On August 18, 2026, Fintech company Razorpay launched ‘Razorpay Vulcan,’ its AI-powered payments foundation model built using NVIDIA and AWS technologies.
According to the company, this is India’s first transformer-based payments AI model, developed to make digital transactions more secure, reliable, and efficient. It will help address challenges such as payment failures, fraud, transaction routing, and checkout issues.
According to Razorpay, an internal study involving 1.5 million customers and over 51,000 merchants was conducted prior to developing ‘Vulcan’. The company stated that the initial model was tested against approximately 3 trillion data points to evaluate decisions related to payment routing, fraud, and risk.
Preliminary results showed a 10% increase in payment success rates and an eightfold improvement in the detection and prevention of international card fraud.
Razorpay stated that its Magic Checkout platform enabled 40% more shoppers to use their preferred UPI apps, resulting in an additional 1 lakh to 2 lakh purchases every month.
NVIDIA’s Pahal Patangia stated that India’s rapidly growing digital economy offers an opportunity to make payments smarter, more secure, and more reliable. Razorpay CEO and co-founder Harshil Mathur said that the aim of this initiative is to make digital payments more trustworthy, especially for those who still prefer cash.
Kiran Jagannath, Head of FSI and Conglomerates for AWS India and South Asia, stated that Razorpay’s AI foundation model is built on Amazon SageMaker and aggregates data from billions of transactions. According to him, this unified AI approach will help improve payment success rates, accelerate model refinement, and ensure robust security for critical payment systems.
Razorpay’s AI model aims to make India’s complex payments ecosystem faster, smarter, and more secure. It could improve payment routing, fraud detection, authentication, and eventually lending as digital commerce continues to grow.
