OCEN was pitched as 'UPI for credit'. Here's why that never happened

When Nandan Nilekani introduced the Open Credit Enablement Network at the Global Fintech Festival on 22 July 2020, the comparison to UPI was not a stretch reporters added later. It was the pitch itself. UPI had taken a fragmented payments market and given it one open protocol that any bank or app could plug into. OCEN was supposed to do the same for lending: a common set of APIs connecting borrowers, lenders, loan service providers and account aggregators, so a loan could be originated as easily as a UPI payment is settled.

Six years on, that comparison is still the most useful way to measure what actually happened, because the two systems now sit at opposite ends of a very specific gap. One of them publishes exactly how much it moves every month. The other does not publish anything at all.

The pitch was a straight line from UPI to credit

Every month, the National Payments Corporation of India puts a number on its website: transactions processed, value settled, by month, going back years. In January 2026 that number was 21.7 billion transactions worth roughly ₹28.33 lakh crore, an all-time high. In February it dropped to about 20.4 billion, a normal seasonal dip that anyone following the series could explain in one sentence. This is routine. An entire layer of fintech analysis exists purely to read NPCI's monthly release.

OCEN has no equivalent page. There is no OCEN dashboard, no monthly release, no aggregate figure for loans originated through the protocol in 2025, 2026, or any year since launch. Searching for one turns up the same handful of pilot write-ups from 2020 and 2021, recycled in listicles about India's digital public infrastructure stack, with nothing newer attached.

That silence is not a minor administrative gap. It is the data point that tells you what stage a piece of infrastructure is actually at, independent of what the original press releases claimed.

What actually shipped: two pilots, both government-anchored

OCEN itself is not a switch the way UPI is. NPCI operates UPI directly, and every transaction passes through infrastructure NPCI controls. OCEN is closer to a specification: an open API standard published by iSPIRT that anyone can implement, with no central operator collecting fees or compiling statistics. That design choice has consequences that show up directly in what actually got built on top of it.

Two pilots define OCEN's first five years, and both needed a government body to anchor them. GeM Sahay ran on the Government e-Marketplace, with GeM itself acting as the loan service provider, offering working-capital loans to MSMEs against confirmed government purchase orders. GST Sahay, the second, came from SIDBI, working with Online PSB Loans Ltd and iSPIRT, using GST filing data together with the Account Aggregator framework to underwrite small working-capital loans for micro enterprises.

Both pilots are real and both have processed real loans. Both also needed a state-owned anchor, GeM in one case, SIDBI plus a public-sector-bank-backed loan platform in the other, to get off the ground. Neither has grown into the open, multi-player marketplace the original pitch described, where any fintech could plug in as a loan service provider and any bank could plug in as a lender, the way any UPI app can already talk to any bank.

That pattern makes sense once you ask who benefits from going first. UPI had a natural anchor tenant problem solved for it: every bank was already a participant in the existing payments system, so plugging into one more rail was a comparatively small lift, and NPCI could lean on its existing relationships with every major bank to get them onto UPI at launch. OCEN had no equivalent anchor. A lender deciding whether to plug into an open OCEN marketplace had to weigh new underwriting exposure to borrowers it had never dealt with against a known, bilateral alternative, a direct partnership with one fintech, on terms it could negotiate and audit itself. Given that choice, a government-anchored pilot with a known borrower pool, MSMEs selling to GeM, or filing GST returns through SIDBI's partner, was the lower-risk way to test the model. A fully open marketplace, where the lender has no prior relationship with either the loan service provider or the borrower base, was always going to be the harder sell.

Payments standardise. Underwriting doesn't.

The reason isn't a lack of ambition or funding. It's that a payment and a loan are different kinds of decisions, and only one of them reduces cleanly to a standard message format.

A UPI transaction is a single, well-defined instruction: move this amount from this account to that one, now, and irreversibly. Once the message format and settlement guarantee are fixed, any bank can plug in without changing how it makes decisions internally. There is nothing left to standardise.

A loan is the output of an underwriting decision, and underwriting is precisely the part every lender wants to keep proprietary. Two banks looking at the same GST filings, the same purchase order and the same Account Aggregator data feed will reach different risk conclusions, price the loan differently, and ask for different additional documents before they say yes. OCEN can standardise the pipes that move that data from a borrower to a lender. It cannot standardise the decision waiting at the other end, and that decision is where most of the actual friction in Indian MSME lending lives. A protocol that solves data movement but leaves underwriting untouched solves a real problem, just a much smaller one than "UPI for credit" implied.

UPI's rise also had a push behind it that OCEN never got. The government waived the merchant discount rate on RuPay debit and BHIM-UPI transactions entirely, effective 1 January 2020, and ran an incentive scheme that reimbursed acquiring banks roughly 0.15% on small-value UPI transactions to keep the rails commercially viable for them despite the zero charge to merchants. That is a direct fiscal subsidy aimed at one specific adoption curve. Credit has no equivalent: nobody is paying a lender's acquisition cost down to zero to make OCEN-based lending cheaper to originate than a bilateral NBFC deal. A protocol with no underwriting standardisation and no adoption subsidy was always going to spread more slowly than one with neither problem.

The Account Aggregator dependency nobody scheduled around

OCEN's underwriting data, where it exists, mostly arrives through the Account Aggregator framework: RBI-licensed AAs pass consented financial data from a Financial Information Provider to a Financial Information User, encrypted, without the AA itself reading it. GST Sahay was built directly on this dependency.

That dependency had its own adoption curve to climb, on its own schedule, regardless of anything OCEN's designers could control. The Account Aggregator technical standards were published in November 2019, but the framework took years to reach the kind of bank coverage that makes an AA-based underwriting flow reliable rather than a coin flip. Sahamati, the industry body that runs AA ecosystem reporting, recorded more than 24 crore consent requests and 18 crore linked accounts by its fourth anniversary in September 2025, with average daily consent processing up 78.6% year-on-year in FY25. That is real growth, but it is recent growth. A credit protocol launched in 2020 was, in effect, betting on a data layer that did not reach this kind of scale until roughly five years later.

None of this is a mark against OCEN's design, and the Account Aggregator framework's pace is not OCEN's fault. But OCEN's launch timeline assumed a data plumbing layer that took until around 2025 to mature, and a five-year wait for your main dependency is not a footnote. It is most of the explanation for why nothing shipped faster.

The guarantee cap RBI has revised twice since 2023

Underwriting risk is one obstacle. Deciding who absorbs the loss when underwriting is wrong is a second, separate obstacle, and OCEN-style lending ran straight into it.

Most fintech-originated lending in India relies on a First Loss Default Guarantee, now formally a Default Loss Guarantee: the fintech loan service provider promises to cover the first slice of losses on a loan book, letting a bank or NBFC book the loan on its own balance sheet while the fintech carries the downside risk it priced. Before the Digital Lending Guidelines of September 2022, this was unregulated, and some arrangements had fintechs guaranteeing up to 100% of losses, structures that let a non-bank effectively run a lending book while a regulated entity contributed little beyond its licence.

The RBI's response was a hard cap. Its circular dated 8 June 2023 fixed default loss guarantee cover at no more than 5% of the outstanding loan portfolio it protects. That closed the 100%-guarantee workaround, but it also reduced how much balance-sheet flexibility a lending NBFC could draw from a fintech partner, which was exactly the flexibility that made OCEN-style multi-party lending attractive in the first place.

Then the rules tightened further before they loosened again. In May 2025, the RBI required NBFCs to exclude these guarantees from their expected credit loss provisioning, on the grounds that some DLG structures resembled synthetic securitisation closely enough to obscure where risk actually sat. The earnings impact was immediate and public: SMFG India Credit's FY25 profit fell 44% after setting aside ₹115 crore in additional DLG-related provisions, Credit Saison India's profit fell 22% after a ₹178 crore provision, and Northern Arc Capital disclosed an overall impact of ₹780 crore. The RBI then partially reversed course, with new Directions issued on 28 November 2025 and formally notified on 13 February 2026: NBFCs can again factor a DLG into ECL provisioning across all loan stages, provided the guarantee is integral to the loan contract under Indian accounting standards and provisioning is recalculated downward every time the guarantee is actually invoked.

What "still a pilot" looks like from the LSP side in 2026

For a fintech weighing whether to build as an OCEN-style loan service provider in 2026, none of this reads as "infrastructure problem solved, go build". It reads as three open variables a business plan has to absorb at the same time: an underwriting relationship that still has to be negotiated bank by bank rather than plugged in once, an Account Aggregator data pipe that works well for some financial information providers and patchily for others, and a guarantee cap whose accounting treatment the regulator has changed twice in three years.

That is a coherent reason most fintechs building credit products in India in 2026 still default to direct lending partnerships, a bilateral agreement with a specific NBFC, negotiated once, rather than an open OCEN-style marketplace with many interchangeable lenders. A bilateral deal lets both sides fix the underwriting split and the guarantee terms in a single contract, instead of leaving them implicit in a shared protocol with no built-in mechanism for resolving a disagreement over risk-sharing after the fact.

That preference now has its own dedicated rulebook. The RBI's Co-Lending Arrangements Directions, 2025, take effect from 1 January 2026 and extend co-lending well beyond the priority-sector lending it was originally built for, covering any lending activity between two regulated entities. Each partner must retain at least 10% of every loan, down from a previous 20% threshold, funds must move through an escrow account, and a partner bank or NBFC can now rely on the originating lender's KYC instead of repeating it. A default loss guarantee of up to 5% sits inside the same framework. Tellingly, the final directions dropped an earlier draft provision that would have let non-regulated entities, including fintech loan service providers, source loans directly. Fintechs remain partners to a bank or NBFC, not independent participants in the arrangement. RBI built a more efficient bilateral model in the same period OCEN's open marketplace stayed at two pilots, and the contrast tells you which structure regulators are actually willing to scale.

None of this means OCEN is dead. GeM Sahay and GST Sahay both still run, and iSPIRT continues to maintain the specification. It means the protocol has settled into a narrower role, government-anchored working-capital pilots, rather than the open multi-sided market the 2020 pitch described.

What would have to be true for OCEN's UPI moment to arrive

A UPI-style inflection for credit would need three things to land at roughly the same time, not in sequence. Underwriting would need to commoditise enough that lenders compete mainly on price and speed rather than proprietary risk models, which is plausible for narrow, repeat-pattern lending such as invoice or purchase-order finance, and far less plausible for unsecured personal or general working-capital credit. The Account Aggregator network would need to close its remaining coverage gaps, a problem that is already years into being solved rather than years from starting. And the guarantee framework would need a stretch of regulatory stability long enough for lenders to underwrite multi-year credit models against it, a low bar by global standards, but one India's digital lending rules have not yet cleared even once since 2022.

None of that is a prediction that OCEN will or will not get there. It is a description of what "getting there" would actually require, measured against where each of the three pieces stood in early 2026. The honest answer to "is OCEN the next UPI" is that nobody, including OCEN's own backers, has published the number that would let anyone check.