All notes
Fintech

TBC Uzbekistan Q3: net profit down 61%, NPL at 11.8%, while payments grew 46.5%

TBC Uzbekistan's Q3 net profit fell to GEL 15.9 million and its NPL ratio reached 11.8%. What it means for digital lending, regulation and engineers building credit systems.

What happened

On 6 November, TBC Bank Group published unaudited IFRS results for Q3 and 9M 2026. At group level it was a record quarter: net profit of GEL 404 million, up 10% year on year (CFOtech). The Uzbekistan segment told a different story.

According to UzDaily, TBC Uzbekistan's Q3 net profit fell 61.2% to GEL 15.9 million (about $6.1 million), from GEL 41.1 million a year earlier. Net interest income dropped 14.5% to GEL 125.0 million, while operating expenses rose 18% to GEL 92.4 million. The cost-to-income ratio climbed from 41.5% to 53.8%, and return on equity fell from 23.3% to 7.8%. For the nine months, profit was GEL 57.3 million, down 39.7% from GEL 95.0 million.

Asset quality: the non-performing loan (NPL) ratio stood at 11.8%, versus 4.7% a year earlier (UzDaily). CFOtech reports the quarterly increase was 1.2 percentage points, which the company attributed to portfolio maturity. Cost of risk fell 1.0 pp in the quarter to 10.8%; for comparison, the group figure was 1.3% and Georgia's 0.6%. Per the investor presentation, the Q2 NPL jump was largely tied to extending the debt write-off period from 270 to 360 days (UzDaily).

Gross loans were GEL 2.37 billion (about $910 million): up 2.8% in the quarter but down 10.1% year on year. Excluding instant cash loans, quarterly growth was 13%. The mix: instant cash loans 53%, business loans 20%, POS and instalments 14%, cards and other 13%. The Tier 1 ratio was 21.4% (minimum 8.5%) and net interest margin 20.1% (UzDaily).

Payments kept growing: ecosystem payment volume of 43.9 trillion soums (+46.5%), 198 million transactions (+39%), 6.1 million monthly active users and 48% more active merchants (UzDaily). Salom cards passed 1.5 million, and a shared Novvot loyalty programme launched across TBC Uzbekistan, Payme and other products (CFOtech).

Group CEO Vakhtang Butskhrikidze said: "In Uzbekistan, lending is gradually recovering as we adapt to new regulatory requirements" (CFOtech). UzDaily reports he expects automatic debt collection rules to take effect in early 2027 (UzDaily).

Context

TBC Uzbekistan is part of the same ecosystem as Payme and one of the country's most visible digital banks, so its numbers work as a barometer for digital lending in the market as a whole.

The central tension in the report: payments are growing fast while lending shrinks. Same group, two markets, and a gap in cost of risk of roughly 18x (0.6% vs 10.8%). That is the real price of digital lending in a young market: you can approve a loan in seconds, but the bill for defaults arrives later.

The regulatory backdrop is tightening too. On 4 November the Central Bank held its policy rate at 14% and raised its year-end 2026 inflation forecast to about 7% (Gazeta.uz). Under Resolution No. 24/18, adopted in September, a warning about financial risks must take up at least 10% of a credit ad and at least 30% of the first page of a loan agreement (UzDaily).

One caveat: the year-on-year NPL comparison is not like-for-like. When the write-off period is extended to 360 days, bad debts stay on the balance sheet longer and the NPL ratio rises mechanically. Part of the move from 4.7% to 11.8% is methodological; how much, I could not determine from public sources.

What it means

For users. Instant loans will likely get harder to obtain, and contracts will carry more warnings. Less convenient, but it protects people from over-borrowing.

For businesses. TBC is shifting its book away from instant cash loans towards business lending, cards and other products (13% growth ex-instant loans). Business customers grew fivefold in a year, with 84% outside Tashkent (UzDaily). That means more credit and payment services for regional SMEs.

For engineers, the lessons are practical:

  • Don't hard-code credit decision rules. Regulation changes often. If limits, debt-burden ratios and product restrictions live in configuration and a versioned rules engine, adapting takes days, not weeks.
  • A metric's definition is code too. Swapping 270 for 360 looks like changing one constant, but it moves NPL, provisions and reporting. You need versioned definitions and the ability to recompute history under old and new methodology (data lineage).
  • Compliance reaches the UI and documents. A "30% of the first page" rule means changes to contract generators, mobile screens and ad templates. Requirements like this are worth checking with automated tests.
  • Collections. If automatic debt collection rules take effect in 2027, banks will need reliable, idempotent integrations for debiting accounts, with a full audit trail: the same amount must never be taken twice.
  • Risk models. A 10.8% cost of risk shows how sensitive scoring is to the cycle. Drift monitoring, challenger models and a fast kill switch are mandatory infrastructure.

My take

I think this report is more nuanced than "TBC is in trouble". The bank shows strong growth in payments and its ecosystem, and its capital buffer sits well above the minimum (21.4% vs 8.5%). The weak spot is the instant-loan model: it delivered rapid growth, but the cost of risk came in higher than expected.

It's to the company's credit that it is saying so openly and reshaping the book. It's also a useful signal for the market: digital lending is not just a speed contest. Behind a "one-tap loan" experience you need strong risk management, clear metrics, and an architecture that can adapt quickly when the regulator moves.

What I'll be watching: the final text of the automatic collection rules, the share of instant loans over the next quarters, and what NPL looks like without the methodology effect. And worth remembering: these figures are unaudited.

Sources: UzDaily — TBC Uzbekistan Q3, CFOtech — TBC Group results, Gazeta.uz — policy rate, UzDaily — credit warnings. This is personal opinion, not investment advice.

// comments

Loading comments...

    // next step

    Let's build something together.

    Message me on Telegram
    aslbek@asldev: ~
    ❯