Long-Term Thesis

Long-Term Thesis - Kaspi.kz JSC

1. Long-Term Thesis in One Page

The long-term thesis is that Kaspi compounds book value at 20%+ per year for the next decade if — and only if — three things hold: (1) the Kazakhstan three-engine flywheel sustains a 50%+ Kazakhstan-only ROE through the next NBK rate cycle and the regulatory ratchet, (2) Hepsiburada plus the Rabobank A.Ş. licence transplant the deposit-funded super-app playbook into a market five times the size of Kazakhstan, and (3) the still-nascent Marketplace VAS engine (advertising, delivery, classifieds) scales from 2.3% of 3P GMV today toward the 4-6% benchmark global super-apps reach. This is not a long-duration compounder if the Marketplace stays a low-take 3P, the Türkiye build-out drags engagement under 10 purchases per consumer per year, or the FY2023 ICFR weakness in the ECL model proves under-reserving rather than mix-shift. The franchise is one of the highest-quality EM consumer engines listed anywhere — 51% consolidated ROE on a 93% retail-tenge deposit-funded balance sheet, book value per share compounding at a 45% CAGR FY2020-FY2025 — but the durable upside requires the moat to either deepen at home or transplant abroad, and only one of those has been proven.

The honest grade is High thesis strength, High Kazakhstan-only durability, Medium reinvestment runway (capped by a 19M-person home market until Türkiye proves out), Medium evidence confidence because the load-bearing data-flywheel claim depends on an ECL model whose internal controls have failed in production once and whose first Section 404(b) auditor attestation has not yet landed. Read this page as the underwriting frame for a five-to-ten-year position, not a catalyst calendar.

2. The 5-to-10-Year Underwriting Map

The driver that matters most is the three-engine integration at the top of the table — every other line in the map either reinforces it (deposit funding, capital allocation, VAS depth) or is a hedge against its single weakness (Türkiye is the geographic hedge, governance maturity is the disclosure hedge). If the Kazakhstan flywheel cracks — through ECL under-reserving, forced unbundling, or founder departure — no amount of Türkiye execution or VAS scaling recovers the equity story. If it holds and one of the two growth legs (Türkiye OR Marketplace VAS) delivers, this is a structural compounder. If both deliver, it is the highest-quality EM consumer-internet asset listed.

3. Compounding Path

The cleanest test of whether the model can compound for another decade is to read returns on capital and book value per share across a full cycle that already includes Covid, two activist shocks, a 220 bps funding-cost step-up, and the Hepsiburada consolidation event. The franchise has compounded through every one of these.

Book value per share compounding at 45% per year for five years is the cleanest single number on this page — it is unitless of consolidation effects, dividend pauses, or metric-changes, and it isolates the rate at which retained earnings produce additional shareholder capital. Even if the next decade halves that rate — to a 20% book-value CAGR — the compounding case stands.

The compounding mechanics are unambiguous in the past data: revenue scaled from ₸47B in FY2015 to ₸4,046B in FY2025 at a ~58% ten-year CAGR while operating margin held in a 75-79% band from FY2019 through FY2024 — a profile only a handful of listed businesses anywhere can match. The FY2025 flatline in net income is the Hepsiburada consolidation effect dressed up as a growth break, not a structural deterioration of the Kazakhstan engine. The cash conversion story is the area where the model gets tested most by the bear case: FCF/NI fell from 1.3× (FY2022-23) to 0.46× (FY2024-25) as the Hepsiburada cash deployment, the dividend pause, and the working-capital draw of fintech loan-book expansion combined. For the long-term thesis to compound, FCF/NI needs to migrate back above 0.7× by FY2027 as Türkiye normalises and dividends remain resumed.

4. Durability and Moat Tests

A 5-to-10-year thesis has to be tested against shocks that are not the next quarter's print. Five durability tests, drawn from the moat and competition work, define whether the franchise survives the decade or merely the cycle.

The asymmetric setup: two of the five tests (integration ROE and brand) are already passing at near-decisive strength, two (VAS scale and Türkiye transplant) are at credible-but-unproven levels with visible improvement trajectories, and one (underwriting/ECL) is the single test most likely to break the thesis. A long-term position underwrites whether the underwriting moat holds — every other test moves the magnitude of returns, but only that test moves the sign.

5. Management and Capital Allocation Over a Cycle

The 5-to-10-year case rests partly on whether the same management team that built the Kazakhstan franchise can deploy capital sensibly across the next decade — through a Türkiye build, a likely second-country expansion, and a normalisation of the in-country regulatory ratchet. The track record reads honestly when separated into three things they did well, two they have not been tested on, and one they have changed mid-flight.

What the record supports. CEO Lomtadze has been in seat 19 years; CFO Mosidze, Deputy Mironov, and Deputy Didenko have all been on the senior team since 2007-2008. The founder duo (Lomtadze 23.04%, Kim 20.78%) owns ~$6.5B of stock between them at current prices and has never engineered material dilution — share count was 191.8M in FY2020 and 190.6M in FY2025. LTIP design is shareholder-friendly to the point of being unusual for a fintech: nominal-cost options on a 5-year vest, full forfeiture on departure, and ~0.7% unvested overhang. Cumulative FY2020-FY2024 dividends were ~₸1,932B (~$4.3B) — meaningful capital return through a heavy growth period. Two IG bond placements in twelve months (March 2025 $650M at 6.250%, April 2026 $600M at 5.900% Baa3/BBB−, 3.5× oversubscribed) demonstrate that institutional credit markets price the credit at investment grade through the worst of the activist-short and rate-cycle period.

What the record does not yet prove. Türkiye is the first true cross-border deployment of company capital — the ₸553B Hepsiburada acquisition (plus ~$168M of additional stake-up to 85.66%, plus the $300M Rabobank A.Ş. capital injection planned at closing) is roughly 17% of equity in a single country call. The playbook (deposit-funded consumer credit + integrated marketplace) is unproven outside Kazakhstan. The post-Türkiye second growth-leg deployment — whether the next decade brings a second-country expansion to Azerbaijan/Uzbekistan/CIS or an over-extension of the integrated model in a market where it does not fit — is the largest single capital-allocation test ahead.

What they have changed mid-flight. FY2025 net income guidance was cut twice (from ~20% to ~15% to 10-12%), the FY2026 guide switched the headline metric from net income to Adjusted EBITDA, and the geographic scope flipped from "ex-Türkiye" to "incl. Türkiye." Each step is individually defensible — Hepsiburada consolidation makes the comparison non-meaningful, and the Adj-EBITDA-to-IFRS-NI gap widened from 20% (FY2023) to 46% (FY2025) primarily because the add-back strips ₸369B of non-fintech interest expense including the cost of the Eurobonds raised to fund Türkiye. The cumulative effect, however, is that the headline 2024 IPO-era guidance is no longer comparable to the 2026 guidance management has chosen to lead with. The historian's credibility score of 6/10 captures the right read: broadly trustworthy on operations, suspect on what management chooses to talk about and when.

The defining capital-allocation question for the next decade is whether the team treats Türkiye as a single, contained capital call or as the first of a series. If HEPS reaches the engagement metrics Kazakhstan demonstrated and Rabobank A.Ş. closes cleanly, the playbook is proven and a measured second-country deployment is value-accretive. If HEPS plateaus and the team responds by deepening the bet rather than absorbing the impairment, that is the classic late-cycle pattern that ends in a goodwill write-down and a multi-year ROE compression. The single biggest tell will be whether the FY2027-2028 capital plan emphasises consolidation of the existing two-country business or signals a third geography.

6. Failure Modes

The red-team failure modes below are the specific, observable conditions under which the long-term thesis breaks. They are not generic "execution risk" — each maps to a disclosure event or a measurable metric.

7. What To Watch Over Years, Not Just Quarters

These are five multi-year milestones that will validate or weaken the long-term thesis. Each is observable in dated disclosures, not in tape action.

The long-term thesis changes most if the first Section 404(b) auditor attestation lands clean for FY2025 and NPL coverage rebuilds above 90% within four quarters — that combination would validate the data-flywheel moat that anchors the entire equity story and remove the single failure mode capable of resetting the moat rating from "narrow and proving" to "narrow and shrinking." Every other multi-year signal moves the magnitude of returns; this one moves the sign.