Razorpay Launches Vulcan, an AI Payments Foundation Model Built With NVIDIA and AWS Support

Razorpay has launched Vulcan, an AI foundation model built specifically for payments, with technology support from NVIDIA and AWS. The Bengaluru-based fintech unicorn said the transformer-based model is designed to act as a common intelligence layer across payment functions rather than relying on separate models for individual tasks.

According to Razorpay, Vulcan has been trained on nearly 3 trillion data points across 4 billion payments and uses around 3,000 signals per transaction. NVIDIA GPUs were used to train and run the model, while AWS cloud infrastructure and Amazon SageMaker supported development, training and deployment.

Vulcan is intended to support payment routing, fraud detection, risk assessment and checkout personalisation. Razorpay said components were tested on its network before launch, and customers including Blinkit, Bachatt and redBus have started using some capabilities in live payment environments.

Razorpay claims Vulcan has improved payment success rates by 8-10% and helped detect and stop 8x more international card fraud. It also said the model identified five times more fraudulent or disputed transactions without increasing alerts, and that Magic Checkout is showing 40% more shoppers their preferred UPI app, helping complete an additional 1-2 lakh purchases every month. Razorpay plans to expand the model into areas including authentication and lending.

Razorpay’s internal study across 1.5 million shoppers and more than 51,000 businesses found payment friction such as failed transactions, delays and drop-offs across metro and smaller markets. The company frames Vulcan as a single AI layer for more payment decisions as India’s digital ecommerce market is projected to reach $350 billion by 2030, according to Razorpay.