Itaú's 24% ROE Is the Real Thing. Half Its First-Half Earnings Growth Is Not.

Results vs. Consensus

A note on units before the numbers. Itaú reports in Brazilian reais on a recurring managerial basis, which strips extraordinary items and reallocates certain lines from the statutory accounts. That is the basis the bank guides to, the basis its disclosed consensus panel is struck on, and the basis used throughout this report. Where a figure is in US dollars it refers to the New York ADR. The two are not interchangeable, and the scorecard below says so explicitly rather than blending them.

Q2 2026 Scorecard

The two consensus rows disagree because they are not measuring the same thing, and the disclosures do not permit a bridge. The real comparison is the first row: a thirteen-house panel of the largest domestic and global brokers covering the name, averaged, in reais, on the recurring managerial result. That panel was expecting R$12,466m and got R$12,407m. Independently, a separate analyst poll was reported at approximately R$12.5bn. Two different panels, 0.3% apart, both effectively on the number.

The dollar rows sit on a vendor-compiled ADR basis that mixes a different earnings definition, a different revenue definition and a translation convention. For scale: managerial operating revenues of R$47,985m convert to roughly US$9.3bn at the company's June reference rate and roughly US$9.5bn at the quarterly average rate, a band that straddles both the dollar consensus figures and the dollar "actual". Nor can currency explain the wider dollar gap, because the real appreciated 5.1% against the dollar year on year, which flatters dollar-reported results rather than depressing them. We report the 6.3% figure because the market saw it, and we decline to call it an earnings miss on Itaú's reporting basis, because it is not one.

Managerial Income Statement

Read the bottom four rows together, because that is where the quarter actually happened. Pre-tax income grew 5.0% year on year. The recurring result grew 7.8%. Minority interests were a headwind, rising 25.7%. The entire difference, and more, is the tax line, which fell 2.5% on a 5.0% larger pre-tax base.

First Half 2026 vs First Half 2025

Quality of Beat

• Revenue: Organic and broad, but decelerating and partly inorganic. Operating revenues rose 4.7% year on year with every major line positive. Two qualifications matter. First, the income statement has reflected the consolidation of Avenue since the first quarter of 2026, where the prior-year comparatives carry it only as an equity-method associate, and the contribution is not quantified anywhere in the executive summary. Some unknown part of the 4.7% is therefore an acquisition rather than growth. Second, the 2.5% sequential revenue gain was outpaced by a 3.3% sequential rise in non-interest expenses, so the operating jaws were negative in the quarter even though they were positive, by 1.6 points, year on year.
• Margins: The headline margin is flat; only the risk-adjusted version improved. Net interest margin with clients was 9.1% consolidated and 10.0% in Brazil, the third consecutive quarter at both levels and 30bps lower than a year ago on each. What improved was the risk-adjusted margin, up 10bps sequentially to 6.2% consolidated and 6.7% in Brazil. That is a real achievement and it is exactly how management frames the business, but it means the margin contribution to revenue growth is coming from volume and from credit cost restraint, not from spread. On the cost side the efficiency ratio was 37.4% consolidated and 35.5% in Brazil, both better than a year ago and both worse than the first quarter, which management attributed to first-quarter seasonality rather than to any loss of discipline.
• Earnings: Below the line, and by management's own design. Beyond the tax ratio, recurring earnings per share of R$1.13 grew 8.7% year on year against 7.8% growth in the absolute result, the difference being a 0.7% reduction in share count. Book value per share rose 4.6% year on year to R$18.92 while distributions fell 20.1%, at a payout against the recurring result of 30.5% versus 41.2% a year ago. The retention is visible in the capital ratio sequentially and the front-loaded 2025 distribution is visible in it annually.

Key KPIs

Three things stand out in that table. The physical network has come down by about a fifth in a year while the loan book grew nearly a tenth, which is the whole efficiency thesis expressed in two lines. Return on assets fell to 1.5% from 1.6% even as return on equity rose, which tells you leverage rose: each unit of equity now carries more assets after a year of capital return. And every capital ratio tells the same story: higher than last quarter, lower than last year.

Segment Performance

Itaú does not publish conventional business-unit revenue and margin tables. It publishes two different cuts, and both are more informative than a standard segment table would be: the credit portfolio by product, and a capital-based "business model" view that allocates average regulatory capital and a return to each activity. We work through both.

Credit Portfolio by Product

The product lines sum exactly to the individuals total, so the mix shift in the right-hand column is clean. Secured lending (mortgage, payroll, vehicle) rose to 55.1% of the individuals book from 53.9% a year ago; unsecured lending (cards and personal) fell to 44.9% from 46.1%. That 1.2-point shift is the de-risking the bank keeps describing, made visible.

Business Model View: First Half 2026 vs First Half 2025

This is the table that explains the valuation argument, and it is the one most investors never look at. Insurance and services produced R$13.4bn of the R$24.7bn first-half result, about 54% of group earnings, on R$53.5bn of average regulatory capital, about 27% of the total, at a 50.0% return. Credit produced 39% of earnings on 63% of the capital at a 15.4% return. Itaú is not primarily a lender that happens to sell insurance; on a capital-weighted basis it is a fee and insurance business that funds itself with an enormous, very well-run loan book earning barely above its cost of capital.

Which is precisely why the fee guidance cut deserves more weight than it received. The line that was downgraded sits inside the segment that generates more than half of group profit at three times the return on capital of the lending book.

Individuals: mortgage overtakes cards

Individual lending grew 1.6% sequentially and 7.8% year on year, the slowest of the three Brazilian blocks year on year, and the composition is more interesting than the rate. Mortgage balances reached R$152.2bn and have now passed the credit card book at R$150.4bn. A year earlier the card book was the larger of the two, at R$141.1bn against R$134.4bn of mortgages, so the crossover happened inside the last twelve months. Payroll lending grew 3.5% in the quarter and 11.7% year on year, with the private-sector sub-product up 14.3% sequentially. Personal loans were flat to down and vehicle loans shrank 3.3% year on year.

"In fact, the mortgage portfolio has now surpassed our credit card portfolio, which has historically been one of our most important portfolios at approximately BRL 150 billion Mortgage lending is a long-term product that fosters strong client loyalty and reciprocity, which is why this strategy is so important for us." — Milton Maluhy Filho, CEO

Management was explicit that personal lending is being cannibalised on purpose, redirecting eligible borrowers into private payroll loans "due both to its pricing advantages and its priority in the repayment structure." On mortgage funding it argued a structural edge: a savings and treasury deposit base that lets it price competitively where other private banks cannot, with R$36bn of twelve-month originations and a 55% share among private banks.

Assessment: This is the highest-quality mix shift available to a Brazilian retail bank, and it is being executed deliberately rather than defensively. The cost is embedded in the fee guide: shifting a client from revolving credit card balances into a payroll loan moves revenue from the fee line into the margin line and lowers the blended yield, which is one reason card issuance revenue is flat while the loan book grows. It is right for returns through a cycle and it mechanically suppresses the very revenue line the Street was asking about.

Very small, small and middle market: the grace-period mechanic

The small-business book grew 1.5% sequentially and 11.6% year on year, and government-backed programme balances within it rose 7.2% in the quarter on originations up 47.3%. This is also the only part of the book where management guided delinquency higher, from 1.9% to 2.0% this quarter with another 10bps expected next, as grace periods on those same government programmes expire.

"Previously, we benefited from these grace periods as the denominator grew significantly without any impact on the numerator. As these grace periods begin to expire, we naturally see this increase in the indicator. Our best estimate is that this indicator should increase by another 10 basis points next quarter, reaching approximately 2.1% in which remains well below levels observed not so long ago, such as in September 2024." — Milton Maluhy Filho, CEO

Assessment: The mechanical explanation is credible and the arithmetic is self-evident: subsidised growth with deferred amortisation flatters a delinquency ratio on the way in and unflatters it on the way out. Two things still deserve watching. The guided path is to 2.1% and then stability, which is a forecast rather than a mechanic, and it is being made in an economy management itself describes as carrying rising household and corporate leverage at a restrictive policy rate. And the growth has been concentrated in exactly the government-guaranteed product whose grace periods are now rolling off.

Corporate: the fastest grower and the pickiest allocator

Corporate lending grew 4.4% sequentially and 10.1% year on year, the strongest sequential advance of the three, and management framed it almost entirely in capital-allocation terms rather than growth terms. It also disclosed that the wholesale book was the source of this quarter's Stage 2 to Stage 3 migrations, and that wholesale credit is "more challenging," with provisions already taken.

"Most importantly, we've been very disciplined in the capital allocation and returns. Very easy to grow portfolio with the wrong returns." — Milton Maluhy Filho, CEO

Assessment: A 4.4% sequential advance in large-corporate lending at a moment when management is publicly describing market pricing as below cost of capital on many deals is a combination worth revisiting next quarter. Either Itaú is winning on relationship and structure rather than price, which is the stated case, or the appetite is drifting. The Stage 2 coverage decline discussed below is the first place that would show.

Latin America: growing into a disposal

Latin America grew 3.0% sequentially and 9.8% year on year to R$253.0bn, and contributed the entire 10bps rise in the consolidated 15-to-90-day delinquency ratio while Brazil held flat. The more consequential fact is not in the quarter at all: the sale of the Colombian and Panamanian retail operations to Banco de Bogotá completed on 31 July 2026, a month after the 30 June balance sheet date.

That transaction moved a loan portfolio of roughly R$9.7bn net of provisions and about R$7.2bn of deposits, taking some 267,000 customers with it, for an approximate net value of R$2.5bn. Corporate banking and treasury remain. The disposal is therefore still inside the June balance sheet and comes out of the September comparison, worth roughly 0.6% of the consolidated book and about 3.8% of the Latin American book. Management sized it on the call as "BRL 10 billion of credit" leaving in July.

Assessment: Strategically clean and sensibly scoped: exit sub-scale retail, keep the corporate and treasury relationships, release capital. The modelling consequence is that reported third-quarter portfolio growth will understate the underlying rate by roughly 60bps, and anyone marking the maintained 5.5% to 9.5% full-year portfolio guide needs to carry that step-down deliberately rather than treat a weak sequential print as deceleration.

Insurance and services: the profit engine, and the line that got cut

Commissions, fees and insurance after retained claims reached R$14.1bn, up 0.4% sequentially and 3.6% year on year, with the first half up 4.4%. Beneath the flat headline the composition moved: asset management up 4.6% in the quarter and 11.0% in the half, advisory and brokerage up 25.3% in the half, insurance and pension results up 12.8% in the half on higher earned premiums, against payments and collections down 0.4% and current-account fees for individuals down 21.4% in the half.

"Commissions and fees and result from insurance operations increased by 0.4% in the quarter, reflecting higher revenues from fund management and lower gains from payments and collections, as expected, due to the commercial adjustment in service package fees." — Itaú Unibanco results release

Assessment: The 21% decline in individual current-account fees is deliberate and disclosed as such, part of repricing packages to deepen relationships. We take management at its word that this is a lifetime-value trade. The problem is one of evidence rather than intent: the account fees have fallen now, the engagement revenue is promised later, and when asked directly whether the One Itaú migration was delivering the expected revenue benefit, the answer was that the benefit shows up in the credit margin rather than in the fee line. That may well be true, and it is not currently visible in a margin-with-clients line running at its guidance floor.

Trading: more capital, lower return

Financial margin with the market was R$931m in the quarter, up 13.6% sequentially and 8.6% year on year, but down 1.7% for the half, which management attributed to a weaker trading desk partly offset by better banking-desk gains. The business-model view shows why this needs watching: trading's average regulatory capital rose from R$5.3bn to R$8.2bn while its return fell from 30.7% to 21.6%.

Assessment: A nine-point return decline on a 55% larger capital allocation is the weakest line in the segment table. It is small in absolute terms, roughly R$0.9bn of a R$24.7bn half-year result, and the quarter's absolute performance was creditable given volatility and an ongoing capital-hedge cost that management defended as the right trade for earnings predictability. But the maintained R$2.5bn to R$5.5bn full-year guide for this line is extremely wide, the half-year run rate of R$1.75bn is tracking the lower portion of it, and management flagged that the fourth quarter carries most of the structured-operation seasonality. This is the guide line with the least visibility and the widest range.

Guidance & Outlook

One of eight lines moved, and it moved a long way. The fee and insurance range dropped by 3.0 points at the floor and 4.0 points at the ceiling, which halves the midpoint from 7.0% to 3.5%. For context, the first half delivered 4.4%, comfortably inside the new band and below the floor of the old one. That is the honest reading of the revision: the original guide had already stopped being achievable, and the half-year print made saying so unavoidable.

Implied second-half ramp: The arithmetic is unflattering in two places and benign in two others. Cost of credit ran at R$20.1bn in the half, an annualised R$40.2bn against a R$41.0bn midpoint, so the maintained range needs nothing unusual and management explicitly guided to "close to the midpoint." Non-interest expenses at 3.9% sit almost exactly mid-range. Financial margin with the market needs R$0.75bn to R$3.75bn in the second half to land in its range, a span so wide it carries no information. The line that does carry information is the margin with clients: the first half grew 4.8% against a maintained full-year range starting at 5.0%, so the second half must accelerate simply to reach the floor of a guide that was not revised. Management did not dispute this, and said the line is running nearer the floor than the midpoint.

Street at: The thirteen-broker panel the company itself polls was at R$12,466m for the quarter against R$12,407m delivered, and a separate poll was reported at roughly R$12.5bn. Consensus and company were aligned going in, and the revision was to a line management had flagged as at risk of the low end in prior quarters, which is a plausible explanation for why a halved fee guide produced a flat share price.

Guidance style: Itaú guides eight separate lines with explicit ranges and revises them mid-year, which is unusually transparent for the sector and cuts both ways. It means a cut is visible and cannot be buried in a mix comment. It also means the bank has published the mechanism by which its bottom line survives a revenue downgrade, and that mechanism is the tax line rather than an operating offset. The posture here is pragmatic rather than conservative: management said plainly it would revise again "whenever we have a better performance in the market," and that it does not give point estimates because "it wouldn't be a point, it's a range."

Key Topics & Management Commentary

Overall Management Tone: Management was composed and unusually forthcoming, volunteering the mechanics behind a guidance cut rather than waiting to be asked, and conceding without prompting that the client margin line is tracking nearer its floor than its midpoint. The least convincing stretch was the gap between the written explanation for the fee revision and the spoken one, which went unreconciled across thirteen questions. Tone on the macro was markedly more cautious than tone on the franchise, with repeated unsolicited warnings about system-wide credit excess, household leverage and non-bank visibility.

1. The fee revision, and the two different reasons given for it

The only guidance change in the print cut the fee and insurance growth range from 5.0% to 9.0% down to 2.0% to 5.0%. The written release attributed it to one cause and the call attributed it to another, and the difference is not cosmetic. Capital-markets volatility is episodic and reverses; a structural step-down in card and account-package revenue tied to activity levels and to the bank's own repricing does not.

"Itaú revised its projections for Commissions and fees and results from insurance operations for 2026. The bank now projects growth between 2.0% and 5.0%. The adjustment in the growth expectation is mainly related to higher volatility in the capital markets than projected at the beginning of the year." — Itaú Unibanco results release

"The only change we made was to commissions and fees and to results from insurance, which, as I mentioned earlier, is closely linked to the level of economic activity. We revised the expected growth range to between 2% and 5%, whereas at the beginning of the year, we expected growth between 5% and 9%." — Milton Maluhy Filho, CEO

On the call the cause was narrowed further, to card issuance: "we have observed some moderation in this line throughout the year as a function of economic activity levels... this is the line where we are making an adjustment." Management also noted it had flagged the risk in prior quarters: "we already saw some risk that performance could trend closer to the lower end of the range."

Assessment: We weight the call explanation more heavily, because it is more specific, because it is corroborated by the actual composition of the quarter (card issuance up 0.4%, payments and collections down, individual account fees down 21% in the half), and because capital-markets volatility is a poor fit for a line whose advisory and brokerage component grew 25.3% in the half. Treating this as a cyclical capital-markets dip risks under-reserving the structural piece.

2. The tax line is doing the work, and management said so

Asked implicitly by its own guidance table how a cut revenue line leaves the bottom line intact, management supplied the answer in the prepared remarks without being asked. It is the single most important passage on the call.

"The second comment I would like to make is not a change in guidance itself, but rather a comment on the position of the guidance. If you recalculate the implied results, I would ask you to consider the effective tax rate at the lower end of the range, which reflects our best current estimate. If you run the math based on those assumptions, you will see that the implied bottom line remains unchanged, despite the revision to fee income and insurance results, assuming the effective tax rate remains closer to the lower end of the range, the bottom line outlook is effectively the same." — Milton Maluhy Filho, CEO

This is consistent with what already happened. Pre-tax income grew 4.9% in the first half; the recurring result grew 9.1%; the realised ratio of tax expense to pre-tax income fell from 31.0% to 28.2%. The forward instruction is to assume that relief persists.

Assessment: Full marks for transparency, and a mark down for earnings quality. A lower tax rate is real money and a real return on the bank's management of deferred tax assets and interest-on-own-capital distributions, so this is not an accounting trick. But it is not operating leverage, it does not compound, and it is substantially a policy variable rather than a franchise variable. A bank growing pre-tax income at 4.9% while the market capitalises it at roughly 2.3 times book is a different proposition from one growing earnings at 9.1%, and the difference between those two descriptions is a tax ratio.

3. Margin with clients: at the floor, and asked about first

The first question of the call went to the gap between a first half at 4.8% and a maintained guide starting at 5.0%, and the last question returned to it. Management's framing is that the range remains its best estimate while conceding where inside the range the business is sitting.

"Margin with the client. If we annualize what we have in the first semester is a challenge that we have for the next 2 quarters, certainly. Today, we are running close to the floor than the midterm, the midpoint because of the effects that I commented with the liabilities that are very solid." — Milton Maluhy Filho, CEO

The attribution was specific and worth recording. Asset spreads are growing in line with the portfolio. The deceleration sits in three other components: the liability margin, which enjoyed an exceptional benefit from the rate-hiking cycle that is now lapping; working capital, where the dividend brought forward to late 2025 left the bank entering 2026 with less capital to monetise; and structured wholesale operations, which management describes as volatile and fourth-quarter weighted. Management also cautioned that a falling policy rate does not pass through quickly, because liability and working-capital positions are hedged long: "Even in a cycle of interest rate, the pass through to the margin is not automatic. There is an temporal gap as the hedges are done for the longer."

Assessment: This is the most important forward watch item in the print, ahead of credit. Three of the four margin components are decelerating for identifiable and largely non-reversing reasons, and the fourth is seasonal. The hedging point cuts against the easy bull argument that Selic cuts will lift the margin: the long hedge that protected the margin on the way up will damp it on the way down. A guide floor that requires second-half acceleration from a line whose drivers are all pointing the other way is the definition of a stretch.

4. Risk-adjusted margin is the metric management chose, and it is the right one

Headline net interest margin with clients was 9.1% consolidated for a third consecutive quarter and 10.0% in Brazil, both 30bps lower than a year ago. The risk-adjusted versions rose 10bps sequentially, to 6.2% and 6.7%.

"As I always say, generating a very high margin only to give it back through credit costs is not a sensible capital allocation strategy. What we have shown consistently is our ability to manage margins with discipline and consistency." — Milton Maluhy Filho, CEO

Assessment: We agree with the framing and note what it concedes. A flat headline margin and a rising risk-adjusted margin together mean the improvement is coming from the credit-cost denominator, not from pricing. That is the correct way to run a bank into a leveraged consumer cycle and it is a thinner source of upside than spread expansion, because credit costs cannot fall indefinitely from 2.7%.

5. Credit quality: the best numbers in the system, and a guided deterioration

Non-performing loans over 90 days held at 1.9% for a sixth consecutive quarter. The 15-to-90-day ratio rose 10bps to 1.8%, entirely from Latin America, with Brazil flat. Cost of credit over the average portfolio has now been 2.7% for six straight quarters while the portfolio grew 9.6%. Within that stability, the individuals ratio rose 10bps and the small-business ratio rose 10bps to 2.0%, with another 10bps guided for next quarter.

"So whatever you see, we are very comfortable with the indicators of credit don't see in that any type of message. Our best estimation is stability." — Milton Maluhy Filho, CEO

Management added two pieces of evidence that are hard to argue with. Write-off policy was unchanged despite the latitude the new provisioning standard affords: "We didn't change. At any point, our policies of write-off for any product." And on relative position: the card book operates at delinquency levels "roughly half of the system levels."

Assessment: Credible, and the strongest pillar in the investment case. The unchanged write-off policy matters because a bank wanting to flatter these ratios had a legitimate accounting route to do so and did not take it. What tempers the read is not the level but the direction of every sub-component that moved, and the fact that this is being delivered into an environment management itself describes as deteriorating at the margin.

6. The Stage 2 coverage decline, explained before it was asked about

Stage 2 coverage fell in the quarter, concentrated in the companies portfolio. Management pre-empted the question in the prepared remarks.

"This quarter, we experienced migrations of clients from stage 2 to stage 3. Typically, clients leave Stage 2 with a relatively high level of coverage when they are ready to migrate. And this affects the overall coverage ratio. Once again, this is essentially a mechanical effect that is fully accounted for in our projections and in our cost of credit, which I will discuss shortly." — Milton Maluhy Filho, CEO

Management emphasised it does not manage by stage but by expected loss, and that stage allocations sit materially above observed delinquency.

Assessment: The mechanic is real: if the best-covered Stage 2 exposures are the ones migrating out, average coverage on what remains falls without any change in risk. The reason to keep this on a list rather than close it is that it was disclosed in the same quarter management described wholesale credit as "more challenging," and both facts point at the same book. A second consecutive quarter of falling companies-side Stage 2 coverage would stop being mechanical.

7. Capital: up against last quarter, down against last year, and both are intentional

CET1 closed at 12.3%, 30bps above March and 80bps below June 2025. Tier 1 was 13.8%, with additional Tier 1 shown at the 1.5% regulatory cap against 1.7% actually available. The sequential build came from about 80bps of earnings retention against roughly 30bps of dividends and interest on own capital and 10bps of risk-weighted asset growth.

"It is worth remembering that we did an early dividend distribution at the end of last year, which meant that we entered 2026 with a highly optimized capital position. We also had the regulatory phase-in effects, which still had an impact during the first quarter, and yet we continue to generate capital with strong quality." — Milton Maluhy Filho, CEO

"As a result, we ended the quarter with a common equity Tier 1 ratio of 12.3% and a very strong and solid capital position with further growth expected, which should allow us to have our traditional discussion regarding additional dividend distributions at the beginning of the following year." — Milton Maluhy Filho, CEO

Assessment: The year-on-year decline is a capital-return decision plus a regulatory phase-in, not erosion, and the sequential rebuild at 30bps a quarter is the proof. The important disclosure is forward: management has signalled an extraordinary distribution discussion in early 2027, and the normalised profitability it prefers to be judged on assumes CET1 at 11.5%, which is 80bps below where it sits. That is a visible and self-identified capital surplus. It is also the clearest potential catalyst in the story, and it is two quarters away.

8. Costs: the cleanest line in the print

Non-interest expenses rose 3.3% sequentially and 3.1% year on year, with Brazil commercial and administrative personnel expense actually declining 0.5% year on year in the quarter. The efficiency ratio reached 37.4% consolidated and 35.5% in Brazil, better than a year ago on both and worse than the first quarter on both.

"Commercial and administrative expenses declined by 0.5% year-over-year and increased 3.2% in the first half of 2026 compared with the first half of 2025 remaining below both inflation and collective bargaining adjustments." — Milton Maluhy Filho, CEO

The sequential deterioration in the ratio was attributed to first-quarter seasonality in personnel accounting and transaction volumes rather than to cost drift, which the expense table supports. The structural driver is visible in the footprint: branches down 19.3% and ATMs down 20.3% year on year, with headcount down 5.5%, while the loan book grew 9.6%.

Assessment: This is unambiguously good and it is the part of the story we would underwrite without reservation. Note the precision of the claim, though: the 0.5% decline is nominal and quarterly, and the same line grew 3.2% across the half. The half-year figure is the one to carry forward. Management declined to guide branch or headcount reduction, which leaves the pace of the remaining efficiency capture undisclosed.

9. ia.i, AI First, and an expense line that is going up

The quarter's flagship launch was ia.i, a conversational generative-AI experience inside the SuperApp, positioned as a step in an "AI First" strategy. Management quantified ambitions rather than results: 20% to 30% better accuracy in credit assessment, 20% to 30% higher first-contact resolution, three times more specialist time with clients, and generative-AI initiatives in use up 88% in the half.

"The launch of ia.i is the embodiment of this vision: a proprietary Artificial Intelligence designed to enhance clients' financial decisions with context and depth, going well beyond generating a high volume of interactions." — Milton Maluhy Filho, CEO

The CFO was more useful on the economics, and more candid about the direction of travel.

"AI is a lever that is important we are implementing. We are bearing the fruits all. We have initiatives that we're doing at the same time for you to do this. It goes through expenses. The AI expenses in the bank will increase, but it generates efficiencies will generate revenues." — Gabriel Amado de Moura, CFO

Assessment: The analogy management drew to its cloud migration is the honest one: spend rises first and the efficiency arrives later. We would hold the AI narrative as neutral to the thesis today. It is credible, it is early, none of the quantified benefits are yet visible in a reported number, and the one disclosed near-term financial consequence is higher expense against a maintained 1.5% to 5.5% cost guide.

10. One Itaú: migration complete, revenue benefit relocated

Management reported the One Itaú migration concluded with 99.3% of clients moved, net promoter scores above 80, more than 20 new products launched, over 70% of clients holding three products, and an 18-point closure of the NPS gap to the leading digital bank. Asked directly whether the migration was producing the expected revenue uplift, the answer was that the benefit appears somewhere other than where the question assumed.

"And here, there are opportunities to grow in credit with the client that you knew -- there is a relationship with quite didn't explore a full bank with the private payer of loan or whatever. That is in the margin was decline. It's in the growth of the portfolio. It's not in the margin of services." — Milton Maluhy Filho, CEO

Assessment: The engagement metrics are strong and the strategic logic is sound. But the answer relocates the payoff from the fee line, which was just cut, to the credit margin line, which is running at its guidance floor. Both cannot be a source of upside at once, and currently neither is demonstrating the uplift. This is the piece of the equity story most reliant on management's word rather than on a reported number, which is precisely why it belongs in the thesis as a bear point rather than a bull pillar.

11. The system, not the bank: credit excess and non-bank visibility

The most striking feature of the call was how much time management spent warning about the Brazilian credit system rather than selling its own position in it, almost all of it unprompted or in response to open questions.

"The new banks, the newcomers brings the operations to the balance but they distribute to the fund. A great deal of the risk goes back there. So these are capital markets, but it's a risk. It's almost a shadow bank that we have low visibility to what is inside the findings." — Milton Maluhy Filho, CEO

Management described an oversupply of credit relative to what the market can absorb, consumers holding five or six cards, and a corporate credit market where it has walked away from mandates it judged to return below its cost of capital. It also argued publicly for a larger Central Bank supervisory budget. Its own posture was stated bluntly: "We lose the market share or if we are going to lose market share or money, we'd rather lose market share."

Assessment: This is the right instinct and it is also a tell. A bank that is this vocal about system-wide excess is a bank that expects the cycle to turn, and the discipline it is describing is the reason to own it through such a turn. It is simultaneously the reason not to pay a premium multiple for near-term growth, because the stated policy is to forgo volume rather than defend share. Growth guidance and this posture are in tension, and management has told you which one wins.

12. Taxation of the system, and the deferred-tax overhang

A question on whether the next administration might press banks for more tax drew an unusually direct structural answer, including the observation that the corporate tax rate for Brazilian banks runs at 45%, and that the sector's large deferred tax asset stock constrains its own behaviour.

"It's true that our system, our industry retains a huge amount of capital in the balance sheet. And why is that? Because it's regulated activity. So the Central Bank requires that for you to make the credit and to have a portfolio the size we have, you need to retain capital. And this capital that you retain at the end of the day has the benefit of but you have, on the other hand, a very high level of tax -- corporate tax rate for banks in Brazil, which is 45%, and there is an increase for financial companies in the consumer finance and also for IPs in the recent years." — Milton Maluhy Filho, CEO

Management's view was that peers may need to reduce interest-on-own-capital distributions to protect deferred tax assets, which would raise their effective rates, and that this is not Itaú's base case for itself.

Assessment: Note the circularity. Itaú's maintained bottom-line outlook depends on an effective tax rate at the low end of its range, and that rate depends in part on interest-on-own-capital deductibility, which is a tax policy instrument, in an election year, in a country with a fiscal problem, for a sector earning 24% returns. Management is right that there is no live proposal. It is also the case that the earnings bridge it has asked the Street to model is the item most exposed to a policy change it does not control. That is a tail risk, not a forecast, and it is the correct place to file it.

Analyst Q&A Highlights

Where the second-half margin acceleration is supposed to come from

The first question of the call went straight to the arithmetic problem in the guidance table, observing that the half-year margin-with-clients growth rate sat below the floor of a range that had been left unchanged. Management did not contest the premise. It separated the margin into asset spreads, liability margin, working capital and structured operations, confirmed that only the first of those is growing in line with the portfolio, and rested the maintained guide on the width of the range plus fourth-quarter wholesale seasonality.

Q: "Question about the margin with the clients. The quarter was good without the offenders of the first quarter, but any accumulative of the semester, the line is a bit below 5% against a guidance that starts at 5% and goes to 9%. The guidance was kept I want to understand where the acceleration is coming from. In the second semester, more volume, more mix margin of liabilities or any relevant own capital. in this account. And the [indiscernible] rate and the cycle turning, how do you foresee the behavior of the margin of liabilities from now on?"
— Bernardo Guttmann, XP Investments

A: "So when we look at the 2 quarters, the expectation is that the range still comprises our best opinion of projections. Of course, it depends on the activity and a series of factors. Nonetheless, we can see some volatility in the margin in the third fourth quarter, but due to the seasonality of the structured operations of the wholesale that tend to be stronger in the fourth quarter."
— Milton Maluhy Filho, CEO

Assessment: The answer defended the range without defending the midpoint, which is the correct read and a lower bar than the unchanged guide implies. Three of four margin components are decelerating for structural reasons and the bridge to the floor is wholesale seasonality in a line management elsewhere called volatile. Treat the floor, not the midpoint, as the planning case.

Which guidance lines carry the most risk from here

A recurring line of questioning through the call asked management to rank its own guidance by fragility, and the final question put it most directly. The answer produced the single most useful concession of the morning, and paired it with an unprompted statement of provisioning sequence.

Q: "The market has changed a lot. It's not normal to see a decision guidance. And the question is the new guidance, given the breadth of the guidance somehow reflects this estimation for 2026. But in the quickly the world is lines that are weaker in the guidance, if there is a mistake an expectation of review, what are the lines that should be more at risk."
— Henrique Navarro, Santander

A: "Margin with the client. If we annualize what we have in the first semester is a challenge that we have for the next 2 quarters, certainly. Today, we are running close to the floor than the midterm, the midpoint because of the effects that I commented with the liabilities that are very solid. [...] The cost of credit, you discussed I wanted to tell you that a bit to the left a bit to the right, our best expectation is should be closer than to the midpoint of the guidance. [...] And the message that I wanted to make it very clear to you is that, first, we do the provisions and then we discuss the profit. It's not from the profit to the provision."
— Milton Maluhy Filho, CEO

Assessment: Management named the client margin as the most at-risk line before anyone else did, which is the behaviour of a team that intends to be believed next time. The provisioning-sequence remark is the more durable disclosure: a bank that says it will accept a weaker quarter rather than under-provision has set an expectation it can be held to, and it is the reason to treat the credit numbers in this print as unmanaged.

What actually drove the fee and insurance revision

The revision drew the most sustained questioning of the call. The answer that came back was a product-by-product account of deliberate strategy rather than the capital-markets explanation the written release had offered: card de-risking and a shift to high-income customers, account-package repricing to remove friction, and a conscious decision to decline capital-markets deals priced below cost of capital.

Q: "I want to see the service line with a review of the guidance, but I wanted to understand more of the dynamics that provoke this review. And what is the strategy of the bank -- about the issuance of credit cards, the lines of payment, the payments of companies, acquirents and even the insurance line, I wanted to understand the service lines up ahead."
— Marcelo Mizrahi, Bradesco BBI

A: "First, throughout the years, we did a derisking important derisking in the portfolio in the income toward the less resilient public, we did a derisking that is very relevant. Looking here in the rearview mirror, we lost marginally revenue we saved important volumes of loss of credit. So the strategy was good. [...] And the big offender are the packages of companies, tariffs that we are reducing directionally. And the objection of this reduction is to remove the friction, increasing the lifetime value and creating long-term value."
— Milton Maluhy Filho, CEO

Assessment: This answer is more convincing than the written explanation and considerably less comforting. Capital-markets volatility would reverse; a repriced fee structure and a de-risked card book will not. The trade is probably right on a lifetime-value basis, and it means the fee line should be modelled at a structurally lower growth rate rather than as a cyclical trough.

Whether a mid-twenties return is defensible while peers reach for growth

The clearest exchange of the call, and the one that produced the number the whole investment case turns on. The question put the competitive risk precisely: a bank earning roughly 26% in Brazil while incumbents struggle to reach double digits creates an incentive for those incumbents to behave irrationally. The answer supplied an explicit cost of equity and an explicit decision rule.

Q: "you're delivering about a 26% ROE in Brazil at a time where a lot of your incumbent competitors are struggling to do double digits, right? And there's concerns about the credit cycle, growth slowing, high interest rate environment? And how do you think about that competitive dynamics because that could create some incentives for some irrationality perhaps from some of your competitors to try to improve their position relative to yours."
— Tito Labarta, Goldman Sachs

A: "And it depends a lot of cost of equity that today we pretty much said at 14.75%. This is where we believe our cost of equity is set, okay? So whenever we are generating at 14.75 plus, we are creating value to the shareholders. If we are operating there and less than that, we are deploying capital in the wrong way. So this is the discipline we have. It's true that whenever you need to show some results, you try to grow fast and growing portfolios and to underwrite credit in a rational way. we wouldn't be doing that because I believe this is not sustainable. And whenever we see some irrationality and it's happening in some segments. We give one step behind and say, that's okay because it's not sustainable."
— Milton Maluhy Filho, CEO

Assessment: A disclosed hurdle rate of 14.75% against a 24.3% consolidated return is the strongest single fact in the bull case, and the admission that irrationality "it's happening in some segments" is the strongest single fact in the bear case. The stated response to competitive pressure is to cede volume, which protects returns and caps growth. An investor is being asked to choose between those two, and management has already chosen.

Whether asset-quality comfort survives a flat-to-worse trend

A pointed line of questioning pushed past the headline stability to the direction of travel, noting that if the ratios are no longer improving then the only remaining question is how much they deteriorate. Management held its ground on the consolidated picture while conceding the specific sub-segments.

Q: "So I want to understand is if you're comfort with the asset quality, well, things are not going to improve a lot. There is a worsening. I know that Itau has a better balance, you are more prepared, but we are concerned are we going to see any levels of worsening or no. This is a scenario of comfort, stability to understand. And if you can explain what happened with the 15 to 90."
— Yuri Fernandes

A: "First information on our side. We didn't change. At any point, our policies of write-off for any product. The 4966 gives you liberty, but we've kept it as is expectation to take the clients write-off, didn't change. We don't do that for provision or using this degree to freedom for that. [...] Our best estimation is stability. We don't have the best estimation for the individuals and the -- that's why individuals for the companies and SMEs, we should worsen basis in the next quarter and then stable in onwards."
— Milton Maluhy Filho, CEO

Assessment: The unchanged write-off policy is the most verifiable thing said about credit on this call, because the new standard offered latitude that was declined. The forward guide is stability at the consolidated level and a further step down in small business before stabilisation. Both can be true, and the consolidated ratio will hold only if the secured mix shift keeps offsetting the small-business normalisation, which makes the mortgage and payroll growth rates an asset-quality variable rather than just a volume one.

Cost contingency if revenue slows, and the price of the AI build

A question framed around the scenario that actually matters to the thesis, revenue growth slowing to single digits, asked what cost levers exist to defend returns, and whether AI is already funded or still to be paid for. The CEO answered on the lever, the CFO answered on the bill.

Q: "there is a scenario where your revenue growth single digit, how much management could you have a cost contingency to keep the ROI. What limits the cost of cost? Is it more institutional, regulatory or simply is a deliberate choice speed of investment, crossing with your revenue, and looking at what we are doing now."
— Daniel Vaz, J. Safra

A: "AI is a lever that is important we are implementing. We are bearing the fruits all. We have initiatives that we're doing at the same time for you to do this. It goes through expenses. The AI expenses in the bank will increase, but it generates efficiencies will generate revenues. So it's [indiscernible] different from all the transformation that we had with cloud before. Its expenses have increased, but you generate the development of the privacy for the efficiency of the processes and the bank as a whole."
— Gabriel Amado de Moura, CFO

Assessment: The honest answer, and a mildly unhelpful one for anyone modelling 2027. Costs rise first and the efficiency arrives later, on the cloud template. The CEO's companion remark that he does not "have a silver bullet" and that efficiency rests on a mapped programme of initiatives rather than one lever is consistent with the expense trend, and it also means there is no large discrete cost reserve to deploy if revenue disappoints.

Non-bank credit growth and what the system cannot see

An open question about system preparedness for a credit shock produced the most forceful answer of the call and the clearest statement of where management thinks the risk actually sits, which is outside the regulated banks.

Q: "Let's go with the credit cycle and see your opinion. How do you see the system being prepared for a crisis economic one. The market changed a lot. The companies, the Capital Markets Day multiplied. And for example, the industry, there is BRL 800 million. That dilutes reason on 1 side, but it causes changes. And we don't see how the capital markets would react to a crisis."
— Eduardo Rosman, BTG Pactual

A: "And then is the unsupervised. The new banks, the newcomers brings the operations to the balance but they distribute to the fund. A great deal of the risk goes back there. So these are capital markets, but it's a risk. It's almost a shadow bank that we have low visibility to what is inside the findings. And when you have a situation of stress who is the owner of the receivables, these are situations that come up, and we are going to have to deal with that well."
— Milton Maluhy Filho, CEO

Assessment: A market leader arguing publicly for a bigger supervisory budget and describing a shadow-banking blind spot is making a competitive argument as well as a prudential one, and that does not make it wrong. The investment implication is asymmetric in Itaú's favour on a two-to-three-year view and negative for near-term volume growth, because the credit being distributed into funds is credit Itaú is choosing not to write.

Bank taxation and the deferred-tax stock

A question on policy risk to the sector's tax burden landed on exactly the variable this quarter's earnings bridge depends on, though it was asked about the industry rather than about Itaú's own guidance.

Q: "Do you differentiate yourselves not only for the higher profitability, but also for being the bank with probably the highest effective tax rate. But when you think about it from the policymaker point of view, the system has probably paying less taxes today than they were before. And with the amortization of the TAs possibly even more. Are you concerned that in the next administration, there could be a pressure for the industry to pay more?"
— Carlos Gomez-Lopez, HSBC

A: "So I think there is no risk of this discussion because it has to do with the level of capital, and I think banks at a certain point they will have to reduce the IOC. It's not our base case. We don't have any issue with that. But I think the market might need to reduce IOC in order to avoid having losses in their tax credit that will reduce capital. I think this is something that we might see, and this will necessarily increase their effective rate."
— Milton Maluhy Filho, CEO

Assessment: The answer is about peers and the exposure is Itaú's own. Management dismissed policy risk and in the same breath described a mechanism by which sector effective tax rates rise. Given that the maintained bottom-line outlook rests on a tax rate at the bottom of its guided range, this exchange deserved a follow-up it did not get. Nobody asked what the bottom line looks like at the top of the range instead.

What They're NOT Saying

1. What Avenue contributed: The income statement has consolidated Avenue since the first quarter of 2026, where the prior-year comparatives carry it only as an equity-method associate. The executive summary quantifies the contribution nowhere, so the organic component of 4.7% revenue growth and 4.4% fee growth is undisclosed. The acquisition was not raised once across the prepared remarks and thirteen questions. For a quarter whose central question is the trajectory of fee revenue, the inorganic share of that line is the number most needed and least available.
2. The two reasons for the fee cut, reconciled: The release says capital-markets volatility. The call says economic activity and card issuance. Thirteen questions were taken and nobody put the two side by side. They have very different implications for whether 2027 fee growth recovers to the old 5% to 9% range or stays near the new one.
3. The bottom line at the top of the tax range: Management instructed the Street to model the effective tax rate at the low end of 29.5% to 32.5% in order to keep the bottom-line outlook unchanged. It did not disclose what the implied result is at the midpoint or the high end, which is the sensitivity that now matters most.
4. Any quantified AI benefit: Accuracy gains of 20% to 30%, resolution gains of 20% to 30%, three times more specialist client time and an 88% rise in generative-AI initiatives in use are all presented as capabilities. None appears as a reported revenue or cost figure, while the CFO confirmed AI expense will rise. The payback period is undisclosed.
5. The pace of the footprint reduction: Branches fell 19.3% and ATMs 20.3% year on year, and headcount 5.5%, which is a material part of the efficiency story. Asked how much further this goes, management said explicitly: "We don't give guidance. on the amount of branches or head count." The remaining runway on the biggest cost lever is therefore unquantified.
6. Where the One Itaú revenue benefit shows up, measurably: The migration is complete and the engagement statistics are strong. Asked whether the expected revenue benefit had arrived, management moved it from the fee line to the credit margin. Neither line is currently showing it, and no figure was offered for either.
7. The third-quarter portfolio step-down, pre-flagged: The Colombian retail disposal completed a month after the balance-sheet date and removes roughly R$9.7bn of loans. The CEO sized it in passing as "BRL 10 billion of credit" in an answer about margin, and it was not presented as a guidance or modelling item in the release or the presentation despite a maintained full-year portfolio growth range.

Market Reaction

• Pre-print setup: The ADR closed at $8.19 on Tuesday 4 August, the day the results were released after the close. It entered the print up 14.4% year to date, up 27.4% over twelve months and up 0.9% over the trailing 30 days, against a 52-week closing range of $6.29 to $9.57, which placed it about 14% below the 52-week closing high. The S&P 500 was up 13.0% year to date on the same date, so the year-to-date outperformance was only about 1.4 points despite the much larger twelve-month gain.
• Pre-print session: The print day itself was the violent one, and it was not company-specific. The ADR fell 3.53% on 4 August while the Ibovespa was essentially flat, down 0.06%. The other large Brazilian bank ADR fell 3.59% the same day and both local bank preferred shares fell between 1.8% and 2.5%, with roughly 0.8 points of the ADR declines coming from a weaker real.
• Reaction session: On Wednesday 5 August, the day of the 9:00 a.m. call, the ADR opened at $8.40, a 2.6% gap, reached $8.49 intraday for a gain of 3.7%, and closed at $8.21, up 0.2%. It surrendered roughly 93% of its peak advance during the session.
• Volume: 26.6m shares against a 19.8m thirty-day average, or 1.34 times normal.
• Relative performance: The local preferred share closed up 0.67% against the closest domestic peer down 0.82%, a spread of about 1.5 points, on an index that was flat. The ADR's 0.2% gain against the Brazil country fund's 0.1% was a much narrower edge, the difference being the currency leg.

The shape of the session matters more than the close. A 2.6% gap higher says the print read well on first pass: an in-line bottom line, a mid-twenties return, a sixth straight quarter of 1.9% delinquency and a capital ratio up sequentially. The fade to 0.2% says the second pass found the fee cut, the sequential decline in return on equity and the fact that the maintained bottom line depends on a tax assumption rather than an operating offset. That is not a rejection of the quarter; it is a repricing of the growth rate inside it.

Two external factors shaped the tape and neither is about Itaú. The Brazilian central bank's rate-setting committee was meeting on 4 and 5 August and announced a 25bp cut to 14.00% after the close on the reaction day, a fourth consecutive cut taking cumulative easing to 100bp since March. The reaction session therefore traded with that decision pending, which is a reason for position-squaring rather than conviction, and goes some way to explaining why a gap-up faded on heavy volume. Separately, United States tariffs on Brazilian goods remained a live overhang on Brazilian risk assets. Itaú is domestically funded and domestically lent, so that channel reaches it through growth and asset quality rather than directly, but it reaches the multiple immediately.

The honest summary of the reaction is that the market took an in-line quarter from a high-quality bank and declined to pay more for it. Given the stock had already risen 27% over twelve months into a print that cut a revenue guide, that looks like a correctly priced response rather than a mispricing in either direction.

Street Perspective

Debate: Is a mid-twenties return on equity a plateau or a peak?

Bull view: The bull case on the Street holds that a 24.3% consolidated return against a 14.75% cost of capital is structural rather than cyclical, resting on a fee and insurance business that earns 50% on 27% of the capital, a physical footprint shrinking a fifth a year, and the best credit book in the system. On this view the return floor has been reset upward and the market is still valuing the bank on a historical average return it has left behind.

Bear view: The bear camp points out that the return fell 50bps sequentially, that return on assets fell to 1.5% from 1.6%, that the headline client margin has been flat for three quarters and down 30bps year on year, and that nearly half of first-half earnings growth came from a tax ratio rather than operations. On this reading the mid-twenties return is a cyclical high struck on an optimised capital base and a favourable tax rate, both of which mean-revert.

Our take: The bears have the better of the near term and the bulls the better of the long term. The level is real and defensible; the rate of change has turned, and every sequential metric in this print agrees on that. A plateau at 24% is an excellent outcome and it is not what a 2.3 times book multiple pays for after a 27% twelve-month run. We would rather own this franchise at a lower multiple or after the growth question resolves.

Debate: Does the fee cut signal a revenue problem or a deliberate mix shift?

Bull view: The constructive reading on the Street is that the fee decline is self-inflicted and value-accretive: account packages are being repriced to deepen relationships, the card book has been deliberately de-risked toward high-income customers, and capital-markets mandates priced below cost of capital are being declined. Revenue moves from the fee line into the credit margin and into customer lifetime value, and the asset-management and insurance components growing 11.0% and 12.8% in the half show the engine is intact.

Bear view: The sceptical reading notes that a guidance midpoint does not halve for reasons of strategy, that the written and spoken explanations for the cut do not match, and that the revenue supposedly migrating into the credit margin is arriving in a line already running below its own guidance floor. On this view the fee cut is the first visible crack in a revenue story that has been carried by loan growth.

Our take: The bears are right about the diagnosis and the bulls are right about the intent. Both explanations given are probably true in part, and the call version is the more structural of the two, which argues for modelling fee growth near the new range rather than treating this as a cyclical trough. The decisive problem is the one neither camp emphasises: the fee line and the credit margin line cannot both be the beneficiary of the same relationship deepening, and right now neither is demonstrating it.

Debate: Does a Selic easing cycle help or hurt from here?

Bull view: The prevailing expectation on the Street is that four cuts and 100bp of cumulative easing mark the start of a genuine turn: lower rates relieve household and corporate debt service, which supports asset quality and revives loan demand, and a bank with the best credit book in the system and surplus capital is the natural way to own a Brazilian recovery into 2027.

Bear view: The counter is that a falling policy rate compresses the liability margin that drove the last two years of revenue, that the deposit and working-capital book is hedged long so the pass-through is slow in both directions, and that at 14.00% policy is still deeply restrictive after four consecutive cuts. On this view easing delivers the margin headwind first and the volume and credit benefit much later.

Our take: The bears are more right than the market currently assumes, and management said as much: "Even in a cycle of interest rate, the pass through to the margin is not automatic. There is an temporal gap as the hedges are done for the longer." The easing cycle is a 2027 tailwind to volumes and a 2026 headwind to the margin mix, which is exactly the wrong order for a guide that needs second-half margin acceleration. We would not underwrite rate cuts as near-term support for this particular line.

Model Update

Aardvark Labs carries no prior model for Itaú Unibanco. The table below sets the initiating assumptions against the company's own guidance, and flags where we sit away from it and why.

Valuation: At 30 June the company's disclosed market capitalisation of R$476.9bn against stockholders' equity of R$208.5bn put the shares on roughly 2.29 times book, with book value per share of R$18.92. The local shares drifted modestly lower into the print, so the multiple entering the results was a little under that. Annualising the quarter's recurring earnings per share of R$1.13 gives roughly 9.3 times on the pre-print local price. For a bank compounding book value at 4.6% a year while returning about 30% of earnings and earning a mid-twenties return, that is a fair multiple rather than a cheap one.

Valuation impact: We initiate without a published price target. The reason is specific rather than evasive: the central uncertainty here is not the multiple but the 2027 earnings base, which turns on whether the fee range is permanently reset, whether the client margin holds its floor once the liability tailwind has fully lapped, and whether the effective tax rate stays at the bottom of its band. Those three resolve over the next two prints. A target struck today would be a view on a tax rate.

Thesis Scorecard Post-Earnings

This is initiation, so the pillars below are established here rather than graded against a prior quarter. Status tags use ON TRACK, AT RISK or BROKEN for bull pillars, and CONTAINED, EMERGING or MATERIALISING for bear points. Each is written so that the next quarter can grade it against a specific number.

Overall: Thesis established. Four bull pillars are confirmed by this quarter's disclosure and rest on reported numbers rather than on management narrative, which is unusual and is the reason this is a Hold rather than an Underperform. Two bear points are already live and both concern the composition of growth rather than the level of returns. The franchise is not in question; the growth rate is.

Action: Hold. We want to own this business and we do not want to pay a full multiple for a quarter whose earnings bridge runs through a tax ratio. The specific signposts that would move us to Outperform are a client margin line reaching the floor of its own guide on operating drivers, a fee line stabilising inside the new range without further revision, and confirmation of the extraordinary distribution flagged for early 2027. The signposts that would move us to Underperform are a second revenue guide cut, small-business delinquency overshooting the guided 2.1%, a second consecutive decline in companies-side Stage 2 coverage, or any policy move against the sector's effective tax rate.