The UPI Compounding Effect

UPI processed roughly 16.5 billion transactions in March 2026 — about 530 million per day, ~6,100 per second. For comparison, Visa processes about 720 million transactions per day globally. India's domestic real-time payments network has crossed the world's largest card network on volume. The implications for the next decade of Indian consumer building are larger than the comparison itself suggests.

Why UPI is structurally different

There are three things about UPI that the global fintech press still underweight, even six years into the boom. One: it is free at the end-user level — no transaction fee, no minimum balance, no subscription. Two: it is interoperable across all 350+ Indian banks and most major wallets, by regulation, not by negotiation. Three: it is open at the API layer — any licensed payments operator can build a UPI app and access the same rails as PhonePe or GPay.

These three properties together are the reason UPI's per-user cost-of-transaction is roughly 1/40th of a card swipe and roughly 1/200th of a cross-border wire. Not 'cheaper.' Two orders of magnitude cheaper. That is what unlocks the long-tail use cases that follow.

The unit cost of money has dropped to near-zero

When the unit cost of a transaction approaches zero, the categories that become viable change. Suddenly you can charge a customer ₹2 for a one-off service, ₹5 for a video, ₹10 for an AI agent invocation. Micro-billing in India is now genuinely possible at retail scale — and that is the substrate every India2-targeted AI-native product is going to run on.

What this does to AI-native products

Three concrete ways the UPI substrate changes what an AI-native company in India can build:

• Inference-priced products. Charging per agent invocation (rather than monthly subscription) becomes viable at India2 prices. We have seen vernacular voice products charge ₹3 per query at scale.
• Conversational checkout. UPI Lite + UPI AutoPay let AI-native shopping experiences complete payment in-flow without a redirect. The conversion lift in early portfolio data is 15–25%.
• Programmatic refunds and credits. Agents that need to compensate users for failed actions can issue UPI credits in seconds. The customer-experience improvement, especially in support automation, is meaningful.

The current frontier: credit on UPI

The 2025–26 evolution of UPI is the addition of credit lines on top of the existing rails. Pre-approved credit at the moment of transaction, settled into a small unsecured credit line. This is what changes UPI from a payments substrate to a lending substrate, and it is where most of the next generation of Indian fintech is being built right now.

“Most countries spend a generation trying to build a real-time, interoperable, near-zero-cost payments substrate. India built one in five years and shipped it as a public good. Indian founders should not take that for granted.”

What we are funding

Three patterns we are actively investing behind on top of the UPI substrate: (1) micro-billing AI-native products that monetise per use rather than per month, particularly for India2; (2) credit-adjacent products that use UPI as the underwriting and settlement layer; (3) workflow agents inside SMB software where UPI is the payment surface that closes the loop. We are not funding 'a UPI wrapper' — the rails are an enabler, not a product. We are funding founders who have internalised that the unit cost of money in India is roughly zero, and who have built the company that fact makes possible.