The MSME credit gap in India has been quoted as somewhere between Rs 25 lakh crore and Rs 30 lakh crore for so long that the number has lost its ability to surprise. What has changed in the last three years is the volume of platforms claiming to address it. By a rough count, there are now over 80 fintechs operating in the MSME lending space in some form - origination, underwriting, marketplace, or balance sheet lender.

Two years into the post-2023 regulatory tightening on digital lending, the dust has started to settle. Some patterns are clear.

Account Aggregator-led underwriting has matured

The most genuinely positive structural development has been the maturation of the Account Aggregator framework for MSME underwriting. Three years ago AA-based loan flows were running at low volumes with patchy data quality. By Q1 2026 the network is processing over 14 crore consents per quarter, and MSME-focused use cases are a meaningful slice of that.

The practical effect is that fintechs originating MSME loans now have access to bank statement data with a fraction of the friction that used to exist. Underwriting decisions that used to take three to five days happen in hours for the cleaner segments. The cost-to-acquire is genuinely down for lenders that have built their stacks on AA from the start.

What hasn’t changed is the quality variation. AA gives you cleaner data; it doesn’t give you better businesses. The lenders that have done well are the ones that built their underwriting models around AA data flows from the beginning rather than retrofitting them. The legacy NBFC players that bolted AA onto existing manual processes have generally seen smaller efficiency gains than they expected.

GST data integration is finally useful

The other foundational shift has been the practical usability of GST return data for underwriting. The data has been technically available for years, but the consent and access frameworks were clunky. By 2026 several lenders have direct or near-direct integration with GSTN, with consent flows that work in under a minute.

This matters because GST data is the single best signal for genuine business turnover in the formal MSME sector. It’s verified by the tax authority, it’s harder to game than self-declared revenue, and it has the granularity to support sector-specific risk modelling. Lenders that have invested in GST-data underwriting models are seeing materially better portfolio performance than those still relying primarily on bank statements and bureau scores.

Where defaults are concentrating

The trouble is concentrated in two segments. First, the unsecured, ticket-size-under-Rs-5-lakh business loans that several fintechs scaled aggressively in 2024-2025. The loss rates here are running well above original modelling assumptions - some platforms are reporting 90+ DPD in the high single digits, which makes the unit economics very challenging at the rates they’re charging.

Second, the supply-chain finance segments tied to specific anchor corporates have shown clustered stress when the anchor’s sector turns. A few prominent platforms with concentration in particular industries learned this the hard way through 2025.

The lower-risk segments - secured MSME term loans, GST-data-driven cash flow lending against established businesses, and trade finance with insurance wraps - have performed broadly in line with expectations. The market is bifurcating.

Regulatory signals

The RBI has been clear in its quarterly commentary that it expects digital lending platforms to operate with the same prudential discipline as bank lenders, even when they’re operating as agents or co-lending partners rather than principals. The default-loss-guarantee structure that some platforms used to push portfolio risk back onto originators has been tightened. First-loss positions are now disclosed and treated as on-balance-sheet exposure for capital purposes.

This has changed the economics of platform lending in ways that aren’t fully reflected in published unit economics yet. Several fintechs that grew rapidly under looser interpretations of the rules are now sitting on portfolios that look less attractive once the capital treatment is applied.

The recent guidance on co-lending arrangements has also tightened the requirements around credit decisioning ownership. Platforms that pitched themselves as origination engines for bank balance sheets have to demonstrate genuine credit assessment capability or step back to a pure marketing role.

Where AI is helping and where it isn’t

The conversation about AI in MSME underwriting has matured. Three years ago every pitch deck had a slide about machine learning models that would unlock new credit pools. The reality has been more measured.

AI is genuinely useful for two things in the MSME space - early-warning signal detection on existing portfolios, and automating the document collection and verification workflow. Both produce real cost savings. AI is less useful, in practice, for substituting hard credit data with alternative signals when underwriting genuinely thin-file applicants. The models that promised this haven’t held up well in default cycles.

The platforms doing serious AI work tend to partner with specialist consultancies for the harder modelling and infrastructure pieces. There’s a small but growing ecosystem of Asia-Pacific AI specialists serving Indian financial services, alongside the larger global players.

The 2026-2027 outlook

Three things to watch over the next eighteen months. First, the consolidation among the smaller fintech lenders is going to accelerate as funding markets stay disciplined and unit economics on unsecured MSME lending stay stressed. Expect at least a dozen meaningful acquisitions or wind-downs by year-end.

Second, the larger platforms with GST-data-driven underwriting at scale are likely to expand into adjacent products - working capital, invoice finance, and embedded credit within ERP platforms. The unit economics here are better.

Third, the regulator’s posture matters more than market dynamics for the next phase. If the current tightening continues, the platforms that survive will be smaller in number, larger in average size, and tightly integrated with bank balance sheets. That’s broadly a healthy direction, even if it’s uncomfortable for the current crop of mid-tier players.