Razorpay Launches Vulcan, India's First AI Foundation Model Built For Payments

Fintech company Razorpay on Tuesday launched Razorpay Vulcan, which it describes as India's first transformer-based artificial intelligence foundation model built specifically for payments. Developed in collaboration with NVIDIA and Amazon Web Services, the model is designed to make digital transactions more reliable, secure and predictable as India's digital payments ecosystem continues to expand rapidly.

What Makes Vulcan Different

Unlike general-purpose large language models built to understand and generate text, Vulcan has been trained specifically to interpret payment behaviour and transaction patterns. The company said the model analyses roughly 3,000 signals for every transaction it processes, drawing on a training dataset built from nearly 3 trillion data points collected across more than 4 billion payments on Razorpay's network. Early components of the model have already been deployed live, testing routing, fraud and risk decisions on real transactions before the full rollout.

Razorpay said the model was developed after an internal study covering 1.5 million shoppers and more than 51,000 businesses, which found that payment-related friction was a widespread issue affecting consumers in both large metropolitan markets and smaller towns alike. Vulcan is intended to address exactly that kind of friction by analysing payment routes in real time and selecting the most suitable one before a transaction is even attempted, spanning authentication, routing, fraud detection and eventually lending decisions as the platform matures.

Early Results

According to the company, the technology has already improved payment success rates by as much as 10 percent and helped detect eight times more international card fraud than previous systems. It has also identified five times more fraudulent or disputed transactions without a corresponding increase in the number of alerts generated, addressing a common trade-off in fraud detection systems between catching bad transactions and overwhelming users with false alarms. Razorpay also said the model has helped 40 percent more shoppers see their preferred UPI app on its Magic Checkout platform, contributing to an estimated 100,000 to 200,000 additional purchases every month.

Industry Reaction

Razorpay co-founder and chief executive Harshil Mathur said the initiative is aimed at making digital payments more dependable for consumers who remain undecided about trusting digital transactions over cash, a persistent challenge in parts of India's economy. Executives from NVIDIA and AWS also weighed in on the launch, describing it as part of a broader shift toward unified, continuously learning AI systems replacing the fragmented machine learning models that have historically powered payment infrastructure, built on Amazon SageMaker and consolidating billions of transaction insights into a single intelligence layer.

Why It Matters

India's payments ecosystem is unusually fragmented by global standards, spanning UPI, cards, net banking, digital wallets and cash on delivery, with transactions routed through hundreds of different banks and payment gateways. A model purpose-built to navigate that complexity in real time could meaningfully reduce failed transactions and fraud losses at scale, both of which remain persistent drags on merchant revenue and consumer trust in digital payments. As India's digital commerce market continues to grow, tools like Vulcan are likely to become increasingly central to how payment platforms compete on reliability rather than just cost or convenience.

This article is an original editorial summary based on publicly reported information. It has been independently written for publication and does not reproduce content from any single source.