For most of the past decade the question about the Southeast Asia digital lending segment was how fast it could grow. That question has been answered, and the answer was: fast, then less fast. The regional loan book went from about $62bn in 2023 to $76bn in 2024 to $90bn in 2025 — growth of 23%, then 19%. The absolute increments held while the percentage decelerated. This is a maturing market, not an exploding one.
The interesting question now is a different one, and the second quarter of 2026 is the first reporting period in which it became visible in the numbers. The market has split into two halves that are behaving nothing alike.
The Southeast Asia digital lending split
The platform lenders are compounding their books at 60–70% while reporting non-performing loans below 1.5%. Sea’s Monee reported $11.1bn of principal outstanding at Q2 2026, up 62.5%, with a 1.0% 90-day NPL flat quarter on quarter, more than 40 million active credit users and roughly 5.3 million first-time borrowers added in the quarter alone. Grab’s gross loan portfolio reached $2.32bn, up 197%, with the financial services segment guided to adjusted-EBITDA profitability in the second half. GoTo’s fintech book stood at Rp11tn, up 58%, at a 0.8% NPL.
The licensed standalone lenders are compounding at roughly 25% with defaults up by half. Indonesia’s licensed peer-to-peer sector is the clearest lens available, because OJK publishes. It reached Rp105.1tn outstanding in June 2026, up 25.9%. Its TWP90 default proxy sat at 4.26% against a 5% supervisory line — easing from 4.42% in May, but well above where the sector ran a year earlier. Platform count fell from 101 in February 2024 to 94. Sixteen platforms are above the supervisory line; seven are still below the minimum equity requirement.
That divergence is the whole investment question in the sector. It is also only partly about underwriting, and the reasons why are where most analysis goes wrong.
Why lending, and not payments
Before the credit question, the southeast asia digital lending question is a structural one, because it explains why every platform in the region reorganised around this. Lending generated roughly $22bn of the region’s $34.6bn in digital financial services revenue in 2025, about 63% of the pool, and compounded at 21% from 2023, well ahead of payments. Merchant discount rates are compressing toward zero in every major market. The implied portfolio revenue yield on digital lending is around 24%.
If an investor wants Southeast Asian fintech exposure with unit economics attached, credit is currently the only place it lives. That is not a preference. It is arithmetic.
The number that makes the point best came out of Jakarta this quarter. GoTo’s fintech segment produced Rp481bn of adjusted EBITDA in Q2 2026, up 447%, overtaking on-demand services at Rp464bn for the first time. Indonesia’s ride-hailing super app now makes more money lending than moving people.
Why the NPL comparison is close to meaningless
Three reasons, and they compound.
Denominator flattery. A 1.0% NPL on a book growing 62% with short tenors is not the same statistic as a 4.26% default rate on a book growing 26%. A large share of the platform books simply has not aged. And nobody in Southeast Asia publishes static-pool or vintage loss curves — not Sea, not Grab, not GoTo, not one of the standalones. Every ratio above is a point-in-time number on a fast-moving denominator.
A structural advantage that is real but narrow. The platform lenders genuinely see what others cannot: Shopee purchase history, Grab ride and driver-earnings data, a captive repayment channel that can offset against a wallet balance or future earnings, and near-zero incremental acquisition cost. That is not marketing. It is also not underwriting skill, and it does not travel.
Scale asymmetry that distorts intuition. Monee’s book alone is roughly 1.7 times the entire licensed Indonesian peer-to-peer industry — 94 platforms combined. One segment of one company. Any comparison between “the platforms” and “the fintechs” is comparing populations that differ by an order of magnitude in almost every respect.
The test that is already running
The most useful thing about Sea’s disclosure is that the company is dismantling its own advantage in public, and the results will settle the argument.
Off-Shopee SPayLater passed 20% of the total SPayLater portfolio in Q2 2026, and 35% in some markets, driven by QR integration and merchant onboarding. Every point of off-Shopee share is a point where Monee is lending without purchase-level data and without a captive collection channel. Sea is deliberately converting itself from a closed-loop lender into an open-loop one, because the closed loop has a ceiling. It is the right strategic call, and it is the single largest risk to the reported credit metrics.
There is already a signal. Monee’s revenue grew 58.9% in the quarter; adjusted EBITDA grew 12.8%. That gap is provisions. Management attributed it to product and country mix — off-Shopee lending and Brazil, both of which naturally carry higher provisions — and said NPL ratios were consistent within segments and countries. That is a reasonable explanation and may well be correct. It is also exactly what a lender would say if vintages were softening. The right response is to take it at face value and watch the spread for two more quarters.
The control experiment nobody cites
FinVolution is the closest thing the region has to a controlled disclosure, because as a US-listed Chinese lender expanding into Indonesia and the Philippines it reports what nobody else does.
In Q1 2026 its international segment produced 29.6% of group revenue from 3.7% of the loan balance. The implied international take rate was 23.1% of volume, against 5.8% in China, four times the yield. International operating profit was RMB45.8m. A 4.8% margin.
The bull case for Southeast Asian consumer lending is usually stated as: look at the yields. FinVolution’s disclosure says the yields are not the constraint. Credit cost and customer acquisition consume the entire premium.
Regulation is compressing yield at the wrong point in the cycle
This is what turns a credit story into a business-model story.
Indonesia is shrinking its own addressable market from several directions at once: a debt-service-to-income cap stepping down to 40% across 2027–28 and 30% from 2029, a minimum borrower income of Rp3m per month, and an order that non-bank, non-multifinance providers exit BNPL entirely by end-2027. Model that as addressable market contraction, not margin pressure.
The direction of travel is regional rather than Indonesian. The Philippine SEC cut the effective monthly rate ceiling from 15% to 12% from 1 April 2026 — narrowly scoped to unsecured loans of P10,000 or less with tenors of four months or less, but that is precisely the payday-adjacent segment where the yields were. Malaysia’s Consumer Credit Act came into force on 1 March 2026 with licensing from 1 June, and the Hire Purchase (Amendment) Act 2026, effective the same day, ends flat-rate and Rule of 78 pricing. Thailand’s Bank of Thailand responsible lending notification took effect on 31 January 2025.
Every one of these pushes the same way: affordability testing, bureau visibility, reducing-balance pricing. Anyone underwriting Southeast Asian consumer credit on 2021–23 unit economics is modelling a regime that no longer exists.
The questions worth asking
For anyone doing diligence on a lender in this market, portfolio NPL is the least informative number available. Six questions produce more:
- Static-pool or vintage loss curves by origination cohort. The highest-value ask by some distance, and “we don’t produce those” is itself an answer.
- Share of borrowers with three or more concurrent lenders. Indonesian lenders now have the SLIK data to answer this.
- Committed versus uncommitted warehouse capacity, advance rates, covenant triggers, all-in cost of funds, and USD hedge ratio.
- Revenue yield modelled against the published rate-cap step-down and the DSR ladder — not against last year’s book.
- KPPU fine provisioning and appeal status, for anyone holding an Indonesian P2P licence. Ninety-seven lenders were fined a combined Rp755bn on 26 March 2026 for agreeing consumer interest rates between themselves — conduct their industry association has defended, and is appealing, on the basis that it tracked OJK’s own guidance.
- Share of book exposed to export-manufacturing employment, by market. That payroll is the implicit collateral behind a great deal of salary lending and paylater in Vietnam, Thailand and Malaysia.
Three companies in the region have solved funding: Sea, Grab and the wallet-backed banks fund themselves from deposits. Everyone else funds through bank warehouses, offshore private credit and parent equity injections — at costs none of them disclose, with hedge ratios none of them disclose, in a market where Indonesian fintech equity funding fell 83% to $77m in 2025 and there is no visible consumer ABS market in Indonesia, the Philippines or Vietnam.
That is the more likely proximate cause of the next failure in this sector. Not a credit cycle. A funding line.