Came up in a Discord discussion this week and the framing matters enough to write down: UPI is not a "government app" or a "private app." It is public-utility infrastructure.
The structure
- NPCI (National Payments Corporation of India) was founded in 2008 by the Reserve Bank of India and the Indian Banks' Association, under the Payment and Settlement Systems Act, 2007 — an Act of Parliament.
- NPCI is a Section 8 not-for-profit company — not a government department, not a private company. Its 56+ shareholder banks include public-sector banks (SBI, PNB, BoB, Bank of India, Canara Bank, Union Bank) and private ones (HDFC, ICICI, etc.). RBI has direct regulatory oversight.
- UPI is the protocol NPCI built. It is open: any bank or fintech can plug in, anyone can write a UPI app, and the per-transaction cost to the user is zero.
The "apps" you actually use (PhonePe, Google Pay, Paytm, BHIM) are built on top of UPI's rails. They are private companies. The rails are public.
Why this is the right model
The lesson is not "government good, private bad" or vice versa. The lesson is: the part that needed to be a public good is a public good, and the part that benefited from competition was open to competition.
If UPI had been built as a private consortium trying to monetize the protocol, you would have ended up with a Venmo-style walled garden — closed APIs, per-tap fees, regional fragmentation. Instead you got a free, open, programmable payment network that processed roughly 12 billion transactions in May 2025 and is being studied as a model by dozens of other countries.
The same shape works for AI infrastructure. The model weights don't have to be a public good; the protocols for model interoperability, evaluation, and safe deployment do. The compute doesn't have to be public; the benchmarks and the safety disclosures do.
Open rails, private apps. That's the bet UPI made. It worked.