Business

Know the Business — Kaspi.kz JSC

Kaspi.kz is a deposit-funded consumer bank, a card + QR acquirer, and a marketplace stitched together by a single super-app used by roughly 83% of Kazakhstan adults. The economics that matter are bank economics — net interest margin on a ₸-funded loan book — re-skinned with software distribution and an ad-tech overlay. The market is pricing Kazakhstan + Türkiye risk and a funding-cost squeeze; the underweighted facts are that the three engines generate near-equal net income and that advertising — the highest-margin overlay — is only beginning to monetise. The thing to get right before owning the stock is whether the Türkiye build-out is a beachhead or a quagmire.

1. How This Business Actually Works

Kaspi sells digital distribution of payments, commerce, and credit through one mobile front-end, and earns a different unit of revenue on each leg of the same transaction. A customer acquired once is monetised three times: a fee on payment volume, a commission (plus ad/delivery fees) on marketplace transactions, and net interest income on consumer/SME loans funded by retail tenge deposits. The cost to acquire a user is paid once — the revenue compounds across three pots — which is why returns on capital are exceptional and why the model is hard to disrupt with a single-purpose competitor.

The three engines map to three different monetisation mechanics, and reading them together is the only honest way to understand the P&L.

The Fintech leg dictates the cycle: it carries credit risk and is funded by retail deposits, so policy-rate moves and credit-quality vintages drive the swing factor. Payments and Marketplace are asset-light in their core form (Kaspi exited the 1P e-Cars line in Q4 2025 because "it was capital-intensive with limited possibility to replicate in other markets"). Group ROE printed in the high-50s percent for FY2025 — multiples above what a global universal bank earns — because the three engines share one customer base and one stack.

The FY2025 revenue jump from ₸2.53T to ₸4.05T is roughly two-thirds Hepsiburada consolidation, one-third organic. Net income grew only ~3% absolute because Hepsiburada runs at an EBITDA loss and Kazakhstan funding costs rose ~140 bps in Q4 alone. Underlying (ex-tax/regulatory/base-rate noise) net income grew 18% — closer to the real economic engine.

2. The Playing Field

There is no listed pure-play super-app substitute for Kaspi, which is why investors get the multiple wrong. The closest analogs are six names, each illuminating a slice of the bundle but none capturing all three: Nu Holdings (digital bank), MercadoLibre (e-com + Mercado Pago + credit), Hepsiburada (now a Kaspi subsidiary), PagSeguro and StoneCo (Brazilian POS + SME credit), and Halyk Bank (the one in-country listed competitor).

Three facts jump off the page. First, Kaspi earns the highest ROE in the group by a wide margin — almost double what Nu earns and roughly 1.6× MercadoLibre — because the three-engine super-app structurally captures more value per customer than any pure-play peer. Second, it trades at the bottom of the multiple range alongside the Brazilian acquirers (PAGS, STNE) and Halyk Bank, which is a Kazakhstan country-risk multiple, not a super-app multiple. Third, MercadoLibre's 43.5× P/E on a 36% ROE versus Kaspi's 7.8× P/E on a 58% ROE is the largest valuation dislocation in the peer set — and the structural answer is that MELI is priced for LatAm runway while KSPI is priced for Kazakhstan saturation. The investment debate reduces to whether Türkiye can re-rate Kaspi toward the MELI/NU bucket, or whether HEPS will drag the consolidated multiple toward the PAGS/STNE bucket.

3. Is This Business Cyclical?

Kaspi has three cycles, not one — and the leg that hurts first depends on which macro variable is moving. The headline net-income line therefore hides what's really happening underneath. In a rate-hike cycle the Fintech leg compresses first (NIM and Cost of Risk). In a consumer-spending downturn the Payments leg shows up first (volume slows, lower-take products grow faster than higher-take cards). In a marketplace supply shock — like the 2025 smartphone import-registration disruption in Kazakhstan — the e-commerce GMV slows but value-added services on the rest of the platform absorb part of the hit.

The defining cycle today is the rate cycle. Kaspi's funding cost rose roughly 220 bps cumulatively into Q1 2026 (deposit rate reached 14.3%). That is the gap between underlying net-income growth of ~18% and reported growth of ~10% in FY2025 — almost the entire compression is funding cost, not credit quality or business model decay. The corollary, which the bull case rests on, is that NBK rate cuts later in 2026 would mechanically reverse this compression and the Fintech leg would re-accelerate without management doing anything.

Read this chart left to right: in Payments, volume grows faster than revenue (take-rate compression); in Marketplace, revenue grows faster than volume (value-added services scaling) but net income lags both (delivery cost on small-ticket frequent items); in Fintech, net income grew slowest because the deposit cost squeeze took a 220 bps bite out of NIM. None of the three is broken — but each is fighting a different battle, and that diversity is part of why the consolidated number is steady.

4. The Metrics That Actually Matter

Revenue and EPS hide the story because they collapse three different mix shifts into one line. The metrics professionals actually watch sit at the segment level — and three of them, in particular, predict whether the multiple expands or compresses.

The honest hierarchy is: Cost of Risk and deposit cost dominate the equity story today, because the Fintech leg is the largest profit pool and the most rate-sensitive. Take-rate (both Payments compression and Marketplace expansion) defines the long-term thesis — VAS is the leg with the most runway because only 7% of merchants pay for advertising. Hepsiburada engagement metrics matter because they decide whether the consolidated multiple gets MELI-like growth optionality or PAGS-like discount.

5. What Is This Business Worth?

The right lens is earnings power × growth runway on a single integrated franchise, not a sum-of-the-parts shopping list. Two-thirds of the profit pool is a deposit-funded EM bank that earns mid-50s percent ROE because the marketplace and payments legs feed its underwriting data and customer base. Trying to value the three legs separately overstates precision: management runs them as one stack, the loan book is funded by deposits gathered through the Super App, and the Hepsiburada acquisition was paid for with the consolidated cash flow.

That said, SOTP is genuinely useful as a stress test because Hepsiburada and the listed e-commerce/payments peers carry materially different multiples and Türkiye's economic profile is different enough that consolidating it inside one P/E multiple distorts both directions.

6. What I'd Tell a Young Analyst

Read every quarterly press release from the segments first, headline last — the consolidated revenue and net income compress three different stories (a payment-mix shift, a marketplace VAS scale-up, and a NIM squeeze) into one number. Track three things in priority order: Cost of Risk (because it is the single canary for the largest profit pool), Marketplace take-rate + VAS yield (because that is the leg with the most runway), and Hepsiburada engagement metrics (because the Türkiye outcome decides whether this re-rates).

What the market may be missing: the price reflects a Kazakhstan country-risk bank, but the actual engine is a three-engine consumer franchise with ~58% ROE on a deposit-funded balance sheet, and management has resumed a ~5-6% dividend yield while investing in Türkiye. The gap between underlying (+18%) and reported (+10%) FY2025 net income growth is almost entirely funding cost — the cycle leg most likely to reverse on the next NBK easing.

What would change the thesis. Bull case dies if (a) Cost of Risk drifts above 3% (vintages cracking), (b) Marketplace take-rate stops expanding and ad/delivery growth flattens, or (c) Hepsiburada order growth stalls back into single digits without engagement convergence. Bear case dies if (a) NBK cuts the base rate and deposit costs lag (mechanical NIM expansion), (b) HEPS reaches Adj-EBITDA breakeven on schedule with engagement metrics improving, or (c) Rabobank A.Ş. closes cleanly and the TR fintech monetisation path becomes credible. The 7.8× P/E is an unstable anchor either way; the question is which catalyst lands first.