Razorpay rolls out AI payments model trained on 4 billion transactions

Razorpay says its India-built payments foundation model improves transaction success and fraud detection. Blinkit recorded a 1–2 percentage-point payment-success uplift in testing, while some merchants now route 50–60% of payment volume through the model.

— Source published Tue, 18 Aug, 2026, 06:00 IST · First seen Tue, 18 Aug, 2026, 06:07 IST · Source The Hindu BusinessLine

What happened

Razorpay launched an India-built AI payments foundation model trained on four billion transactions, improving merchant payment success and fraud detection.

Key facts

  • 4 billion transactions used for training
  • around 3,000 signals per transaction
  • 1.5 million transactions tested
  • more than 51,000 businesses tested
  • 8-10% improvement in payment success rates
  • 8x increase in international card fraud detection
  • 5x increase in fraudulent or disputed transaction identification
  • 50-60% of payment volumes routed through the model by some merchants
  • 1-2 percentage point payment-success-rate improvement at Blinkit

Why this matters

Retail, fintech and commerce platforms should view Razorpay as a strategic payments partner or competitive benchmark as AI-led authorization optimization becomes a differentiator.

What to watch

  • Whether Blinkit's uplift holds after rollout beyond the initial test cohort and across UPI, cards, wallets, and net banking.
  • The percentage of Razorpay merchant payment volume routed through the model, particularly among large enterprise accounts.
  • Reported changes in fraud chargebacks, false-positive declines, and payment-processing costs alongside success-rate improvement.
  • Evidence that issuers or payment networks grant better approval performance to model-routed transactions.
  • Rival product announcements, pricing changes, or merchant migration disclosures in India's payment-gateway market.
  • Regulatory scrutiny around AI decisioning, data localization, consent, and fraud-liability allocation.
  • Razorpay will pursue published merchant case studies segmented by payment method, issuer, geography, and transaction size to substantiate uplift claims.
  • Quick-commerce and high-frequency merchants will test routing more payment traffic through the model during peak-hour and failed-payment-heavy cohorts.
  • Razorpay will package AI routing, fraud controls, retry logic, and checkout optimization into higher-value enterprise payment bundles.
  • Competitors including Indian gateways, PSPs, and bank acquirers will accelerate proprietary authorization-rate and fraud-model launches.
  • Merchants may redirect part of saved payment-failure leakage into promotions, delivery capacity, or customer retention rather than pure margin expansion.