The Great Banking Divide

The central contradiction

On 19 October Indian private banks weakened even as PSU banks rallied sharply following Q2 FY27 results. That provides a timely contradiction, but market divergence alone does not prove that public-sector banking has become a superior long-term franchise. Private banks have historically commanded better deposit franchises and risk processes; PSU bank earnings now benefit from lower credit losses and healthier capital. Relative valuations are pricing different durations of improvement.

The sector story must reconstruct four components: risk-adjusted net interest income, sustainable non-interest fees, normalised credit cost and capital consumption. Headline loan growth is distorted by large foreign currency non-resident funding flows; a bank with 30% reported loan growth may only have ≈24% under a disclosed comparable adjustment.

The critical investor question is which group can compound book value per share while distributing capital at an attractive ROE after the provision tailwind fades and FCNR-supported balance-sheet growth normalises.

Dated news & chronology

June 8 2026: the RBI opened a special US dollar / INR swap facility for eligible FCNR(B) deposits with specified 3-5-year tenor. The Ministry of Finance subsequently described very large dollar inflows. The RBI advanced closure of the deposit-mobilisation window to 31 August. Deposit mobilisation therefore reflects a policy-induced funding opportunity, not solely organic branch CASA franchise improvement.

16-17 October: PNB, HDFC Bank, ICICI Bank and Axis Bank reported quarterly earnings, joined by Union Bank on 19 October. On 19 October, Reuters recorded PSU bank sector outperformance (≈+2.5%) versus weakness in HDFC/Axis/ICICI and a broad Nifty decline. The earnings chronology precedes Monday's sector performance, but cross-market causality remains multi-factor.

Rupee volatility, global energy costs and changes in FX-derivatives reserve/hedging rules also affect marginal funding costs. Any blanket claim that private banks lost their deposit franchise should be rejected without average CASA and risk-adjusted cost evidence.

Industry value chain

Banks are levered maturity transformers. Value is created by gathering durable low-cost liabilities, investing them in risk-priced assets, generating fees and managing lifetime credit costs below risk-adjusted spreads. A high reported NIM may coexist with excessive customer-acquisition cost or elevated future credit loss.

PSU banks can benefit from operating leverage on branch infrastructure and decaying legacy NPAs. Private banks can benefit from stronger fee ecosystems, digital distribution and risk segmentation. But public ownership does not mechanically create weaker underwriting, just as private ownership does not guarantee superior risk adjustment.

The appropriate peer panel uses common definitions and periods: average earning assets, disclosed NIM and cost of deposits; operating cost/assets; credit-cost net of write-backs; slippage and recovery rates; risk-weighted assets and CET1; P/B, sustainable ROE and cost of equity. The present review cannot certify complete comparability of bank definitions.

The business machine

Axis disclosed Q2 period-end deposits +21% reported vs +17% after excluding the specified portion of FCNR(B) backed by international-branch advances; advances +23% vs +19% on the same disclosed measure. Its published NIM was 3.36%, versus 3.47% excluding the FCNR(B) impact, a difference of 11 basis points. These are issuer-defined alternate figures, not universally adjusted sector measures.

Axis reported ₹1,018 billion of FCNR(B) deposits, including ₹438 billion backed by international-branch advances and ₹449 billion backed by standby letters of credit. Those categories should not be summed into a claim of risk-free growth without determining economic borrower, currency basis and asset-liability maturity.

ICICI was reported to have headline loan growth around 30% versus approximately 24.1% excluding certain FCNR-backed international loans, and deposit growth 27.1% versus around 17% excluding diaspora deposits. Exact numerator, denominator and eligibility differ from Axis; treat cross-bank comparisons as indicative until both full IR decks and footnotes are reconciled.

Management & governance

Axis Q2 FY27 standalone PAT ₹7,003 cr (+38% YoY, -2% QoQ); NII ₹15,371 cr (+12% YoY, +5% QoQ); core operating profit ₹10,998 cr (+11% YoY); reported NIM 3.36%; standalone annualised ROE ≈13.3%; GNPA 1.16%. Loan growth is strong, but operating expenses rose 9% YoY and treasury/non-interest income introduced volatility.

HDFC Bank PAT ₹19,510 cr (+4.7% YoY), NII ₹34,619 cr (+9.7% YoY), loans +16.3%, deposits +18.8%, credit/deposit around 96%. Trading gains declined to ≈₹780 cr from ₹2,390 cr a year earlier. Its headline profit growth understates some recurring earnings momentum, but a normalisation must not casually reverse losses or provisions.

ICICI Q2 PAT ₹14,843 cr (+20.1% YoY); NII grew around 17.7%. PNB PAT ≈₹5,532 cr (+12.8%), NII ≈₹11,601 cr (+10.8%) and operating profit ≈₹7,532 cr (+4.2%), while provisions for NPAs fell strongly. Union Bank PAT ≈₹5,537 cr, estimated ROA 1.38% and ROE 17.21%; one source reported gross advances growth of 18.53% against deposit growth of 6.87%. PNB/Union numbers must be re-verified against full original statements.

Reported earnings bridge

Axis: report NII, fees, operating expenses and core PPOP, then remove treasury volatility; normalise credit cost to a five-year through-cycle range, adjust tax and compute sustainable PAT. A provision-driven PAT jump without equivalent NII or PPOP progress is lower-persistence than recurring margin/fee growth. The current quarter offers enough data to identify the question, not enough to resolve it fully.

PNB: reported PAT growth +12.8% and PPOP +4.2% implies leverage to provisions, recoveries or tax rather than proportionate core operations; evaluate slippage trends against gross and net recoveries. Union: a reported 30% PAT improvement and >17% ROE are valuable, but check whether reversal of credit provisions or non-operating income explained the delta.

HDFC: treasury gain YoY decline of ₹1,610 cr is material. At an illustrative 25% tax rate, the after-tax earnings difference would be ≈₹1,207.5 cr, all else equal. This is an analytical counterfactual only: changes in trading income may coincide with hedge and ALM effects and must not be added directly to reported PAT to claim normal earnings.

Cash conversion & capital

Distinguish period-end deposit growth from average deposits. For 3-5-year FCNR(B) inflows, determine actual rupee-equivalent funding cost after swap, the maturity repricing of underlying assets and whether hedges are eligible for discounted RBI treatment. Deposits can rise sharply while core Indian low-cost CASA growth remains slower.

PSU banks have more room for provision normalisation only until credit costs approach a sustainable floor. Indian Bank's Q2, for example, included a reported NSE share-listing gain; such treasury/capital-market items are not proof of superior lending spreads. Compare all banks excluding such exceptional and one-off gains, not just private-bank outliers.

Asset quality should be analysed through new slippages and write-offs versus recoveries, not just GNPA levels. A decline in GNPA following rapid growth can result mechanically from denominator expansion. The credit cycle should be stress-tested using borrower segment, unsecured retail share and corporate concentration.

Peer corroboration

For banks, price-to-book relative to sustainable ROE and cost of equity is more robust than current-year P/E. A stylised steady-state relation is justified P/B = (ROE - g)/(Ke - g), where g is sustainable equity growth below cost of equity. This formula is an analytical identity under strong assumptions, not a plug-in fair-value model.

The sensitivity to credit losses is large: on ₹10 lakh cr of average advances, 25 bp extra credit cost reduces pre-tax earnings by ₹2,500 cr; after 25% tax, approximately ₹1,875 cr. Such a move can eliminate a sizable portion of a one-quarter provisioning-driven PAT improvement.

If a bank earns 16% ROE and retains 70% of earnings, rough book-value growth is 11.2% before new issuance, OCI and regulatory deductions. At unchanged P/B and no capital redistribution, stock returns roughly follow book compounding plus dividends. An investor demanding 18% annualised requires either a higher sustainable ROE, a valuation rerating or faster book growth.

FY27–29 operating outlook

Do not forecast precise bank EPS without starting book equity, risk-weighted assets, loan mix, marginal funding cost and credit costs in a uniform data model. Instead model three archetypes with earnings drivers: PSU-credit-normalisation bear/base/bull; deposit-led private compounding; and FCNR-sensitive private funding mix.

Illustration: a bank with ₹1,00,000 cr book equity and a sustainable 12% / 15% / 18% ROE generates ₹12,000 / 15,000 / 18,000 cr annualised PAT before changes in book, tax and capital structure. At payout ratios of 20%, 30% and 40%, retained earnings differ substantially. A 100 bp ROE difference on this equity base represents ≈₹1,000 cr profits, before compounding.

A ranking across PNB, Union, SBI, HDFC, ICICI and Axis should emerge from a granular FY29 ROE and market-implied ROE bridge. We deliberately do not assign share-price targets without all live price/book inputs and corrected cross-bank accounting bases.

Expectations vs price

The 19 October divergence can reflect weaker-than-expected private-bank margins, technical profit-taking after relative performance, PSU earnings upside and low prior expectations. It may not represent a rational long-term reversal of franchise quality. The price driver must be broken into revisions of sustainable PAT, long-term ROE and valuation multiple, alongside sector fund flows.

The most publishable variant insight may be that the FCNR scheme inflated both deposits and advances, but did not create equivalent shareholder economics. Separately, certain PSU-bank earnings accelerated because impairment costs moved down. Neither should be treated as permanent without policy and credit-cycle analysis.

The incremental buyer for PSU shares is likely underwriting sustained mid/high-teen ROE at moderate P/B; the incremental buyer for private shares may require proof of fee/retail franchises and adjusted deposit growth. The difference between those investor requirements drives a potentially temporary valuation gap.

Bear / base / bull cases

Bear for PSU relative trade: credit slippages reaccelerate, asset-quality recoveries plateau and PPOP grows slowly. Bear for private banks: marginal funding stays high, FX-swap economics roll off and liability competition compresses NIM. Base: PSU provisions remain low but plateau, private growth adjusts as FCNR effects normalise. Bull: private franchise growth reasserts with better spreads; PSU operating leverage supports structurally higher sustainable ROE.

Q3 FY27 proof: normalised deposit growth, average funding cost and FX-deposit deployment spreads; bank-specific core PPOP. FY28 proof: slippage/credit-cost trend through policy normalisation; CET1/RWA capacity. FY29 proof: sustainable ROE compared with FY26 actual and price paid today.

Kill signals for a "PSU structurally superior" headline include narrowing risk-adjusted ROE spread versus private peers, divergence driven entirely by provisions or non-core income, deterioration in CASA/funding cost, or a hidden need for equity recapitalisation.

Risks, proof & kill switches

NOT CLEARED as a comparative financial master: fully reconciled FY26 and H1FY27 disclosures for every shortlisted bank, PNB/Union primary slides, ICICI FCNR reconciliation to disclosed note, common P/B market pricing, consistent CET1 and credit-cost definition and full management call transcripts remain incomplete.

Ask: What share of headline growth is FCNR-funded? What is all-in post-swap spread? How much treasury gain did the quarter generate? What is PPOP before recoveries? What is the three-year through-cycle slippage expectation? How does deposit beta change when FX facilities expire? What is the explicit capital impact of the proposed asset growth?

Publishing standard: tables must disclose exact denominator, average versus period-end, operating versus reported profit, one-off gains, and whether a number is management-adjusted. We should not declare a stronger banking model from one earnings weekend.

What would have to happen by FY29?

These are explicitly analytical scenarios, not management guidance or a prediction. Change the scenario to inspect the year-by-year implications. The result depends on the variables stated in the source dossier.

Evidence, provenance and publication gate

Every numerical claim should be tied to a dated, original source wherever accessible. Missing primary PDFs and unsupported supplier/customer economics are explicitly visible below rather than silently upgraded to facts.

Reported factManagement statementAnalytical inferenceForecast / sensitivity

Unresolved publication proof: Common bank-by-bank FY26–H1 FY27 balance-sheet bridge, average CASA, credit-cost normalisation and matched current tangible P/B. The HTML is a finished presentation of an evidence-qualified working dossier—not a replacement for missing due diligence.