or private service can build on, usually covering three functions:

• Identity: proving who someone is, online, at low cost.
• Payments: moving money between any two accounts in real time.
• Data exchange: letting a person share verified records with consent.

The distinguishing features are open standards and interoperability. A private bank, a fintech and a government department all connect to the same rail on the same terms. That is different from a government building an app, and different from a private platform that owns its users.

India's layers, at 2026 scale

The figures come from the UIDAI dashboard, the DigiLocker homepage, the DBT Mission and Sahamati, the Account Aggregator industry body, which the RBI recognised as a self-regulatory organisation in June 2026. The growth curve is steep. In September 2022, Business Standard reported UPI at about 7.3 billion transactions a month; four years later it handles more than three times that. The same report noted that close to 1,000 government schemes already used Aadhaar authentication.

For the payments layer in more depth see our guide to UPI and its global expansion; for the data layer, our explainer on open banking and Account Aggregator.

How the layers compound: a worked example

The value of DPI is not in any single layer but in what happens when they combine. Account Aggregator lending is the clearest case, because Sahamati publishes both volume and value.

• In FY26, AA-enabled lending reached ₹3.82 lakh crore across 3.68 crore loans.
• That is an average of about ₹1.04 lakh per loan (₹3.82 lakh crore divided by 3.68 crore). These are small loans: working capital, personal loans, two-wheeler finance.
• Sahamati says this was 8.4% of retail and MSME lending by value. Working backwards, that implies a total retail and MSME lending pool of roughly ₹45 lakh crore for the year, of which AA already touched a meaningful slice.
• By volume, AA's share was higher, 11.8%, which confirms that it is used disproportionately for smaller tickets, exactly the borrowers who were hardest to underwrite on paper.

Now add the other layers to the Coimbatore loan. Aadhaar eKYC replaces a branch visit. DigiLocker replaces document collection. AA replaces bank-statement PDFs, which are easy to forge. UPI or NEFT replaces a cheque. Each step cuts cost by a modest amount; together they make a ₹1 lakh loan economical to underwrite. That is the compounding effect policymakers abroad are trying to reproduce.

India did not build a fintech; it built the roads, and let thousands of firms run vehicles on them.

Who else is building it

India's approach is now one model among several, and the export channels are multiplying.

• Identity. MOSIP, the open-source identity platform set up at IIIT Bangalore in 2018, reports 32 country engagements and 14 national rollouts. Countries listed include Ethiopia, Morocco, the Philippines, Sri Lanka, Togo, Uganda, Sierra Leone and Burkina Faso.
• Campaigns. The 50-in-5 campaign, launched on 8 November 2023, aims for 50 countries to design, launch or scale at least one DPI component within five years. It lists 41 members, from Brazil, Singapore and Norway to Bangladesh, Ethiopia, Sri Lanka and Togo.
• Convening. The Global DPI Summit in Cairo on 1-3 October 2024 drew representatives from more than 100 countries, and its outcome statement named seven priorities including universal safeguards, financing and interoperability standards for components such as digital wallets.
• Payments links. UPI connections abroad, starting with the UPI-PayNow linkage with Singapore, extend the payments layer across borders; our piece on Project Nexus covers the multilateral version.
• G20 follow-through. The Delhi declaration also welcomed India's plan for a Global DPI Repository and took note of a proposed One Future Alliance to fund and support DPI in low- and middle-income countries.

Brazil is the other heavyweight. Pix is a payments DPI in all but name, built by a central bank rather than a ministry. Our explainer on Pix and Drex compares the two approaches.

Why the development case is strong

The World Bank's identity and G2P team sets out the stakes: 850 million people worldwide lack any ID, and around 5 billion lack an official digital identity usable online. During COVID-19, among 85 countries with data, those with existing digital ID records and data-sharing platforms reached over three times as many beneficiaries as those that had to collect data from scratch. In India, research cited by the same team found leakages in employment and pension programmes fell 41% after biometric payments were introduced.

India's welfare system is the domestic proof. Direct Benefit Transfer works only because three layers meet: Jan Dhan gave nearly every household an account, Aadhaar gave every account a verifiable owner, and NPCI's mapper lets a ministry pay an Aadhaar number without knowing the bank. On that combination the government has routed about ₹53.8 lakh crore. A Finance Ministry assessment published by PIB in 2025 credited the system with cutting subsidies from an average of 16% of central spending in 2009-2013 to 9% in 2023-24, even as the number of beneficiary records grew sixteen-fold. Those figures blend genuine leakage savings with people dropped by verification failures, but the scale effect is not in dispute.

That is the strongest argument for DPI. It lowers the fixed cost of reaching people, for governments and companies alike, and the benefit is largest for those who were once too expensive to serve.

The critiques that deserve answers

• Exclusion. When a rail becomes mandatory, failure to authenticate becomes failure to receive. India's move to Aadhaar-only wage payments for rural employment in January 2024 is the standard example; we cover it in our piece on Direct Benefit Transfer.
• Privacy and surveillance. Shared identity across every service makes linking records easy. Consent-based designs such as Account Aggregator are a partial answer; independent oversight is the rest. UNDP and partners now promote a Universal DPI Safeguards Framework for exactly this reason.
• Concentration risk. A UPI outage is a national event. The more services depend on one rail, the higher the bar for uptime and cyber resilience.
• Who captures the value. Open rails do not guarantee open markets. On UPI, a handful of apps carry most volume, which is why regulators keep returning to market-share caps.
• Exportability. India's stack grew alongside a near-universal ID and a central government with unusual reach. Countries with weaker state capacity may find the identity layer, not the payments layer, is the binding constraint.

What this means for payments

For anyone in payments, DPI changes where competition happens. When identity, data and settlement are public utilities, firms compete on credit, service and user experience rather than on owning the rails. That is good for users and hard on incumbents whose moat was distribution. ONDC is the attempt to apply the same logic to commerce: instead of a buyer app and a seller being locked inside one marketplace, both sides connect to a shared protocol and any participating app can show any participating seller. Whether that works as well for physical goods, with returns, logistics and disputes, as it does for money is still an open question, and ONDC has published far less performance data than UPI or Account Aggregator. Payments people should watch it closely, because a commerce protocol that settles over UPI would push even more retail flows onto public rails. It also creates a template for cross-border work: if two countries each have instant payment DPI, linking them is a far smaller project than building correspondent banking relationships one bank at a time.

What to watch next

• Sahamati's first year as an RBI-recognised SRO, from June 2026: watch for conduct rules on consent design and data retention.
• The 50-in-5 deadline in 2028. The campaign has 41 of its target 50 countries; how many actually ship a live component is the real test.
• MOSIP rollouts. Movement from 14 national rollouts toward the 32 engaged countries will show whether open-source identity scales outside India.
• India's Union Budget, expected 1 February 2027, for funding of ONDC, DigiLocker and any international DPI commitments under the One Future Alliance idea.

This article is for research and general information. It is not legal, financial or investment advice.

About this researchMuffinPay InsightsIndependent analysis of stablecoins, CBDCs, cross-border payments and digital public infrastructure. Every figure is linked to its source. Read our editorial standards →