A 7×-earnings super app yielding 8–9%, with a free option on 85 million Turks

1The thesis in one paragraph

Kaspi.kz is the dominant super app of Kazakhstan — payments, shopping, lending, travel, groceries and government services inside one application used by 25 million people, two to three times every day. It earns roughly $2 billion of annual net income, returns most of it to shareholders, and trades at about seven times earnings with a dividend yielding 8–9% at the newly raised rate. That is a valuation normally reserved for a shrinking bank in a bad neighbourhood.

The discount exists for reasons that are real but, I believe, mostly temporary: a Kazakh bank-tax rise, high domestic deposit costs, and an expensive expansion into Türkiye that has not yet produced profit. My base case is a 2–3x over three to five years as earnings resume growing and the multiple normalises.

2What just happened: Q2 2026

Results landed on 10 August. They were, in a word, mixed — and worth reading honestly. The market took them well: shares rose about 3.4% after hours. Below is the complete picture rather than the flattering half.

The good: the dividend was raised 18%, and management framed it as confidence in the long-term outlook. Boards do not raise dividends when they are worried. e-Commerce is compounding hard — constant-currency GMV +28% with revenue +35%, which means Kaspi is capturing more of each sale, not less. The third-party take rate rose 160bps to 16.1%, and advertising-plus-delivery revenue grew 49%. That is the high-margin layer, and it is the fastest-growing thing in the business.

The bad: revenue growth halved from +31% to +15%. Net income was flat again — roughly zero profit growth for three consecutive quarters. Payments stayed weak at +5%. And first-half marketplace GMV is running near 17% against a ~20% full-year guide, which implies the second half has to accelerate.

3Why profit is stuck — and why that may be ending

Three forces have been compressing the margin, and it matters enormously which of them are permanent:

• Deposit funding costs — reversing now. Kazakhstan’s high rates pushed funding costs up 220bps year-over-year, to 14.3%. This is the largest single driver of flat profit. In August, Kaspi cut its three-month deposit rate by 100bps — about a third of the deposit base, and the first cut in over two years. Repricing takes time to flow through, but the catalyst is specific, dated, and already in motion.
• The bank tax — permanent. Kazakhstan raised the rate on banks from 20% to 25% effective January 2026. This does not reverse. It is a permanent haircut and should simply be priced in.
• Türkiye — deliberate investment. Hepsiburada is a thinner-margin business, so consolidating it mechanically lowers group margin, and management has guided the Turkish operation to roughly breakeven for 2026. It depresses reported numbers by design, not by accident.

One of these is reversing on a known timeline, one is permanent but modest, and one is a deliberate investment. None of them is customers leaving or the product failing. That is why I read flat profit as a squeeze rather than a decline.

4The one thing that genuinely worries me

Payments revenue growth has slowed every single year: roughly +37% and +30% in 2022–23, then +22%, +12%, and now +7% and +5% in the last two quarters, with segment EBITDA flat. Payment volume still grew 13% — consumers have not left. Kaspi is keeping a smaller share of each transaction while spending on new payment tech.

The charitable reading is that this is deliberate: compress the payments take rate to defend share and pre-empt regulators, and monetise through higher-margin marketplace advertising instead. The uncharitable one is that competition and regulation are permanently eroding pricing power in the oldest, most profitable engine.

I lean towards the first, because marketplace take rates are rising at the same time — which is what deliberate mix-shifting looks like. But I hold the view loosely. If payments EBITDA is still flat-to-shrinking a year from now, that is a thesis problem, not noise.

5Valuation

At roughly $90 per ADS: about seven times earnings, around 3x EV/EBITDA, a dividend yielding 8–9% at the new rate, and a return on equity near 49%. The balance sheet holds net cash, debt-to-equity is around 0.12, and a $600M five-year bond was placed in April 2026 at 5.9%.

For context, and approximately: Nubank trades near 15x forward earnings, MercadoLibre near 40x, Sea Limited above 50x. Kaspi’s return on equity is higher than all three, and it is the only one paying a dividend. It trades at roughly half the cheapest of them.

6Insider signals — both directions

Buying: CEO and co-founder Mikheil Lomtadze — already the company’s largest individual shareholder, with essentially his entire fortune in the stock — purchased additional shares with personal money in April 2026, alongside senior management, at prices near today’s. He has held through every drawdown in the company’s history, including the January 2022 unrest.

Selling: co-founder and Chairman Vyacheslav Kim has been a net seller — roughly $23M over the past twelve months per public filings. Two founders, opposite directions. I don’t think it kills the thesis, since the CEO runs the business day-to-day and is accumulating. But anyone showing you only the buying half is selling you something.

7Türkiye: the free option

Kaspi owns about 86.7% of Hepsiburada, a real, operating Turkish e-commerce platform that already accounts for roughly half of group e-Commerce GMV. It also owns Hepsi Bank (formerly Rabobank A.Ş.), acquired in July 2026 — which, when bought, had no meaningful retail operations. It is a licence, not a business. A first shopping loan is now being piloted inside Hepsiburada, with the broader fintech rollout planned from 2027.

Why it could be very large: 85 million people against Kazakhstan’s ~20 million, and Kazakhstan is saturated. Turkish shoppers buy far less frequently than Kazakh consumers do, and closing even part of that engagement gap is the entire opportunity. Commerce platform plus banking licence is precisely the combination that worked at home, and the hard regulatory part is already done.

Why it could disappoint: Turkish inflation is running above 30% and the lira keeps setting record lows, so revenue can grow strongly in lira and still shrink in dollars. Policy rates near 37% make consumer lending — the profitable part of the playbook — very hard until rates fall. Trendyol, backed by Alibaba, is an entrenched competitor of a calibre Kaspi has never had to fight at home. And roughly $1.5B of capital is committed with the payoff years away.

8The honest bear case

If I am wrong, this is most likely why:

9What would prove me right or wrong

10How I am approaching it

Base case: 2–3x over three to five years, driven by earnings resuming growth as funding costs fall, plus multiple normalisation from ~7x towards 10–15x, with 8–9% collected throughout. This requires no heroics — only that the profit squeeze proves temporary.

The moonshot: Türkiye works, engagement transfers, and this becomes a materially larger company. I assign low odds and a long timeline — think a decade, not a year. If a position needs the moonshot to make sense, the position is wrong.

Sizing: this is an emerging-market position, not a core holding. Kazakhstan concentration, currency and political risk are real and unhedgeable. Size it so a bad Türkiye year or a tenge devaluation cannot force you out at the bottom — the entire thesis depends on being able to wait, and forced sellers never get paid.