heading · body

Earnings · LTF · NBFC (retail lending)

L&T Finance — a retail NBFC betting its future on an AI underwriting engine

L&T Finance Limited

period Q2 FY26 → Q4 FY26 (+ Nov-2025 Investor Day) added 2026-06-09 score 7/10
earnings-call nbfc retail-lending LTF india

The Pulse

L&T Finance is the Larsen & Toubro group’s retail NBFC, and it has just completed a remarkable makeover: from a lumpy, loss-prone wholesale/infrastructure lender into a 98%-retail franchise (rural microfinance, two-wheelers, tractors, personal loans, home loans, and now gold) with 2.8 crore customers. FY26 was a two-front year — clawing out of a microfinance collection crisis triggered by Karnataka’s February-2025 debt-relief ordinance, while proving that its home-built AI underwriting engine, “Project Cyclops,” structurally lowers credit losses. Both went well: microfinance collections are back to pre-crisis 99.8%, profit hit a record ₹3,003 crore, and the Cyclops-underwritten two-wheeler book shows bad loans of 2.8% versus an industry 7.1%. But the one number that matters most — return on equity — is still stuck at ~11%, and management missed its flagship 2.8-3% return-on-assets target, pushing it a year to FY27. At ~22x earnings and ~2.35x book, the stock is priced for the AI bet to pay off and the returns to finally catch up.

The Business

The model is a granular retail spread business: borrow cheaply (a AAA rating from all four domestic agencies, plus a sovereign-level international rating, thanks to L&T’s ~66% ownership), and lend across a wide retail product set at a healthy ~24% financing margin. The marquee products are rural group/microfinance loans (~25% of the book and the highest-yielding), two-wheeler finance, farm equipment, fast-growing digital personal loans (sourced through partnerships with Cred, Amazon Pay, PhonePe, Google Pay), home loans, SME, and a brand-new gold-finance arm (acquired June 2025, scaled from 130 to 330 branches in nine months). The retail book grew 26% to ₹1.2 lakh crore.

What it claims as its moat — and where it genuinely differs from peers — is technology, specifically Project Cyclops. This is an in-house AI credit-underwriting engine (a million lines of code, 55+ algorithms, decisions in under three seconds, zero downtime in 15 months) now live across two-wheelers, farm, SME and personal loans, with home loans and microfinance to follow. The proof points are striking: the two-wheeler bad-loan rate at 10 months on the Cyclops-underwritten book is 2.8% against a 7.1% industry average; festive-month loan bounces ran 7% versus an industry 20-22%. If real and durable, that’s a genuine cost-of-credit edge. Surrounding it are a portfolio early-warning engine (“Nostradamus”), an SME co-pilot (“Helios”), and a 2.2-crore-download app. The structural caveat: this is still an unsecured-heavy lender (~44% of the book), exposed to the rural/microfinance cycle that just bit it — which is exactly why management is now pushing into secured products (gold, micro-loans-against-property, home) to reach a 60:40 secured mix.

How Management Thinks

Under MD & CEO Sudipta Roy, the register is “risk-first, tech-first,” delivered with unusual candor and a fondness for quoting Drucker and Lenin. The candor is real: management openly admitted missing the 2.8% return-on-assets goal and blamed the Karnataka microfinance shock for delaying the whole industry’s recovery six months; conceded Cyclops “had a couple of bugs in the original stages”; and acknowledged a “back book underwritten by legacy algorithms” still drags. The CFO corrected an analyst’s wrong write-off number to his face. That willingness to own misses is a good sign.

The strategic discipline shows in the microfinance handling. Roy says he sensed the crisis coming in late 2023 and pre-emptively cut repeat loans, raised income thresholds, and added ~1,000 collectors before it hit — and the company adopted the regulator’s eventual guardrails (loan caps, three-lender limits) back in 2021-22, ahead of the rules. It explicitly refuses to chase microfinance growth (“15% is a safe speed; we’re not chasing 25-30%”). On provisioning it is deliberately conservative: it built a ₹975 crore macro-prudential buffer, used it sparingly, and in the FY26 annual model refresh raised its baseline Stage-1 coverage to a permanent 0.80% (double the regulatory minimum) — building cushion in good times. As Roy put it, “the microfinance crisis is a very good example of how that helped.”

Capital allocation is steady: ~66% L&T ownership and no equity dilution, a consistent ~23-27% dividend payout, the holding-company structure collapsed into a single listed lending entity, opportunistic deployment (the gold acquisition “cut time-to-scale by 24 months”), and a refusal to enter the home-loan rate war (pivoting to higher-yield loans-against-property and gold instead). A new payments platform is launching in FY27 — a fee-and-data play to watch, though also a step into a competitive new arena.

Where It’s Going

The trajectory is “recovery done, now deliver the returns.” Management closed its “Lakshya 2026” plan (hitting its retailisation, growth and asset-quality goals but missing the return target) and launched Lakshya 2031: 20%+ book growth, credit cost at or below 2%, return on assets of 3.0-3.2%, and return on equity of 16-18%. The near-term promise is reaching 2.8% return on assets by exit-FY27 and credit cost of 2-2.2%, both riding on Cyclops-underwritten loans “seasoning” through FY28 to prove the lower-loss thesis at scale. Tailwinds are real — the GST 2.0 cut and a strong monsoon drove record festive demand, gold is scaling fast, and falling rates are cutting funding costs.

The genuine tensions: the ~11% return on equity has to climb to the mid-teens for the valuation to make sense, and much of that uplift depends on (a) Cyclops genuinely lowering credit costs as the book ages and (b) the slow release of ~₹7,000 crore of capital still stuck in legacy security receipts and wholesale loans (a 2-4 year, court-dependent drag). The microfinance book remains the recurring event-risk, and the whole “AI moat” thesis, however impressive in early data, is unproven across a full credit cycle.

The Four Checks

  1. Quality & moat (gate). An improving business with a contested moat. The conventional edges — L&T brand, AAA funding, broad retail distribution — are real but ordinary. The distinctive claim is Project Cyclops, and the early data (2.8% vs 7.1% industry two-wheeler delinquency) genuinely suggests a credit-underwriting edge that could lower losses structurally. But it’s an unsecured-heavy book exposed to the microfinance cycle, and the AI advantage hasn’t survived a full downturn yet. Verdict: a credible emerging moat, not yet a proven one — which makes the remaining checks carry real weight.

  2. Returns on incremental capital & runway. The open question, and currently the weak spot. Return on equity is ~11% and ROCE ~8.4% — mediocre for an NBFC, and the recovery from the FY23 wholesale-cleanup trough has stalled at ~11% for three years. The runway is real (retail credit, gold build-out, cross-sell into 2.8 crore customers), and management’s own targets (3% ROA, 16-18% ROE by FY31) would transform the picture — but that’s a forward promise resting on the Cyclops thesis and the legacy-capital release. Today, incremental capital earns only adequately.

  3. Capital allocation for the stage. Rational and disciplined. No dilution, a consistent dividend, conservative provisioning that demonstrably cushioned the microfinance shock, a sensible structural simplification, and an opportunistic gold acquisition that bought scale cheaply. The refusal to chase the home-loan rate war or headline growth is sound. The one thing to watch is the new payments-platform foray — strategically logical for data and fees, but a fresh competitive investment with uncertain returns.

  4. Price. Demanding relative to delivered returns. At ~22x earnings and ~2.35x book against an ~11% return on equity, the market is paying a quality-NBFC multiple for an average-NBFC return — the gap is the Cyclops-and-recovery bet. If the FY27 return targets land and the AI edge proves out, the valuation is justified by the trajectory; if returns stay stuck near 11% (as they have for three years), it looks expensive. You’re paying up for the promise, with limited margin of safety if execution slips.

Sources

  • Concall transcripts read: Q2 FY26 (16 Oct 2025), Investor Digital Day (6 Nov 2025), Q3 FY26 (19 Jan 2026), Q4 FY26 / full-year (27 Apr 2026).
  • Annual reports read: FY24, FY25, FY26 (high-signal sections).
  • Financial snapshot: screener.in, fetched 2026-06-09 (FY26 revenue ₹17,914 Cr, PAT ₹2,983 Cr, ROE 11.2%, P/E 21.9x, P/B 2.35x, dividend yield 1.06%).
  • Research dumps (not published): vault/Sources/Earnings/L&T Finance Ltd/.
  • Notes: an initial fetch for ticker “LTF” mis-resolved to “L T Foods Ltd” (the Daawat rice company) and was discarded; this digest uses the correctly re-fetched L&T Finance data. Snapshot returned blank GNPA/NNPA (consolidated gross Stage-3 ~2.88%, net ~0.96% per the concalls; consolidated credit cost ~2.6%). The 2.8-3% ROA target was missed in FY26 (~2.4%) and pushed to exit-FY27.