MELI Trans: Revenue tops $10bn; margins still give way to investment.

Dolphin Research Trans of MELI FY26 Q2 Earnings Call

I. Key Takeaways

1) Revenue and profit. Net revenue topped $10 bn for the first time, up 50% YoY. OP was $683 mn with an OPM of 6.7%, roughly flat QoQ but down 550 bps YoY, reflecting a deliberate choice to prioritize engagement, growth, and scale over near-term margin.

2) Credit portfolio and asset quality. Scale: Credit book reached $16.4 bn, +75% YoY. Delinquencies: 15–90 DPD at 7.0% for the total book and 4.6% for cards, both near historical lows; Brazil broadly flat YoY. Spread: NIMAL improved from 18% in Q1 FY26 to 21% in Q2, with gains across the three core markets, led by Brazil consumer credit as provisioning normalized and spreads recovered.

3) Cash flow and capex. Adj. FCF was $214 mn this quarter. This was delivered after higher capex of $441 mn and $2.1 bn deployed into the credit book.

4) QoQ margin bridge: broadly flat, driven by opposing factors. Positive: Credit profitability improved, especially Brazil consumer credit, as the temporary spike in Q1 provisions normalized and spreads reverted. Positive: Strong scale effects in Brazil diluted fixed costs; OpEx ratio was diluted by 250 bps QoQ. Negative: Brazil e-commerce investments (lower take-rates in select verticals and PIX payment subsidies). Negative: Acquiring margin compression, mainly in Mexico, due to chip cost inflation lifting device costs and a large inventory build with upfront loss recognition on below-cost devices. Negative: Higher energy prices raised logistics costs, partially passed through and partially absorbed.

5) High-growth, high-margin engines fund reinvestment. The credit book grew 75% YoY, ads grew 73% YoY, and acquiring also expanded. The company is reinvesting scale benefits into free-shipping thresholds, 1P, and CBT. Investments follow strict discipline: each initiative carries engagement and growth targets with a clear path to profitability, and spend is tuned to outcomes.

II. Details from the Call

2.1 Management Remarks

1) Brazil free-shipping threshold: one-year readout. In Q2, items per buyer in Brazil rose 19% YoY, despite a steady inflow of new buyers who typically purchase less than average initially, indicating behavior change rather than just user growth. Conversion in Brazil jumped by 110 bps YoY in a step-change that has persisted for a full year. DAU/MAU inflected; post-threshold cut, DAU growth has exceeded MAU growth every quarter. New-buyer cohorts since the threshold cut show higher items purchased after one year, broader category coverage, and better retention vs. prior cohorts.

2) 'Ecosystemic users' were the focal theme. Defined as users active on both the marketplace and Mercado Pago, not just one side. These users drive materially higher GMV, broader category mix, and deeper Fintech usage. Their profit contribution is multiples of the sum of 'commerce-only' plus 'Fintech-only' users, underpinning the current level of investment.

3) Structural shifts in credit. High growth with stable asset quality reflects underwriting discipline at scale and a multi-year mix shift toward lower-risk users. Consumer and merchant lending have been deliberately moving up the risk curve, and cards are issued only to lower-risk users. Tight integration with the marketplace provides a large pool of high-quality, high-engagement users to underwrite, supporting continued expansion of the book.

4) Ecosystem flywheel and investment philosophy. A larger, more engaged marketplace increases the odds of building LatAm's largest digital bank and a bigger ads business; in turn, a stronger Mercado Pago makes the marketplace more compelling. These are not two parallel businesses but a single flywheel with mutual reinforcement. Invest with discipline only when unit economics are validated; as the flywheel spins, engagement, loyalty, and scale rise structurally. For nearly three decades, the core playbook has been to invest in removing friction and deepening engagement, then let compounding work.

2.2 Q&A

Q: Gross margin compressed more than expected QoQ. How much of the pressure from acquiring, memory costs, and higher shipping can be passed through via pricing? Were these headwinds embedded when you guided 'flat vs. Q1' margins in early May? (Goldman Sachs)

A: The best lens is QoQ OPM, which was broadly flat, driven by offsetting forces. On the positive side, credit margin improved as last quarter's temporary provision spike normalized and the business reverted to attractive spreads. On the negative side, first was Brazil e-commerce investment previously flagged: lower commissions in select verticals and PIX subsidies for consumers, a strategic lever with proven positive effects but some margin compression. Second was acquiring margin compression: part from industry chip inflation raising device costs, which is likely persistent and will be monitored, and part from a one-off effect in Mexico as we materially restocked devices to capture strong growth; because devices are sold below cost, we must recognize losses upfront when purchasing, so inventory builds compress margins temporarily and should be absorbed through the rest of the year. Third was energy-driven logistics cost inflation; some has been passed through, some absorbed, leaving a small but not material drag. We are not raising POS device prices for two reasons: payback remains on track, and competitors have not raised prices, so increasing prices amid rapid growth and share gains would not make sense.

Q: Brazil active seller growth accelerated to 29%. What was the prior baseline? Do new sellers drag profitability via higher promo uptake or mix shifts? Is there a seller 'maturity curve' akin to buyers? (Morgan Stanley)

A: Cutting commissions in Brazil goes back to e-commerce fundamentals: buyers choose the widest selection, best price, fastest delivery, and best installments. Securing the right assortment is strategic, which is why we invest in 1P and CBT, and why we reduced commissions again this quarter as in 2024 and 2025. This lever has been validated repeatedly: each cut accelerates the number of effective sellers. Together with Brazil's improved value proposition, successful seller growth has clearly accelerated over the past year, especially since last quarter's action. From a unit economics and profitability standpoint, the seller acceleration has no notable adverse impact; while there are programs to help sellers scale, they do not materially affect margins.

Q: Early read on the newly launched gamification and points program in Brazil? (Bank of America)

A: It is early; we have only run A/B tests for a short period. The impact so far is positive, with higher engagement among exposed users, but nothing concrete to share yet.

Q: What behavioral impact has the agentic shopping pilot had on search, conversion, frequency, ad CTR, and ecosystem behaviors? (Bank of America)

A: Two big things on agentic for commerce. First, AI-enhanced search is already delivering positive results in marketplace conversion and items sold; on the ads side, more AI in search improves context, enabling better ad selection and placement, higher CTR, and thus higher revenue. Second, the shopping assistant remains in A/B testing with partial rollout, and while we have no shareable metrics yet, early signals are encouraging. From a broader engagement perspective: the larger and stickier the marketplace, the greater the chance to build LatAm's largest digital bank; conversely, a stronger Mercado Pago makes the marketplace more attractive by offering the best installments and payment options where people shop. Few companies globally operate at this scale at the commerce–Fintech intersection in LatAm, creating a distinctive flywheel that is hard to replicate regionally, which is why we emphasize engagement this quarter—the investments on both sides make the flywheel spin faster.

Q: How is the ramp of Argentina credit cards? Are early results satisfactory? (JP Morgan)

A: We are very optimistic on Argentina cards. Issuance started around last Sep, so three quarters in we see strong demand and adoption and usage, particularly card-on-platform payments, which contributes to Mercado Pago volumes within the marketplace. It is early to discuss payback; the first cohort is only nine months old, but repayment behavior is in line with expectations with no surprises. Despite some institutions' concerns on Argentina's credit cycle, we remain pleased with our issuance given our reach in that market—most people use Mercado Pago daily or monthly—allowing us to select lower-risk users. Overall this reinforces both the marketplace and Mercado Pago in Argentina.

Q: Investors fear a deterioration in Brazil's credit cycle from 2H to 2027. Any signs of material worsening by product today? What proactive measures can you take if the cycle turns? (Itau BBA)

A: We see no slowdown or deterioration in Brazil's credit book now. Delinquencies are roughly in line with a year ago and better than last quarter, near historical lows, so we see no worsening, and we remain conservative in underwriting. On the outlook, we have operated through down cycles in Brazil and are currently navigating tough macro in Argentina; in both cases we remain cautious and have reduced limits or tightened credit when needed. Confidence in our models is higher than in the past and they are performing better. At a higher level, delinquency trends across the region and products are actually near historical lows almost everywhere, reflecting both our risk-first origination policies and our technology—this is a tech company with extensive tech deployed in underwriting, and that combination is key to success.

Q: With the Brazil free-shipping cut now lapping one year and a higher base, will you add more investment to consolidate share? What's the next frontier for Brazil e-comm (quick commerce, social commerce, pharma)? (Bradesco BBI)

A: The one-year results in Brazil are excellent. Item growth rose 56% vs. 26%–26.5% a year ago. The most striking number on engagement is conversion—despite much higher traffic, conversion still rose 110 bps YoY, which is very large at MELI's scale. DAU growth outpaced MAU, users bought across more verticals, engagement deepened, purchases increased, and transaction frequency rose; buying frequency (transactions, items per buyer) in Brazil climbed ~20%. We do not base investment decisions on the base effect but on the intrinsic value of the areas: strategic relevance to the marketplace, contribution to engagement and growth, and reinforcement of market leadership to capture long-term opportunities. Importantly, there was no one-off step-up followed by flatlining; effects continued to improve sequentially, and Q2 comped against last year's heavy changes and still grew well. We are not optimizing for growth per se and will not spend just to sustain a high growth print; we will do what is necessary to enhance the consumer value proposition. Some directions we like and may test, but we will not pursue step-changes in spend or strategy merely to stack on top of last year.

Q: EBIT margin was roughly in line QoQ, with investments offset by better NIMAL in consumer credit. Is this margin level sustainable into 2H? Would you trade off if the environment worsens? (Cantor Fitzgerald)

A: QoQ, the main driver was improved Brazil consumer credit, plus strong scale effects in Brazil diluting fixed costs—OpEx ratio was diluted by 250 bps QoQ, and we chose to reinvest those gains elsewhere. This philosophy stands. Some businesses are highly profitable and growing fast: the credit book grew 75% YoY, ads over 70% YoY, and acquiring as well. We are reinvesting the scale benefits into lower free-shipping thresholds, 1P, and CBT. Looking ahead, investments will be highly disciplined, each initiative with engagement and growth targets and, more importantly, a clear profit path; we manage to those and tune spend intensity accordingly.

Q: You rolled out Claude to employees last quarter. How is usage and spend ramping, and where are early benefits? (Cantor Fitzgerald)

A: AI investment rose by approx. $80 mn YoY this quarter, with strong results. On the consumer side, multiple initiatives are paying back: Mercado Pago's AI agent and MELI's seller-side systems are scaling well. As disclosed in the shareholder letter, usage of the app orchestrator grew 66%, lifting engagement and onboarding more sellers to ads, helping ads grow 73% YoY. AI tools on search across our five largest countries raised costs due to LLM fees, but the combined incremental sales and ad conversion benefits already exceed costs, yielding positive ROI. On productivity, customer support is illustrative: four years ago we had 10k agents vs. 7k today, while business volume has tripled, because 90% of interactions require no human. Engineering productivity gains are even larger: ~20k developers now use AI; a year ago AI assisted humans, today hand-written code is the exception and most code is AI-generated. This shows up in P&L: R&D as % of revenue fell from 8.4% to 7.2%, even including incremental AI costs. We are optimistic and disciplined on cost; per-token costs continue to decline.

Q: How is the affiliates program scaling, what gaps remain, and will you invest more? (XP)

A: Affiliates are scaling well with improving efficiency. In Q2, affiliate GMV mix rose in every market, including Mexico, where we also trimmed some prior couponing. Buyer quality is key: affiliate-acquired buyers retain materially better than non-affiliate, indicating the channel drives repeat users, which was the goal when we launched and accelerated affiliates in Q2 2025. One year on, unit economics and S&M as a % of revenue are broadly stable, with strong advances in product, engagement, and the number of affiliates. Overall, economics are improving and seller co-funding is rising; we are constructive across markets.

Q: Roughly half of Mexico's direct margin compression came from acquiring. How much was memory chip inflation that will be absorbed, vs. deliberate CAC? (FT Partners)

A: Mexico acquiring margin compression mainly reflected two items: higher device purchases and inventory, which require upfront loss recognition since devices are sold below cost, and higher memory chip costs. The vast majority came from these two, not price cuts—we did not cut prices.

Q: Mexico GMV slowed on tax changes. Are those headwinds fully reflected now, including in Q2? (FT Partners)

A: The tax changes we discussed last quarter indeed weighed on growth, compounded by softer macro and lower consumption during the World Cup, with a bigger impact in Jun–Jul. Even so, growth remains solid: our share continued to rise YoY and by more than major competitors. Versus traditional retailers, the structural growth opportunity in the country favors us—e-commerce is taking share. Longer term, Mexico may be the most favorable ecosystem market given low access to financial services and Mercado Pago's leadership in digital banking and Fintech acquiring. We are driving digitization, spinning the flywheel and strengthening both Fintech and commerce. We remain bullish despite near-term challenges; these do not change Mexico's long-term growth trajectory or earnings power.

Q: Token costs rose by $80 mn YoY while R&D ratio fell by ~1 ppt, implying positive ROI. Can this trend continue to benefit the P&L? (Raymond James)

A: To clarify, the ~1 ppt YoY decline in R&D ratio is not mostly from AI. We have diluted R&D for years via better management and productivity; that trend continues even after including most AI costs. On AI specifically, we see organization-wide productivity gains, with some initiatives showing direct ROI—customer support and AI in search—and others still in testing. We are optimistic that, given our tech management approach, data, and position as a tech company, we can scale AI to materially expand revenue generation and sharply improve operating efficiency. Strategically, we view AI as an accelerator of bigger opportunities, not merely a cost line to optimize. AI is speeding up long-term shifts we already target: more personalized discovery, lower-friction transactions, more customized credit, and sharper underwriting. With 27 years of proprietary data across commerce, payments, credit, and logistics, we are well positioned; AI is essentially adding fuel to a tech-first organization. It also aids cost efficiency: FY26 may be the first year in many where the engineering team does not expand, driven by sustained developer productivity gains.

Q: Please unpack the most mature card cohorts and the NIMAL trajectory over the next few years. (Raymond James)

A: Looking at older cohorts, primarily in Brazil where we started first, cohorts typically reach NIMAL breakeven in 12–18 months, a very stable pattern, and profitability improves thereafter. Accelerated issuance inherently increases investment, explaining why card NIMAL is down YoY—near breakeven a year ago vs. -2.5% now—mainly because issuance sped up from 1.6 mn cards in Brazil a year ago to 2.6 mn this quarter. We are confident in payback. Payback also extends beyond the card itself: NPS rises upon issuance, users are more likely to become ecosystem users with higher engagement and profit contribution, helping us achieve 'principality' with them, a key marker for becoming one of the region's largest digital banks. Moreover, Mexico's payback is even better than Brazil due to a superior industry model, while Argentina is early but promising.

Q: Provisions and short-dated delinquencies improved, but 90+ DPD ticked up. How to read this? Is risk cost improvement seasonal and likely to reverse QoQ? What's the 2H risk appetite, especially in Brazil? (UBS)

A: Overall NIMAL is very healthy, with 15–90 DPD at cycle lows. For 90+ DPD, note it mixes all products, so the signal is noisy, and it varies with growth—higher originations mechanically reduce the ratio, while slower originations lift it. Specifically, late last year issuance/loan pacing changed; some early-vintage borrowers ran slightly above expected defaults, and we then slowed that product, so 90+ DPD saw a small uptick. This is not concerning in absolute terms and matters less for profitability than the 15–90 DPD bucket. Also, faster-growing, shorter-duration products naturally show higher delinquency—good customers roll off quickly while defaulters remain in the book up to 360 days—creating some distortion. The book remains healthy with solid profitability, so this should not be overstated.

Q: How do lower Brazil policy rates affect each segment? (HSBC)

A: In acquiring and credit, small moves in Brazil rates are not significant—markets, competitors, and we adjust pricing to Selic. Legacy loans benefit marginally as they were priced at higher rates, but this is edge-only and resets when new pricing cycles start. On the marketplace, the impact can be somewhat larger given less frequent repricing and widespread installment pricing, marginally improving the marketplace take rate. Typically, OPM improves slightly when rates fall and compresses when rates rise, as we do not reprice at every move.

Q: What is the share of ecosystem users in Brazil or overall marketplace, and how much runway remains? (BTIG)

A: The two-sided ecosystem enables users to be active on both platforms, and we want the market to grasp their importance: ecosystem users have 70% higher marketplace GMV, 90% higher TPV, and double the assets under management, with much higher engagement and profitability. They are also the fastest-growing user type, up about 37% YoY. We do not disclose their exact share, but this is a very important and fast-growing cohort, and many of our commerce and Fintech investments aim to convert more users into ecosystem users.

Q: CBT progress on profitability and cost efficiency? How is utilization at the China FC, and will you add other low-cost locations? (BTIG)

A: We are very pleased with CBT's trajectory. CBT GMV grew about 60% YoY, with triple-digit growth in Brazil, Argentina, and other markets, and above-average growth in Mexico, our largest CBT market. Notably, volumes from the China fulfillment center rose 170% QoQ—we built capacity first, and volume is following. CBT improves service levels and selection at attractive prices; more supply drives more demand, which further attracts supply. The China FC also sped up delivery and reduced cancellations, lifting NPS and boosting retention and repeat purchases. Unit economics have improved QoQ for several quarters, driven by scale dilution of costs and progressing along the learning curve to pull the right levers. We are satisfied and encouraged, though CBT remains early with much to build.

Q: The card book is now over $7 bn with substantial potential spreads. Are we nearing an inflection to profitability? How will net interest income and NIMAL trend over the next two years, and how much acceleration is left? (Citi)

A: Card evolution follows the established pattern, with each cohort progressing toward profitability on schedule. The timing of the inflection depends on the cohort mix: issuance has accelerated, so cohorts older than three years are a small share, with most of the book from the past two years, implying average maturity needs more time. Confidence in underwriting quality is what allowed us to speed up issuance, which in turn lowers the book's average maturity. Thus the inflection depends on both future issuance pace and the natural maturation of existing cohorts. Strategically, beyond the card's own profit path, cardholders are 2–3x more likely to become ecosystem users, buy more on MELI, and move more of their financial lives to Mercado Pago. Cards are strategically important to both Mercado Pago and MELI.

Risk Disclosure and Statement:Dolphin Research Disclaimer and General Disclosure