The new foundation model aims to improve payment success rates, strengthen fraud detection, and create a unified AI layer for India’s rapidly growing FinTech ecosystem.
In India’s digital economy, even a small percentage of failed online transactions can translate into millions of unsuccessful payments every month. As businesses and consumers increasingly rely on UPI, cards, wallets, and net banking, ensuring that payments are both reliable and secure has become one of the FinTech sector’s biggest priorities.
Against this backdrop, Razorpay has introduced Vulcan, an AI-powered foundation model designed specifically for payments. Rather than functioning as another chatbot or language model, Vulcan has been built to understand transaction behavior and optimize payment decisions in real time across multiple payment workflows.
A foundation model built for payments.
Unlike conventional machine learning systems that address individual problems separately, Vulcan is designed as a single intelligence layer that supports payment routing, fraud detection, risk assessment, and checkout personalization simultaneously.
The model has been trained on nearly 3 trillion data points generated from 4 billion payment transactions, allowing it to identify patterns across India’s diverse payment ecosystem. According to Razorpay, every transaction is evaluated using around 3,000 behavioural and contextual signals before payment decisions are made. The technology was developed in collaboration with NVIDIA and AWS, which provided the computing infrastructure required to train and deploy the large-scale model.
Why India’s Payments Ecosystem Needs AI?
India processes billions of digital transactions annually, but payment failures remain a persistent challenge. Network congestion, bank downtime, authentication issues, and fraudulent activity often result in abandoned purchases and lost revenue for merchants. Vulcan seeks to address these inefficiencies by learning from historical payment behaviour instead of relying solely on fixed rules. This enables the system to recommend better payment routes, detect suspicious activity earlier, and personalize checkout experiences based on customer preferences.
Razorpay says components of Vulcan have already been deployed across its payments network, delivering measurable operational improvements during early implementation.
Reported outcomes include:
- 8–10% higher payment success rates for participating merchants
- Eightfold improvement in identifying international card fraud
- Five times better detection of fraudulent and disputed transactions without increasing false alerts
- 40% more customers seeing their preferred UPI application during checkout, contributing to an estimated 1–2 lakh additional monthly transactions
Razorpay’s launch of Vulcan reflects a broader shift in India’s FinTech landscape, where AI is becoming integral to payment infrastructure rather than an add-on feature. As digital commerce expands across sectors, payment providers are investing in technologies that can improve transaction reliability while reducing fraud and operational complexity.
The company plans to extend Vulcan’s capabilities into areas such as payment authentication, lending intelligence, and advanced routing, signalling a future where AI models continuously learn from transaction data to make digital payments faster, safer, and more efficient.
