Prepared remarks
Hello, everyone. Good evening, and thank you for joining MercadoLibre's conference call for the quarter ended June 30, 2026. I'm Richard Cathcart, Senior Director of Investor Relations. And I'm joined this evening by Ariel Szarfsztejn, our CEO; Martin de Los Santos, our CFO; and Osvaldo Giménez, President of Fintech. Tonight's format will be slightly different. I'll start by handing over to Martin for a few high-level opening remarks, and then we'll head straight into the Q&A. Before we do that, I would like to remind you that management may make or refer to forward-looking statements and non-GAAP measures. So please refer to the disclaimer on screen, which is also available in our earnings materials on our Investor Relations website. Please note that this call is being recorded, and a replay will be made available on our Investor Relations website as well. With that, I'll pass the call over to Martin de Los Santos, our CFO.
Good afternoon, and thank you for joining us. MercadoLibre delivered another strong quarter in Q2 2026. Net revenue surpassed $10 billion for the first time, growing 50% year-on-year. Income from operations was $683 million, with a margin of 6.7%, broadly in line with last quarter and the result of a deliberate choice to continue prioritizing investment in long-term engagement, growth and scale over near-term profitability. Before turning to a few details on this quarter's results, I want to spend a few minutes on what we've learned one year on from the decision we took last year to lower the free shipping threshold in Brazil. In Brazil, items per buyer grew 19% year-on-year in Q2 despite the fact that we've been adding large numbers of new buyers, who typically start out purchasing far less than an average user. That's a sign of changed behavior, not just a bigger audience. It shows existing users engaging more deeply with us, not simply more people showing up. Conversion in Brazil is up 1.1 percentage points year-on-year. And this wasn't an incremental gain. It was a step change, and it has proven sustainable for a full year. The ratio of daily to monthly active users has inflected too, with daily actives growing faster every single quarter since the free shipping threshold was lowered and new buyer cohorts who joined after the change are a year on, purchasing more items across more categories with higher retention than the cohorts that came before them. This is not an isolated case. Across our business, we see the same pattern. Users who engage more deeply with us become dramatically more valuable. The clearest evidence is what we call ecosystemic users, those who use both our Marketplace and Mercado Pago, not just one or the other. These users generate meaningfully more GMV, purchase across a much wider range of categories and engage far more deeply with our Fintech products. Most importantly, they are dramatically more profitable. Contribution profit per ecosystemic user is multiples of the sum of a Marketplace-only user and a Fintech-only user. That is why we keep investing the way we do. We are changing behavior and building habits we believe will drive this business' profitability for years to come. With that strategic context in mind, I now would like to turn to three topics we believe are top of mind for investors and where we think it is worth spending a few additional minutes. First, our credit business. Our credit portfolio reached $16.4 billion in Q2, growing 75% year-on-year. We have achieved this growth alongside solid asset quality across the portfolio, which reflects our disciplined approach to risk management as we scale and the shift towards lower-risk users in recent years. Our 15- to 90-day NPL in Q2 was 7.0% for the total portfolio and 4.6% for the credit card specifically, both close to historical lows. NIMAL improved from 18% in Q1 2026 to 21% in Q2 2026, with gains in our three largest markets. This includes Brazil, where spreads in our consumer portfolio recovered as provisions normalized. NPLs in Brazil were broadly stable year-on-year. These results are not coincidental. They reflect the deliberate move up market in our consumer and merchant credit portfolios and the scaling of our credit card, which we only offer to lower-risk users. This is highly synergistic with our Marketplace, where we have a large base of high-quality engaged users to draw from as we continue to grow the book. Second, a quick word on margins. This quarter's EBIT margin of 6.7% was down 550 basis points year-on-year as we continue to prioritize long-term strategic investments over short-term profitability. These investments are consistent with the areas of focus we have described in previous quarters. On a sequential basis, our margin was broadly stable. This is a function of stronger profitability in credit, particularly in the Brazil consumer portfolio, where provisions normalized after a spike in Q1. This was offset by margin compression in acquiring, primarily in Mexico and incremental investments in Commerce. Finally, on cash flow. In Q2, we continue to see the underlying cash generation strength of the business. We generated $214 million in adjusted free cash flow for the quarter, even after absorbing higher capital expenditure of $441 million and investing $2.1 billion into the growth of our credit book. Our credit book continues to be very profitable, and it also plays a key role in engagement across the ecosystem. It's this overall strength, strong cash generation, robust profitability and a healthy balance sheet that gives us the confidence to keep investing at the pace we are. With that, we'll open it up for your questions.
Questions and answers
The first question comes from Irma Sgarz with Goldman Sachs.
Thanks also for all the detail in the shareholder letter on the engagement metrics across the ecosystem, very interesting detail that you provided there. However, I'll just be obliged to ask you about margin for one second. There was a bit more quarter-over-quarter gross margin compression than we had perhaps expected. So thank you for also laying out the sources that drove this in the shareholder letter. Now when I think through each of those, the first one, I think, was well mapped. The second and the third one felt a little bit more incrementally new. So my question would be somewhat twofold. Firstly, how much scope do you see to adjust pricing to pass these cost pressures through to consumers? I think there's some in acquiring, there's some related to memory costs, but you also do mention higher shipping costs that weren't fully offset by revenue. So I was wondering if that's gas prices or other pressures in the network and whether you saw scope to pass that through to customers? And then secondly, whether those two incremental headwinds, either in nature or magnitude were something that you were already factoring in when you last spoke to us in early May and you had sort of suggested that you were expecting a similar margin level consistent with Q1.
Irma, it's Martin here. Thank you for your question. I think the best way to answer your question is to look at sequential margins. As you can see, it's pretty much in line with Q1, but there are two things moving in opposite directions there. The first one was an improvement in margins in the credit business, as we discussed in the letter. Last quarter, there was a spike on provisions. It was a temporary spike. We explained that last quarter, and now it has come back to normal spreads with a very profitable business. So that contributed to margins. That was offset, and I will go in order of importance. It was offset by some investments that we have made in Brazil in Commerce. We discussed this in the previous quarterly call, where we lowered certain take rates in certain verticals, and we offered discounts to consumers buying and paying with PIX. That was a strategic initiative that we did in Brazil. We have done that in the past, and it has proven to have positive results, but it generated some margin compression. In addition to that, there was some margin compression on the acquiring business. Two things there. One is the fact that the devices have an increase in cost because of higher cost of chips in the industry. So that's something that's there to stay, and we will monitor that situation as well. And then there was a one-off charge because we, in Mexico, restocked a significant volume in terms of devices. That's because we are growing very fast in Mexico. We see a big opportunity to continue growing. And as you know, as we sell those devices at a loss, we need to book the losses upfront. So when we increase the stock, there's a one-off compression on margin. So that will compensate throughout the rest of the year. And the last point I think you're addressing is the energy cost. We've seen some cost increases in terms of logistics because of energy costs. Some of them we pass on to our users and some we elected to absorb ourselves. So there is some margin compression. It's not significant, but there's a slight margin compression there as well.
And building on Martin's comment regarding POS devices and us not increasing the prices, that was driven by two things. On the one hand, the payback periods continue to be in line with what we expected, and therefore, there was no need to do that. And also our competitors did not raise their prices. So it didn't make sense given how fast we're growing and how fast we're gaining market share to raise our prices.
The next question comes from Andrew Ruben with Morgan Stanley.
I'm interested in some dynamics around Brazil seller growth. You mentioned the acceleration to 29%. I don't think you've mentioned that one before, so just to level set what that was growing before the select discounts. And more fundamentally, is there a profit drag from these new sellers? Maybe they were more likely to take up your promotions or maybe something structural about the new seller mix? And really, what I'm trying to understand here, is there an element of maturation on your platform where a new seller starts to sell more over time, the seller economics change over time, the way we think about buyer economics? Again, I think that was a new call out the growth. So I'd like to understand how that plays in.
Andrew, Ariel here. Great to hear you. So yes, I think what we did in Brazil with take rates goes back to the basics of e-commerce, right? So when consumers are deciding where to shop, they are basically looking for the broadest selection at the best possible price with the fastest shipping and the best financing. And clearly, getting the right selection is a key part of our strategy. That's why we've been doing 1P. That's why we've been doing CBT. And that's why this quarter, as we did back in 2024 and 2025, we decided to lower seller take rates. We've already proven that lever, and we've consistently seen that every time we lower take rates, we get an acceleration in effective or successful sellers in our platform. Clearly, given everything that we've been improving in terms of value proposition in Brazil, the trend of growth in successful sellers has accelerated over the last year and particularly over the last quarter with the initiative that we just mentioned. In terms of unit economics and profitability coming from the acceleration of sellers, I would say there's nothing big to call out. So while there are some programs in order to get sellers to scale and so on, there's nothing really affecting our margins on that one.
The next question comes from Bob Ford with Bank of America.
Ariel, in the letter to shareholders and in the comments, you touched on some very compelling engagement and ecosystemic behaviors. What's your early read on your new gamification and points program in Brazil? And could you also touch on some of the behavioral implications from your agentic shopping pilot in terms of search conversion, frequency, ad click-through rates and your ability to promote more ecosystemic behavior from the engagement that you get agentically?
Bob, so I think it's too early to comment on the gamification initiatives and points. We've only rolled them out as an A/B test for some time. So it's early to actually make a definitive point. I would say the impact so far is positive. We see increased engagement coming from those users who do engage with the proposal, but nothing to share yet. On the bigger picture, I would say, agentic, there are two big things we're doing in Commerce with agentic. On the one hand is improving our search through AI, and that's proving to have positive results, both in terms of conversion rates in the Marketplace, items sold and so on, but also on our advertising platform—the more AI we deploy to search, the better context we provide, the better ads we are able to pick, choose and display to our consumers and the more CTR that we get from those ads, so higher revenue. On the second initiative, which is our shopping assistant, we are just A/B testing that one. So nothing to really share in terms of engagement and results, but we are very excited with the early results we are seeing on the shopping assistant, which is live for some consumers in the Marketplace. Taking a step back and picking up your point on engagement, I would say that we look at engagement with a broader view. So the bigger and the more engaging our Marketplace become, the better chances we have of building the largest digital bank in Latin America. So the bigger the Marketplace and the better the Marketplace, the better our Mercado Pago application works. And conversely, the better Mercado Pago becomes, the more appealing our Marketplace becomes to our users, basically because users find the best financing and payments alternative right at the place where they are already shopping. And basically, we see very few companies anywhere in the world who operate at this intersection of Commerce and Fintech at the scale that we do it in Latin America. And that's really creating a unique flywheel that is very difficult to replicate for any other player across the region. So we are very satisfied, and that's why we've been so precise on highlighting engagement this quarter. We think this is a unique competitive advantage that we have. And basically, the results that every single one of the investments that we have been making both across Commerce and Fintech are playing in favor of making that flywheel turn faster and faster.
The next question comes from Marcelo Santos with JPMorgan.
Could you discuss a bit how the ramp-up of credit card in Argentina is going? Are you happy with the early results? Is it progressing well? Just anything would be great.
Marcelo, we are very excited with how the credit card in Argentina is going. As we know, we only started issuing cards three quarters ago, roughly around September last year. And we have seen a lot of demand, and we are seeing significant adoption and use and it's significant, for example, how much people are using the card to pay on the MercadoLibre platform. It's contributing to the amount of payments with Mercado Pago means of payments within the platform. Still early to talk about payback periods because it's only nine months in the first cohort. However, we are glad to see that the payments are in line to what we expected when we started issuing the cards. So there have been no surprises there. And even in an environment where some other financial institutions are concerned about the credit cycle in Argentina, we are very happy with the issuance we have had in Argentina. We think that given our penetration in the market where the majority of the country use Mercado Pago every day or every month and every day, we have been able to cherry-pick those users, which we deem to be less risky. So in general, I would say we are very happy and that this strengthens both the Marketplace and Mercado Pago and the presence that Mercado Pago has in Argentina.
The next question comes from Rodrigo Gastim with Itaú BBA.
Just a quick question here on the credit cycle in Brazil. Investors are quite concerned about the potential deterioration during the second half of the year and 2027. So just trying to understand here two parts of my question. So number one, if you understand that so far, so good. In other words, when you look at most of the products at the NPL of these products, so far, no important deterioration or signs of deterioration. That's the first part. And the second one is, what can you guys proactively do to protect yourselves from an eventual deterioration in the credit cycle, something that you cannot control. So what are you doing here? That would be my question.
Rodrigo, yes, I confirm that so far, we are not seeing any deceleration or deterioration of the credit book in Brazil. If you look at NPLs, they are roughly in line to what they were a year ago, and they are even—NPLs are even better than they were last quarter. So they are, I would say, nearly bottoms of NPLs that we have had. So we don't see any deterioration. I think that we have been very, I would say, conservative in terms of whom we issue credits to. And with regards to what we expect for the future, I would say that we have been through a downgrade cycle in the past in Brazil, and we have been through adverse macro conditions in Argentina. And in both cases, we have been cautious. And when we thought that we had to curtail the number of available lines or the lines of credit, we have done so. At this point, we are confident that NIMALs are working better than they were in the past.
Just to add on Osi's comments, I think taking a step back, what you see in terms of credit and deterioration, it's actually the opposite. I mean if you look across the region, across products, we are almost at our all-time low NPLs for every single product and region. And I think that's a testament to, a, our risk management policies and how we are able and we are prioritizing risk at the moment of issuance; and b, the power of our technology, right? We are a tech company, and we've deployed lots of technology in underwriting and the combination of the two has been vital for us to be successful.
The next question comes from Pedro Pinto with Bradesco BBI.
My question is about now that we completed one year since the implementation of the lower free shipping threshold in Brazil, which has been pretty effective in GMV acceleration in Brazil. Very clearly mentioned in the letter. Now the comps get tougher, would you guys plan additional investments to continue consolidating the market? What would be the next frontier for Brazil commerce at this point? I don't know if it's quick commerce, social commerce, pharmacy category. What should we expect as a priority strategy-wise for Brazil commerce from this point onwards for momentum to persist?
Pedro, it's Martin here. We are—as we described in the letter, I think the results of lowering the free shipping threshold in Brazil after one year are amazing. I mean we are seeing tremendous growth. If you look at items growing at 56%, they were growing 26% or half of that rate a year ago. Engagement with the platform is incredible—the number that among most amazing is conversion. I mean we have been growing significantly in terms of traffic. But on top of that, we grew our conversion by 1.1 points year-on-year. At the scale of MercadoLibre, that's enormous in terms of volume. But also vibrancy and frequency on our platform, daily active users growing much faster than monthly active users. We're getting more users also buying in more verticals or becoming more engaged in different verticals within the ecosystem, buying more times with more frequency. In Brazil, frequency of purchase increased by 20% in number of transactions or number of items per user. So all the metrics that we wanted to influence were there. Of course, like you said, comps become tougher, but we don't make investment decisions based on comps, obviously. We make them based on the merits of the areas where we're investing. We're making sure that those are strategic to our marketplace. They contribute to engagement and to growth, and they contribute to strengthening our market position, our leadership position to capture this long-term opportunity that we have in front of us.
Yes. So just complementing Martin here. We have not seen a step change from one day to the other. We lowered our free shipping threshold and then steadily saw metrics in terms of engagement, conversion and so on. The different effects that we've seen across our Marketplace have been consistently improving quarter after quarter. And with that, you can see the numbers from Q2, right? The results from this quarter already reflect a lot of the changes that we've implemented last year, and we're still growing very, very nicely. So picking up on Martin's point, we are not optimizing for comps. It's not that we will be deploying one thing or the other in the future just for the sake of keeping growth high. We will just find whatever we think is needed in order to continue improving the value proposition for our consumers. Of course, there are things that we will test and so on, but it's not that we are making a step change in investments or strategies just for the sake of compounding versus last year changes.
The next question comes from Deepak Mathivanan with Cantor Fitzgerald.
Just wanted to follow up on EBIT margin. EBIT margin came in pretty much as you anticipated sequentially in Q2. It does seem like some of the investments were offset by improvements in NIMAL on the consumer credit side. Can you comment about the sustainability of margins at these levels in the second half? Are you committed to making trade-offs if potentially operating environment somewhat changes and becomes unfavorable? And then perhaps a related one on AI cost. You rolled out Claude to employees last quarter. Can you talk about the usage and how the spend is ramping? What are some of the ways you're seeing benefit in the early days?
It's Martin here. I think if you look at margin sequentially, as I mentioned before, the main driver of margin improvement was the improvement that we saw on the portfolio of consumer credits in Brazil. And in addition to that, we've seen some scale—very strong scale in terms of growth in Brazil that enabled us to dilute fixed cost. As you've probably seen over the past several years, we have been diluting OpEx very nicely. In this particular quarter, we diluted OpEx by 2.5 points quarter-on-quarter. So that's also contributing. And we elected to reinvest that margin into other areas of the business, as I explained before. And I think the philosophy continues to be the same. We have some areas of the business that are delivering very strong profits and are growing very fast. If you look at our credit portfolio growing at 75% year-on-year, our advertising business growing 70-plus percent year-on-year, the acquiring business as well, we are continuing to deliver scale because of the growth that we are delivering consistently. So we are electing to reinvest in areas of the business, as we discussed before, lowering the free shipping threshold, 1P, CBT and so on. That philosophy will continue to be the same going forward. We invest in a very disciplined manner. We have engagement and growth targets for each initiative that we take on. More important than that, we have a clear path to profitability for those initiatives and we measure against those, and we leverage and we graduate the intensity of investments based on those results. So that's the first part of your question. Then you asked about AI. We are very excited about AI. We invested more this quarter compared to a year ago—we invested about $80 million on AI this quarter compared to a year ago—and we are seeing very strong results. Let me break it down in a couple of ways. Consumer-facing, we have several initiatives that are paying out very nicely. We talked about the Mercado Pago AI agent, the Seller Assistant on MercadoLibre—those continue to scale very nicely. In this quarter, we disclosed in the letter, the ad orchestrator that is increasing the engagement with our users grew by 66% in usage of that tool. That is a way to get more sellers to our advertising platform and is helping us to grow the advertising business by 73% year-on-year. I think Ari mentioned the AI tools that we deploy on our search engine in the five largest countries. And it's important to mention that—of course—that increased the price because we had to pay for LLMs. But when you put together the incremental volume that we sold plus the incremental conversion and advertising, it more than pays the cost of that initiative. So it's an initiative that has a positive return on investment. That's a good example of AI contributing to profits. Then on the productivity side, in the past we talked about customer service. Four years ago, we used to have 10,000 reps on customer service. Today, we have 7,000 reps, even though the business grew by three times during that period. And that's because 90% of the interactions are done without a human participating in the issue. In product development, that's tremendous productivity gains. We have 20,000 developers that are using AI. A year ago, they were helped by AI. Today, human written code is the exception. The majority of the code is generated or assisted by AI. And you can see that also flowing through our P&L. Product development scaled from 8.4% of revenues to 7.2% of revenues year-on-year, even though it has this incremental cost of AI within it. So we are very, very optimistic about it. We are seeing the results. We manage it with discipline. We are focusing on costs as well. Cost per token continues to come down, but we're seeing very positive results in terms of investments on AI and MercadoLibre.
The next question comes from Danniela Eiger with XP.
I'd just like to hear a little bit more about how you're seeing your affiliates program, what you're doing to scale it, current gaps that you see? And if you can expect any heavier investments in this front going forward?
Dany, this is Ariel. So the affiliate channel is actually scaling very, very nicely and becoming more efficient simultaneously. So we think it's actually working. The affiliate GMV share grew across every market in Q2, including Mexico, where we dialed down some of the couponing that we deployed in the past. So buyer quality is also a very important metric that we look at when evaluating the program. Affiliate buyers show materially higher platform retention than non-affiliate buyers. So the channel is not only driving one-time purchases, it's also acquiring users who come back later to our platform and that's really exciting as well. This is exactly what we wanted to achieve when we launched and accelerated investments in affiliate back in Q2 2025. Year on year, unit economics and sales and marketing as a percentage of sales are broadly flat. Simultaneously, we are making a lot of progress in the product, in engagement, in the number of affiliates and so on. So we're happy. Economics are improving. Seller co-funding for affiliate is growing. Extremely excited and positive on the outlook for the different markets.
The next question comes from Craig Maurer with FT Partners.
I wanted to just ask quickly about Mexico. Two points. In terms of direct contribution margin compressing for Mexico, roughly half of that was from acquiring. How much of that was the device cost pressure related to memory chip inflation that will eventually abate versus deliberate customer acquisition cost investment? And second, in terms of GMV decelerating with the tax reform headwind, is that headwind fully in the run rate now in Q2? That would be helpful to know as we think about modeling the back half of the year.
Craig, with regards to acquiring margin in Mexico, it was mostly related to both increasing the amount of devices we bought— increasing inventory as we sell the devices at a loss. Whenever we buy inventory, we need to book the loss upfront. And the rest was driven by, as you mentioned, the increase in memory chip cost. So I'd say the majority of the compression in Mexico is related to those two issues and not to a decision to lower our prices, which we did not.
Craig, on Mexico commerce and demand dynamics, the tax reform that we explained last quarter is definitely a headwind to our growth. That, together with some headwinds coming from a weaker macroeconomic environment and lower consumption during the World Cup, did weigh on our growth somewhat. But still, if you see our numbers, our growth rate was pretty robust. We continued gaining market share year-over-year, and we even gained more than our main competitor. Also when you compare our performance with traditional retailers, you can see how the structural growth opportunity in the country is playing to our favor. The market is growing. We are a large slice and physical retail is clearly being challenged by e-commerce. Mexico is perhaps the market where our ecosystem could play out the most in our favor in the long run because of the lack of access to financial services and Mercado Pago's positioning as the leading digital bank and the leading Fintech acquirer. We are driving digitalization in the country, and that's pushing the flywheel, which is strengthening both Fintech and e-commerce. So we remain optimistic with the market, although we do see near-term challenges, which were a bit deeper in June and July with the World Cup. There's nothing that is actually impacting the huge secular trend of growth and opportunity that we see in Mexico and, more importantly, the long-term earnings power that we find in that market.
The next question comes from Josh Beck with Raymond James.
I wanted to go back to the AI costs because I think you've shared a little bit more than we've generally heard on earnings calls. I think you said the cost of tokens was up $80 million year-over-year, but you're saving almost 1% of revenues in terms of product development. So I'm curious, as we look forward, could this be a trend that continues and maybe provides more ROI and benefit to the P&L? That's one question on AI costs. The other is on the credit card NIMAL breakdown you gave on credit card versus other—it was helpful. Could you double-click on the most mature credit card portfolios, how they look and how we should think about the evolution of credit card NIMAL in the years ahead?
Thank you for your question. Let me clarify the product development scale I mentioned. Most of that doesn't come from AI, just to be clear. We have been scaling product development for many years now, and that trend continues as we become more productive and more efficient in the way we manage our product development. What I tried to explain is that even though we include a majority of the AI cost within that line, that line continues to scale. Regarding the AI investment, we can see productivity gains throughout our workforce and certain initiatives where we can measure the actual ROI. The example was customer service or the initiative on AI applied to search—there are initiatives where we're experimenting. We are very optimistic. Given how we manage technology, the data we have and our position as a technology company, we should be able to take this new technology, scale it, and it will help us significantly to scale our revenue generation and make us more efficient in how we manage the business.
Let me take one step back on AI to add a complementary comment to Martin. Although we care about the bill and proactively manage every cost lever to make it more efficient, strategically we think of AI as an opportunity and accelerator of our larger opportunity, less as a cost line to optimize. AI is accelerating the secular shift that we are already trying to capture. Discovery is becoming more personalized. Transactions are becoming more frictionless. Credit becomes more tailored to any given consumer. We can underwrite better and so on. We are convinced that the 27 years of proprietary data we have accumulated across commerce, payments, credit, logistics positions us well to capture the opportunity that AI is bringing. This is fuel applied to an organization already wired into technology and the idea of using every technological lever to maximize results. In parallel, AI is contributing to cost efficiency. 2026 is probably the first year in many in which we are not growing our engineering headcount, and that's coming from AI driving developer productivity up consistently.
Josh, regarding your question on credit cards and older portfolios: what we continue to see mostly in Brazil, which is the market where we started first, is that each cohort typically reaches NIMAL breakeven after 12 to 18 months. That has been pretty consistent and then continued improving profitability. As we accelerate issuance, you see some compression on the NIMAL of the credit card portfolio between last year and this year. It was nearly breakeven a year ago and minus 2.5% now. But that is driven mostly by the fact that we were able to accelerate the speed of issuance. A year ago, we issued 1.6 million cards in the quarter. This quarter, we issued 2.6 million cards in Brazil. Therefore, we are confident that we are investing and we have certainty about the payback. Beyond the payback in the card itself, whenever we issue cards, we see a lift in Net Promoter Scores from those users. They are more likely to be ecosystemic users with higher engagement and higher profitability on the platform. It's one of the key contributors to gaining principal relationships with users as a step toward building one of the largest digital banks in the region. We are excited with how the credit card is evolving in Brazil. In Mexico, paybacks are even better than in Brazil because the economics of the industry work better there. It's still early to tell about Argentina, but the initial impression is very good.
The next question comes from Kaio Prato with UBS.
I have one question. It's more a follow-up on the credit side. I'd like to double-click on the asset quality of your credit book. We saw an improvement in provisioning levels this quarter with the cost of risk reducing quarter-on-quarter. We saw good trends on the short-term NPLs as well. But as your duration is short, it's tricky to look only at short-term NPLs. At the same time, the 90-day NPL had quite a meaningful deterioration this quarter. How should we read that? Was the improvement in cost of risk seasonal and should it be higher sequentially? And how would you describe your appetite toward the second half of the year, especially in Brazil as well?
One second, please. In general, Kaio, I would say our NIMALs are super healthy. NPLs are at the lowest point for 15 to 90 days. When it comes to 90 days, this metric mixes all products, so the signal is not so clear. It moves up and down depending on how fast we are growing. When we originate loans, the ratio improves automatically. When we slow down, it deteriorates a little bit. Part of what happened is that there was a change in the pace we were issuing at the end of last year and some early borrowers defaulted a little bit higher than we expected. Then there was a slowdown in issuance of that product because of that deterioration. So you see a little jump in the over-90-days bucket. But it's not something I would be concerned about. It's a small number. The 15- to 90-day bucket is significantly more relevant for profitability.
If I may complement, keep in mind that some products growing very fast now are relatively short-duration. Those products tend to have higher NPLs because the good payers get out of the portfolio quickly while defaulters remain for longer. So there might be some distortion because of that. But again, as Osvaldo mentioned, nothing to worry about. The health of the portfolio continues to be very profitable.
The next question comes from Neha Agarwala with HSBC.
Could we talk a bit about the impact from the lower policy rates expected in Brazil for your different business segments? What kind of impact can we expect from lower rates?
Neha, I would say that both on the acquiring and credit business, there is no significant impact by a small change in the policy rates in Brazil. Typically, the market, our competitors and ourselves adjust those rates depending on what's happening with the policy rates. There is marginal improvement in the credit you have already issued because you priced them with a higher rate, but it's really marginal. Once we get to the new standard, we adjust prices. There is probably a little more impact on the Marketplace because typically, we don't adjust as frequently the parcelada sem juros prices on the Marketplace. So margin in the Marketplace could improve when rates come down.
The Marketplace typically benefits slightly when interest rates come down because we don't adjust installment offers as often. When rates go up, we lose margin because we don't reprice immediately.
The next question comes from Marvin Fong with BTIG.
I apologize if these were asked before jumping between calls here. I appreciate all the information on ecosystemic users. Could you give an idea of what percentage of your users in Brazil or across the entire marketplace are ecosystemic users? I'd like to get an idea of how much future growth is possible there. And second, on cross-border, could you update us on your strategy to improve the profitability and cost efficiency of that product? I know you opened a facility in China. How much is that being utilized? Might you open additional facilities in low-cost regions?
Thanks for your question. Regarding ecosystemic users, we wanted to make sure it's clear the importance of users that are active on both our Marketplace and Mercado Pago. We wanted to give color—an ecosystemic user has 70% more GMV on the Marketplace and 90% more TPV, double the assets under management—very engaged and more profitable. We mentioned the growth rate of those users is the highest of any user type—growing 37% year-on-year. We don't disclose the actual share of users, but wanted to be clear these are important users, growing very fast. A lot of our investments aim to convert more users on both sides to become ecosystemic.
Marvin, on cross-border trade, we are extremely satisfied with the trajectory. CBT GMV is growing approximately 60% year-over-year with triple-digit growth in Brazil, Argentina and other markets and above-average growth in Mexico, our largest market for CBT. More interestingly, the volume coming from our Chinese fulfillment center is growing 170% quarter-over-quarter. We built the capacity and the volume is following. With CBT, we're providing a better service and more choice to our consumers—adding selection at attractive prices. The more supply we get, the more demand we get. The warehouse in China has enabled us to improve delivery speed, reduce cancellations, which is driving NPS positively and pushing retention and repurchase up. On unit economics, we continue to see sequential improvements in margins, consistent for a few quarters now, driven by scale and the learning curve that allows us to improve every dimension of the business. Very satisfied, encouraged, but still early days for CBT and we have many things to continue doing.
The next question comes from João Soares with Citi.
I wanted to double-click on the credit card profitability trajectory. Given the level where it is now—above $7 billion—and the yields and spreads on this product, is it fair to say we are reaching a level where we could expect a profitability inflection? Any color on where we should start seeing the NII and NIMALs for this product in the next couple of years would be helpful.
João, we are super excited with how the credit card is evolving and continue to see the same patterns in terms of reaching profitability for each cohort. Reaching a turning point is mostly related to the size of each cohort. We have been accelerating issuance, so cohorts older than three years are a small part of the portfolio; the majority has been issued in the last two years. We need more time for the average portfolio to mature. As we are confident in how we are issuing, we have accelerated issuance, which slows the average maturity of the portfolio. It's mostly a combination of issuance pace and cohort evolution.
If you step back, the strategic importance of the credit card goes beyond the direct profitability shown in the letter. People who use the credit card have two to three times more chances of becoming ecosystemic. They buy more on MercadoLibre and bring more of their financial life into Mercado Pago. So the credit card has a clear path to profitability, but it's also a strategic product for both Mercado Pago and MercadoLibre.
This concludes the question-and-answer session. I would like to turn the conference back over to Ariel Szarfsztejn for any closing remarks. Please go ahead.
So I want to take this opportunity and go back to one of the things I was saying at the beginning of the call. We are building something quite unique globally, an ecosystem of commerce and financial services that compounds on itself. The bigger and the more engaging our Marketplace become, the better our chances of building the largest digital bank in Latin America and the bigger our advertising business will become. Simultaneously, the better Mercado Pago becomes, the more appealing our Marketplace becomes for our consumers. MercadoLibre and Mercado Pago have become a daily habit for Latin Americans across the region, but the opportunity ahead of us is even bigger. These are not two businesses running in parallel. This is one flywheel with each side making the other one more valuable. As we said in the letter, we invest in it carefully with discipline. Only when we see that the economics are working do we increase intensity. If we get that flywheel right, the result is structurally higher engagement, loyalty and scale, and that is what will maximize our long-term profitability. The underlying philosophy with which we are operating has not changed in almost three decades: invest in eliminating friction, deepen user engagement and let it compound. Our conviction in this approach, the evidence supporting it and the ecosystem behind it are all stronger than they have ever been. We are convinced this is the right way to create the best long-term proposition for our consumers and through this to maximize shareholder value in the long run as well. Thank you very much, and looking forward to connecting with you next quarter.
This brings to a close today's conference call. You may disconnect your lines. Thank you for participating, and have a pleasant day.