Razorpay Launches Vulcan AI Model Trained on 4 Billion Payments
Razorpay has launched Vulcan, an AI Payments Foundation Model built for India’s digital payments ecosystem. The company says the model uses data from about 4 billion payments and nearly 3 trillion data points.
The new model aims to make digital payments more reliable, secure and predictable. It focuses on payment routing, fraud detection, risk analysis and checkout personalisation.
Razorpay Builds AI for Digital Payments
Razorpay developed Vulcan to address common problems in India’s payment ecosystem.
A transaction can move through UPI, cards, net banking, wallets or other payment methods. It can also pass through different banks and gateways.
This creates many possible payment routes. A payment may fail even when the customer has enough money.
Vulcan aims to improve those decisions in real time. The model scores possible routes and selects the path that Razorpay expects to have the highest chance of success.
Vulcan Uses Billions of Payment Data Points
Razorpay says Vulcan learns from approximately 4 billion customer-to-merchant payments each year.
The company says the model works with nearly 3,000 signals per transaction. These signals help the system identify patterns across the payments ecosystem.
Razorpay built both the model architecture and training data in-house. The company also says Vulcan is not a large language model. Instead, it uses transformer technology to understand patterns linked to money movement.
One AI Model for Multiple Payment Challenges
Razorpay says traditional payment systems often use separate machine-learning models for different tasks.
One model may handle payment routing. Another may focus on fraud. A third may support checkout decisions.
Vulcan takes a different approach. It creates one shared intelligence layer for several payment decisions.
The model currently supports four main areas:
- Payment routing: It selects routes that are more likely to complete successfully.
- Fraud detection: It looks for fraud patterns across merchants and transactions.
- RTO risk intelligence: It identifies potentially risky Cash on Delivery orders.
- Checkout personalization: It recommends payment methods that are more likely to work.
Early Tests Show Higher Payment Success
Razorpay has already tested parts of Vulcan across its payments network.
An internal study covered 1.5 million shoppers and more than 51,000 businesses. Razorpay reported an 8% to 10% improvement in payment success rates.
The company also reported stronger fraud detection. Its early results showed eight times more international card fraud detected and stopped.
Razorpay also reported five times more fraudulent or disputed transactions identified without increasing the number of alerts.
These figures come from Razorpay’s internal testing. They have not been independently verified.
NVIDIA and AWS Power the Model
Razorpay built Vulcan with technology from NVIDIA and AWS.
NVIDIA’s accelerated computing supports model training and large-scale processing. AWS provides cloud infrastructure for development, training, and deployment.
Razorpay also uses Amazon SageMaker to support model training and inference.
The scale of the project highlights the growing role of AI infrastructure in financial services.
AI Could Reshape Payment Infrastructure
The launch shows how fintech companies are moving AI deeper into payment infrastructure.
For merchants, better payment routing could reduce failed transactions and lost sales. Stronger fraud detection could also reduce payment losses.
For consumers, the goal is simpler. Payments should work with fewer delays and fewer failed attempts.
Razorpay wants Vulcan to keep learning from transactions across its network. The company plans to expand its use across more payment decisions in the future.
Razorpay’s Bigger AI Strategy
Vulcan also fits into Razorpay’s wider push into AI-powered payments.
The company has already explored agentic payments, AI app monetisation and AI-powered access to payment data. Its latest foundation model takes that strategy deeper into the payment infrastructure itself.
Razorpay says it ultimately wants more payment decisions to run through the continuously learning model. That could include authentication, routing, fraud detection and lending.
What Razorpay Vulcan Means for Fintech
Razorpay’s Vulcan launch marks a shift from using AI for individual payment tasks to building a shared intelligence layer.
The model combines payment signals from across the ecosystem. It can then use those signals to make decisions about routing, fraud and checkout.
If the technology performs at scale, it could help merchants improve payment conversion while reducing fraud and transaction friction.
For India’s fast-growing digital payments market, that could make AI an increasingly important part of the payment infrastructure itself.
About Razorpay
Razorpay is an Indian payments and financial technology company that provides payment and business-banking solutions for businesses. The company was founded in 2014 by Harshil Mathur and Shashank Kumar, alumni of IIT Roorkee.


Comments
No comments yet. Be the first to share your thoughts.
Leave a comment