픽페이(PICS) 2026년 2분기 실적 발표 요약

픽페이는 2026년 2분기 매출과 대출 포트폴리오의 빠른 성장을 기록했으며, 영업 레버리지 효과에 힘입어 조정 순이익이 급증했습니다. 포트폴리오 성숙에 따라 자산 건전성 지표, 특히 민간 급여 담보 대출의 건전성이 핵심 주시 항목으로 유지되었습니다.

핵심 요약

• 총매출은 전년 동기 대비 67%, 전분기 대비 17% 증가한 41억 헤알을 기록했습니다. 파생상품 및 헤지 회계를 제외한 경영 매출은 37억 헤알이었습니다.
• 조정 순이익은 브라질의 기술 세제 혜택인 '레이 두 벵(Lei do Bem)' 정책의 지원 등에 힘입어 전년 동기 대비 135%, 전분기 대비 67% 증가한 2억 8,300만 헤알을 기록했습니다.
• 대출 포트폴리오는 전년 동기 대비 99%, 전분기 대비 14% 확대된 319억 헤알을 기록했습니다. 저위험 대출과 성숙 단계의 신용카드가 분기 포트폴리오 증가분의 86%를 차지했습니다.
• 수익 다각화가 지속되었습니다. 담보 및 부분 담보 대출, 수수료, 커미션, 플로트(float) 및 헤지 회계를 포함한 무위험 또는 저신용위험 원천에서의 매출 비중이 전년 동기 63%에서 71%로 확대되었습니다.
• 초기 연체율은 8.4%에서 7.5%로 개선되었으나, 포트폴리오 성숙에 따라 90일 이상 연체 대출 비율은 9.8%로 상승했습니다. 경영진은 연말까지 이 비율이 10%대 초반으로 이동할 것으로 예상하고 있습니다.
• 픽페이는 2026년 3분기 개별 기준 가이던스로 경영 매출 약 40억 헤알, 순이자이익 약 21억 헤알, 조정 순이익 약 2억 6,500만 헤알을 제시했습니다.

핵심 재무 데이터

사업 및 영업 실적

픽페이의 이번 분기 말 총 계좌 수는 전년 동기 대비 10% 증가한 7,040만 개, 분기 활성 고객 수는 4,540만 명을 기록했습니다. 고객당 평균 매출(ARPAC)은 전년 동기 대비 52%, 전분기 대비 14% 증가한 92헤알을 기록했습니다. 헤지 회계를 제외한 ARPAC은 83.3헤알이었습니다.

수익 구조는 더욱 다각화된 원천으로 계속 전환되었습니다. 담보 대출 매출은 전년 동기 대비 158% 증가한 10억 헤알에 달했습니다. 무담보 대출 매출은 40% 증가한 12억 헤알을 기록했고, 비신용 매출은 57% 증가한 19억 헤알을 기록했습니다.

카드 TPV는 전년 동기 대비 40%, 전분기 대비 12% 증가한 195억 헤알을 기록했습니다. 신규 소비자 대출 취급액은 전년 동기 대비 78% 증가한 48억 헤알에 달했습니다.

민간 급여 담보 대출 포트폴리오는 360만 건 이상의 계약에 걸쳐 72억 헤알에 달했습니다. 경영진은 이 상품이 매력적인 한계 ROE와 더 강력한 교차 판매 효과를 계속 창출하고 있다고 밝혔습니다. 민간 급여 담보 대출을 이용하는 고객의 평균 ARPAC은 픽페이 전체 평균 고객보다 8.9배 높으며, 교차 판매 비율은 전체 고객 평균보다 30% 높습니다.

중소기업(SMB) 활동도 확대되었습니다. 픽페이는 2026년 상반기 동안 월평균 85,000개의 신규 SMB 계좌를 확보하여 전년 동기의 27,000개에서 증가했습니다. 공급망 금융 신규 취급액은 1분기 6억 9,300만 헤알에서 10억 5,000만 헤알로 늘어났습니다.

AI는 고객 상품과 내부 효율성 모두에 기여했습니다. 픽페이는 2025년 10월 이후 직원 수가 정체 상태를 유지함에 따라 당초 예상했던 2026년 인원 10% 증가는 발생하지 않을 것이라고 밝혔습니다. 전체 임직원의 약 90%가 사내 AI 플랫폼을 활발히 사용하고 있으며, 토큰 비용은 연초 대비 70% 감소했습니다.

픽페이는 8월 3일 Cover(현 CAV) 인수를 완료했습니다. 경영진은 인수된 보험 사업 부문이 2026년 8월부터 12월까지 8,000만 헤알에서 1억 헤알의 순이익을 기여할 것으로 기대하고 있습니다.

경영진 가이던스

픽페이의 2026년 3분기 가이던스는 CAV의 기여분을 제외한 개별 기준 사업실적을 반영합니다.

경영진은 순이익의 전분기 대비 하락 예상이 영업 악화 때문이 아니라, 2분기에 집중되었던 '레이 두 벵' 혜택 이후 유효세율이 정상화된 데 따른 것이라고 설명했습니다.

픽페이는 3분기 조달 비용이 CDI 대비 약 94%–95% 수준으로 회복될 것으로 예상합니다. 경영진은 또한 위험조정 순이자마진이 12%–12.1% 부근에서 대체로 안정적인 수준을 유지할 것으로 예상합니다.

경영진은 3분기 CAV 인수에 따른 자본 비율 영향이 약 150bp일 것으로 추정하며, 픽페이가 2026년 말 총 자본 비율 약 14%, 보통주 자본(CET1) 비율 12%–12.5% 수준으로 연도를 마감할 것으로 예상하고 있습니다.

리스크 및 주요 주시 분야

90일 이상 연체 대출 비율은 9.8%로 상승했고, 스테이지 3 대출은 12.9%에 달했습니다. 경영진은 이러한 상승이 유사 위험군 내 건전성 악화라기보다 포트폴리오 성숙과 일부 민간 급여 담보 대출 부문에서의 의도적인 추가 위험 취하에 주로 기인한다고 설명했습니다.

스테이지 3 대손충당금 적립률은 77%에서 74.1%로 낮아졌습니다. 회사는 이것이 충당금 적립 기준 완화 때문이 아니라 FGO 보증으로 지원되는 재조정 대출의 부도시 손실률(LGD) 하향 가정을 반영한 결과라고 설명했습니다. 경영진은 향후 분기 동안 적립률이 70%대 후반 수준으로 회복될 것으로 예상합니다.

분기 위험비용은 3.7%에서 3.9%로 증가했습니다. 대손충당금 설정 비용은 9억 7,400만 헤알에서 12억 헤알로 늘어난 반면, 전체 포트폴리오 충당금 적립률은 13.9%로 안정적으로 유지되었습니다.

경영진은 대출 빈티지가 성숙함에 따라 장기 연체 지표에 추가적인 압박이 가해질 것으로 예상합니다. 90일 이상 연체 대출 비율은 2026년 말까지 10%대 초반으로 이동하여 현재 스테이지 3 비율에 수렴할 수 있습니다.

애널리스트 Q&A 주요 내용

• 부채 재조정: 이 프로그램으로 2분기 신용 비용이 약 5,900만 헤알 감소했으며, 90일 이상 고정이하여신(NPL) 비율이 약 117bp 낮아졌습니다. 경영진은 3분기에도 규모는 작지만 추가적인 수혜가 있을 것으로 예상합니다.
• 자금 조달: 픽페이는 자본시장 수단 및 기타 조달 채널을 통한 다각화를 추진하는 동시에 지속적인 예금 성장을 예상하고 있습니다. 경영진은 향후 몇 분기 동안 조달 비용을 CDI의 95% 수준으로 유지하는 것을 목표로 하고 있습니다.
• 민간 급여 담보 대출 리스크: 경영진은 리스크 관리 체계를 유지하면서 선택적으로 더 높은 리스크를 지닌 고객군으로 확장을 추진하고 있다고 밝혔습니다. 민간 급여 담보 대출의 연환산 위험비용은 10%대 중후반 수준으로 설명되었습니다.
• 영업비용: 2분기에는 3분기에서 조기 집행된 마케팅 비용 3,000만 헤알이 포함되었습니다. 경영진은 마케팅 비용이 전분기 대비 줄어들고 효율성 비율이 더욱 개선될 것으로 예상합니다.
• 상각 정책: 신용카드 및 개인 대출에 대한 픽페이의 상각 정책은 360일로 변함없이 유지됩니다.
• ARPAC 성장: 경영진은 기존 고객층에 대한 신용 및 보험 상품의 침투율 심화가 주요 수익화 동력으로 유지될 것으로 예상합니다.

실적 발표 전화회의 전체 전문

전체 실적 발표 컨퍼런스 콜 녹취록

경영진 발표

Operator

Good evening, everyone, and welcome to PicPay's Second Quarter 2026 Earnings Conference Call. Joining the call today are Eduardo Chedid, Chief Executive Officer; Andre Cazotto, Chief Financial and Investor Relations Officer; and Danilo Caffaro, Vice President of Consumer Banking.

Please note that this presentation may contain forward-looking statements and non-GAAP financial measures. Please refer to the disclaimer on the screen and to the earnings materials available on the Investor Relations section of PicPay's website for additional information. This call is being recorded, and a replay will be available on the company's website shortly after the conclusion of the call. At this time,

I would like to turn the call over to Eduardo Chedid, Chief Executive Officer of PicPay.

Eduardo Simoes

Thank you, operator, and welcome, everyone. This is our third earnings call as a public company and I'm proud to share another quarter of strong execution across our platform.

Before we get into the results, I want to say a few words about our CFO transition. As we announced in early August, Andre Cazotto has succeeded Rodrigo Couto as our Chief Financial Officer. This transition is the result of a planned succession process, and I'm confident in the strength and [ of ] our leadership team. Rodrigo played a key role in a critical phase of PicPay's evolution, strengthening our finance organization, leading our Sarbanes–Oxley preparation and being instrumental in our successful IPO in January.

He has been a tremendous partner, and I'm glad he will continue working with us as special adviser through year-end. Cazotto brings over 20 years of experience in payments and financial services and has been with PicPay since 2021 leading the capital markets work stream for our NASDAQ listing, Investor Relations and M&A. He has deep institutional knowledge and strong relationships with our financial stakeholders. Cazotto, I'm confident you are the right person for this role. Welcome and best of luck as we enter this new chapter together.

Andre Cazotto

Thank you Eduardo. It's a privilege to step into this role at such an exciting moment for the company. I have spent the past few weeks working closely with Rodrigo and our teams to ensure a seamless transition. What stands out to me is the strength of our financial foundation and the discipline with which this business operates. I'm excited to lead the next phase of PicPay's financial journey.

Eduardo Simoes

Thank you, Cazotto. Let's jump into the second quarter results now. I'm proud of what we delivered in the second quarter. This slide tells the story in one picture with big guidance on virtually every metric. Credit portfolio came in at BRL 31.9 billion, 3% above the high end of guidance. Cost of risk came in at 3.9%, aligned within our guidance range. Revenues reached BRL 3.7 billion, 3.6% above guidance and net interest income was BRL 2 billion, 5.4% above our guidance range. But the real story is in the profitability.

Gross profit came in at BRL 1.25 billion that's 8.4% above guidance, driven by the operating leverage. And adjusted net income reached BRL 283 million, 15.5% above guidance reflecting strong top line momentum and continued cost discipline. That's the story. We delivered on our commitments across the board with particularly strong [ bets ] on the profitability metrics that matter most.

Let me start with our operating metrics, which are scaling with consistency. Total accounts reached BRL 70.4 million, up 10% year-over-year and 3% sequentially. Quarterly active clients grew to BRL 45.4 million, reflecting sustained engagement across our base. Consolidated TPV came in at BRL 167.6 billion, 27% above the prior year and 7% higher sequentially. Wallet and banking TPV reached BRL 142.6 billion, up 19% year-over-year and 6% quarter-over-quarter.

Total cash in was BRL 136.4 billion, growing 17% versus a year ago and 9% sequentially. On average, more than BRL 45 billion per month. Deposits grew to BRL 35.8 billion, up 45% year-over-year and 10% higher than last quarter. This is a strong signal of increasing trust and principality in our franchise. And active insurance policies reached BRL 11.1 million, 63% ahead of last year and 9% above Q1 as our insurance vertical continues to scale rapidly across every metric, consistent sequential growth on top of already strong comparables.

Turning to financials. And this is where the monetization engine really shows its power. Total revenues reached BRL 4.1 billion, a 67% increase year-over-year and 17% higher than last quarter. That's the top line growing fast. But let me highlight what's underneath. Excluding derivatives and hedge accounting, managerial revenues were BRL 3.7 billion up 59% year-over-year and 17% sequentially. That acceleration is driven by secured and partially secured credit origination, deeper card engagement and a richer fee-based product mix.

ARPAC grew to BRL 92 per active client, 52% above where we were a year ago and 14% ahead of Q1. Excluding hedge accounting, ARPAC was BRL 83.3 showing that even on a like-for-like basis, we are monetizing each client significantly more. Gross profit came in at BRL 1.25 billion up 48% year-over-year and 14% higher sequentially. The gap between revenue growth at 67% and cost growth at a fraction of that is the operating leverage this model was built for and that leverage shows up clearly in our unit economics.

Cost to serve was 21.3% per active client up 13% year-over-year, but only 5% sequentially. It's worth noting that this figure includes BRL 0.70 per client of opportunistic investments in marketing campaigns for seasonal events that we brought forward from the third quarter. Excluding this anticipation, cost to serve would have been BRL 20.6 representing only a 1% sequential increase.

Let me put that in perspective. Revenue per client expanded 67% year-over-year, while cost to serve grew just 13%. For every real, we invest in serving our clients, we are generating over BRL 4 in revenue. That's the leverage embedded in this model. Adjusted earnings before taxes reached BRL 291 million, up 174% year-over-year and 17% sequentially. This reflects a business that is scaling efficiently and translating top line growth into bottom line results. Adjusted net income was BRL 283 million, up 135% year-over-year and 67% above last quarter.

The sequential jump from BRL 169 million to BRL 283 million reflects strong top line momentum, continued cost discipline and the positive tax benefit from Brazil's Lei do Bem incentive program for technology companies. I want to spend a moment on this slide because it captures a planned structural shift in PicPay's revenue mix. Total revenue of BRL 4.1 billion is broken down as follows: 29% from unsecured credit, 24% from secured and partially secured products, 24% from fees and commissions and 23% from float and hedge accounting. The key number, 71% of our revenues are now driven by no or lower credit risk streams, float, hedge accounting, fees, commissions and secure and partially secured credit. That's up from 63% just 12 months ago.

Let me say that again, we are growing total revenue 67% year-over-year, while simultaneously building a fundamentally more resilient business. A more diversified revenue mix, combined with a higher share of collateralized credit revenues allows us to balance growth across more mature collateralized portfolios while using intentional risk as a lever, growing through small and progressive limits on cards, buy now, pay later on loans and selectively expanding into slightly higher risk clusters within private payroll loans, all of this while maintaining the same risk appetite and targeted risk-adjusted returns.

Looking at the 3 revenue engines individually over the last 5 quarters, secure credit revenues reached BRL 1 billion, up 158% year-over-year and 23% sequentially. Trajectory from BRL 391 million to BRL 1 billion in 12 months tells the story of our payroll loan franchise reaching meaningful scale. Unsecured credit revenues came in at BRL 1.2 billion, up 40% year-over-year and 11% above last quarter, growing at a strong deliberate but measured pace. Noncredit revenues hit BRL 1.9 billion up 57% year-over-year and 19% higher sequentially. This is fees, commissions, float, hedge accounting, insurance and acquiring, all capital-light, all compounding quarter after quarter. Three engines, three growth vectors and each one getting stronger.

On returns, let me walk you through the two charts on this slide. First, adjusted net income, BRL 283 million, up 135% year-over-year and 67% sequentially. This represents a significant acceleration in profitability as we scale the business. Second, adjusted ROE, 20.2%, up from 15.5% in the previous quarter. Both metrics benefited from the positive impact of Lei do Bem, our R&D tax incentive program, which contributed to the strong quarterly performance.

Moving to credit. PicPay card TPV was BRL 19.5 billion, up 40% year-over-year and 12% sequentially. Card engagement continues to deepen as our maturing vintages drive higher spend per user. Consumer loan origination reached BRL 4.8 billion, up 78% year-over-year and 7% above last quarter. Total credit portfolio reached BRL 31.9 billion, up 99% from a year ago and 14% higher sequentially. The consumer book represents 93% of the total with SMBs and others comprising the remaining 7%.

On our audiences and ecosystem business unit, we've built a portfolio that lets our users solve most of their daily needs within PicPay. More reasons to use that every day drives higher engagement, which creates opportunities to cross-sell financial products and increase customer lifetime value. From shopping and food delivery to travel, entertainment, telecom and urban mobility, we cover the key journeys of everyday life.

One standout example is iGaming. In just 1 year, we built a high-margin business with over 2.7 million clients across lucky numbers, national lotteries and themed World Cup games, all integrated into our ecosystem. This broader everyday ecosystem increases our relevance, deepens engagement and strengthens the financial relationship with our customers. On our small and medium businesses segment, we are seeing real momentum across the board. New small and medium business accounts, reached 85,000 per month in the first half of 2026, up from 27,000 in the first half of 2025, a threefold acceleration. Supply chain finance is scaling fast.

Origination hit BRL 1.05 billion in the quarter from BRL 40 million in the last quarter of last year and BRL 693 million just a quarter ago. The trajectory is clear and the unit economics are attractive. We're also rolling out tap on phone to individual consumers turning 70 million paid users into potential merchants. It's a distribution play that uniquely positions us in the payments value chain. And we just launched our marketing AI agents. SMBs now can create self-serve ads and our platform identifies the most relevant customers within the merchant geographic footprint and delivers the ads to them. First week results, 10,000 [ options ] 1,500 campaigns and 1.7 million individuals reached, AI-powering SMBs to boost sales through our base of more than 70 million customers.

Danilo, please tell us more about our highlights on consumer finance products.

Danilo Caffaro

Thanks, Eduardo. I'm pleased to share an update on our progress and priorities. Our focus remains simple: serve customers well, build products, people value and grow with discipline. Our day-to-day banking business continues to evolve, reflecting growing customer trust and deeper engagement across payments, credit and everyday benefits, supported by disciplined execution, thoughtful risk management and a strong customer experience.

Our investment platform now offers more than 280 products, including investment funds and fixed income. We also launched a brokerage platform that allows customers to buy and sell stocks through our app. We are gradually rolling out the Epic segment to existing customers. The offer reached 23% penetration of the eligible base this quarter. Epic credit cards account for 14% of total card TPV and 80% of the user base is actively using benefits such as Amazon Prime, Einstein telemedicine and same [ Podar ] toll tags.

In Brazil, convenience matters. Whether paying a bill, using telemedicine or passing through a toll, the experience should be quick and reliable. AI agents are also becoming central to our strategy. We are the first Brazilian bank with an official plug-in in both the Claude and OpenAI ecosystems. We are also rolling out second-generation WhatsApp and in-app agents with more tools, memory, Internet access and sequential multistep execution. This reinforces our [ Atlas ] strategy, solving broken journeys wherever our users need us. with contextual and relevant products and services.

Turning to credit. We continue to gain market share by increasing our share of wallet across the products used by our customers. We reached 6.4% in private payroll loans, 2.8% in personal loans, 1.6% in card TPV and 1.2% in the credit card portfolio. We still believe we have significant room to grow.

Moving to portfolio growth. Our credit portfolio grew BRL 3.9 billion in the second quarter. 86% of that growth came from lower-risk loans and mature credit cards. New cards also almost doubled their contribution to portfolio expansion compared with last quarter, reflecting our progressive limits approach and the maturation of newer card cohorts.

Moving to underwriting strategy and cohort performance. We continue to execute our underwriting strategy across two complementary objectives: performance optimization and growth optimization. Progressive limits are becoming a larger share of the portfolio as the cohorts mature. NPL creation in the credit card portfolio is trending better than in the same period last year across both strategies and remains relatively stable versus recent quarters even after considering seasonality. Cohort performance across both strategies has remained relatively stable in recent quarters, reflecting the resilience of our models and our active risk management approach.

In private payroll loans, we resumed increasing originations in growth clusters after regaining confidence in the product's operational maturity and implementing new features since the fourth quarter of 2025. This is increasing the growth strategy mix. Newer cohorts reflect the deliberate incremental risk assumed to accelerate growth while remaining within our approved risk appetite and targeted risk-adjusted returns. Although we see no relevant early signs of credit deterioration within the same risk groups, we expect portfolio indicators to reflect additional intentional risk taking in private payroll loans and cohort aging and maturation in the coming quarters. These indicators include 90-plus NPL Stage 3 and cost of risk as a percentage of the total portfolio.

As new originations become a smaller share of the outstanding portfolio, their dilution effect on these metrics will naturally decrease. Cazotto will provide further detail on these dynamics in the next session.

Now a deeper dive into our private payroll loans operation. We reached a portfolio of BRL 7.2 billion this quarter with more than 3.6 million contracts and well-diversified employer risk. Expected marginal ROEs remain attractive supported by risk-adjusted pricing and credit-related revenues. We are also seeing better ARPAC and cross-selling indicators for these clients, supporting other revenue streams. We remain confident in our ability to scale this operation with healthy ROEs and risk-adjusted returns.

Now I will pass it to Andre Cazotto, our CFO, to cover our financial results.

Andre Cazotto

Thank you, Danilo. Now let me go over the evolution of our delinquency metrics and explain the dynamics behind these curves. On the left-hand side, we show our early NPL, define it as loans between 15 and 90 days past due. After reaching 8.4% in the first quarter, early NPL improved to 7.5% in the second quarter. A quarter-over-quarter reduction driven by a favorable seasonal effect in the period, combining with improving performance in more recent vintages.

On the right-hand side, NPL over 90 days increased to 9.8% in the quarter while Stage 3 reached 12.9%. These two metrics need to be interpreted together. NPL over 90 days is fully captured within Stage 3. Meaning the loans driving that metric are already classified as credit impaired and provisioned accordingly. Stage 3 is the broader classification as it also encompasses other [ Credipar ] exposures that may not be at be more than 90 days past due, but have already been identified as deteriorated.

In other words, there is no additional credit risk sitting outside Stage 3. It's all already recognized it and provision it within that bucket. The increase in these later-stage metrics is primarily driven by portfolio aging. As our products and vintages mature, a larger portion of the book naturally migrates into later stages of delinquency. A mechanical and expected dynamic in a rapidly growing portfolio, not a sign of deterioration. It's also worth noting that these metrics will continue to be influenced by our deliberate strategy of intentional risk taken in private payroll loans. A conscious portfolio decision where we are comfortable assuming higher delinquents in exchange for meaningfully better risk-adjusted returns over the life of the product.

As the portfolio continues to season and this strategy matures, we expect both NPL over 90 days and Stage 3 to gradually converge toward a more stable level. It's also important to highlight that Stage 3 portfolio is already more than 75% provisioned reflecting a robust level of coverage against expected losses and reinforcing the adequacy of our provisioning framework.

Moving to the next page. This slide breaks down the key drivers behind the sequential movement in both NPL over 90 days and Stage 3 from the first quarter '26 to the second quarter '26. Starting with the NPL over 90 days, which moved from 80.9% to 9.8%, a net increase of 93 basis points. The primary driver was portfolio aging which contributed to 318 basis points, reflecting the natural seasoning of earlier vintages flowing to later delinquency stages. This was partially offset by the [ Disney Holo ] program, which contributed in 117 basis points improvement.

Seasonality added 50 basis points, consistent with typical patterns for the period. It's also worth noting that lower pace of new originations relative to prior years, generated a smaller dilution effect on the metric. Meaning the denominator grew less rapidly contributing to the upward pressure on the ratio. Product mix and other factors contributed modest offsets of 34 basis points and 7 basis points, respectively. For Stage 3, which moved from 12.7% to 12.9%, a net increase of only 27 basis points. The drivers are broadly similar but with one important distinction.

Aging contributed 184 basis points, materially lower impact than the 318 basis points observed in NPL over [ '19. ] This is not a constant. Stage 3 is a pre-NPL metric capturing credit deterioration earlier in the cycle. As a result, the aging dynamic that is still fitting NPL over 90 days has already been partially absorbed in the Stage 3 in prior quarters resulting in a lower incremental aging effect. Seasonality added 41 basis points, while the [ Desal ] program offset 46 basis points. Origination offset 117 basis points and product mix and others provided additional offsets of 26 and 9 basis points, respectively.

Taken together, these waterfall charts reinforce our earlier message. The NPL dynamics we are observing are mainly driven by vintage maturation seasonality in our deliberate strategy of intentional risk taking private payroll loans and not by a deterioration in the underlying quality of our portfolio.

Moving to the next slide. On the left-hand side, Stage 2 plus Stage 3 formation continue to improve. Declining to 4.9% in the second quarter compared to 5.1% in the previous two quarters. On the right-hand side, Stage 3 formation declined to 3.65% in the second quarter. from 3.9% in the previous quarter. This improvement was mainly driven by the impact of the [ Desire ] negotiation program. Most of the loans renegotiated under the program were still on our balance sheet as were less than 360 days past due.

The renegotiated exposure totaled approximately BRL 520 million on a gross basis. Considering an average discount of approximately 50%, the outstanding balance was reduced by around BRL 260 million. This reduction directly lowered the balance contributing to Stage 3 formation and was, therefore, the main factor behind improvement in the ratio to 3.65%. Excluding the impact of the [ inhale ], Stage 3 formation would have been around 4%, broadly in line with the previous quarters. This underlying level also reflects the natural aging of the portfolio. Our products and vintages continues to mature.

Now let me walk you through the portfolio classification by stage and our coverage levels. Stage 3 remained stable at approximately 13% of the total credit portfolio in the second quarter. In terms of coverage, we continue to see comfortable levels with coverage for Stage 2 plus Stage 3 at 62.7% and Stage 3 coverage at 74.1%. Stage 3 coverage decreased from 77% in the first quarter to 74.1% in the second quarter. This reduction was primarily related to the [ Disinhala ] renegotiation program.

Loans renegotiated under the [ Sinhala ] benefit from FGO guarantee, the operations guarantee fund covering 50% of the outstanding exposure. This guarantee increases the expected recovery on these loans and consequently reduces the LGD applied to those exposures. Since a lower LGD results in lower provisional requirements, the inclusion of these loans mechanically reduced the overall Stage 3 coverage ratio. Therefore, the reduction from 77% to 74% does not reflect the deterioration in portfolio quality a change in our provisioning standards or a change in our risk appetite. It's primarily a mix effect resulting from the lower LGD of the [ Sinhala ] portfolio supported by the FGO [ GGR ].

As this effect normalize, we expect Stage 3 coverage to move back toward the high 70% range in the coming quarters. On credit risk management, our three key metrics, loss absorption, cost of risk and portfolio coverage, collectively paint a picture of well-controlled and increasingly well provisioned book. Our loss absorption ratio reached 56.5% in the second quarter, comfortably within our internal guidelines of 40% to 60%.

Moving to quarterly cost of risk which came in at 3.9% in the second quarter. We think the 3.7% and 3.9% guidance range we provided at the beginning of the quarter. The sequential increase from 3.7% in the first quarter was primarily driven by the natural aging of our private payroll one portfolio as earlier vintages continue to season and flow through the provisional cycle, a mechanical and expected dynamic given the rapid growth of this product over the past several quarters. This increase was partially offset by a BRL 59 million positive impact from the [ Sinhala ] program, which represented approximately 5% of our total cost of credit in the quarter.

Finally, on credit loss allowance expenses and total coverage, CLA expenses reached BRL 1.2 billion in the second quarter, up from BRL 974 million in the first quarter, consistent with the pace of portfolio expansion. More importantly, total portfolio coverage held stable at 13.9%, unchanged from the prior quarter, reinforcing the adequacy of our provisioning levels as the book continues to scale. The combination of stable coverage and growing absolute provision balances reflects a disciplined and consistent approach to credit risk management.

Moving to the next slide on operating leverage. The trend speaks for itself. Net revenues reached BRL 4.1 billion in the second quarter, up 17% quarter-over-quarter and 67% year-over-year compared to BRL 2.5 billion we reported in the second quarter of last year. Over the same period, adjusted operating expenses, which exclude stock-based and expenses, grew to BRL 955 million, increasing a fraction of the pace of the revenue growth. The result is a continued and consistent improvement in our adjusted efficiency ratio, which declined to 44.8% in the second quarter, down from 46.9% in Q1. A 210 basis point sequential improvement. A key driver of this dynamic is AI. Its impact on our operations is already tangible and measurable.

Our head count has been flat since October 2025 and the projected 10% increase we had originally anticipated for 2026 will not materialize. Productivity gains are translating directly into margin expansion rather than incremental hiring. We expect AI to be a major and accelerating driver of our operational leverage going forward, making the efficiency trajectory you see on this slide, not a ceiling but a floor.

Moving to financial margin expansion. Net interest income reached BRL 2 billion, up 18% quarter-over-quarter and 65% year-over-year. Our net interest margin came at 19.4%, growing from the 18.7% reported in the first quarter. Margin from credit products reached BRL 2.1 billion, growing 18% sequentially and 81% year-over-year. This metric captures the full economic contribution of our credit operations including revenues from products directly tied to the credit origination such as credit card interchange and credit insurance, while excluding cash remuneration and derivative revenues providing a cleaner view of the true margin generated by our lending activity.

Net interest margin from credit products came in at 27.8% and growing from the 27.2% reported in the first quarter. Equally important is the margin from credit products after losses, which reached BRL 980 million in the second quarter up 14% quarter-over-quarter and 68% year-over-year. The net interest margin after losses held stable at 12.1%.

Moving to funding on the next slide. Our funding base grew 10% quarter-over-quarter, reaching BRL 35.8 billion in the second quarter, up 45% year-over-year from BRL 24.8 billion in the second quarter, '25. The modest sequential increase in the cost of funding from 94% to 96.2% of CDI is largely explained by the issuance of our new [ Fiji ] in May 2026. A securitized structure, backed by our FGTS portfolio through which we raised BRL 1.2 billion.

More recently, in July and August, we executed additional capital markets transactions. Raising funds through promissory notes and debt security issuances, consistent with our strategy of continuously diversifying our funding sources. These transactions further strengthen our balance sheet and enhance our capacity to sustain the rapid growth of our credit portfolio in a disciplined and cost-efficient manner. We will continue to mobilize multiple funding channels, spanning digital platform deposits third-party platforms, creditory funds and capital markets instruments, actively seeking the most efficient funding alternatives available to support our growth ambitions.

On the capital side, we maintain a solid capital position with a total capital ratio of 17.6% and a common equity Tier 1 ratio of 15.6% in the second quarter. It's worth highlighting that approximately BRL 450 million, equivalent to roughly 1.7 percentage points of our total capital and common equity Tier 1 ratio remains held at our holding company in the Netherlands and has not been injected in the operating entity. With the acquisition of cover now closed, we expect the capital consumption of approximately 150 basis points in Q3 even after absorbing this impact, we remain comfortably above our internal capital appetite thresholds.

And we expect to close the year with a total capital ratio of approximately 14% and a common equity Tier 1 ratio in the 12% to 12.5% range. Levels that provide meaningful headroom above regulatory requirements and fully support our growth ambitions.

Finally, on the next slide, we're now providing guidance for the third quarter of 2026. As with our previous guidance, these figures reflect PicPay's stand-alone operations and exclude any contribution from [ over ]. We expect our total credit portfolio to reach approximately BRL 34.7 billion. Quarterly cost of risk is expected to remain within the 3.9% to 4.1% range.

On the revenue side, managerial revenues are expected at approximately BRL 4 billion, and net interest income should reach approximately BRL 2.1 billion. Gross profit is guided at approximately BRL 1.3 billion. On profitability, we expect strong pretax earnings expansion. IFRS earnings before taxes is guided at approximately BRL 360 million. 34% higher sequentially and adjusted EBITDA at approximately BRL 378 million, up 30% from the second quarter 2026.

The net income level, however, it's important to provide context on the sequential dynamics. IFRS net income is expected at approximately BRL 255 million down 5% sequentially and adjusted net income at approximately BRL 265 million, 6% below the second quarter. This decline is not driven by any operational deterioration, quite the opposite. The second quarter, we benefited from a significant positive impact from the late [ Ben ], a recurring tax incentive that this year was heavily concentrated in the second quarter, materially reducing our effective tax rate in the period. In Q3, our effective tax rate normalized back to levels consistent with the first quarter of the year.

With that, I will now hand the call back to Eduardo Chedid for his closing remarks.

Eduardo Simoes

Thanks, Cazotto. Before my closing remarks, I want to highlight a milestone that deserves attention on its own. After obtaining approval from the insurance regulator, the antitrust authority and the Central Bank, the acquisition of cover was finalized on August 3. This is not just an M&A transaction. It's a strategic acceleration of our insurance ambitions. Cover brings a full-service insure platform with over 100 products a senior executive team with over 20 years of track record in insurance and established distribution channels that complement our own. The economics are compelling, and we expect a meaningful incremental contribution to PicPay's bottom line from August to December 2026.

But what really excites me even more is the strategic fit and opportunities in the coming years. For the insurance products that cover sales through our channels, we will now capture the full economics and be able to develop more customized products for our client base. Furthermore, around 70% of cover's business is done with high-quality partner distributors, and we expect that channel to keep delivering. With cover, we now have the product development speed the underwriting expertise and the distribution reach to turn insurance into an even more meaningful recurring earnings stream. We're maintaining covers independence and strengthening its partnerships. This is just the beginning of a new phase, and it's already marked by a change of brand.

Cover is now CAV. Let me leave you with 6 points to summarize where we stand. First, the [ macro ] -- while delinquency remains elevated, recent [ bans ] point to stabilization. The economic scenario continues to offer important support for credit quality. The labor market remains highly resilient with unemployment near historic lows and more than 103 million people employed.

Real wage income reached approximately BRL 380 billion as of June up 3.6% year-over-year. Notably, net formal job creation was concentrated in income brackets earning up to 2x the minimum wage with more than 160,000 new positions generated in that segment while economic activity is showing gradual deceleration as expected under contractionary monetary policy growth remains positive. Our scenario does not contemplate an abrupt employment deterioration, but rather a progressive normalization the labor market at historically strong levels. This combination of elevated employment resilient income and moderate [ pension ] reduces the risk of a systemic deterioration in households repayment capacity and positions PicPay well for the quarters ahead.

Second, asset quality. Our portfolio remains resilient by design, supported by greater exposure to secure and partially secured products disciplined underwriting and robust risk management following our credit fundamentals of a balanced portfolio loss absorption ratios between 4% and 6% and ROEs above 30%. The increasing NPL over 90 days reflects portfolio aging and intentional risk taken in payrolls not deterioration. Early delinquency improved to 7.5%. Cover ratios are robust and the underlying quality of our origination remains strong. All of this while maintaining the same risk appetite and targeted risk-adjusted returns.

Third, private payroll loans. This product is scaling with attractive economics. We have reached more than 3.6 million contracts since inception very healthy marginal ROEs and stable over 30 days NPL metrics on both the standard and the growth portfolios. That is supporting profitable growth in partially secured lending. Fourth, noncredit revenue. It's up 57% year-over-year, underscoring the strength of our broader platform monetization beyond credit-related revenue streams. Fifth, small and medium businesses. This segment is gaining scale, relevance and customer traction with increasing potential to contribute meaningfully to future growth. And sixth, cover, the acquisition accelerates our insurance ambitions, creating opportunities to devalued products and penetration, capture additional economics within our customer base and also through distribution partners. It should unlock a meaningful and recurrent contribution to earnings growth.

Finally, we beat guidance on all major metrics this quarter. We are confident in our ability to deliver sustainable, profitable growth and create long-term value for our shareholders. We'll move with urgency but never at the expense of quality or trust. Thank you, and we'll now open the line for questions. Operator?

Operator

[Operator Instructions] The first question comes from Mario Pierry from Bank of America.

질의응답

Mario Pierry

Let me ask you 2 questions. First one on the [ Zinhola ]. I think you made it clear, right, the 117 basis points benefit to NPL and about a 5% reduction in the cost of credit. So that's about BRL 59 million. Were there any other benefits from the [ Zinhola ]? And my understanding is that the program was extended, right? So should we expect further benefits in the third quarter from the debt renegotiation program? And then I'll ask the second question later.

Andre Cazotto

All right. Thanks, Mario. Thanks for your question. Let me try to reinforce the messages that we just shared in our conference call. So yes, in terms of cost of risk, the [ this ] generated a positive impact of approximately BRL 59 million which is equivalent to around 5% of our total cost of credit in the quarter.

On the NPL over 90 days, the program reduced the ratio by approximately 117 basis points partially offsetting the impact from portfolio aging and the seasonality. This also helped the Stage 3 formation, just like we said in the conference call, approximately BRL 20 million of loans were renegotiated on a gross basis, applying an average discount of 50% basically reduced the outstanding balance by around BRL 260 million directly lowering the balance contributing to stage reformation. So as a result, the ratio declined from 3.9% to 3.65% excluding the [ Zolastage ] formation, let's say that the ratio will be close to 4%, broadly in line with previous quarters. For Q3, yes, we are expecting some additional positive impact from this [ enroll ], but more limited. I think that we had a much higher impact in the second quarter.

Mario Pierry

Okay. I think that's clear. Now one thing that surprised us on the results was the funding cost it came a little bit higher than what we had in our models. When we look at your deposits, right, they're growing slower than your loans, and you talked about you are issuing other sources of funding. Can you just talk a little bit about the ability to continue to grow deposits from your existing clients?

Andre Cazotto

Mario, thanks for your question. Yes, I think we're very comfortable with the level of deposits that we're capturing in our platform. We are expanding our capabilities. Like we said in the conference call, the slightly increase in our cost of funding from 94% to 96.2% is primarily driven by the issuancy of our [ fiji ] that's basically backed by our FGTS portfolio. We're also accessing other funding capabilities in capital markets.

We are increasing, let's say, the principality of our customer base. So our platform more and more is getting more transactional. We continue to grow the cash in around 20% year-over-year. So more and more, we are converting more cash into deposits. So we feel very comfortable and keep growing our deposit franchise going forward. We're expecting to keep, let's say, the funding cost around 95% of CDI in the coming quarters. So we do believe that we can continue to grow access new lines of funding and still deliver a very healthy cost of funding in our operations.

Operator

Our next question comes from Dan Dolev from Mizuho.

Dan Dolev

Great results. Congrats Cazotto on the new role. Very much looking for you in this role.

I have a question on AI. You showed a lot of very exciting products on the genetic side or AI. Can you maybe, Cazotto or Eduardo, can you comment a little bit about those products, we found it very interesting.

Eduardo Simoes

This is Eduardo. I could comment, but I think that I'll pass that to Danilo as he is actually heading the AI initiatives here, who'll be able to give you a more, let's say, a deeper understanding on all the dimensions we're going through.

Danilo Caffaro

Well, first of all, I think we have -- our AI strategy is actually tacked on 2 different pillars. The first one is customer-facing products. The second one, I comment a little bit is around our AI operational capabilities. So for the agents that we mentioned throughout the presentation was more focused on the first one, right, for the customer-facing products. And that's the ones that we have the goal actually to empower our customers anywhere they need it.

So we mentioned the agents for consumers that are actually on our second generation, and they are actually able to not only answer questions, but actually execute tasks like pay bills, same PIX transactions, managed savings, renegotiations and so on. All of that, of course, with user confirmation in every transaction. But it's actually more than 70 tools now that we are releasing to our customers. And the agent is actually capable of executing multiple sequational tests for the customer, right? And then also leads now multiple channels.

So not only on our app, but also on WhatsApp. And as we mentioned, we were the first Brazilian bank to actually have it available on both Tropic and OpenAI official pllug-in store, our Plugin. So that's for the consumer. We also mentioned around small and medium business agent -- marketing agent that is actually able to create and distribute ad campaigns for our small and medium business clients to our customer base based on geolocation of the business. That's the one that we just launched, and we shared some of the first week results. But we also have agents for our internal operational capabilities, right?

So from the beginning of the year, what we did, we actually developed our own proprietary platform, our AI harness around some of what we think is the key in order to extract value of the -- for the AI agents. So we built a platform that has our model routing, cashing, a lot of governance layers. And because of that, we were actually able to reduce our token costs by 70% from the beginning of the year to now.

And that actually enables us to maintain access to the best frontier models without scaling total token costs because of that, right? And we are using different areas. So we are using credit. So we're just rolling out our proprietary foundation model for personal loans underwriting. That's a model that we expect to have something around 15% to 20% benefit from the previous one. We also have agents and people using our internal platform for product development. So they are supporting coding, designing, quality assurance.

Nowadays, approximately something around 90% of our employees are actively using our AI platform with most of them using daily. -- we have like from the beginning of the year, we are up 30% of employees that are contributing with deployments and real deployments for prepaid products. And most of them such as myself, wouldn't be able to contribute without AI, right, without actually coders, but now that's possible. So we are having more and more people contributing with real products.

And the number of deployees actually doubled from the beginning of the year because of that productivity. And we are doing some very good stuff internally as well, and we hope to benefit from that also on leveraging our operational efficiency.

Operator

Our next question is from Gustavo Schroden with Citi.

Gustavo Schroden

Congratulations on another quarter of solid results and congratulations Cazotto for the new role. Let me concentrate the first question that I have on the private payroll loan. You've continued to grow this product at a very strong pace. Even though it's considering a secured product, we've seen a delinquency trends worsening and in some cohorts, the NPL ratio is starting to get closer to what we see in unsecured personal loans. It is according to the Central Bank data, right? So I mean, how are you thinking about the risk reward balance in this private [ per ] loan today and are you considering being more selective or adjusting your risk appetite in this product going forward.

So -- and why do you think that the consolidated data from the Central Bank is pointing to this faster deterioration in this product? So this is my first question and then I do my second question later.

Eduardo Simoes

Gustavo, it's Chedid here. First of all, I think that we haven't seen any deterioration at the risk at the same risk profile. What we have seen in our case, it's there in the presentation, is that we're actually opening intentionally to riskier profiles, which, on average, you will see the NPLs going up, but not a deterioration on the same risk profile.

If we look forward, and that's -- I'd say that this is kind of philosophy we've been adopting for all [ products ]. Every gain that we are actually getting from our new models, we're actually not, let's say, deploying that into further growth but basically maintaining origination, but with the gains of the models so that asset quality remains, let's say, in control. If you look at the Central Bank data, I think it also is a reflect of -- if you look at the previous product, the only -- let's say, it only cater for very large companies.

Now that this is a product that -- and that's mainly due to the new way of doing it. People are actually extending that to also smaller companies. And that's a benefit of the centralized system. So as you are actually getting more companies and more employees of those, let's say, smaller companies. It's very hard to compare the previous product with what you have now going on.

Andre Cazotto

Sorry, a just to complement here on this delivery strategy of taking incremental risk in very specific and selected customer segments. It's very important to highlight that our risk framework remains unchanged. So we continue to target the loss absorption ratio between 40% to 60%, and our ROE is above 20%. So that's very important to highlight.

Gustavo Schroden

Okay. Cool. Just a follow-up here -- 2 follow-ups on this private payroll loan. Have you seen an improvement on the operational issues that we saw in a few months ago? And if you -- I mean, if you can share with us what is the cost of risk level that we have in this product?

Eduardo Simoes

Gustavo, on the operational issues, I'd say that we went through kind of three stages, right? So the first stage where we had huge operational issues. In the beginning, we were seeing FPDs around 17%, then we went through a cycle, a second momentum, basically where we've diminished originations very much so that we could see the operational issues being solved. Some of them were solved by the centralized system. Some of them were soft by workarounds that we have implemented ourselves. At the same time, we also -- I think we're in the third or fourth different concept for the evolution of the concession model, which also helped us on basically getting to the third phase, which is expanding the product. If you're mentioning any, let's say, large games from core to this one, I wouldn't say that. And if we look at guarantees as well as the automatic [ tankage ].

Payroll we linkage. We're still not underwriting as if they were meaningful. And so that means that we still think that those were not meaningful enough so that we could take that into consideration.

Andre Cazotto

And in terms of the cost of risk, it's basically in line with other, let's say, public peers that published this number recently. So we can say that around mid to high teens in annual basis.

Gustavo Schroden

Just still on asset caller, just to finalize here. What are your expectations for the trajectory of NPLs stage -- sorry, 90 days NPLs and Stage 3 over the next quarter. So should we expect some further normalization as the portfolio matures? Or do you believe current levels are already broadly representative of the underlying credit performance?

Andre Cazotto

Let's say that we are still expecting NPLs to continue to be impacted by the aging effect, right? So we are expecting by the end of this year, the NPLs over 90 days could be more around, let's say, low teens. So basically converting to something similar that we have on our Stage 3 over total credit portfolio. Remember that the Stage 3 is a pre-NPL metric and it's pretty much absorbing, let's say, all the credit impaired than we had in the model.

So basically, we believe that could convert to a level similar to what we have currently on Stage 3 over the total credit portfolio by the end of this year. But again, we are not seeing deterioration. It's basically the portfolio aging. The growth that we have on the payroll loan that is still maturing. We were, let's say, earlier doctors of this product, probably the second company prepared to operate private payroll ones in Brazil. So naturally, that portfolio continues to age and impact this metric going forward.

Operator

Our next question is from Ricardo Buchpiguel from BTG Pactual.

Ricardo Buchpiguel

Hi, everyone. Thanks for the opportunity of making questions. I have just one follow-up here on private payroll. With the increasing concerns about the macro environment, higher unemployment has been -- become an important risk we have been discussing with some investors on private payroll loan, particularly because it has like a higher duration.

So could you comment to what level of unemployment would you become more concerned about the profitability of the product? What would be like a more -- a level where the profitability would be closer to breakeven in your view, depending on a rise in employment?

Eduardo Simoes

Okay. Thanks for your question. I think that I'm going to answer, I would say, not as you were expecting, but trying to get to the same answer, right? So if you look at our credit approach, it's primarily based on the loss absorption indicate, right? So for private payroll loans, in order for those vintages to breakeven, we could withstand an increase of up to 70% in the products delinquency rates. So meaning that my expected losses could actually grow 17% and I will still be on a breakeven condition.

Now talking about the unemployment. And if you look ahead, market consensus and focused projections currently -- they actually expect unemployment to remain pretty stable and quite healthy and basically growing from 5.4% to around 6% and through 2027, which kind of reinforces our evolved structure of floor under household income rather than any sudden labor market deterioration.

Even if we look at the most pessimistic scenarios in the focused survey, unemployment will peak at levels back around what we saw through 2024. Something between 6.5% and 7.2%, meaning that even under stress scenarios, we're talking about historical levels that didn't mean a heavy deterioration on credit or households or household repayment.

Obviously, we keep dynamically looking at those projections. And as I told you, we are currently using gains from our concession model, more to actually keep the levels of originations than actually growing originations. So that's how we feel about it.

Ricardo Buchpiguel

That's super clear. And if I may do a second question, if you could comment what's your expectation for the bottom line in 2026 now that you have -- that they will be consolidating cover, any sense on how much cover could eventually contribute in the second half of the year will be very helpful for us here?

Eduardo Simoes

Okay. Let's talk about the cover acquisition, right? And cover now it's called [ CAF ]. So [ CAF ] will be consolidated from August 3. Our expectation is something around -- something between BRL 80 million to BRL 100 million in net income contribution for the August, December period. So that's basically what we are sharing on cover for those 5 months of the remaining of the year. And well, we also share third quarter guidance. So that's about what we can share right now.

Operator

Our next question is from Dan Perlin from RBC.

Daniel Perlin

I just had -- I had a little bit of a follow-up on the ARPAC. It remains very strong here again and your monetization rate continues to improve. I wonder if you could just kind of revisit the strategy like the go-forward strategy and maybe how some of that dovetails into the product road map and mix shifts that you're seeing in the business, clearly, it's moving in the right direction, but I'm just making sure I understand the cadence as to how that progresses from here.

Eduardo Simoes

So I think it's more or less the same story moving forward. So it is still basically driven by more penetration of our products and mainly that instead of being new clients, but heavily concentrated on cross-selling those products and mainly, let's say, credit and insurance products into our user base. This is what's primarily driving growth in ARPAC at the same time, you can see that our cost to serve is growing at a much lower pace, growing at 52%, while you have cost to serve growing at a 13% rate year-over-year.

And most of that growth in cost to serve mainly driven by the adoption of new profit. So I'm just trying to give you other proof points that this is what's actually driving all of that RPO growth. If you look at more mature cohorts that you will see also ARPAC more than doubling, if you compare to the average ARPAC, which just reinforces the thesis, which is cross-selling more of those products, especially credit products will be the key driver for further increasing ARPAC ahead.

Daniel Perlin

Great. And then just real quickly on [ cover ], I heard you on the contribution from August to December in terms of net income. Is there just -- is there a revenue number that you're also attributing to that, that we could just make sure we're level setting appropriately in the model?

Eduardo Simoes

Maybe we can share that with you everyone else. [ Leer ] not -- we don't have that figure right now, but we can share later.

Operator

Our next question is from Neha Agarwala from HSBC.

Neha Agarwala

Hi, I actually have three questions, quick ones. First one on the operating expenses, there was a bit of a jump in 2Q, I believe there were some extraordinaries some extra marketing expenses that you undertook in 2Q and your guidance. implies the sequential decline in 3Q. Could you just shed a bit more color on the trend for OpEx growth that we should expect going forward? And what were the one-offs in 2Q?

My second question is on risk-adjusted margins. So on the reported numbers, it went down 20 basis points. But if you adjust for the [ Desenrola ], benefit it probably is around 11.3% in risk-adjusted margins. What should we -- what trajectory should we assume in the coming quarters and where should this is adjusted margin stabilize for you? And my third question is on the write-off policy. Could you remind us right of policies? And has there been any change lately to that.

Andre Cazotto

So Neha, thanks for your question. Let me start from the last one. Our write-off policies remained unchanged, 360 days for both credit cards and personal loans. On the risk-adjusted NIM, we're expecting stabilization for Q3 compared to what we did in the second quarter, around 12.1%. And in terms of efficiency, we had anticipation of BRL 30 million in marketing expenses this quarter. We decided to anticipate because of the workout, we took a decision to accelerate some initiatives on marketing for very specific products like the iGaming platform they have and other initiatives that we saw the opportunity to accelerate.

So we should expect some, let's say, better benefit from lower marketing expenses compared to the second quarter. In terms of the overall efficiency, we are in the very health trend. As you can see in the quarter, our efficiency ratio reached around 44%, coming down more than 200 basis points sequentially. We're expecting that trend to continue going forward. We are seeing AI accelerating our operating leverage opportunities.

Head count is pretty much flat since October '25. If you remember, we were expecting to grow head count by around 10% the year. It's not happening because of AI and all the initiatives that we have. So we believe that we can deliver our efficiency ratio around low 40s 30s by the end of this year, contemplating many initiatives that we have including AI opportunities on the let's say, personnel expenses, but also on tech expenses as well.

Neha Agarwala

Perfect. I just have a quick follow-up there. On the risk-adjusted margins, you mentioned you should expect to be around 12.1%. So if you exclude the [ Desenrola ] benefit from my calculations, it's around 11.3%. So you expect a rebound in 3Q and for it to stay around the 12% range. Is that right?

Andre Cazotto

Correct. We're expecting in Q3, risk-adjusted means to being the same, let's say, pretty much flattish sequentially. We do have some impact from this [ hall ] in Q3 as well. But like we said a bit more limited compared to the second quarter. But yes, we are expecting this ratio to be around 12%, 12.1% pretty much in line with the previous quarter.

Neha Agarwala

And what would be the driver for that? Because based on your guidance, cost of risk will continue to enter up quarter-on-quarter. So what would be -- and deposit costs will probably be around the same level, not much improvement based on your comments earlier. So what would be driving the improvement of the stability in NIMs, right, excluding the [ Zennoa ] impact?

Andre Cazotto

Yes. Basically, a mix effect. We are, like we said, growing slightly lower on board, let's say, segments that are okay in terms of capturing incremental risk. So we do have this risk-adjusted strategy in the business. Funding costs should be slightly lower compared to this quarter. So we did it 96.2% of CDI in the quarter. We're expecting probably Q3 to be more in the range of 94%, 95%. So it's another, let's say, improvement that we can see on this risk-adjusted earning.

Operator

Our next question is from Craig Maurer from FT Partners.

Craig Maurer

Again, congratulations, Andre. I just wanted to ask with the growth you're seeing in payroll loans, what are the attach rates you're seeing in other products once you've made those loans, credit cards, other offerings?

Eduardo Simoes

This is Eduardo here. First of all, if you look at our, let's say, the average ARPAC on clients with that product. It's 8.9x higher than of the average client in PicPay. And that's driven by both the product itself, but also, let's say, the attachments, as you said, that he basically gets it. And if you look at the gross selling, it's 30% higher at what we have on for average customers. And that includes many different products but on average, 30% higher than our average client. So it's also, let's say, it poses not only ARPAC part but also adoption of other products.

Operator

[Operator Instructions] The question-and-answer session is over. We would like to hand the floor back to Mr. Eduardo Chedid for the company's final remarks.

Eduardo Simoes

So guys, thanks again for being with us in our third call. We remain confident here on the year-end results. We finalized the Cover acquisition, which was an important milestone, not only for this year, but for the coming years as well. And well, let's see if we can surprise you next quarter again. Thank you.

Operator

PicPay's conference is now closed. We thank you for your participation and wish you a nice evening.