This is investment research, not personal financial advice.

Judo Capital Holdings Limited (ASX: JDO) opened 24% lower on 25 June 2026 and fell as much as 46%, to A$0.83, inside the first hour of trade, after the small-business lender told the market before the open that three deteriorated loans had forced it to lift its FY2026 cost of risk to A$116-122 million and cut full-year profit guidance (Judo June 2026; Market Index 2026). By late morning the shares were changing hands near A$0.92 — down about 40% from the prior close of A$1.53, a market capitalisation of about A$1,031 million, and the lowest multiple of book value since the company listed in 2021. A bank that had spent four years selling itself as a clean, fast-growing, high-margin compounder erased roughly A$680 million of equity value in a session on a provision worth a fraction of that.

That gap between the size of the cash hit and the size of the repricing is the story. The question this deep-dive sets out to answer is the one the tape posed in a single morning: is the market's reaction correct — is the June update three idiosyncratic borrowers that a growing book can absorb, or the first sign of broader SME credit migration that resets what Judo can earn through the cycle?

What Judo told the market on 25 June

The announcement was titled, flatly, an update on asset quality and trading performance. Its substance was a downgrade. FY2026 profit before tax was guided to A$163-169 million, down from the A$180-190 million the bank had reaffirmed only four months earlier at its half-year (Judo June 2026; Judo 1H2026). FY2027 PBT was framed at A$210-220 million, roughly 16% below the A$255 million the market had pencilled in at the midpoint (Market Index 2026).

The driver was specific and named as such: increased specific provisions across three exposures "in different sectors of the economy," surfaced by a customer-by-customer review of the book undertaken earlier in the year. FY2026 cost of risk was reset to A$116-122 million — about A$20 million above the prior assumption, which is almost exactly the size of the PBT cut. The bank also flagged that 90+ days past due and impaired loans would reach about 3% of gross loans by 30 June, a step up from the 2.43% it reported at June 2025 (Judo June 2026; Judo 2025).

Crucially, almost everything else in the update pointed the other way. Second-half FY2026 net interest margin was running above 3.2%, ahead of the ~3.15% the bank had guided; front-book lending margins were stable at about 4.2% over BBSW; new term-deposit costs had fallen to 76bps over the bank bill rate; and the cost-to-income ratio had dropped below 48.5% (Judo June 2026). The management overlay was left unchanged at 94bps of gross loans — the three provisions are specific names, not a drawdown of the macro buffer. This was not a margin collapse or a funding scare. It was a credit-quality and credibility event.

Why a A$20 million provision erased A$680 million of value

A A$20 million pre-tax provision is worth roughly A$14 million after tax — about 1.2 cents per share against a book that was, on the same announcement, still guided to grow FY2026 PBT by around a third year-on-year. On the one-year cash arithmetic alone, a 40% fall is wildly out of proportion; the market was plainly not capitalising the provision.

What it was repricing is the durability of Judo's return on equity. For four years the equity story rested on a simple bridge: a high net interest margin (close to 3%, far above the majors), a cost-to-income ratio falling toward the low-40s as the platform scaled, and — the load-bearing assumption — a "normalised," contained cost of risk that let most of that pre-provision spread drop to the bottom line. Management's stated at-scale target is a return on equity in the low-to-mid teens (Judo 2025). The June update did not touch the margin or the cost line. It went straight at the credit assumption, the one input the bull case could least afford to lose. Reset the through-cycle loss rate and you reset the at-scale ROE; reset the ROE and, for a bank, you reset the multiple of book the shares deserve. That is the mechanism by which a small provision and a single sentence about three borrowers can move A$680 million: it is a change in the assumed slope of the entire earnings model, not a change in this year's earnings.

The bank underneath the downgrade

Judo Bank is Australia's only listed bank built solely around small and medium enterprise lending, run on a relationship-banker model that deliberately prices for risk rather than competing on rate with the majors. It funds itself through a fast-growing direct term-deposit franchise, wholesale markets and, increasingly, capital-relief securitisation. It listed in November 2021 and reached its first full-year statutory profit in FY2022. The compounding engine is straightforward to describe and harder to prove: originate SME loans at a wide margin, fund them ever more cheaply as the deposit brand matures, keep the cost base roughly flat as the book scales, and let operating leverage plus a disciplined loss rate lift ROE from a low base toward the teens.

The multi-year record shows the scaling clearly — and the credit tension just as clearly.

Metric FY2022 FY2023 FY2024 FY2025 FY2026e
Statutory NPAT (A$m) 15.6 73.4 69.9 86.4 ~115 (est.)
Gross loans & advances (A$bn) 6.1 8.9 10.7 12.5 ~14.6 (est.)
Net interest margin (%) 2.08 3.29 2.94 2.93 ~3.05 (est.)
Cost of risk / avg GLA (%) 0.52 0.74 0.72 0.66 ~0.87 (est.)
90+ DPD & impaired / GLA (%) n/d 1.09 2.31 2.43 ~3.0 (est.)
CET1 ratio (%) n/d 16.7 14.7 13.1 ~12.4 (est.)
Return on equity (%) 1.2 5.1 4.5 5.3 ~6.8 (est.)
Tangible book value/share (A$) ~1.27 ~1.30 ~1.37 ~1.44 ~1.54 (est.)

Sources: FY2022-FY2025 figures from Judo's annual reports and results releases (Judo 2024; Judo 2025; Judo 2023); FY2026e is an author estimate built from the 25 June guidance and is labelled throughout. Tangible book value per share is an author calculation: reported shareholders' equity less intangible assets, divided by shares on issue. The reported NIM in FY2023 (3.29%) reflects the tail of cheap Term Funding Facility funding; the step down to ~2.9% in FY2024-FY2025 is the normalisation as that funding rolled off, not a franchise failure.

The loan book has more than doubled in three years and statutory profit has compounded off a tiny FY2022 base. But ROE has never cleared 5.3%, because the bank remains over-equitied relative to its mature state and because the loss rate has run hot for a scaling lender. The table makes the bear's case before a single word of argument: 90+ DPD and impaired loans have risen every year since FY2023, from 1.09% to 2.31% to 2.43% and now to about 3%. The June "three exposures" sit on top of a three-year uptrend, not a flat line.

High margin, high credit cost: the trade-off the model is built on

The right return metric for Judo is not ROIC — for a bank, capital is the raw material, and the discipline is return on equity measured against the cost of equity. The owner-economics question is distributable earnings: profit left after funding the capital the growing book consumes. On that test Judo has, to date, distributed nothing — it has retained all earnings and still drawn its CET1 ratio down from 16.7% to 13.1% to fund growth (Judo 2023; Judo 2025). That is normal for a sub-scale compounder, but it means reported profit overstates owner cash: almost every dollar earned has been reinvested in regulatory capital behind new loans.

The model's defining feature is the margin-for-risk trade. Judo's reported NIM near 3% is double the majors', and its front book runs at ~4.2% over BBSW. The price of that spread is a structurally higher loss rate. Judo's cost of risk has averaged roughly 0.7% of gross loans across FY2023-FY2025 and is now guided near 0.87% for FY2026. The benchmark is stark: National Australia Bank, the dominant business lender, ran a FY2025 credit impairment charge of about A$833 million on a loan book many multiples larger — on the order of 0.1% of loans — for a cash ROE of 11.4% (NAB 2025). Judo earns five to eight times the major's credit loss rate by design, and the entire equity thesis depends on the pre-provision margin more than paying for it. The June update is unsettling precisely because it lands on the variable the model is least able to absorb a surprise in.

So the analytical core is a single spread: at-scale ROE equals the pre-provision return on equity (wide and improving) minus the through-cycle credit charge on equity (the contested input). Management's low-to-mid-teens target embeds a normalised loss rate closer to 0.5%. The realised numbers have not been near 0.5% in any recent year.

Three loans, or a vintage turning? The asset-quality read

This is the crux, and it is genuinely two-sided. The contained reading has real support. The three provisions are specific, named to different sectors, and surfaced by a deliberate file-by-file review rather than by a macro deterioration. The system around Judo is not flashing red: the RBA's March 2026 Financial Stability Review found business loan arrears still low, supported by a firm labour market, even as it warned that domestic credit growth had accelerated and that lending standards needed to stay prudent (RBA FSR 2026). NAB's FY2025 result showed only a 7bp rise in its business "default but not impaired" ratio — a tick, not a wave (NAB 2025). Judo's own book is substantially secured against business and property collateral, so a 90+ DPD and impaired ratio of 3% materially overstates ultimate loss; the 94bp management overlay sits on top of specific provisions, intact. On this read, the selloff capitalises a timing-and-disclosure event as though it were demand destruction.

The migration reading is equally legitimate, and the table is its evidence. A lender that grew gross loans from A$6.1 billion to A$12.5 billion in three years has a book dominated by FY2022-FY2024 vintages now reaching the age at which SME loans typically sour. 90+ DPD and impaired loans rising for three consecutive years — and a "customer-by-customer review" that, by management's own account, was needed to find the problems — is exactly what early-stage vintage seasoning looks like before it is named as such. If the through-cycle loss rate is really 0.8-0.9% rather than 0.5%, the at-scale ROE target is unreachable on the current margin, and the three exposures are not the exception but the sample.

The honest position is that one update cannot settle it. What it does is shift the burden of proof: until June, the market gave Judo the benefit of the 0.5% assumption; now it is demanding the bank prove it. The next two reporting periods — the FY2026 result in August 2026 and 1H2027 in February 2027 — are where the 90+ DPD trend either flattens near 3% or keeps climbing, and that is the single most important number in the entire thesis.

Capital and funding: the part of the result that held

If the credit line was the bad news, the capital and funding lines were the reassurance the price ignored. CET1 stood at 12.6% at the half and about 13.2% pro forma after the A$750 million capital-relief securitisation priced on 29 May 2026 at 171bps over BBSW — a 102bp improvement on the bank's inaugural 2023 deal at 273bps (Judo 1H2026; Judo May 2026). The transaction frees regulatory capital without moving the loans off the income statement, and management estimates a 25-30bp benefit to FY2027 ROE. On the back of it, Judo cut its CET1 management range to 11-12% and explicitly raised the prospect of capital management initiatives — a buyback — if the share price stayed soft (Judo June 2026).

That is an unusual posture for a bank in the middle of a credit scare: not raising capital, but signalling it has surplus capital and cheaper funding. Deposits reached A$10.9 billion, up 21% year-on-year, with the cost of new term deposits falling as the RBA eased through the cycle (Judo 1H2026; RBA 2026). The survivability question for Judo is therefore not solvency or liquidity in any near-term sense; it is whether the credit charge quietly compounds to the point where it consumes the operating leverage the bank is banking on. A bank with 12%+ CET1, a 94bp overlay, falling funding costs and securitisation optionality is not fragile. It is, however, now being asked to demonstrate that its loss rate is cyclical rather than structural.

What the A$0.92 price now assumes

A bank is valued on the spread between its sustainable ROE and its cost of equity, not on a DCF of accounting cash flow. The cleanest frame is justified price-to-book: P/B ≈ (ROE − g) / (k − g). With tangible book value per share around A$1.50 (an author calculation from reported equity), a cost of equity of roughly 12.5% for a single-name, higher-beta SME bank, and long-run growth of ~4.5%, the price does the talking. At A$0.92 the shares trade at about 0.6x tangible book. Solving the relationship backwards, a 0.6x multiple implies a sustainable ROE of only about 9.4% — barely above the cost of equity, and a full three to four points below management's low-to-mid-teens target.

In other words, the market has not merely docked a year of earnings; it has repriced Judo from "a compounder approaching a teens ROE" to "a lender that will struggle to clear its cost of capital." Building each scenario from the at-scale ROE and a matched justified multiple, rather than around today's price, gives the following per-share ranges (all on tangible book of ~A$1.50, growing):

  • Severe downside (A$0.40-0.60): the three exposures are the leading edge of a seasoning wave, the credit cycle turns, through-cycle ROE settles near or below the ~6% cost-of-equity floor, and the multiple compresses to 0.3-0.4x.
  • Bear (A$0.65-0.85): cost of risk is structurally ~0.85%, NIM drifts toward 2.9%, and at-scale ROE stalls near 8.5% — a ~0.5x multiple.
  • Base (A$1.05-1.30): the exposures are largely idiosyncratic, cost of risk normalises to ~0.65%, at-scale ROE reaches ~10.5%, supporting ~0.75x book.
  • Bull (A$1.70-2.05): the reset proves a one-off, operating leverage and securitisation accretion carry at-scale ROE to ~13%, and the shares re-rate to ~1.1x a growing book.

The post-event price of A$0.92 sits between the bear and base ranges — closer to bear. The two variables that move everything are the through-cycle cost of risk and the cost of equity the market applies to a bank whose credit narrative has just been broken; margin and growth are second order from here.

The verdict, and the dates that settle it

So, is the reaction right? On the one-year cash, no: a roughly A$20 million provision against a book still guided to grow PBT by about a third does not justify a 40% fall, and the parts of the franchise that drive value — NIM above 3.2%, falling funding costs, 12%+ CET1, buyback optionality — held or improved in the same release. Read literally, the selloff capitalises a deferral-and-disclosure event as if it were structural demand destruction.

On the durable economics, the reaction is more defensible than it first looks. The market was pricing Judo as a clean-credit compounder, and the June update is the third consecutive year of rising arrears plus an admission that problems had to be hunted for file by file. Repricing the sustainable ROE toward 9-10% is not obviously wrong; it is a refusal to keep extending the 0.5% loss-rate assumption on trust. The most accurate observation is that the tape has over-reacted to FY2026 and is now pricing close to the bear case on the through-cycle question — a question the evidence in hand cannot yet resolve in either direction. The price implies the credit reset is permanent; the franchise data implies it need not be; the 90+ DPD trend is the tiebreaker, and it is still pointing the wrong way.

The crux resolves on a clear calendar. The FY2026 result in August 2026 carries the actual realised cost of risk, the June-30 arrears print, and any further single-name provisions. The 1H2027 result in February 2027 shows whether 90+ DPD flattens near 3% or keeps climbing, and whether the 2H26 margin strength persists. A buyback, if announced, would signal management's own confidence that the capital is genuinely surplus. Those are the disclosures that turn this from an argument into a fact.

Confidence, gaps and source notes

Confidence is high on the event mechanics and the FY2022-FY2025 financial history, which come from Judo's own results releases and annual reports (Judo 2024; Judo 2025; Judo 2023) and the 25 June, 29 May and 17 February 2026 announcements (Judo June 2026; Judo May 2026; Judo 1H2026). Confidence is medium on the precise intraday share-price path and the pre-announcement consensus, which are drawn from market commentary (Market Index 2026), and on the prior FY2026 cost-of-risk assumption, which is inferred from the size of the guidance cut rather than separately disclosed.

These source notes flag the principal estimates. The FY2026e column is an author estimate built from the 25 June guidance (PBT A$163-169 million, cost of risk A$116-122 million, 90+ DPD ~3%) with NPAT derived at an assumed ~31% effective tax rate; it is not an audited row. Tangible book value per share is an author calculation from reported shareholders' equity less intangibles divided by shares on issue. The justified P/B, the implied ~9.4% sustainable ROE, distributable earnings and the scenario ranges are author calculations, with the cost-of-equity (12.5%) and growth (4.5%) assumptions stated so a reader can substitute their own. The peer comparison to NAB (NAB 2025) is context for the SME credit-cost premium, not a like-for-like valuation. Macro framing relies on the RBA's March 2026 Financial Stability Review and cash-rate record (RBA FSR 2026; RBA 2026) and APRA's ADI statistics (APRA 2026); these describe a system in which business arrears remain low, which is itself evidence in the idiosyncratic-versus-systemic argument. Verification is partial: figures are sourced to filings read for this article but not independently re-audited, and the centrepiece forward numbers are explicitly estimates.

References