The Next Phase of India's NBFCs Will Be Won on Infrastructure, Not Arbitrage

For most of the last two decades, India's non-banking financial companies grew in the gaps that banks left behind. They lent where banks would not go, priced risk banks could not model, and moved at a speed a branch network could not match. That story is now over, not because NBFCs have stopped growing, but because the gaps themselves are being formalised. NBFC credit today stands at roughly 16.7 percent of India's GDP and about 27 percent of scheduled commercial bank credit.¹ The regulator has stopped treating the sector as a shadow. It is treating it as a pillar.

I run a company that builds the core lending systems on which NBFCs and microfinance institutions operate, so my view is from the engine room rather than the boardroom. From there, four shifts will decide who thrives over the next decade.

Regulation is converging with banking, and compliance is becoming a data problem

The scale-based regulatory framework was the first signal. What has followed in 2026 makes the direction unmistakable. Expected credit loss provisioning, now mandated for commercial banks from 1 April 2027 with NBFCs expected to follow,² replaces incurred-loss thinking with forward-looking models defensible to auditors and supervisors. A consolidated recovery conduct framework takes effect on 1 January 2027³ and reaches into how every field agent and every call is managed. Cybersecurity and technology risk directions that once applied only to banks now apply to NBFC boards.

None of this can be handled by a compliance officer with a spreadsheet. ECL needs clean, historical, loan-level data with consistent staging logic. Recovery conduct needs every borrower touchpoint logged and auditable. NBFCs built on fragmented tools, an LMS here, an accounting package there, collections on WhatsApp and Excel, will find the cost of regulation is no longer a fee. It is a re-architecture.

The balance sheet is becoming a network

The revised co-lending directions in force since January 2026 changed the economics of partnership: minimum retention cut to 10 percent from 20, default loss guarantees of up to 5 percent, and the framework extended to every regulated entity.⁴ A mid-sized NBFC can now originate far more than it can fund. Add securitisation, Account Aggregator rails, and the Unified Lending Interface moving from pilot to systemic rollout,⁵ and the NBFC of the future looks less like a lender and more like an orchestrator, originating, underwriting, distributing and servicing across multiple balance sheets at once.

The constraint on this model is not capital. It is plumbing. Co-lending only works when two institutions can agree, in real time and at the loan level, on who owns, collects and provisions what. NBFCs whose systems cannot expose loan data cleanly to a partner bank will be locked out of the cheapest funding the sector has ever had.

Underwriting is moving from collateral to cash flows

India's digital public infrastructure has quietly changed what it means to assess a borrower. GST filings, consent-based bank statement sharing, UPI histories and now ULI mean an MSME or microfinance customer can be underwritten on what they earn rather than what they own, precisely the segment where bank models are weakest. But it also raises the bar: cash-flow lending demands models retrained continuously and decisioning explainable to a regulator increasingly asking how an algorithm arrived at a decline.

AI and automation are moving from the pilot deck to the production floor

Every NBFC board has seen an AI presentation by now. What separates 2026 from 2023 is that the use cases have become boring, which is when technology starts to pay. RPA closes the gap between origination and disbursement, handling document collection, bureau pulls and sanction letters without a maker-checker queue of humans. Document intelligence reads bank statements and ITRs in seconds. Voice bots handle first-level collections in a dozen languages, routing sentiment flags to humans only where needed. Generative AI drafts credit appraisal memos from structured data.

Underneath, machine learning models drive early warning signals from repayment behaviour, monthly-updating propensity-to-default scores, fraud detection across applications sharing a device or bank account, and dynamic collections strategies. The cumulative effect is not marginal efficiency. It is a change in institutional shape: an NBFC that once needed one credit officer per hundred files can run several hundred, and servicing a Rs 50,000 loan starts to look viable rather than charitable.

The caveat is that AI is only as good as the data it sees. A model fed by three systems that disagree on the outstanding balance will confidently produce wrong answers at scale.

Nothing above works without an end-to-end connected system

The sector's real technology debt is not the absence of AI. It is the absence of a single, connected record of the loan from lead to closure. In most NBFCs I visit, origination lives in one system, the loan book in another, accounting in a third, collections in a fourth, and reconciliation between them is a monthly exercise conducted by people. Every regulatory ask, every co-lending partner, and every AI model then becomes a project of stitching systems together after the fact.

An end-to-end connected core, where onboarding, KYC, underwriting, sanction, disbursement, servicing, collections, accounting and regulatory reporting sit on one data spine, changes the economics of everything above. ECL staging becomes a query, not a quarter-end scramble. Co-lending reconciliation becomes an automated feed. A new AI model trains on the truth rather than one department's version of it. A board sees a real-time portfolio view rather than a fortnight-old extract. When the regulator issues a new direction, the change is made once and propagates to every workflow.

This is the difference between an institution that absorbs change and one that is consumed by it. I believe the next wave of consolidation in the sector will be driven less by capital and more by operating architecture.

What this means for the long tail

The Union Budget's call to consolidate public-sector NBFCs into fewer, larger entities is a signal for the private sector too.⁶ Scale will matter, but not as it did in 2018. The dividing line will be between institutions built to absorb regulatory change, partnership complexity and data-driven underwriting, and those running on patchwork. A Rs 500 crore NBFC with a modern, integrated stack will be more resilient than a Rs 5,000 crore one with fifteen disconnected systems, because the former adapts in weeks and the latter in years.

The regulator has laid the rails. Co-lending, ULI, Account Aggregator, the recovery framework and ECL together describe a sector that is transparent, interoperable and accountable. AI, automation and cash-flow underwriting are the engines; a connected core is the chassis they bolt onto. The next phase of Indian NBFCs will be won by those who treat technology as the balance sheet's operating system, not a cost centre attached to it.

Sources

1. RBI Deputy Governor S.C. Murmu, 7th CII NBFC & HFC Summit, 3 Sept 2026 (rbi.org.in). ↩
2. RBI Governor's statement, 1 Oct 2025; final ECL directions, 27 Apr 2026 (rbi.org.in). ↩
3. RBI Responsible Business Conduct amendments, Aug 2026 (rbi.org.in). ↩
4. RBI (Co-Lending Arrangements) Directions, 2025 (rbi.org.in). ↩
5. Reserve Bank Innovation Hub, ULI updates (rbihub.in). ↩
6. Union Budget 2026-27 speech (indiabudget.gov.in). ↩

This article was also shared on Instagram.