Cuộc họp công bố kết quả kinh doanh Q2 2026 của Nu Holdings (NU): Lợi nhuận ròng đạt 1,1 tỷ USD

Nu Holdings (NYSE: NU) báo cáo quý đầu tiên đạt lợi nhuận ròng hơn 1 tỷ USD, nhờ tăng trưởng dư nợ cho vay, biên lợi nhuận điều chỉnh theo rủi ro mở rộng và đòn bẩy hoạt động tiếp tục phát huy hiệu quả. Ban lãnh đạo cũng nhấn mạnh mức độ gắn kết sâu hơn của khách hàng, giấy phép ngân hàng mới tại Mexico và việc mở rộng sử dụng nền tảng AI nuFormer.

Điểm tin chính

• Lợi nhuận ròng đạt kỷ lục 1,1 tỷ USD, tăng 49% so với cùng kỳ năm ngoái và 17% so với quý trước. Tỷ suất lợi nhuận trên vốn chủ sở hữu (ROE) duy trì ở mức 33%.
• Tổng doanh thu tăng 39% so với cùng kỳ năm ngoái lên gần 5,9 tỷ USD, trong khi lợi nhuận gộp tăng 43% lên 2,4 tỷ USD.
• Danh mục tín dụng tăng 37% so với cùng kỳ năm ngoái và 5% so với quý trước lên 39,4 tỷ USD. Dư nợ thẻ tín dụng đạt 26,0 tỷ USD và cho vay tín chấp đạt tổng cộng 10,3 tỷ USD.
• Biên lãi thuần (NIM) điều chỉnh theo rủi ro mở rộng lên mức kỷ lục 12,4% từ mức 9,5% trong quý 1, chủ yếu nhờ thu nhập từ tín dụng, chi phí tín dụng thấp hơn và tối ưu hóa bảng cân đối kế toán.
• Nu đã phục vụ 139 triệu khách hàng. Tỷ lệ hoạt động tăng so với quý trước lên 83,5%, ARPAC đạt 17 USD và tỷ lệ hoạt động tại Brazil lần đầu tiên vượt 86%.
• Ban lãnh đạo tiếp tục dự báo tỷ lệ hiệu quả trung bình cả năm đạt khoảng 20% và coi vùng NIM điều chỉnh theo rủi ro hiện tại là có thể duy trì trong tương lai gần, đồng thời nhấn mạnh rằng 12,4% không phải là mức đáy.

Dữ liệu tài chính cốt lõi

Trừ khi có tuyên bố khác từ ban lãnh đạo, các tốc độ tăng trưởng được trình bày trên cơ sở so với cùng kỳ năm ngoái và loại trừ biến động tỷ giá hối đoái.

Kết quả kinh doanh và hoạt động

Tăng trưởng tín dụng và chất lượng tài sản

Tăng trưởng tín dụng tiếp tục diễn ra trên diện rộng. Dư nợ thẻ tín dụng tăng 35% so với cùng kỳ năm ngoái lên 26,0 tỷ USD, cho vay tín chấp tăng 45% lên 10,3 tỷ USD và cho vay có thế chấp tăng 30% lên 3,1 tỷ USD.

Ban lãnh đạo cho biết đà tăng trưởng so với quý trước đang bình thường hóa sau một giai đoạn mở rộng đặc biệt mạnh mẽ, trong khi các tiêu chuẩn thẩm định tín dụng vẫn không đổi. Công ty cho rằng sự cải thiện ở nhóm nợ quá hạn sớm là do yếu tố mùa vụ thông thường, bị bù trừ một phần bởi việc chủ động mở rộng sang các nhóm khách hàng có tổn thất dự kiến cao hơn nhưng mang lại lợi nhuận điều chỉnh theo rủi ro dự báo tốt hơn.

Tỷ lệ nợ quá hạn trên 90 ngày tăng lên 6,9%, phản ánh sự chuyển dịch theo mùa vụ của các khoản nợ quá hạn sớm từ quý 1 và sự thay đổi trong cơ cấu danh mục. Ban lãnh đạo tuyên bố chưa thấy bằng chứng nào về sự suy giảm trên diện rộng đối với tín dụng tiêu dùng.

Dự phòng tổn thất tín dụng tăng từ 6,1 tỷ USD lên 6,6 tỷ USD. Tăng trưởng danh mục đóng góp 342 triệu USD, trong khi việc chủ động mở rộng rủi ro đóng góp thêm 170 triệu USD. Tỷ lệ bao phủ nợ xấu trên 90 ngày ở mức 244%.

Tiền gửi và mở rộng biên lợi nhuận

Tiền gửi đạt 45,3 tỷ USD, bao gồm 36,4 tỷ USD tại Brazil, 5,7 tỷ USD tại Mexico và 3,3 tỷ USD tại Colombia. Tiền gửi tại Mexico giảm nhẹ do Nu tiếp tục tối ưu hóa chi phí vốn. Tỷ lệ dư nợ trên tổng tiền gửi (LDR) của quốc gia này duy trì ở mức 35%.

Chi phí tiền gửi tổng thể bằng 88% lãi suất liên ngân hàng, không đổi so với quý trước và thấp hơn 3 điểm phần trăm so với cùng kỳ năm trước.

Thu nhập từ tín dụng đóng góp 178 điểm cơ bản vào mức tăng NIM điều chỉnh theo rủi ro so với quý trước, trong khi chi phí tín dụng thấp hơn đóng góp thêm 115 điểm cơ bản. Ban lãnh đạo cho biết chương trình tái thương lượng nợ Desenrola chiếm khoảng 5% chi phí tín dụng và đóng góp một phần nhỏ vào việc cải thiện biên lợi nhuận.

Brazil: thương mại hóa sâu hơn và phân khúc khách hàng

Nu cho biết khoảng 60% khách hàng thuộc phân khúc đại chúng tại Brazil sử dụng Nu làm tài khoản chính. Ban lãnh đạo coi mối quan hệ tài khoản chính này vừa là lợi thế về doanh thu vừa là lợi thế tín dụng vì chúng tạo ra dữ liệu hành vi phong phú hơn và giúp Nu có thứ tự ưu tiên thanh toán cao hơn từ phía khách hàng.

Phân khúc Ultravioleta có gần 1 triệu khách hàng. Doanh số giao dịch và tài sản lưu ký của phân khúc này lần lượt tăng 41% và 37% so với cùng kỳ năm ngoái.

Vào tháng 7, Nu đã ra mắt Chroma, một gói dịch vụ trả phí dành cho phân khúc Supercore. Gói dịch vụ này bao gồm các sản phẩm tín dụng nâng cao cùng các đặc quyền ngân hàng và phong cách sống nhằm khuyến khích khách hàng tập trung nhiều hoạt động tài chính hơn với Nubank.

Nu cũng phục vụ 6,8 triệu doanh nghiệp nhỏ, trở thành tổ chức tài chính lớn nhất Brazil tính theo số lượng khách hàng doanh nghiệp, theo ban lãnh đạo. Công ty ước tính hiện tiếp cận khoảng một phần ba thị trường đó.

Mexico và Colombia

Các cơ quan quản lý Mexico đã phê duyệt giấy phép ngân hàng cho Nu, cho phép công ty tiếp cận các năng lực bao gồm trả lương trực tiếp qua tài khoản, bảo hiểm tiền gửi cao hơn và nguồn vốn rộng lớn hơn cho các sản phẩm tín dụng.

Mexico đạt 16 triệu khách hàng vào cuối tháng 7. Ban lãnh đạo cho biết 35% khách hàng đã mở tài khoản ngân hàng đầu tiên với Nu và 52% nhận thẻ tín dụng đầu tiên thông qua công ty.

Tỷ lệ thâm nhập khách hàng của Nu tại Mexico đạt 16,5% dân số trưởng thành. Ban lãnh đạo so sánh mức này với tỷ lệ thâm nhập tại Brazil vào năm 2020 và lưu ý rằng ARPAC của Mexico ở cùng giai đoạn là 12,3 USD, so với 5,6 USD tại Brazil.

Ban lãnh đạo cho biết Mexico đã hòa vốn trong sáu năm, so với tám năm đối với Brazil. Công ty mô tả Mexico là cơ hội tăng trưởng dài hạn lớn, trong khi Colombia cũng được cho là đang hoạt động tốt hơn kỳ vọng của công ty.

AI và năng suất hoạt động

Nu tiếp tục mở rộng nuFormer, nền tảng AI tập trung dành cho thẩm định tín dụng, hỗ trợ khách hàng, đề xuất sản phẩm và tiếp thị. Nền tảng này đã được triển khai trong danh mục thẻ tín dụng và cho vay tín chấp tại Brazil, đồng thời nhân rộng sang Mexico. Nền tảng cũng đang được thử nghiệm cho thẻ tín dụng SME và khách hàng tại Colombia.

Ban lãnh đạo cho biết các trợ lý AI hiện xử lý hơn 60% các cuộc trò chuyện hỗ trợ khách hàng tại Brazil, với đánh giá từ khách hàng bằng hoặc cao hơn mức của nhân viên tư vấn là con người. Hơn 100 chiến dịch tiếp thị cũng đã sử dụng tính năng định vị mục tiêu dựa trên nuFormer.

Số lượng nhân viên đạt khoảng 10.400 người. Ban lãnh đạo không kỳ vọng có sự tăng hoặc giảm đáng kể trong ngắn hạn, nhưng tin rằng AI có thể giúp lực lượng lao động tăng năng suất từ hai đến năm lần trong vài năm tới.

Triển vọng từ Ban lãnh đạo

• Nu tiếp tục dự báo tỷ lệ hiệu quả cả năm 2026 trung bình khoảng 20%.
• Ban lãnh đạo kỳ vọng NIM điều chỉnh theo rủi ro sẽ duy trì trong cùng vùng chung như mức 12,4% của quý 2 trong tương lai gần, nhưng tuyên bố rõ ràng rằng mức này không nên được coi là mức đáy.
• Công ty có kế hoạch duy trì khoản đầu tư liên quan đến việc gia nhập thị trường Mỹ trong tương lai trong phạm vi 100 điểm cơ bản đối với tỷ lệ hiệu quả của mình.
• Ban lãnh đạo dự kiến việc xây dựng dữ liệu tín dụng đặc thù tại Mỹ và đạt được niềm tin đầy đủ vào các mô hình của mình sẽ mất khoảng 12 đến 30 tháng sau khi gia nhập thị trường.
• Nu kỳ vọng hoạt động cho vay qua lương tư nhân tại Brazil sẽ tăng tốc khi các hệ thống, quy trình tài sản đảm bảo và hành vi của người vay trở nên hoàn thiện hơn. Ban lãnh đạo tin rằng sản phẩm này có thể trở thành một phần quan trọng và chống chịu tốt hơn trong danh mục đầu tư trong 18 đến 24 tháng tới.

Rủi ro và các yếu tố cần theo dõi

• Tỷ lệ nợ quá hạn trên 90 ngày tăng 35 điểm cơ bản so với quý trước lên 6,9%, mặc dù ban lãnh đạo cho rằng mức tăng này là do yếu tố mùa vụ và cơ cấu danh mục hơn là sự suy giảm tín dụng trên diện rộng.
• Việc chủ động mở rộng sang các nhóm khách hàng có rủi ro cao hơn đã làm tăng nợ quá hạn sớm và bổ sung 170 triệu USD vào khoản dự phòng. Ban lãnh đạo cho biết các nhóm này dự kiến sẽ tạo ra lợi nhuận điều chỉnh theo rủi ro cao hơn.
• Nu đưa ra giả định về sự suy giảm khi thẩm định cho vay và duy trì các đệm rủi ro bổ sung, nhưng ban lãnh đạo thừa nhận sự thận trọng tiếp diễn đối với kinh tế vĩ mô.
• Tiền gửi tại Mexico giảm nhẹ trong quý thứ hai liên tiếp theo chiến lược tối ưu hóa nguồn vốn của công ty.
• Các mô hình tín dụng tại Mỹ có thể tái sử dụng nền tảng công nghệ và chuyên môn rủi ro của Nu, nhưng ban lãnh đạo cho biết việc phát triển dữ liệu và thử nghiệm đặc thù cho quốc gia sẽ mất thời gian.

Điểm nhấn phần Hỏi & Đáp với chuyên gia phân tích

Tính bền vững của biên lợi nhuận điều chỉnh theo rủi ro: Ban lãnh đạo cho biết kết quả của quý 2 là nhờ đà tăng trưởng cho vay trước đó, số dư sinh lãi được cải thiện, xu hướng tín dụng theo mùa vụ và hiệu quả tín dụng cốt lõi tốt hơn dự kiến. Desenrola chiếm khoảng một phần ba mức cải thiện so với kỳ vọng trước đó của công ty.

Tác động của AI đối với công tác thẩm định: Nu cho biết các mô hình có sự hỗ trợ của AI mạnh hơn so với các mô hình hồi quy logistic truyền thống, nhưng nhấn mạnh rằng việc kiểm định mô hình và giám sát tín dụng vẫn không thay đổi. Ban lãnh đạo đã liên hệ các mô hình mạnh hơn với khả năng thực hiện mở rộng rủi ro có mục tiêu với lợi nhuận điều chỉnh theo rủi ro dự kiến cao hơn.

Tăng trưởng lợi nhuận dài hạn: Ban lãnh đạo đã xác định ARPAC cao hơn tại Brazil, Mexico và Colombia, việc phân khúc khách hàng được cải thiện thông qua Chroma và Ultravioleta, cùng việc thương mại hóa cơ sở khách hàng SME là những động lực tăng trưởng chính. Các nhóm khách hàng đã trưởng thành tại Brazil hiện tạo ra ARPAC ở mức khoảng 25 USD.

Cho vay qua lương tư nhân: Nu đang từng bước tăng tốc phát hành khoản vay khi trở nên tự tin hơn với các hệ thống tài sản đảm bảo và chất lượng người vay. Ban lãnh đạo cho biết những người vay ban đầu có rủi ro cao ngoài dự kiến, nhưng hiện công ty đang thấy nhiều khách hàng có rủi ro thấp hơn sử dụng sản phẩm này để tái cơ cấu các khoản nợ đắt đỏ.

Mở rộng sang Mỹ: Nu kỳ vọng hạ tầng AI và tín dụng của mình có thể chuyển giao giữa các thị trường, nhưng có kế hoạch bắt đầu bằng việc thử nghiệm và thu thập dữ liệu trước khi mở rộng quy mô cho vay. Số lượng nhân viên làm việc tại Mỹ có thể tăng từ khoảng 1% lên 2% tổng số nhân sự, chủ yếu để tiếp cận nguồn nhân tài chuyên môn, đặc biệt là trong lĩnh vực AI.

Toàn văn biên bản cuộc họp kết quả kinh doanh

Toàn văn cuộc gọi công bố kết quả kinh doanh

Phần trình bày của ban lãnh đạo

Operator

Good evening, ladies and gentlemen. Welcome to Nu Holdings conference call to discuss the results for the Second Quarter of 2026. A slide presentation is accompanying today's webcast which is available in Nu's investors simulations website, www.investors.nu in English and www.investidores.nu in Portuguese. This conference is being recorded, and the replay can also be accessed on the company's IR website. This call is also available in Portuguese. [Operator Instructions] [Foreign Language] [Operator Instructions]

I would now like to turn the call over to Mr. Guilherme Souto, Investor Relations Officer at New and Holdings. Mr. Souto, you may proceed.

Guilherme Souto

Thank you, operator, and thank you, everyone, for joining our earnings call today. With me on today's call are David Velez, our Founder, Chief Executive Officer and Chairman; and Rob Livingston, our Chief Financial Officer.

All financial metrics discussed and presented today reflect our mini geo P&L framework, which we introduced in the Q4 2025. These managerial measures are important to how we manage the business but are not financial measures as defined under IFRS and may not be comparable to other companies. A full reconciliation report to the most directly comparable IFRS figures is available in our managerial P&L reconciliation report and in the appendix to this presentation. Unless otherwise noted, all growth rates discussed today are presented on a year-over-year FX neutral basis.

Today's discussion may include forward-looking statements, which are not guarantees of future performance and involve risks and uncertainties. Actual results may differ materially from those expressed or implied. Please refer to the forward-looking statements disclosure included in the earnings presentation for additional information.

With that, I will now turn the call over to David. Please go ahead, David.

David Velez-Osomo

Hello, everyone, and thank you for joining us today. 13 years ago, we started with a simple hypothesis. A bank built on technology with no branches and no legacy to defend could serve hundreds of millions of people better and at a fraction of the cost. Today, I'm proud to announce that in the past quarter, for the first time, we generated more than $1 billion in net income. This milestone is a result of our customer obsession translated into an earnings generating formula. It is also a testament to the tremendous work of our team here at Nubank 13 years later, that hypothesis continues to play out exactly as we envisioned. Our customer base reached 139 million customers, including almost $118 million in Brazil more than $5 million in Colombia. And in the end of July, Mexico just reached 16 million customers. Engagement continued to deepen alongside that growth.

Our activity rate expanded sequentially to 83.5%, and while Brazil surpassed 86% for the first time. The combination of more customers and deeper engagement continues to drive monetization with RPAC reaching $17. Together, they generated $5.9 billion in gross revenue, while maintaining a highly efficient operating model with an efficiency ratio of 20%. This operating leverage allows us to continue investing in our 3 core markets. Brazil, Mexico and Colombia, while laying the foundation for our international expansion. That is why we have always been by optimizing for the long term. It is why we can continue building for the next decade while delivering a quarter like this one.

Let me walk you through both starting with Brazil. Brazil remains our largest growth opportunity and most of it lies within our existing customer base. The mass market alone represents roughly $30 billion in industry gross profit. We already serve most of the segments, and we're the primary account for approximately 60% of those customers. Even so, there is significant room in deepen those relationships and capture more of that profitability. That is possible because of the capabilities we have built over the past 13 years. They allow us to expand financial access while delivering a better customer experience, lower costs and increasingly personalized products.

As we built one of the leading financial services brands in Latin America for the mass market, we find ourselves attracting millions of higher income resilience that, unfortunately, we were not able to serve well at the time. In 2021, we launched Ultravioleta, a high-income focused brand and product or nearly 1 million Ultravioleta customers have significantly higher purchase volumes and assets under custody than the rest of our portfolio with both continuing to grow strongly, up 41% and 37% year-over-year, respectively, in Q2 2026. However, we have realized that there is a meaningful segment between mass market and high income that we could also be serving better. We call this segment Supercore. And in July, we launched Chroma, a subscription-based tier for our Supercore customers. A segment with an even larger profit pool than high income and one where we already have significant penetration.

Chroma gives them a dedicated experience, enhanced credit offerings and a broader set of banking and life style benefits, designed to reward customers for concentrating more of their financial lives from Nubank. That includes new cell, a free ChatGPT subscription, accelerated savings products and other benefits across our own ecosystem partners. Our goal is to develop primary banking relationships, and Chroma is a significant step in this direction for this segment.

Of course, the opportunity also extends beyond consumers. We already serve 6.8 million small businesses making Nubank, the largest financial institution in Brazil by number of business customers. Though we still reach only about 1/3 of that market. This is how we see the exceptional growth in Brazil. continuing to expand our customer base while increasingly serving a larger share of our customers' financial lives through better product and segmentation.

Now let me turn to our other core markets. Earlier this month, Mexican regulators approved our banking license in the country, and we're happy to be born as the largest digital bank in Mexico with more than 16 million customers. That completes our transformation from a credit-first Fintech into a full-scale digital bank, and it unlocks capabilities we did not have before. Payroll direct deposit strengthened primary banking relationships and customer engagement. Higher deposit insurance increases confidence in holding balances with us. Those deposits fund a broader credit offering while allowing us to expand into new products and customer segments over time.

Financial inclusion has been a defining part of the journey. For 35% of our customers, we were their first bank account, for 52% their first credit card, Today, our customers live in 98% of Mexico municipalities with nearly 80% outside the country's major cities, demonstrating how technology lets us reach customers everywhere. But what excites us most is what comes next. Mexico remains at an earlier stage of digital financial adoption. Bank account penetration has increased from 44% to 63% over the past decade. Just 85% of Mexicans still preferred to pay in cash. Yet the pace of change is accelerating, digital payments in Mexico continue to compound year after year. In the first half of this year, transfer below $5 grew more than 60%. And today, nearly half of all transfers in the country are less than $25. These are everyday transactions and a clear sign that cash is steadily giving way to these payments.

In June, the Central Bank introduced new rules that every financial institution must implement by the end of the year. The objective is to simplify the experience across different payment rails. Going forward, consumers will see a standard interface and follow the same steps regardless of who they are paying or how they choose to pay. Since these rules are mandatory for the entire financial system, they strengthen network effects and should further accelerate digital payment adoption.

We have seen this movie before. In Brazil, the regulatory agenda foster competition and digital innovation, expanding financial inclusion, driving everyday usage and ultimately accelerating credit adoption. Pix is the clearest example. We have put a simple and seamless experience from the very beginning, became the market leader in transaction volume and turned that into primary banking relationships. That environment toward exactly the digital model we have built. We believe Mexico is following a similar path. We can already see it in our numbers. Today, we reached 16.5% of Mexico's out of population, essentially the same penetration we had in Brazil in 2020. But the cohorts are monetizing earlier. At the same stage, ARPAC in Mexico is $12.3 against $5.6 in Brazil, higher income per capita better unit economics in the credit card product and higher interest-earning balances, all at a lower cost to serve.

Mexico is Brazil's playbook running faster and with the benefit of the scale we have today. That's how we broke even in 6 years in Mexico compared with 8 years in Brazil. To recap, customer behavior, technology and regulation are now all moving in the same direction. Taken together, they create one of the most compelling opportunities we have ever seen in Mexico. As more financial activity moves onto our platform, we build deeper customer relationships, gain better underwriting insights and expand our ability to serve a larger share of our customers' financial lives. For the first time, we now have the full set of capabilities to capture that opportunity in Mexico. And Brazil and Mexico run on the same technology stack and increasingly on the same brand.

Let me show you what that means. About a year ago, we introduced or flotation model for financial behavior. Since then, we are focused on one objective, building a single AI platform that powers business and customer decisions across the Nubank. That work spans every layer of the stack. We increased and upgraded our own GPU fleet, giving us full control of the compute layer. We expanded our architecture research efforts, and we continue building on one of our greatest advantages more than a decade of transaction history across more than 100 million customers in 3 countries. That research is unlocking compounding against inefficiency and model quality. We recently advanced the former to a hybrid linear attention the same architectural approach behind front models like KemiK3 and Quon3.5, and we trained with Moen, the same class of optimizer powering today's most efficient, large language models.

By decoupling the former core backbone for specific downstream decisions, any improvement to the central model can instantly upgrade performance across all our business lines with a cost of retraining. The latest generation quadrupled context length, trainining speed and inference speed while reducing the cost of running models in production. As we've scaled pretraining the base models, understanding of how our customers behave has become deep enough to change or we build every model on top of it. To give you 1 example, today, we can achieve the same predictive performance with 20 million fine-tuning data rows that previously required over 400 million, cutting development cycles from weeks to days. The platform now reaches nearly every decision we make.

We first deployed nuFormer in our flagship credit portfolio in Brazil. Through 2025, we replicated the model in Mexico, demonstrating that the platform generalizes across markets. During the first half of this year, we extended it to unsecured lending in Brazil and to the next generation of our core credit models. We're now testing it in credit cards for SMEs and for our Colombian customers. But underwriting is only one application. Today, AI agents handled more than 60% of customer support conversation in Brazil with customer ratings at or above human parity. Beyond underwriting and customer support, we're using artificial intelligence to optimize decisions across credit, deposits and growth, moving from predictive outcomes to determining the actions that maximize value on the real world constraints and have the same understanding of transactions that predict credit risk also predicts what a customer wants next. It allows us to recommend the products that maximize long-term customer value, personalize the app experience and move toward our vision of an private banker.

nuFormer is also improving how we grow. As the model learns our representation of how every customer behaves, we use it to put each campaign in front of the customers most likely to find it useful and more than 100 campaigns have already run this way. One platform powers underwriting, customer support, optimization and growth. every improvement we make benefits every application built on top of it. We're incredibly excited about the progress Nubank has had to date, levering AI as its informative technology and have strong confidence our approach will be a meaningful differentiation going forward.

Before we turn to our financial results, I want to say a few words about our CFO transition. As we announced in early June, Rob Livingston has succeeded Guilherme Lago as our Chief Financial Officer. Lago spent 7 years with us, 5 of them as CFO, and he handed over the role at the strongest moment in our history with our first $1 billion quarter. He has been an incredible partner, and I am glad we will keep working together in his new role as special adviser. Rob has spent the past few weeks working alongside Lago and our teams, and we're very excited to be able to counter with his significant experience. Rob, welcome. Over to you.

Rob Livingston

Thank you, David. It is a privilege to step into this role at such an important moment for the company. Since joining Nubank, I have spent time with our teams across the organization. What has impressed me most is the customer obsession, the consistency of the business model and the discipline with which it has been executed. I'm excited to help lead the next phase of Nubank's journey. And today, I'm pleased to walk you through our Q2 2026 financial results.

Let's start with our consolidated credit portfolio. Portfolio reached $39.4 billion, up 37% year-over-year and 5% sequentially. Growth remained broad-based. Credit cards increased 35% year-over-year to $26 billion. Unsecured lending grew 45% to $10.3 billion and secured lending increased 30% to $3.1 billion. Sequential growth remained solid while normalizing after a period of exceptionally strong expansion, origination does not expand in the straight line, and we see that as a healthy dynamic. Throughout the quarter, our underwriting framework remained unchanged and growth remained strong relative to the broader market. As we'll discuss in the next few slides, we're comfortable with the quality of the portfolio and the performance of the vintages we're originating.

Now turning to deposits. We ended the quarter with $45.3 billion in deposits, up 18% year-over-year and 6% sequentially, recovering the seasonal outflows we discussed last quarter. Brazil closed to $36.4 billion, Mexico was $5.7 billion and Colombia at $3.3 billion. In Mexico, deposits declined modestly again this quarter, reflecting our ongoing deposit optimization strategy. This continues to improve our cost of funding while maintaining ample liquidity with our loan-to-deposit ratio still at just 35%. Our cost of deposits was 88% of the interbank rate essentially unchanged from last quarter and 3 percentage points lower than a year ago. Overall, we're pleased with both the growth and pricing of our deposit franchise across all 3 markets. As always, our objective is not simply to maximize deposits but to build a resilient funding base that deepens customer relationships supports profitable growth and strengthens the long-term economics of the business.

Net interest income reached $3.7 billion, up 9% and net interest margin expanded 180 basis points to 22.9%. That is the result of what we laid out last quarter. The growth we put on the books, a mix weighted further towards unsecured lending and the deliberate risk expansions we made. Cost of credit reduced quarter-over-quarter to $1.7 billion. Desenrola, the government debt renegotiation program impacted this number by just about 5%. And even more important than the accounting impact the program allowed us to help nearly 1.8 million customers renegotiate past due balances and get their finances back on track. With margin up and cost of credit down, risk-adjusted net interest margin expanded to a record 12.4%, up from 9.5%.

I'm going to walk you through the drivers of this expansion in risk-adjusted net interest margin in more detail. That brings us to the risk-adjusted net interest margin bridge, and I want to focus on the expansion between Q1 and Q2. Credit income was a primary driver of risk-adjusted NIM expansion, contributing 178 basis points to the quarter-over-quarter increase compared to 152 basis points in Q1. This acceleration was driven by our strong loan growth in cards and unsecured lending in Q1, reflected in our improving loan-to-deposit ratio. Lower cost of credit contributed a further 115 basis points. The majority of the improvement in cost of credit came from the expected seasonal patterns we observed and disciplined underlying business performance rather than onetime items. Float income and funding costs both remained broadly neutral. Together, these dynamics explain the strong expansion in risk-adjusted NIM this quarter and continue to demonstrate the resilience of our underlying unit economics, supporting the sustainability of current levels going forward.

Now let me turn to asset quality. As expected, our NPL metrics continue to follow their normal seasonal pattern. 15- to 90-day delinquencies improved 16 basis points to 4.8%. That improvement reflects several underlying dynamics, which I'll unpack on the next slide. 90-plus delinquencies increased 35 basis points to 6.9%, broadly reflecting the seasonal migration of first quarter early delinquencies into the 90-plus bucket. Taken together, these metrics are consistent with the seasonal dynamics we expected and continue to support our view that the underlying quality of the portfolio remains robust.

Looking at the drivers of early delinquency, this bridge explains why the 15- to 90-day ratio improved sequentially. Seasonality reduced the ratio by 37 basis points. Against that, our intentional risk expansions in the first half of the year added back 24 basis points. Those were deliberate decisions to serve cohorts with higher expected losses, but which also generate higher risk-adjusted returns as you've seen in our risk-adjusted margin performance. Product mix and the remaining drivers are broadly neutral. We don't see any evidence in our portfolio of a broad-based weakening in consumer credit, yet we remain vigilant as always. Altogether, the 15- to 90-day ratio improved 16 basis points during the quarter.

The allowance bridge tells a similar story from the balance sheet perspective. The allowance increased from $6.1 billion to $6.6 billion. The largest driver by far was portfolio growth, contributing $342 million. Under IFRS 9, we recognize expected credit losses at origination. So growth increases the allowance before the associated interest income is earned. The intentional risk expansions we just discussed contributed another $170 million, while all other movements were immaterial, including Desenrola, which primarily affects recoveries rather than the ECL allowance due to the accounting treatment of renegotiated loans.

Together, these 2 bridges reinforce the same message. The quarter's credit dynamics were driven by growth, seasonality and disciplined risk expansion not by any deterioration in the underlying quality of the portfolio. Our approach to provisioning and coverage remains disciplined and consistent. Starting with the chart on the left, we built allowances equal to 113% of new 15-plus delinquency formation during the quarter, broadly in line with our historical averages. The chart on the right reinforces the same point. Total coverage over NPL90+ stood at 244%, meaning we continue to carry allowances equal to almost 2.5x or 90-plus balances. This provides a strong balance sheet cushion and remains consistent with the levels we've maintained over the past several years even as the portfolio has continued to grow. Together, these 2 metrics reinforce an important point, while the portfolio continues to grow across products and customer segments, our provisioning philosophy remains disciplined and consistent through the cycle.

Now one final point on credit risk, and this is an important one. As David mentioned, Nubank leads the Brazilian market in primary banking relationships. That leadership reflects the trust customers place in us. Combined with the analytical rigor of our underwriting models and the quality of the data generated through those relationships, it creates a structural edge in credit. You can see that clearly on this slide showing the percent of credit card outstandings that are 90-plus days past due in Brazil. Across every income band, our credit risk performance has been steady. This strong and stable performance is driven by customers who have Nubank as their primary banking relationship. The delinquency measures of these customers is roughly half the portfolio average.

What's the implication? That customer primacy is not only a growth and revenue advantage, it is also a credit advantage. A deeper relationship gives us richer behavioral data, strengthens our underwriting and place this Nubank at the top of our customers' payment hierarchy. Together, these factors produce consistently better credit outcomes. That is why we continue to view customer primacy as a key pillar of our credit superpower.

Now turning to our income statement. Gross revenues reached nearly $5.9 billion, up 39% year-over-year. Gross profit reached $2.4 billion during the quarter, up 43% year-over-year and 25% sequentially. As credit normalized in line with its expected seasonal pattern, its contribution to gross profit increased 41% this quarter, all fees represented 25% and flow 34%. These shares naturally fluctuate from quarter-to-quarter. What matters is that all 3 components continue to grow in absolute dollars, reflecting the diversification of our business model. Looking ahead, we remain confident that credit,, fees and float will continue to complement 1 another as drivers of long-term growth profit growth.

Turning to operating leverage. Net revenues surpassed $4 billion for the first time, reaching $4.1 billion, up 8% sequentially. Operating expenses totaled $806 million, up 20% quarter-over-quarter as real estate and marketing expenses shifted from the first quarter into the second, will alongside our continued investments in international expansion. As a result, our efficiency ratio ended the quarter at 19.5%. As we discussed last quarter, the 17.6% reported in Q1 was not a run rate. Roughly 2/3 of that improvement reflected temporary timing effects, which reversed as expected this quarter.

Looking ahead, we continue to expect the efficiency ratio for the full year to average about 20%. More importantly, our long-term view remains unchanged. And we will continue to invest while maintaining operating leverage as we scale.

To conclude, Net income reached $1.1 billion for the first time in Nubank's history, up 17% from the first quarter and 49% year-over-year. More importantly, we delivered that result while sustaining a record 33% return on equity and continuing to invest across our 3 markets and in our long-term opportunities. I believe this quarter reflects the strength of the business model, David described earlier. The investments we've made in technology, AI, underwriting and customer experience continue to translate into profitable growth at scale. I'm excited to be part of this next chapter, and I look forward to continuing to build on this momentum. And with that, we'll open the call for your questions.

Operator

[Operator Instructions] I would like to turn the call over to Mr. Guilherme Souto, Investor Relations Officer. .

Guilherme Souto

Thank you, operator. Could you please open the line for Mr. Tito from Goldman Sachs. .

Phần hỏi đáp

Unknown Analyst

I understand a little bit the different role impacts Rob, you mentioned a few different numbers. I think you said maybe 5% of provisions, but not sure if that necessarily impacted the bottom line? And just thinking because our prior calls, you had mentioned risk-adjusted margin getting back 10.8%, well above that this quarter, very good performance here for sure. But just to understand, was there an impact from this enroll out on that risk-adjusted margin given the different parts? And then I have a follow-up after that.

Rob Livingston

All right. Thanks so much, Tito. Let me take that and good to hear from you again. Yes, I did mention that this enroller had an impact of about 5% on our cost of credit. That's the main metric that we're looking at there. And so as a result, it did also have an impact on our risk-adjusted net interest margin. If you think about the progress that we made from last quarter to this quarter of almost 3% expansion in the metric, the majority of that did come from lending growth, A.nd this was really due to the strong growth that we saw in Q4 last year, Q1 last year and the matriculation of that into revenues in Q1. Now the cost of credit also contributed 115 basis points. And you're right, that this enroller would be part of that. It's a minority of the impact. The majority did come from seasonality, but also just really solid underlying credit performance.

Unknown Analyst

Okay. No, very helpful, Rob. And I think also, just to think in terms of the context, what everybody is worried about is going into next year. I mean you mentioned overall credit quality trends look good, you feel comfortable. But given the macro that we're seeing in -- how do you think about the growth outlook maybe going into 2027 and this level of risk-adjusted net?.

David Velez-Osomo

David here. So as we've said a few times to investors, we don't take a directional view necessarily on the economy. Our base assumption when we underwrite a loan is that the future will be worse than the past. The things will actually be much worse than everything we have seen. So a lot of -- everything, every single underwriting decision already assumes a deterioration by default. It has a pretty significant cushion in terms of what do we need to see for that decision to continue to be NPV positive. Obviously, we're in an environment where there is a lot of caution. We are actively looking at every single sign that we have. So far, we don't really see any significant or structural deterioration in our numbers and we continue to operate with very significant cushion in this environment.

We are also in a position where we -- while we have a large consumer base, we're still a very small percentage of the market. As we've said in last call in this call, we have 7% market share of that profit pool. So we're still a small player in that big market, and we get to cherry-pick our customers. Cherry pick them with loans and products that have very short-term duration, which gives us a huge amount of ability to react quickly, have a lot of conviction on a lot of the underwriting capabilities as we mentioned. And then primarily, we have this huge advantage, as Rob mentioned, of being the largest primary bank accounting in the country today.

Over 60% of our mass market customers use us their primary bank accounts. So that's a huge advantage because, as Rob mentioned, positions as effectively as being senior in the credit stack of a customer. And so when you combine strong analytics, significant cushion being the primary bank account and having that seniority then there is a lot of conditions for us to continue growing at a very good pace. I won't necessarily give you a specific number of growth that we continue to see the conditions to continue growing and taking share as we use a lot of these levers to do that very effectively.

Rob Livingston

Yes. And to your question on net interest margin or risk adjusted net interest margin for the foreseeable future, we see it as being in the same region as where we are today. We think that it's sustainable.

Guilherme Souto

Operator, could you please open the line for Mr. Jorge Kuri from Morgan Stanley.

Jorge Kuri

Hi, everyone. Thanks for the opportunity to ask questions and congrats on the great numbers. I wonder -- I guess to go back to the risk-adjusted NIM at 12.4%. And maybe tie this to the usage of AI and the sophistication on your credit line increases and overall, your ability to take on more risk with lower losses. And now that you've seen maybe a full year vintage of people that you improve their offer with AI models and that you've been tracking them. Would you mind sharing some of the KPIs that you've seen? And to what extent they are tied to this 12% risk-adjusted margins because you evidently seem very confident about this being the new level. Not only you said it Rob right now, but also in an interview with Bloomberg earlier. And in the past, we have seen volatility in that number based on mix. And so I guess, yes, I just want to get to a little bit more of the KPIs that are driving that? To what extent is the AI models and get more comfort on that being a floor from here.

Rob Livingston

Yes. So thank you so much for the question, Jorge. I think a few things to say there. The first is that it certainly is the case that our sort of generated models and assisted models are more powerful than traditional logistic regression models that is incontrovertible. And we are tracking them though in the exact same way that we would have tracked our historical models. We're looking at the degree of predictability, the variance at the low end and the high end of the predictive range as well as the outcomes across both back testing as well as forward testing of that model in production.

So the macro point is that our risk approach and our credit monitoring hasn't actually shifted in this dynamic. And in fact, we are very happy to continue to have that same level of discipline going forward. I would correct 1 thing you said, though, where you said that 12% is a floor. I didn't say that it was a floor. I said we'd be in that ballpark. And so I don't want to overcommit there. But what we are seeing is that our strategy that has been partially enabled by stronger models to make intentional risk expansions that produce more risk-adjusted margin is paying off, and that's what we're seeing for the foreseeable future.

David Velez-Osomo

I think, Jorge, the additional factor always need to take into account is the increasing LDR and what LDR brings to the business model. This is something we mentioned a few times that if you look at our balance sheet, it continues to be very unlevered. You've seen this in Slide 16, how that LDR has evolved over the past few quarters. Q1 had a significant growth, but then a large cost of credit, mainly because of seasonality. As we go into Q2, we start seeing the benefits of a lot of the growth and the sort of optimization of the balance sheet. And there's a significant opportunity going forward.

We will continue to optimize the balance sheet, obviously, as we continue to grow our credit portfolio. And that just will simply be reallocating a lot of deposits that they are earning CDI or refi rate towards a much higher-yielding assets. And that obviously evolves directly into margin and into ROE. So that's a very strong dynamic that is also happening within the business model.

Guilherme Souto

Operator, could you open the line for Mr. Eduardo Rosman from BTG Pactual.

Eduardo Rosman

Congrats on the numbers. I have a question for Dave regarding I think we read recently that you became a part of the Board of OpenAI. So it would be great if you could share with us how do you believe you can help open AI, but more importantly, how this experience might help you here at Nubank?

David Velez-Osomo

Sure. Thanks, Ed. So obviously, this is not OpenAI's earnings call. So I won't get in too much of a lot. But -- but I think effectively, we discussed very openly that internally and externally that we think artificial intelligence is the most important technology transformation in our history. And it will be one of the most technological impactful shift in any business in any industry around the world. So this is a global trend and a very powerful trend.

Businesses are going to see significant transformation. And it's early days, but we're seeing it very clearly inside Nubank. We've discussed here today a lot of the different applications from credit and underwriting, but even customer-facing and and we're in the middle of a significant transformation across our organization around how we're using, putting AI in front and center as a technological trend and what will deliver as effectively an advantage.

So from that perspective, for me, getting closer to a company like OpenAI obviously provides a very interesting insight. I personally also, I think it's a great opportunity to make sure that some of these great AI companies build something great for humanity, and I have a huge amount of respect for the OpenAI team and the way they are executing this mission. So I think it's a win-win. And clearly, I mean, it's early days, but I hope I can be able to contribute significantly to the way the organization is executing.

Guilherme Souto

Operator, could you please open the line for Mr. Pedro Leduc from Itau BBA.

Pedro Leduc

Thanks, everybody. Two questions. The first, a little more homework technical. The portfolio that you now have with clients under the government renegotiation program that appeared in Stage 3, maybe in your personal loan book. And I'm assuming with a little bit less expected default loss given the fall given the coverage -- just that's the homework question as I'm trying to interpret here the movements. And then the -- and the second, a little bit back to business. In the prepared remarks when going over the unsecured lending, you mentioned the pace. You've made some comments around that. If you can give us a little more color and also 1 on your latest update on how you are on the payroll, including private payroll.

Rob Livingston

Okay. Well, let me start with the technical question there. There was a small impact on expected credit losses, as I mentioned in my prepared remarks, but it wasn't material, less than $10 million. And it does appear in Stage 3 of lending is where it shows up. I do think that we are going to see a little bit more impact from Desenrola in Q3, but we've already seen more than 4/5 of that hitting us or benefiting us in Q2. And I hope that answers your question.

Pedro Leduc

The portfolio that you now have under the program, it is a Stage 3 portfolio or starts in Stage 1.

Rob Livingston

Right. So that -- no, it's in a stage 3 portfolio.

David Velez-Osomo

Yes, sure. And on private payroll, we are accelerating month-over-month. We are slowly getting more comfortable with the product. There has been a significant progress in how the product is set up in the Brazilian market, the way companies are able to get the collateral, the way the systems are working. So as we've said many times, we found as metric. And as you made bet to go too fast too quickly on a product that has so many question marks. We think we're getting close to a system that makes a lot of sense, and we are accelerating. And ultimately, the lowest cost provider and whoever treats the customer best will win this market. And so we think we're extremely well positioned to be one of the leading players in this market over the next 18, 24 months. So we have -- and we think it's a good thing for the market.

The other part that is also starting to change slightly is counter to most people's intuitions, the first people that started taking these loans were very high risk. These were not the use cases that you would expect of people refinancing -- low-risk people refinancing high-cost debt. It was actually very high-risk customers. And so from that perspective, just didn't make a lot of sense for us to be opening the door for that as we understand the true level of risk. And there is a bit of a change in behavior. Customers actually -- good customer actually seeing the opportunity to refinance then this becomes a much more attractive product for consumers.

We would love to do the trade. There's been a lot of conversations that we are fearful of cannibalizing ourselves so that we don't want it to be successful. We will be super happy being able to refinance all of the customers that want to lower interest rates for a product that has that collateral, it would be very beneficial for us to have a more diversification in our portfolio. It would add more resilience. It would be less cyclical. So strategically, it's a product that we are very we think in the long run will be very good, and we are very well positioned to do it. We just are going at right pace, and we're getting more comfortable by the month.

Guilherme Souto

Operator, could you please open the line for Mr. Uri Fernandez from JPMorgan. I have a question regarding the over $1 billion net income, and congrats on that, David. .

Unknown Analyst

I was checking here in Banco, 1 of the leading banks in Brazil. And when I look to the retail operation, it was around $1.1 billion, right? So you are very close to that. And my question is how to continue increasing this net income, David? I know your ROE are higher, you have better cost to income you have this efficiency tailwind. But you are getting very big, right? So if you can help us understand if this is just an RPA normalization you have like your mature cohorts coming or it's about new products? Or is this about Mexico. So trying to congratulate you on the $1 billion, but I also ask how to keep growing this sizable profitability?

David Velez-Osomo

Sure. A couple of points. You are right that at some point, we're not there yet. At some point, we're going to run out of Brazilian customers. We have been saying that for about 4 years. We continue to get close to 1 million customers in Brazil every quarter. And so we're very -- sorry, every month, and so we're in very good shape in terms of user count. But there'll be a time where the number of Brazilian customers will decrease T.hen the opportunity is ARPAC. And as you'll see in Slide 6, the ARPAC expansion is pretty significantly. We've gone from 13 to 17. A lot of the incumbent banks, if you look at the ARPAC there are $40 million to $45 million we don't think necessarily will get to 40 or 45 because there's a lot of fees that we don't charge. There might be a lot of products that we don't offer.

But we'll certainly -- there is significant upside from the '17 and above. And when you look at all the cohorts customers that have been with us for 7, 8 years, they're already in the mid-20s ARPAC in the higher ARPAC. So from that perspective, there's going to be -- a lot of the opportunity in Brazil is to continue increasing ARPAC. Then we have the opportunity in Mexico and Colombia. As we've said, we think Mexico -- our base case for Mexico is a business that could be 60%, 70% of Brazil, if the utilization in Mexico happens in a real-time payment system works could be as big of Brazil. It's a lower population but has 30% higher income per capita and the ARPAC that we're seeing in Mexico are equal or above Brazil. So significant opportunity in Mexico.

There is also a significant opportunity in Colombia. We are -- our business there is significantly overperforming and we're very happy with the opportunity there. So there is a huge amount of avenues of growth going back a little bit to Brazil. What we do see is that we need to have a better segmented portfolio, and that's why we announced Croma. Now we have 3 core segments and value propositions to serve better other segments like super core and high income, where we already have a lot of customers, but we're not serving them well.

We have a low share of wallet. We gave them -- they came for a credit card. We gave them a very low credit limit. And because of the sophistication and improvement in our models, we're finally able to improve or create an underwriting capability for these segments and the value proposition for a lot of the products. And then finally, SME, we highlighted here on Slide 7. This is a blue ocean. This is a big opportunity. We are already the largest SME player in Brazil, over 6.5 million SMEs. We're just beginning to monetize that entire base and the cost structure advantage that we have, especially for the small businesses is pretty significant. So net-net, yes, over $1 billion in net income, but we're looking at a gross profit pool of $100 billion from that perspective and a lot of opportunities still to grow even in our core market like Brazil.

Unknown Analyst

Super clear a if I may, just a follow-up on Croma. Do you have -- can you share any market share you have today and any goal you have for this segment?

David Velez-Osomo

I can tell you that we already have 3 out of 5 Brazilians in this bracket as customers of new. So it's not our opportunity necessarily we have to go out in the market and acquire these customers and spend a lot of money on marketing. They're already inside our base. We are -- we just haven't treated them as good as they deserve. We just haven't given them the product set and the bundle that they should need. And so that is the opportunity. There's a huge opportunity to increase the share of wallet within those 3 to 5 Brazilians that represent -- that exist in that base.

Guilherme Souto

Operator, could you please open the line for Mr. the line for Mr. Joe Elafrom Autonomous.

Unknown Analyst

Thanks, I noticed that the number of employees is down, it's down from 10,500 to 10,400 which doesn't sound like a big change, but it had been growing pretty quickly up until now. what are your hiring plans? And how is AI allowing you to use the workforce more effectively?

David Velez-Osomo

Sure. I mean I think -- we -- as you might remember, we announced getting back to the office end of last year. And that announcement caused a number of -- a meaningful amount of people to decide not to work in Nubank anymore. So there was some attrition because of that. We have rehired effectively a lot of that attrition. And so you end up being something about flat. Looking forward, we are seeing a huge amount of productivity increase with AI, and we're very excited about the potential that, that creates. But the list of things that we also want to do, the list of things that we want to build is also infinite. And so it just opens up a larger opportunity of things that we can try.

So net-net, I don't see us significantly increasing that number. I also don't necessarily see us decreasing the number feels more or less right, but it's certainly 10,400 employees that will be 2, 3, 4, 5x more productive over the next few years as we really integrate more AI with that. And then obviously, the output of that head count will be much larger than we were able to provide today. And so we're very excited about that.

Unknown Analyst

And then staying on headcount, thinking about the expansion employees in the U.S. are pretty expensive. You're moving into the U.S.? How far do you see the head count shifting towards the U.S.?

David Velez-Osomo

I don't think it's going to be a significant change in the way we are distributed today with our majority of employees are in Brazil and Latin America. We are hiring more in the U.S. and specifically in certain areas where we are able to find a certain level of talent and experience that we just cannot find in Latin America, especially around AI. So we will increase the number of head count we have in the U.S., but it will not move the needle.

I mean it might go up from 1% to 2% total, and that's sort of the level of changes. But it was obviously, it's talent that we would be adding that would be very impactful. And then as we launch U.S. as a market, and we start growing that market, then there'll be more hiring in the U.S. and hopefully, we can be very productive and efficient as we launch our market and relying a lot on a lot of the AI capabilities that we're using.

Guilherme Souto

Operator, could you please open the line for Mr. Mario Pierry from Bank of America. Let's move on for the next one. Could you please open the line for Mr. Daniel Vaz from Safra, please.

Daniel Vaz

David, on your Slide 11, you show deposit and credit financing, price optimization is still in testing with -- so I was wondering on the credit card financing, where is the biggest price for you there? I mean, is to reprice the existing revolvers, maybe you're using personalized rates to convert more transactors who never did credit card finance before. So you want to offer them a cheap interest rate there. So trying to understand where is the biggest prize there for you in credit card finance? And second, on deposits, right? So your loan-to-deposit is very low, as you mentioned. So how should we read that primarily as a funding cost level? So do you want to bring your funding costs down, so you stimulate people to put deposits there on your platform or doesn't have to do anything with that. So you want to bring more deposits, maybe pay more with for people who doesn't have deposits today. So kind of understand that where is the biggest price for an AI applicable in these 2 businesses?

David Velez-Osomo

Sure. So -- just as a reminder, one of the most important metrics for us is Net Promoter Score, NPS or a number of different other metrics around product quality. And we seek to -- we think that the way our model works is that if we build the very best product in the market, then customers will come and financial results will follow. So a lot of this optimization is not necessarily about minimizing cost, but it could also be about optimizing quality. And so the sense or the opportunity is that whenever we have a price, be it a price for a loan or a credit product or be it yields that we offer in a deposit or truly any other product that has a price then every single customer will going to have a price which will maximize that equation of quality and cost, being able to offer higher quality at a lower cost. And so that is the analytical exercise that we're increasingly investing in.

Specifically on the deposit, we just get to an equation whether we -- it would be our decision to decide if we want to optimize cost. Then we would be able to optimize the funding cost, but we would know specifically what would be what we're giving up in terms of quality and competitiveness in that opportunity. So I wouldn't necessarily think in the short term that this will drive an improvement in funding costs because we might decide to reinvest all of those gains back into the customer, especially in countries like Mexico, where we're so early and Colombia and Brazil in some of these segments.

We want -- we are still very much on day 1, we're still very much on the challenging mode. We're not in the mode of optimizing for net income, we're increasing earnings. We're investing a lot in growth. We're investing a lot in improving our products and our customer experience. And a lot of what we can do with this new model is being able to make a better decision as we trade off quality with cost for that quality. .

Guilherme Souto

Operator, could you please open the line for Mr. Mario Pierry from Bank of America, please. .

Mario Pierry

Congratulations on the quarter, it definitely was better than what we were expecting -- but 2 questions here from my part. So on the previous call, right, in the first quarter results, you guys talked about net interest -- risk-adjusted margin going back to the low levels of the second half of last year by the end of this year. So we were talking about 10.8% to 10.5%, and you jumped to 12.4%. So I'm trying to understand where is the surprise coming from what you guys were expecting? Is it that your credit models are better than you're able to grow faster than you expected? Or like because it is a big beat versus what you guys were expecting.

And then my second question is a little bit more technical, and I appreciate you guys showing the slide on Page 20 that shows the NPLs by income. But I was trying to reconcile that slide on Page 20, Page 17. Because on Page 17, you show that NPLs have some seasonality. And then when I look at this slide on Page 20, a it doesn't appear like there's much seasonality on that data. In fact, right, if we look at your NPLs, I think you showed July 25. So now has improved for every income segment that you showed. But then when we look at the overall NPL, we actually deteriorated 40 basis points. So -- does it mean that the entire deterioration that we're seeing and all the seasonality that we're seeing is coming primarily from your unsecured personal loans because again, credit cards are 65% of your loan book and that is not showing any deterioration at all.

Rob Livingston

Yes. Thank you so much for those questions. I'll start with the second 1 first. So on Page 20, as you're looking at the credit performance that we have relative to other banks in Brazil. Keep in mind that this is credit card only, of course, and the other graph is for the whole company, and it's smooth. So that's in the nice type at the bottom of the page here is that we're taking a rolling average, and that's why you're not seeing the seasonality that does actually exist in these numbers.

Going back to your first question, though, around the overperformance of risk-adjusted NIM. If you recall at the time of the Q1 earnings, Desenrola was not clearly laid out at that time. And so about 1/3 of that benefit relative to what we were expecting is coming from Desenrola, but 2/3 are coming from 2 things. One is just really solid credit performance across the board and in some cases, better than expected. And the other is the increase in our balances that we're earning yield in Q2 at the very beginning. That was driven by the growth in continuing to ramp up. And so that did slightly come in better than we expected, and we're happy to have it now. .

Mario Pierry

Okay. But just let me follow up then even -- okay, I see the footnote here that says you smoothed out the trends. But the starting point is higher than the current point for your NPLs? And when I look at your overall NPL, it's higher. So again, is it a deterioration primarily coming in the unsecured personal loans?

Rob Livingston

So it's -- I wouldn't quite characterize it as that as much as I would characterize it as a mix shift that we are seeing. And so what you see on Page 20 is essentially a disaggregated view of the portfolio by income bracket. And when you look at it that way, you do see this steady, if not decreasing performance across all 3 segments in Brazil. But when you look at our overall portfolio, there are shifts in terms of where we are lending money and it's primarily those shifts rather than deterioration within unsecured lending that's driving that increase over time. Does that make sense?

Mario Pierry

A little bit. But yes, we can follow up later. That's fine.

Rob Livingston

So think of -- yes, it's more of a mix shift rather than a dynamic where it's all coming from 1 product segment. And we can talk about it later .

Mario Pierry

Yes. And I think that the problem that I have is the -- it's not like seasonal, right? It's more like a changing mix, the volatility in NPL is more because of a change in mix rather than seasonal. .

Rob Livingston

Well, the seasonal is still there. It's more of this idea that if you look at Slide and you look at 90 plus over the past 2 years, the general trend is upwards, and that's being driven by the mix. That was my point.

Guilherme Souto

Operator, please could you open the line for Mr. Craig Maurer Murer from FT Partners.

Craig Maurer

Question specifically for Rob. Having -- with your background, and new moving into the U.S. You spent a lot of time today discussing your data advantages in terms of lending, credit is a super power and so on and so forth. How do you think your models will hold up in the U.S. considering the change in demographics and how much legwork do you have to do to rebuild those models before you can have the same degree of confidence?

Rob Livingston

It's a really good question. And I think it dovetails nicely with our commitment that we've made that we are not planning to spend more than 100 basis points in our efficiency ratio on the U.S. entry. And the reason is it will take us some time to build up the same confidence in our credit risk models in the U.S. as we have in Brazil and Mexico and Colombia, where we've been operating for years.

The way to think about it is that the platform, the new performer platform for credit models and the credit risk expertise that we have in the company will translate very, very quickly across the border, but the actual data richness in building the experience of foundational testing and having the models in place that are specifically tooled for the U.S. market will take somewhere between 12 and 30 months, depending on the degree of maturation of those curves.

So our priority at the beginning of our entry into the U.S. market when that happens, will be to test learn, build out our data set and then be ready to expand once we have that same level of confidence there that we do in our core markets.

Guilherme Souto

Okay. With that, we -- sorry, we now surpassed 60 minutes of this session. So we are now concluding today's call. On behalf of Nu Holdings and our Investor Relations team, I want to thank you very much for your time and participation in our earnings call today. Over the coming days, we will be following up with the questions received by our platform that attempted, we were not able to ask to make questions tonight. So please do not hesitate to reach out to our team if you have any further questions. Thank you, and have a good night.

Operator

The Nu Holdings conference call has now concluded. Thank you for attending today's presentation. You may now disconnect.