Open in interactive viewer → charts, metric popovers & call review

Kaspi Wants 3,000 Corner Stores — and a Rate Cut It Has Been Promising for Three Quarters

A currency-driven GMV guide-down masks double-digit volume growth, while the real strategic news is a reverse-franchise grocery model bolted onto Kaspi's dark-store network
KSPI · Earnings Call · 2026-11-09

The store is the new warehouse

For four straight quarters, dark store has been Kaspi's capex story: expensive, one-city-at-a-time boxes built to feed e-grocery delivery. This quarter the keyword slipped to No. 2 in the company's own lexicon, but the concept got a second life. The pilot of partner stores — 42 converted independent shops in Almaty — is the first genuinely new business line the company has launched since Türkiye, and management did not hide its enthusiasm.

So basically, if you just think about it, during the day, we are delivering our items and our goods to consumers through our e-grocery. And then during the night, every morning, we sort of deliver the SKUs to the partner stores, which have been converted to our standard.

Mikheil Lomtadze, CEO · 2026-11-09
That is the whole thesis in one sentence: an asset that was previously idle from midnight to dawn becomes a wholesale distribution hub for a fragmented retail channel. Management claims the dark stores already turn inventory roughly twice a month, and the night shift multiplies that throughput with essentially no incremental real estate. The economics look deliberately thin-but-scalable: partners keep ~7% of turnover, Kaspi carries the refit (~$225,000 of capex per store, funded from cash flow), and the pilot ran at a 15% EBITDA margin with an ~18-month payback. The target is 300 stores across nine cities next year, against a grocery market the CEO sizes at roughly $9 billion and 10,000-plus potential neighborhood stores. What matters is that this is a *distribution* business dressed as retail. Each converted convenience store is a node that can later support fast delivery — the company says 30-minute delivery testing starts this month. And it explains why dark store stayed on the call even as E Grocery receded: the warehouse was always the moat, the fridge was never the point.

A guidance cut that is not a demand cut

The headline is ugly: full-year GMV growth guided to ~12% in constant currency from ~20%. The explanation is almost the opposite of a demand warning. “Full year GMV growth now expected in constant currency to be around 12% versus previous expectation of 20%, and that is driven entirely by the sort of lower ticket size resulting from sharp KZT appreciation.” — David Ferguson, Head of Investor Relations · 2026-11-09 Items sold rose 24% year-on-year in Kazakhstan e-commerce while the price of those items fell 8%. In fragmented, import-heavy categories — sports, travel, electronics, autos — prices dropped 15-17% as a stronger tenge let hundreds of thousands of merchants pass on landed-cost savings. That is a deflation story, not an engagement story, and the engagement numbers back it: purchases per consumer up 27% to 16.3, take rate up 210bps to 16.3% on the e-commerce marketplace. Two details deserve a raised eyebrow. First, the smartphone vertical that dominated three prior calls as a supply disruption has quietly vanished from the narrative — the category is now growing 20% year-on-year, and Apple's price increases are doing the work. Second, when Gabor Kemeny probed a sequential marketplace take-rate decline (12.1% in Q1 to 11.9% in Q3), the answer was seasonality and promotional timing. In May, the same IR voice was adamant that take-rate dynamics had “nothing to do with [seasonality]” — David Ferguson, Head of Investor Relations · 2026-05-11. Both can be true; still, it is the kind of small inconsistency worth tracking into next year, especially with the e-commerce platform described as a "mature business." Note also how thin the Q&A was — two questioners, one line drop-out — for a $17 billion company announcing its biggest new venture.

The funding-cost trade, deferred once more

Cost of funding hit 14.9%, up 150bps, with deposits growing 21% — faster than the loan book — because Kazakh savers spent the summer locking in peak rates. That is a classic early-cycle pinch: customers reprice up instantly, the bank reprices down slowly. Kaspi trimmed its 3-month deposit (roughly a third of savings) by 100bps in August and another 100bps in October, about 2% total. “So total around 2% and the full effect on the savings will come in 2027, and there will be some effect in the 4Q.” — Mikheil Lomtadze, CEO · 2026-11-09 Compare that to August, when shareholders were told that “any rate cuts this year just should give you increased confidence about earnings growth next year” — Mikheil Lomtadze, CEO and Co-Founder · 2026-08-10. The promised inflection has been nudged forward a full twelve months. EBITDA was "broadly flat" and net income only just crossed back into positive territory "for the first time in a long time." The company is not wrong — with liquidity revenue at ~24% of payments revenue and Funding costs still climbing, the sensitivity is mechanical — but the pattern of patience now runs three calls deep, and investments in Türkiye being "significantly" lower in 2027 is the same 2027 promise made in March. The offset is shareholder returns. The board approved a KZT 1,000 dividend, and David Ferguson left the door open: “if you see earnings growth, you see upside risk to cash returns, and that can take the form of either dividends and/or buybacks, but that's a conversation for next year” — David Ferguson, Head of Investor Relations · 2026-11-09. Note the backdrop — strong core deposits are now a global discussion topic, with deposit competition and re-pricing showing up across banks in this quarter's tape.

Kasper gets its first commerce numbers

Last quarter the AI assistant was a speed story ("2x faster to find the product"). This quarter it is a money story: 1.4 million users, conversion 1.5x text search, average GMV per order 2x text search. That is the first credible evidence that a Kasper-style assistant can be monetised inside a transactional app rather than run as a cost centre — and it sits squarely on a global wave of Shopping Agent experimentation that most retailers are still describing aspirationally. The visual search build and a next-generation model are slated for later this year. So where does that leave the story? A flat-EBITDA quarter with a cut headline number is not what the bull case advertised. But the underlying mechanics — volume up 16-24%, take rate expanding, a grocery format that recycles an existing asset base, and a rate cycle that must eventually pay out — are all pointed the same way. The question investors have to price is no longer whether the operating leverage arrives, but when, and whether management buys back stock with it before the tenge does the work for them.