Razorpay’s new AI foundation model, Vulcan, utilises transformer technology to improve payment reliability, fraud detection, and customer experience as the Indian fintech prepares for its public listing.
Razorpay has unveiled Vulcan, a transformer-based AI foundation model that it says is designed to make digital payments more reliable and more secure by scoring payment routes in real time and spotting fraud patterns that may only become visible once a system is live across multiple merchants. The company says the model sits at the centre of a broader push to use AI across checkout, dispute handling and fraud control as the Indian fintech prepares for a public listing.
According to Razorpay, Vulcan was built with NVIDIA and AWS technology, but remains proprietary to the company, with its architecture and training data owned in-house. Rather than behaving like a large language model that processes text, the firm says Vulcan is built to learn how money moves through payment rails, using signals from the full ecosystem and refining itself with every transaction it sees.
Razorpay says the model has been trained on about 3 trillion data points drawn from 4 billion payments, with roughly 3,000 signals analysed per transaction. In early live deployments, the company says it has seen an 8% to 10% lift in payment success rates, an eightfold increase in international card frauds detected and stopped, and a fivefold rise in fraudulent or disputed transactions identified without increasing the number of alerts. It also says more shoppers are now being shown their preferred UPI app at checkout, helping complete an extra 100,000 to 200,000 purchases a month.
The launch also underlines how quickly AI is being folded into payments infrastructure. Stripe has already publicised its own payments foundation model, saying it uses self-supervised learning to analyse transaction patterns and improve outcomes across the payment lifecycle. Separately, in October 2025, Razorpay joined NPCI and OpenAI on an agentic payments pilot aimed at bringing AI assistants into shopping and checkout flows through India’s UPI network.
The timing matters for Razorpay. The company filed confidential IPO papers in June and is widely expected to seek a valuation below its last private mark of $7.5bn. Founded in 2014 by Harshil Mathur and Shashank Kumar, the company now offers payment acceptance, disbursals, lending and business banking services, and says it has raised more than $800m from investors including Y Combinator, Tiger Global Management, Ribbit Capital and Lightspeed.
Disclaimer: This article is intended to inform and educate, not to recommend or endorse any financial product, investment or strategy. Please consider your own financial circumstances and seek professional advice where appropriate before making financial decisions.
