Razorpay vs Cashfree vs PayU: Build or Buy in 2025?

You have a working product, a GST number, and customers who want to pay online. Now you are comparing Razorpay, Cashfree and PayU. All three claim to solve the same problem, and all three will look similar in a demo. The real decision is not just which logo goes on your checkout. It is whether you buy a payment gateway, or build your own payment layer. That choice hits founders, finance leads, and CTOs differently, and it usually shows up later as leaked margin, manual reconciliation, or a checkout that fails on Friday night.

Separate buying a gateway from building one There are two separate decisions. The first is which payment aggregator to buy. The second is whether to build anything on top. Most Indian SMBs should buy a gateway and keep their engineering focused on the product. Building a full payment gateway from scratch means an RBI Payment Aggregator licence, a minimum net worth of ₹15 crore at application (rising to ₹25 crore), PCI DSS compliance, escrow or nodal account operations, settlement teams, and ongoing regulatory reporting. That is not a typical SMB project.

A thin payment orchestration layer is different. That is code between your checkout and one or more gateways. It can route UPI to one provider, cards to another, retry failed transactions, and fall back when a gateway has an outage. That can be worth building at scale. But it is not the same as becoming a payment gateway.

What Razorpay, Cashfree and PayU actually differ on The feature lists overlap heavily. All three support UPI, cards, netbanking, wallets, webhooks, refunds, and payment links. The differences show up in operations.

• Razorpay: strong checkout experience, dashboard, payment links, subscriptions, invoices, and a broad app ecosystem. Often a good default for SaaS, D2C, and SMB teams that want to move fast.
• Cashfree: API-first, strong payouts, bulk refunds, and marketplace or lending use cases. Pricing can be competitive, especially if you negotiate with volume.
• PayU: enterprise relationships, high-volume merchant support, international cards, EMI options, and banking integrations. Onboarding can be slower, and pricing is usually negotiable.

The real differences are less visible: authorisation rates by issuer and category, webhook reliability, refund turnaround time, settlement report quality, support response times, and how quickly they resolve a stuck settlement. A gateway that saves you 0.1% but breaks reconciliation every month is not cheaper.

The numbers that matter Take a business collecting ₹40 lakh per month online. At a published rate of 2% + 18% GST, the gateway fee is ₹80,000 plus ₹14,400 GST, or ₹94,400 per month. That is about ₹11.3 lakh per year. If you negotiate a 0.25% lower rate, you save ₹10,000 plus GST per month, or ₹1.2 lakh plus GST per year. At ₹1 crore monthly volume, the same 0.25% difference is ₹25,000 plus GST per month, or ₹3 lakh plus GST per year.

Also check settlement cycle. T+1 versus T+2 matters if you are paying suppliers or running ads daily. Chargeback fees typically range from ₹100 to ₹500 per dispute. Refund fees, setup fees, annual maintenance charges, minimum volume commitments, and rolling reserves all change the effective cost. GST input credit is available if you are GST-registered, so keep invoices clean. Ask your CA whether the gateway fee attracts TDS under Section 194H or 194J; rates have changed in recent years.

When a build makes sense In our experience, the case for a thin routing layer usually appears when you are processing roughly ₹50 lakh to ₹1 crore or more per month, you have multiple gateways, and you have at least one engineer who can own payments. At that point, you can build a small service that chooses a gateway based on payment method, issuer, amount, and live success rates. You can retry failed UPI collect requests on a second provider. You can fall back automatically when webhooks stop arriving. You can also centralise refunds, settlements, and reconciliation in one place.

Below that volume, start with one primary gateway and one backup. Use the backup manually for large orders or when the primary has an outage. Revisit the build decision every quarter.

How to run a two-week bake-off Do not choose from a sales deck. Run a controlled test.

• Get test and production keys for two gateways.
• Route 10-20% of real traffic to each, keeping the checkout experience identical.
• Track authorisation rate, latency, failure reasons, webhook delivery, refund TAT, and settlement report accuracy.
• Raise a support ticket with each provider and measure first response and resolution time.
• Reconcile one full settlement cycle manually. Do order IDs match? Are fees itemised?
• Review the contract for setup fees, minimum volume, chargeback fees, refund fees, rolling reserve, and notice period.
• If your monthly volume is above ₹50 lakh, ask for better pricing. Ask what UPI, cards, and netbanking cost separately.
• After 30 days, keep the winner as primary and the runner-up as backup.

Common mistakes include choosing on brand alone, ignoring GST and TDS, hardcoding one gateway with no fallback, and not making webhooks idempotent. Under the DPDP Act, payment data, logs, and consent records need care. Do not store card data unless you are tokenised and compliant.

Where to start Start with a simple spreadsheet: monthly online volume, UPI versus card mix, settlement needs, refund volume, and engineering capacity. Buy a primary gateway, add a backup, and build only the thin layer that your volume and operations justify. If you need help wiring a gateway-agnostic checkout, reconciliation jobs, or a finance dashboard that shows fees and settlements in one place, Maysan Labs builds those systems.