The Gross Margin Stopped Falling, the Operating Margin Did Not: MercadoLibre De-Rates Below Where We Initiated

Results vs. Consensus

Q3 2026 Scorecard

MercadoLibre publishes no revenue, margin or EPS guidance, so the only pre-print bar is sell-side consensus. The EPS bar sat between $9.27 and $9.34 across three providers; $9.30 is used below, and the quarter is a mid-single-digit percentage miss at every point in that range.

Year-over-Year Comparison

Sequential Comparison

Q1 2026 columns for the expense ratios are derived from the filed six-month statement less the filed Q2 three-month statement, and the quarterly revenue, operating income, net income and EPS figures sum exactly to the filed nine-month totals of $29,764M, $1,970M, $1,331M and $26.25.

The Sequential Margin Bridge, and the Line Nobody Mentioned

Management attributed the sequential operating margin decline to two things: a net shipping headwind of 60 basis points, split 40 in cost of revenue and 20 in general and administrative, and 40 basis points of additional sales and marketing. That totals 100 basis points of named headwind against an actual decline of 43. The offsets are worth itemising, because one of them is not in the shareholder letter or on the call.

The reconciliation closes to two basis points, which is as tight as this exercise gets. The fourth line is the one that matters: without 30 basis points of relief from revenue-based taxes, a function of country and product mix rather than of logistics, the gross margin would have fallen roughly 40 basis points sequentially rather than 12. The headline that the gross margin has stabilised is true of the reported number and about half true of the operating number underneath it.

Quality of Beat and Miss

Revenue: The 0.6% beat is the smallest of the three quarters we have covered, against 4.9% in Q1 and 4.1% in Q2, and it comes with the fifth consecutive quarter of FX-neutral deceleration: 49%, 47%, 46%, 43% and now 41%. Currency was a tailwind again but a smaller one than in Q2, adding 4.3 points to the headline against 6.9 points last quarter. The composition is the interesting part. Product revenue, which is first-party inventory, grew 65.6% against 42.1% for service revenue, so the fastest-growing revenue line is the one with the structurally lowest margin, and it has risen from 12.59% of the total a year ago to 14.37% now. That mix shift is a permanent drag on reported gross margin that has nothing to do with shipping.

Margins: The gross margin is the number the quarter turns on and it behaved the way a stabilising series behaves rather than the way a recovering one does. At 40.78% it is the lowest in the eighteen quarters back to Q2 2022, and 253 basis points below a year ago, but it is only 12 basis points below Q2 and the 276 basis point sequential collapse did not repeat. The operating margin is the number the market traded on, and at 6.29% it fell for a third consecutive quarter. What changed is the cause. In Q1 the driver was provisions, 390 of the 595 basis points of year-over-year compression. In Q2 it was gross margin, 276 basis points sequentially. In Q3, 72% of the sequential decline is operating expenses, and inside operating expenses it is entirely sales and marketing and general and administrative: a deliberate Brazilian affiliate-marketing step-up that management can dial down quarter by quarter, and recruitment costs for fulfillment capacity that management expects to partly reverse. Those are better reasons to lose margin than a structural pricing giveaway or a credit miss, and they are the first time in this cycle that the marginal pressure has come from a line management described as calibrated rather than committed.

EPS: The $8.83 is a 5.1% miss and it flatters the quarter. Income from operations fell 6.5% year over year, and that became a 6.4% increase in net income because the effective tax rate fell 835 basis points to 25.46% and foreign currency losses narrowed by $30M. The direction of the tax line is the right one for modelling purposes: at 25.46% against 23.10% in Q2 the rate moved back toward the 26% to 28% range we argued for in August, and the nine-month rate is also 25.5%. But it is still eight points below the year-ago quarter, so the year-over-year EPS comparison is not an operating comparison. On a constant-currency basis management disclosed income from operations of $694M against $724M, a 4.0% decline, which is the cleanest single read on the operating trajectory in the release.

Segment Performance

MercadoLibre reports by country and measures each one on direct contribution, which carries shipping, warehousing, carrier costs, provisions, cost of goods, collection fees, funding costs, local salaries, marketing, hosting and artificial intelligence expenses, but excludes all corporate cost. It is the cleanest view of where the investment is landing.

Direct Contribution by Country

Q1 2026 columns are derived from the filed six-month segment table less the filed Q2 three-month table, and the derived Q1 total revenue of $8,845M ties exactly to the filed figure. In every column the total direct contribution margin less unallocated costs as a share of revenue equals the reported operating margin.

Brazil: 56% of revenue, and all of the problem

Brazilian revenue grew 49.7% in dollars, the fastest of the three major markets, and 41% FX-neutral, broadly stable against Q2's 42%. Direct contribution grew $4M, or 0.8%. The margin fell 197 basis points sequentially to 7.98% and 387 basis points year over year, and it is now 17 basis points below where it stood in Q1. The 180 basis point sequential inflection that was the single best data point in last quarter's print has been given back in full. In absolute dollars Brazilian direct contribution fell from $550M to $479M, a 12.9% sequential decline on revenue that rose 8.5%.

The operating detail is better than the margin. FX-neutral gross merchandise volume grew 36% and sold items 45%, with management describing the two-year growth rate as accelerating on both measures even after fully lapping the June 2025 free-shipping threshold cut. Items per buyer grew 16% year over year against a single-digit pace before the change. Daily active users grew faster than monthly active users again. Cost per shipment fell 5% year over year in local currency and is 14% below the last full quarter before the threshold came down. Sold items in Brazil have grown more than 40% for five consecutive quarters.

"In Brazil, our cost per shipment in Q3'26 was 5% lower YoY in local currency and 14% lower than in Q1'25, the last full quarter before we lowered the free shipping threshold (as shown on the chart below). This is despite higher energy prices in Q2 and Q3, and the cost of new capacity that came online in Q3 ahead of peak season." — Shareholder letter

Assessment: Brazil is working on every operating metric and paying for it on every financial one. The honest reading is that the country is carrying three separate costs at once: the structural shipping-revenue giveaway from the threshold cut, which management has now sized at roughly 200 basis points of GMV since Q1 2025; the one-off cost of a 20% sequential increase in fulfillment footprint; and a deliberate affiliate-marketing step-up. Only the second of those is explicitly temporary. The first has no disclosed recovery path and the third is a choice management says it will keep making. Brazil is where the 2027 margin question gets answered, and nothing in this quarter answers it.

Mexico: the only margin that improved, for a reason that will not repeat

Mexican revenue grew 47.1% in dollars and 36% FX-neutral, and direct contribution margin rose 115 basis points sequentially to 14.24%, the only segment to improve. The cause is identified: the point-of-sale device restocking charge that weighed on Q2 reversed, which management sized at roughly 30 basis points of consolidated cost of revenue. Year over year the margin is still down 423 basis points, and it remains 317 basis points below the Q1 level.

The volume picture deteriorated. FX-neutral GMV growth slowed to 21% from 26% in Q2, 28% in Q1 and 35% in Q4 2025, a third consecutive deceleration and now a cumulative 14 point slowdown in three quarters. Items sold grew 25%, faster than GMV, so average selling prices are falling in Mexico too. Unique active buyer growth held at 21%, which is the reassuring part: consumers are still arriving, they are simply spending less. Management attributed the slowdown to the tax reform flagged earlier in the year and to a weaker consumption environment, and claimed share gains both sequentially and year over year.

"In Mexico, FX-neutral GMV grew 21% YoY and items sold 25%, a slower pace than in recent quarters, as the tax reform we flagged earlier this year and a weaker consumption environment continued to weigh on demand. Even so, we gained market share QoQ and YoY, and unique active buyer growth remained broadly stable at 21%, so the shift of consumers to our platform from physical retail continued even as growth in their spending slowed." — Shareholder letter

Assessment: The Mexican margin recovery is a reversal, not an improvement, and management said as much. The device-cost pressure itself has not gone away; only the inventory build has stopped. Underneath, the second-largest market is decelerating faster than any other and has now done so for three consecutive quarters while its contribution margin sits 423 basis points below last year. Mexico was the growth offset to Brazilian margin investment in the Q1 and Q2 arguments. It is no longer performing that role.

Argentina: the margin premium is eroding slowly and management will not discuss it

Argentine revenue grew 26.2% in dollars, the slowest segment, and 42% FX-neutral, which is itself the fourth consecutive quarter of FX-neutral deceleration from 97% a year ago. Direct contribution margin of 33.72% slipped 16 basis points sequentially, the smallest move of any segment, but is down 597 basis points year over year and 203 basis points since Q1. At more than a third of revenue dropping to direct contribution, Argentina produced 41% of group direct contribution on 17% of revenue.

FX-neutral GMV growth actually accelerated to 40% from 38%, helped by a September extension of free and fast shipping on lower-priced items to all delivery methods within the managed network for loyalty members, which lifted free-shipping penetration to levels similar to Brazil and Mexico and produced a record purchase frequency in the month. Items sold grew only 19%, so the gap between volume and units is wide: Argentine growth is coming from price and mix in a high-inflation economy rather than from units.

Asked directly whether the margin premium is structural or converging, management declined to engage, and the refusal is itself informative. It also contained the first acknowledgement in three quarters that a competitor has taken share anywhere.

"And although some of the Asian competitors out there have gained some market share at the beginning of this year and the end of last year, we've seen the situation a bit more stable. And we are very bullish on our ability to continue improving our value proposition, removing frictions in the marketplace for consumers to continue choosing MercadoLibre as the primary destination for shopping and for financial services in the country." — Ariel Szarfsztejn, CEO

Assessment: Argentina is the single largest source of group profitability per revenue dollar and the least discussed. A 597 basis point year-over-year contribution margin decline in a segment that generates 41% of direct contribution on 17% of revenue deserves more analytical attention than it is getting, from management or from the Street. The new loyalty-linked shipping investment is explicitly modelled on Brazil, which means the Brazilian margin pattern is now being deliberately exported into the group's highest-margin market. Management framed the near-term cost as real but limited because it is confined to loyalty members. That framing is untested and unsized.

Other Countries: small, fast and suddenly unprofitable

Revenue grew 63.0% to $502M, the fastest growth in the group, while direct contribution margin collapsed 508 basis points sequentially to 8.96% and 500 basis points year over year. The segment covers Chile, Colombia, Peru, Uruguay, Ecuador, Costa Rica, Bermuda, China and the United States, and the margin has now nearly halved from 16.12% in Q1. No commentary was offered in the letter or on the call, and no analyst asked.

Assessment: At 4.7% of revenue this does not move the group, and a 508 basis point swing on $502M is $25M. But the China fulfillment centre that drives cross-border trade sits in this segment, and cross-border GMV growth was disclosed at 60% FX-neutral in Q2 and not at all in Q3. The combination of a halved margin and a withdrawn growth metric in the same segment is the kind of thing worth a question that nobody asked.

Revenue Composition: Commerce versus Fintech

The Growth-Rate Ladders, FX-Neutral

Assessment: This table contains the most important structural fact in the print and it is not the one the stock traded on. FX-neutral commerce revenue growth has accelerated for three consecutive quarters, from 37% to 39% to 40% to 41%, which is the direct answer to whether the Brazilian investment is producing revenue as well as volume. Over the same stretch fintech revenue growth has decelerated for four consecutive quarters, from 65% to 40%, a 25 point slowdown. The two crossed this quarter, the first time in the five quarters the letter tabulates that commerce grew faster. That crossover reframes the investment case: the money is going into commerce and commerce is responding, while the engine that was supposed to fund it is cooling. It also explains part of the gross margin arithmetic, because commerce carries shipping cost and fintech does not.

Key Operating KPIs

Unit Economics Underneath the KPIs

The disclosed aggregates allow several ratios the company does not print. All are computed from the quarterly KPI table and the Note 6 revenue disaggregation, and none is management's own per-user figure, which uses a different base.

Assessment: Three findings here matter. First, average selling price fell 4.6% sequentially in dollars while items per buyer rose 3.3%, which is the low-ticket mix shift management describes, visible in the arithmetic rather than in an adjective. Second, and against the prevailing narrative, the commerce services take rate rose 60 basis points sequentially and 51 basis points year over year to 20.21% of GMV, despite the buyer discounts and seller take-rate reductions still being in force. Advertising, membership and classified fees sit in that line and are growing faster than GMV, so this is mix rather than pricing, but the simple claim that MercadoLibre is giving away its take rate is not what the disaggregation shows. Third, the acquiring yield proxy fell again, to 3.18% from 3.48% in Q1 and 3.53% a year ago, which is the compression a question on the call raised and management attributed to moving up market. The implied funding cost has been steady at roughly 5% of the average portfolio for three quarters, which rules out funding as a driver of the NIMAL decline and leaves product mix and asset yield as the explanation.

The Credit Book

Assessment: Every flow measure improved and every stock measure deteriorated, which is the same split as last quarter, but the balance has shifted. On the flow side, cost of risk fell for a second consecutive quarter to 31.3%, from 33.0% and 36.8%, provisions fell to 55.5% of credit revenue from 56.0%, portfolio growth decelerated to 9.2% sequentially from 12.5% and 16.4%, and the gap between portfolio growth and revenue growth narrowed from 38 points at Q1 to 25 at Q2 to 17 now. Most importantly, the off-balance-sheet pipeline we have flagged for two quarters finally decelerated hard: unused agreed card lines grew 6.0% in the quarter against 18.1% and 32.1%. Because expected losses on committed lines are provisioned as they are drawn, that deceleration is the single best forward indicator in the credit disclosure, and it turned this quarter.

On the stock side, the 91 to 360 day bucket rose again to 19.33% of the book, as it has in every quarter of 2026, though by 64 basis points against 112 last quarter. Total past due reached 28.71% and allowance coverage 27.09%, both the highest of the four quarters. And a new figure appeared: renegotiated loans jumped from 1.5% of the portfolio at June to 2.2% at September, which is roughly $393M against $246M, an increase of about 60% in a single quarter after being flat through the first half. That happened in the exact quarter management confirmed participating in the Brazilian government debt-renegotiation programme from late July, plus a credit-bureau discount campaign, and twice declined to size the benefit. The 15 to 90 day ratio management cites as stable at 7.0% is also the ratio most directly flattered by moving delinquent balances into renegotiation. We are not asserting that the stability is manufactured. We are observing that the one disclosure which would rule it out was requested twice and withheld.

Key Topics & Management Commentary

Overall Management Tone: Management was more specific and less defensive than at Q2, volunteering a quantified reconciliation of the sequential margin decline in the prepared remarks rather than waiting to be asked for it. The posture on credit was the most confident of the three quarters, and the posture on margin timing was the most explicitly non-committal: the only forward commitment offered on the entire call was that part of the capacity cost comes back in Q4. Where management was least convincing was on sizing: a direct request for the dollar impact of the Brazilian debt-renegotiation benefit, and questions on the affiliate step-up and the Argentine margin trajectory, were each met with a framework rather than a number.

1. The Gross Margin Decline Stopped, 72 Basis Points Short of the Bar

The question we set for this quarter was whether the 276 basis point sequential gross margin collapse of Q2 was a step down to a new level or the second leg of a slide. The answer is the former. Gross margin printed 40.78% against 40.90%, a 12 basis point decline, with the Brazilian pricing investments now in the base for two full quarters and no provision reversal left to spend. On the specific test, the series stabilised.

It stabilised at the wrong level. We said an upgrade needed 41.5% or better; 40.78% is the lowest gross margin in the eighteen quarters back to Q2 2022, and 253 basis points below a year ago. Nine-month gross margin is 41.7% against 45.1%. The filing attributes the decline to higher shipping operating costs and higher cost of goods sold, partly offset by lower revenue-based taxes, and adds the free-shipping threshold reduction for the nine-month period.

There is also a mix effect running underneath that has nothing to do with shipping efficiency. First-party product revenue grew 66.8% year over year against 41.6% for services and is now 14.37% of total revenue against 12.59% a year ago. The filing says plainly what that means for the future.

"In the future, our gross profit margin could continue declining if we maintain the growth of our first-party business, which has a lower pure product margin due to marketing initiatives, or building up of our logistics network." — Quarterly filing, gross profit margins

Assessment: A stabilised gross margin at 40.8% is materially better news than a third consecutive leg down, and it removes the specific scenario that would have forced a downgrade. It does not restore the pillar it broke. The bull case needed the unit-cost curve to pull the gross margin back up, and what happened instead is that the curve stopped pulling it down. Those are different outcomes with different implications for 2027 earnings power, and the difference is roughly 70 basis points of group margin.

2. The Pressure Rotated Into Lines Management Controls

Thirty-one of the 43 basis points of sequential operating margin decline sit in operating expenses, and inside operating expenses the entire adverse movement is sales and marketing at plus 44 basis points and general and administrative at plus 18. Product development improved 24 basis points and provisions improved 7. That is a different quarter from Q1, where provisions were 390 of 595 basis points of year-over-year compression, and from Q2, where gross margin was the whole story.

Sales and marketing reached 11.56% of revenue, the highest of the four quarters we can measure, and the driver is named: a deliberate increase in affiliate-marketing spending in Brazil. Orders through the channel grew 240% year over year and the network now exceeds 12 million affiliates across the region. The filing adds detail the letter omits, attributing the year-over-year increase to $276M more online and offline marketing, $74M more buyer-protection expense, and $40M more salaries on a 35% increase in sales and marketing headcount.

"We calibrate this spend quarter-by-quarter, and in Q3, we decided to invest more in our affiliate program in Brazil to strengthen our presence on social media. Orders through the channel grew 240% year-on-year." — Martin de los Santos, CFO

Assessment: This is the most favourable composition of margin decline in the cycle so far, and it is the main reason the rating does not move down. A 44 basis point drag from a channel management describes as calibrated quarter by quarter, operating inside a stated 11% to 12% range it has held for years, is a reversible cost in a way that a pricing giveaway is not. The 35% headcount increase in the same function is the part that is not reversible, and it is not discussed.

3. Shipping: The Withdrawn Metric Returned, and the Aggregate Got Sized for the First Time

Last quarter the Brazilian unit shipping cost series disappeared from the letter and the call in the exact quarter gross margin fell 276 basis points. It is back. Cost per shipment in Brazil fell 5% year over year in local currency and is 14% below Q1 2025, the last full quarter before the free-shipping threshold came down, and that is despite higher energy prices and the cost of new capacity.

More useful still, the aggregate give-up that the unit-cost series was always being used to offset has finally been quantified. Net shipping margin, defined as shipping charges net of shipping costs, has fallen by around 200 basis points as a percentage of GMV since Q1 2025. That is the first time in three quarters of coverage that the two halves of the shipping argument have been presented on the same basis.

"As a result, our cost per shipment improved with scale – as shown by the chart above – but our net shipping margin (shipping charges net of shipping costs) fell by around 200bps as a percentage of GMV since Q1'25. As intended when we lowered the threshold, growth was led by lower-priced items that stimulate frequency and online buying habits but generate less revenue per item sold." — Shareholder letter

Management also gave the one unit-economics milestone that matters most: in September, free and slow shipments in the R$19 to R$79 range were variable-contribution positive as a whole for the first time. At Q2 the equivalent claim was that the range was positive across half of the ASP bands within it.

Assessment: This is a real improvement in disclosure quality and it goes directly to the pillar that was downgraded in August. The 200 basis point figure is also the most damaging single number in the letter when read honestly: GMV of $22.8bn in the quarter means roughly $456M a quarter of shipping margin has been transferred to buyers, against an operating income base of $676M. The company has now told us the size of the subsidy and told us the unit cost is falling 5% a year. At that rate the subsidy does not get recovered by unit-cost improvement inside any horizon a model can carry. It gets recovered by the frequency it buys, which is exactly what management argues, and which remains unprovable from outside.

4. The Peak-Season Capacity Build, and the Only Forward Commitment on the Call

Fulfillment square metres rose almost 20% sequentially ahead of peak season, against 10% in the same quarter a year ago, and the associated recruitment costs added 20 basis points to general and administrative. Thirteen Metro fulfillment centres have opened in the last twelve months, the number of sorters and robots has more than doubled, robotics is now live in Mexico as well as Brazil, and picking operations where robotics has been deployed show 70% productivity gains against manual processes. Shipments per route reached a new high.

In answering a direct question on which of the Q3 costs reverse, management gave the single forward commitment of the call.

"In this particular year, we increased our capacity quarter-on-quarter by 20%. Given the size of our logistic infrastructure is a lot compared to a year ago where we grew it by only 10%. So there's a significant growth in our footprint, and that creates some headwind in terms of cost in this particular quarter. So that, we expect that part of that to come back in Q4." — Martin de los Santos, CFO

On the energy component management was explicit the other way: some of the cost was passed through in pricing, not all of it, and the pressure is expected to remain. On the pace of future capacity additions, management said increases continue every quarter but not at the Q3 rate.

Assessment: This is now the single testable commitment in the file, and it is testable in ninety days. "Part of that to come back in Q4" plus peak-season volume leverage is what the FY26 consensus needs, and the step it needs sits inside the +38 to +304 basis point range of the last two years. If Q4 operating margin does not expand sequentially at all, the capacity explanation fails and the margin story reverts to structural. That is the cleanest binary the coverage has had.

5. Credit: Every Profitability Measure Improved and the Forward Pipeline Finally Turned

The credit portfolio reached $17,884M, up 62% year over year and 9.2% sequentially, the slowest sequential growth of the four quarters. Cost of risk fell to 31.3% from 33.0% and 36.8%, a second consecutive improvement. Provisions fell to 55.5% of credit revenue. The 15 to 90 day non-performing ratio held at 7.0%, which management described as broadly stable year over year and sequentially and close to all-time lows despite industry-wide weakness in Argentina and Brazil.

The new fact is the off-balance-sheet pipeline. Unused agreed card loan commitments reached $14,892M, but grew only 6.0% in the quarter against 18.1% in Q2 and 32.1% in Q1. We have flagged this balance for two quarters as the forward pipeline of the provision charge, because expected losses are booked against committed lines as they are drawn. Its growth has now collapsed from a third to a sixteenth in two quarters.

"Our 15 to 90 NPL was 7%, stable both versus last quarter and a year ago, and we continued growing at 62%. And beyond Brazil, I'd say in Argentina, NPLs have improved across all products, close to historical lows despite weakness in the industry." — Osvaldo Gimenez, President of Fintech

Management also disclosed, under questioning, that it has become more cautious in Brazil on the riskier segments, describing the pullback as marginal and specific, and restated the priority order: risk management ahead of growth.

Assessment: On the flow measures this is the strongest credit quarter of the three we have covered, and the deceleration in unused commitments is a genuine positive that nobody asked about and nobody wrote about. The mechanism that produced the Q1 provision shock is visibly winding down. What is not winding down is the loss curve on what was already originated: originations still grew 70% in the quarter against a 62% portfolio, and nine-month write-offs are $2,180M against $1,205M.

6. Desenrola, the Renegotiated Book, and a Number That Was Asked For Twice

Management disclosed for the first time that the company participated in Desenrola, the Brazilian government debt-renegotiation programme, from the end of July until the programme closed, and additionally joined a credit-bureau discount campaign over the same period. It confirmed a positive impact on the quarter and refused twice to quantify it, on the grounds that it is small relative to the portfolio.

"Let me start with Desenrola. As I mentioned, it's fairly small. It's not a relevant number. So we have not disclosed it, and I think it's very, doesn't really move the needle." — Osvaldo Gimenez, President of Fintech

The quarterly filing contains a figure the call did not. Renegotiations rose from 1.5% of the loans receivable portfolio at June 30 to 2.2% at September 30, which is $393M against $246M. They were 1.6% at December 31, so the share was flat to slightly down through the first half and jumped 70 basis points in the quarter Desenrola ran.

Assessment: Two things follow. First, the renegotiated book grew about 60% in a quarter in which the portfolio grew 9.2%, which is not a small number by the only comparison available. Second, and more important for the read on asset quality, renegotiation moves balances out of the delinquency buckets. The 15 to 90 day ratio held at 7.0% and the 91 to 360 day bucket rose by its smallest increment in three quarters, both in the quarter when roughly $147M of incremental renegotiation happened. We cannot separate the underlying trend from the programme effect, and that is the point: management was asked for the figure that would allow the separation, by two different analysts, and did not provide it. Treat the Q3 asset-quality improvement in the early buckets as unverified until the Q4 ladder prints without a programme running.

7. The Credit Card Cohort Ladder, Disclosed for the First Time

Credit cards reached $8,467M, or 47.34% of the portfolio, against 45.22% at the start of the year, and the card has absorbed almost $450M of investment over the last twelve months. Management disclosed the Brazilian cohort structure in detail for the first time: cohorts issued between 2021 and 2024 are NIMAL-positive and account for more than half the Brazilian card portfolio, the 2025 cohort is roughly breakeven in its second year, and the 2026 cohort is in its early loss-making phase. Card NIMAL improved sequentially in Q3 even though issuance reached a record.

"Brazil, where we have issued cards the longest, shows the full pattern today: the cohorts issued between 2021 and 2024 are NIMAL positive and account for more than half of the Brazilian card portfolio; the 2025 cohort is roughly NIMAL breakeven in its second year; and the 2026 cohort is in its early, loss-making phase." — Shareholder letter

Blended NIMAL nonetheless fell 60 basis points sequentially to 20.1% and 90 basis points year over year, which management attributed entirely to product mix as the card grows faster than merchant and consumer lending.

Assessment: The cohort ladder is the disclosure we asked for in August and it substantially answers the question of when the card stops diluting. If more than half the Brazilian book is already NIMAL-positive and the 2025 vintage is at breakeven in year two, the drag is a function of issuance pace rather than of cohort economics, exactly as management says. The mix explanation for the blended decline is arithmetically thin, however. Card share of the portfolio rose only 46 basis points sequentially; on the product-level NIMALs carried from our Q2 work that mix shift costs roughly 20 basis points of blended NIMAL, against an actual decline of 60. Mix accounts for about a third. The residual is not explained, and the credit yield on the average portfolio fell 240 basis points sequentially while the implied funding cost held flat, which points at asset yield rather than at mix.

8. Advertising Decelerated 11 Points in the Quarter It Mattered Most

Advertising revenue grew 51% FX-neutral and 58% in dollars, against 62% FX-neutral and 73% in dollars in Q2. Growth broadened in Q2; in Q3 management noted slower growth in Argentina and Mexico. The product news was heavy: the AI advisor now manages campaigns for 56,000 sellers, almost four times the number three months earlier; a new branding product that predicts purchase intent from first-party data launched across six sites; ad load is moving from fixed slots to dynamic decisions; and a connected-television partnership announced in October adds a fourth streaming platform from early 2027.

No absolute dollar figure was disclosed, for the third consecutive quarter.

Assessment: Advertising is the most credible structural offset to gross margin compression in the entire business, because it is the highest-margin revenue line and it rides on volume the company is already paying to acquire. An 11 point deceleration in FX-neutral growth is therefore the quietest bad news in the print, and it arrives in the quarter the offset was most needed. It also partly explains the commerce services take rate finding: the take rate rose 60 basis points sequentially, but the ads line inside it is growing more slowly than it was. Without a dollar disclosure the contribution cannot be modelled, only inferred, and the inference just got weaker.

9. Mexico: a Third Consecutive Deceleration in the Second-Largest Market

FX-neutral Mexican GMV growth of 21% is down from 26%, 28% and 35% in the three prior quarters. FX-neutral Mexican revenue growth of 36% is down from 38%, 39% and 41%. Items sold grew 25%, ahead of GMV, so the slowdown is in ticket size rather than in units, and unique active buyer growth held at 21%. Management named two causes, a tax reform flagged earlier in the year and a weaker consumption environment, and claimed sequential and year-over-year share gains.

The fintech story in Mexico is the opposite. Mercado Pago is now the largest digital bank in the country by active users at more than 18 million, growing above 40%, and holds the largest credit portfolio among fintechs. A merchant business account launched in August.

Assessment: Mexico was the part of the thesis that was supposed to be uncomplicated: a less competitive market, an earlier-stage e-commerce penetration curve, and a contribution margin well above Brazil's. It is now three quarters into a deceleration with 423 basis points less contribution margin than a year ago, and the only sequential margin improvement it produced this quarter was a reversal of a one-off. Watch the Q4 GMV growth rate: a fourth consecutive deceleration would make this a market problem rather than a tax-reform problem.

10. The Brazilian Work Week: a Structural Cost Nobody Has Sized

A constitutional amendment approved in Brazil in October moves the country from six days on and one off to five days on and two off, and shortens the legal work week from 44 hours to 40 with no salary adjustment. For a company whose largest cost line is logistics labour in Brazil, this is a direct and permanent input-cost increase. Management had pre-empted the schedule change, reorganising Brazilian logistics to a five-and-two pattern at the same 44 hours in August, and expects no material cost impact from that part. The hours reduction is the open item.

"On the second change in the country, we are preparing for the next stages where the working week will be gradually reduced. I think it's a bit early on that regard. Clearly, we need the hours in order to process all the demand that is flowing through our network, which, in the end, will mean that we will need more employees in our logistics operation." — Ariel Szarfsztejn, CEO

Management pointed to scheduling efficiencies from the five-and-two reorganisation, natural workforce turnover allowing experiments with compensation structures, and automation as the offsets, and said it would share more next quarter. Automation volume within fulfillment has tripled, in Brazil and Mexico, though management described the scale as still relatively low.

Assessment: A 9% reduction in legal hours with no pay adjustment, phased in, against a logistics workforce that must process more than 40% annual growth in items shipped, is a new structural cost in the single market that drives 56% of revenue and all of the current margin pressure. It is unsized by management, unmodelled by consensus, and absent from every sell-side action we reviewed on the reaction session. It does not change this quarter and it is not a reason to downgrade. It is a reason to be slower to underwrite a 2027 margin recovery, and it is the most important new item in the file.

11. Agentic Shopping: a Strategic Refusal

Asked whether third-party shopping agents would be allowed to transact against MercadoLibre inventory, management answered that they would not. The company has decided not to provide its inventory to any external platform, and is instead building its own agent, already live to a small portion of users in the three largest markets, on 27 years of first-party data.

"But for now, we have decided not to provide our inventory to any external platform. Our priority is still to build the best agentic experience inside MercadoLibre, taking advantage of the 27 years of first-party data that we have, our unique combination of selection, pricing, logistics, financing, loyalty and so on." — Ariel Szarfsztejn, CEO

Assessment: This is the clearest strategic statement on the call and it is a bet that direct app traffic survives agentic intermediation. The logic offered is sound for a market where trust is built below the funnel through logistics and financing rather than at the search layer, and MercadoLibre is better positioned to make that bet in Latin America than almost any retailer is in the United States. It is still a refusal to participate in a channel it cannot control, taken before the channel has demonstrated scale anywhere. Early results were described as promising and too early to share, which is the one place on the call management asked for credit without evidence.

12. AI: a Disclosed Return, and 72 Basis Points of Product Development Leverage

AI investment rose by almost $30M sequentially. Against that, product and technology development fell to 6.93% of revenue from 7.65% a year ago, 72 basis points of leverage, and the filing attributes the dollar increase to $72M more technology maintenance and AI and $56M more salaries. Management argued the returns are measurable in four channels and gave figures for three: the revenue uplift from a language model in search more than covers the model cost; customer support runs today's volume with roughly 7,000 representatives against a multiple of that without AI; the fintech assistant held 19 million conversations in September; technology team output has doubled in a year; and AI has freed roughly 20% of hours worked in Finance. Cost per token fell again and open-source models are expected to become a large part of the mix.

"So when you put the cost of our developers plus the cost of AI, we continue to scale this particular quarter, we scaled 70 basis points relative to last year. So we are very comfortable with the investments that we're making." — Martin de los Santos, CFO

Assessment: This is the cleanest pillar in the thesis and it strengthened again. A company absorbing $30M more sequential AI spend while taking 72 basis points out of the product development ratio is demonstrating the thing most companies only assert. The engineering-headcount commitment made in Q1 is holding. It is also, at roughly $77M a quarter of ratio improvement against last year's level, nowhere near large enough to offset what shipping and marketing are taking out.

13. Cash Flow, a Second Bond, and a Balance Sheet Still Levering

Adjusted free cash flow was $285M, against $214M in Q2 and negative $56M in Q1, after $503M of capital expenditure and $2.1bn of credit portfolio investment partly offset by $1.5bn of fintech funding. Nine-month adjusted free cash flow is $443M against $718M a year ago. Capital intensity rose again to 4.68% of revenue from 4.34% and 3.06%, although it is 14 basis points below the year-ago quarter. A second investment-grade bond raised $1bn in September.

Net debt on the company's own definition reached $8,028M, up 25.0% in the quarter and 71.5% since December. Total debt is $16,154M against $11,392M at the start of the year. Deferred tax assets rose 55.9% year to date to $2,402M, which is the balance behind the low effective tax rate. Treasury shares rose by 662 for the nine months, identical to the figure for the six months, so no shares were repurchased in Q3. Capital allocation was not discussed on the call and no analyst asked.

Assessment: Free cash flow is improving sequentially and still running at 62% of last year's nine-month level. The important number is net debt, which has risen 71.5% in nine months to fund a credit book growing faster than the business that underwrites it. That is a defensible use of leverage while cost of risk is falling and asset quality in the early buckets is holding, and it is the mechanism by which a consumer-lending business becomes a balance-sheet story rather than a platform story. At 0.21x trailing twelve-month revenue and 2.0x annualised Q3 adjusted EBITDA, there is no distress here, and there is also no capital returning to shareholders in any quarter of this cycle.

Guidance & Outlook

MercadoLibre has never published quarterly or annual revenue, margin or EPS guidance, and restated that position explicitly on this call. There is therefore no prior-guide or new-guide comparison available for any line. What exists instead is a set of qualitative forward statements, and this quarter's set is more specific than Q2's.

"The approach is to manage for the long term. [...] We are very disciplined in the way we invest. We set out targets of both engagement as well as path to profitability on all of the investment areas that we have in front of us. And when we see that they are performing well, we pursue them. If they're not, we reiterate and in some cases, we stop them." — Martin de los Santos, CFO

What the Published FY2026 Consensus Actually Requires

With no guidance to anchor on, the only forward arithmetic available is what consensus implies for Q4 against nine months of actuals. FY2026 consensus sits at $41,790M of revenue, $1,933M of net income and $38.18 of EPS across sixteen analysts, and has not moved in the three months since our Q2 note. Nine-month actuals were $29,764M, $1,331M and $26.25.

Implied sequential ramp: the FY2026 number needs Q4 revenue of $12.0bn, which is a smaller sequential step than Q4 delivered in either of the last two years (+18.2% in 2025, +14.1% in 2024), and an operating margin of roughly 7.1% to 7.6% depending on the tax rate. That is a sequential margin step-up of 84 to 129 basis points.

Street at: the implied step sits inside the historical Q3 to Q4 range, which was +304 basis points in 2024 and +38 basis points in 2025. Gross margin, by contrast, fell slightly into Q4 in both of those years, by 54 basis points in 2024 and 11 basis points in 2025, so the entire implied lift has to come from operating-expense leverage on the seasonal revenue step rather than from gross margin. Peak-season volume plus the partial reversal of the capacity cost that management named is a coherent path to it.

Guidance style: there is none, which is itself the single largest structural difference between MercadoLibre and every comparable platform at this scale. The company will not give a number, will not give a timeline, and will not give a target margin. Investors are asked to underwrite a reinvestment cycle on the strength of engagement disclosures the company chooses which of to publish each quarter. That is a governance discount we think is real and that the 40x multiple does not price.

Analyst Q&A Highlights

Twelve analysts asked questions, against fourteen last quarter, with one line failing and being re-promoted at the end of the call. Four exchanges went to credit and two directly to the operating margin, so the balance of attention shifted back toward margin from the credit-heavy Q2 call.

Whether the Margin Inflection Is Now Within Line of Sight

The second question of the call put the bull case to management directly: with the margin drags named, the affiliate step-up described as tactical and every engagement metric strong, it should follow that the inflection is close. The question then asked the only thing that could falsify that, which is whether another investment bucket the size of the Brazilian low-ASP push is coming, and extended to quick commerce specifically. Management declined the framing on inflection timing and answered only the falsification half, confirming no new large bucket in the short term and naming the Argentine loyalty shipping expansion as the exception.

Q: "So with the overall top line metrics and engagement KPIs also remaining so strong and putting this all together, it does feel like the inflection point for margins should be coming within the line of sight here. So my question is, are there any other potential new larger investment buckets perhaps of the size of what you did in Brazil in terms of lower ASPs that you're looking to lean into or perhaps a combination of a couple of different buckets that you're looking to lean more aggressively into from here?"
— Irma Sgarz, Goldman Sachs

A: "In the short term, I don't see significant new areas of investment. We announced recently this quarter in September, expansion of our free shipping program in Argentina, mostly to our [MELI+] users. So that could be an area of investment, but similar to what we have done in Brazil. So I would say that there will not be a significant deviation to what you have seen."
— Martin de los Santos, CFO

Assessment: The answer is more valuable than the question's premise. A commitment to no new large investment bucket is the necessary condition for any margin inflection, and management gave it. What it did not give is the sufficient condition, which is a timeline. The Argentine exception is not trivial: it exports the Brazilian playbook into the segment carrying 41% of group direct contribution, and the only sizing offered was that the cost is limited because it is confined to loyalty members.

Which of the Q3 Margin Drivers Reverse and Which Stay in the Model

The most useful exchange on the call came from a question that framed the margin as the thing weighing on sentiment and asked simply which costs are temporary. Management responded with a line-by-line split that was more specific than anything in the letter: 60 basis points of net shipping headwind divided 40 in cost of revenue and 20 in operating expense, driven mainly by a 20% sequential footprint expansion against 10% a year ago, plus 40 basis points of affiliate investment. It then produced the only forward number of the call.

Q: "I wanted to return back to the operating margin. You called out a number of drivers that weighed on the margin in the quarter. And look, I was wondering if any of those are going to necessarily reverse? Or are these costs that are going to be in the model until we grow over them next year. So any help there would be great because the margin is clearly what continues to weigh on sentiment around the stock."
— Craig Maurer, FT Partners

A: "For the most part, the shipping contraction was because of expansion of our footprint, logistic footprint. We do this every third quarter moving into peak season in the fourth quarter. In this particular year, we increased our capacity quarter-on-quarter by 20%. Given the size of our logistic infrastructure is a lot compared to a year ago where we grew it by only 10%. So there's a significant growth in our footprint, and that creates some headwind in terms of cost in this particular quarter. So that, we expect that part of that to come back in Q4."
— Martin de los Santos, CFO

Assessment: This is the quarter's single commitment and it is falsifiable in ninety days. Note what is carefully absent: no figure for how much comes back, and an explicit statement in the same answer that the energy component is expected to remain. Read with the seasonal arithmetic, consensus needs 84 to 129 basis points of sequential margin expansion in Q4 and this answer is the mechanism by which it happens. If Q4 operating margin does not expand sequentially at all, the capacity explanation fails and the margin decline reverts to structural.

Whether the Brazilian Affiliate Step-Up Is Tactical or Structural

The opening question challenged the word "tactical" directly, arguing that if social discovery is a durable shift in consumer behaviour then the spending that captures it is structural, not a quarter-by-quarter dial. The answer conceded the substance while defending the word: the intensity was tactical, the channel is strategic, and the programme has been running for several quarters at a scale now exceeding 12 million affiliates. Management closed on unit economics improving over time rather than on the spending reducing.

Q: "And then related to that, you framed the Brazil affiliates investment as tactical. I'm curious that framing as opposed to something more structural if the market is evolving in this manner."
— Andrew Ruben, Morgan Stanley

A: "If I can complement, Andrew, we mentioned tactically the intensity. This particular quarter, we intensified a little bit of investment, but it's a strategic initiative for us to develop this channel. We have been doing it for several quarters now. And as I mentioned, we have reached 12 million affiliate, members of affiliate program. So it's an area of investment that we will continue to pursue. And as we improve the unit economics, it should contribute to margins eventually."
— Martin de los Santos, CFO

Assessment: The question was right and the answer confirms it. "Tactically the intensity" is a distinction that does not survive contact with a model: if the channel is strategic and spending continues, the 44 basis points is a new base rather than a one-quarter bulge, and the path to margin runs through improving the channel's economics rather than through turning it off. The co-funding by sellers mentioned elsewhere in the letter is the mechanism that would do it, and it is not sized either.

Whether the Brazilian Credit Cycle Requires Slowing Down

The dominant line of questioning on the call came from multiple analysts pressing the same point from different angles: Brazilian system-level credit metrics are deteriorating, MercadoLibre's book is growing 62%, and the two cannot both continue. Management's answer was the most confident of the three quarters we have covered, and it contained a concession that had not been made before, which is that the company has already pulled back at the margin on the riskier segments.

Q: "So, so far, you have been doing quite well on the metrics. But like when we see countries like Brazil, we see deteriorating trends in the system, you are getting larger. How are you seeing this going forward? I mean, I think the environment is getting a bit tougher. Do you feel the need to step a bit on the brakes or on the MELI side so far so good, and you can keep a good speed?"
— Marcelo Santos, JPMorgan

A: "I'd say that so far so good. We have seen very good results on the credit front across countries. I think you specifically asked about Brazil, I imagine. I would say that given the information we have and our discipline on risk management, we have been able to continue to grow, but also reducing risk at the margin basically. So we have been able to continue to grow across most segments. On the most risky segments, we have been a little bit more cautious. And what we are seeing is stability across products."
— Osvaldo Gimenez, President of Fintech

Assessment: The concession matters more than the reassurance. Three quarters of saying the book is healthy has become the book is healthy and we have tightened at the margin, which is consistent with portfolio growth decelerating from 16.4% to 12.5% to 9.2% sequentially and with unused card commitments growing 6.0% against 32.1% two quarters ago. The underwriting discipline claim is supported by the flow data. What it does not address is the stock of already-originated loans now seasoning through the 91 to 360 day bucket, which rose again, as it has in every quarter of 2026.

Sizing the Government Debt-Renegotiation Benefit

Two separate analysts asked for the dollar impact of the Brazilian government debt-renegotiation programme, one framing it as the only way to establish the recurring level of provisions and net interest margin. Management confirmed participation from late July, confirmed a positive impact, confirmed additional participation in a credit-bureau discount campaign over the same period, and declined both requests.

Q: "And second, if you can share with us if it's possible to share the nominal impact from the Desenrola to your numbers, just to understand what would be the recurring level of provisions on NIMAL."
— Kaio Penso Da Prato, UBS

A: "Let me start with Desenrola. As I mentioned, it's fairly small. It's not a relevant number. So we have not disclosed it, and I think it's very, doesn't really move the needle."
— Osvaldo Gimenez, President of Fintech

Assessment: The question identified exactly the right problem and did not get the answer. The quarterly filing shows renegotiated loans rising from 1.5% of the portfolio to 2.2% in the quarter the programme ran, which is roughly $393M against $246M and about a 60% increase while the portfolio grew 9.2%. Renegotiation moves balances out of the delinquency buckets, so the stable 7.0% early-stage ratio and the slowest increase in the late-stage bucket in three quarters both arrive with an undisclosed assist. "Doesn't really move the needle" is an assertion the filing does not obviously support, and the analyst asked for the one number that would settle it.

Whether the Blended Net Interest Margin Decline Is Only Product Mix

A question returned to a phrase in the letter suggesting deterioration in the net interest margin of products other than the card, asking which products and which countries. The answer was a flat denial on the non-card book, attributing the entire blended decline to the card growing faster than merchant and consumer lending, with a secondary mention of deliberately lower interest rates charged to some Argentine customers where spreads allowed it.

Q: "You mentioned during the letter, yes, yes, it's likely deterioration in NIMAL of some products other than cards product that happened and why in which country, if you could just give a little bit more details here."
— Rodrigo Gastim, Itau BBA

A: "And with regards to the other products, nothing has changed because overall, the change in NIMAL is due to the change in the relative size of the product with credit cards growing faster than merchant and consumer loans and merchant and consumer loans continue to be super healthy."
— Osvaldo Gimenez, President of Fintech

Assessment: The mix explanation does not carry the full weight placed on it. Card share of the portfolio rose 46 basis points sequentially, from 46.88% to 47.34%. On the product-level net interest margins carried from our Q2 work, a shift that size costs roughly 20 basis points of blended margin, against an actual decline of 60. Mix accounts for about a third. Meanwhile the credit yield on the average portfolio fell 240 basis points sequentially while the implied funding cost held flat at roughly 5%, which points at asset yield rather than at mix. The Argentine rate reduction mentioned as a secondary item may well be the residual, in which case it is larger than its billing.

How Quickly the New Fulfillment Capacity Fills, and Whether 2027 Needs Another Step

A question put the capacity build in modelling terms: a 20% sequential increase in fulfillment square metres hit both gross margin and general and administrative this quarter, so the useful questions are how fast it fills and whether the associated cost dilutes away in 2027 or recurs. The answer declined to commit on the fill rate beyond a step up in Q4 and made clear that capacity additions continue every quarter, just not at the Q3 pace.

Q: "Fulfillment square meters increased almost 20% quarter-over-quarter ahead of peak season, creating pressure in both gross margin and G&A. How quickly do you expect that capacity to be filled? And should the associated costs largely dilute in '27? Or does the current growth outlook require another similarly large capacity setup next year?"
— Lucas Esteves, Santander

A: "In terms of the speed at which that capacity fills up, it depends. Obviously, there will be a step-up function in Q4. But then over time, it will continue to fill up, and that obviously helps in terms of unit economics. But as you know, we'll continue to invest in new facilities throughout the year, maybe not at the pace that we're growing in Q3 because of peak season. But every quarter, we do increase capacity to keep up with the growth of the market."
— Martin de los Santos, CFO

Assessment: This is the honest answer and it is unhelpful for the bull case. Capacity cost that recurs every quarter because the network must keep growing 40% is not a one-off that dilutes; it is a permanent feature of the cost base whose intensity varies with the growth rate. The Q3 spike is genuinely larger than normal, at 20% against 10%, so some of it does reverse. The structural point is that a business adding fulfillment footprint continuously cannot also be a business whose logistics costs dilute predictably, and the answer declined to pretend otherwise.

The Brazilian Work-Week Amendment and 2027 Logistics Cost

The last question of the call raised the Brazilian constitutional amendment approved in October, which moves the country from a six-on, one-off schedule to five-on, two-off and shortens the legal work week from 44 hours to 40 with no salary adjustment, and asked for the expected 2027 logistics cost impact and how much automation can offset. Management separated the two changes: the schedule reorganisation was pre-empted in August at unchanged hours and is expected to be cost-neutral or better, while the hours reduction will require more people.

Q: "Most of the topics I wanted to discuss were already answered, but maybe it would be very helpful if you could share what are your thoughts on the 6x1 workweek schedule approved in Brazil recently. So the labor journey has been, is expected to change. And I would like to know what is the expected impact on your logistics costs for 2027 maybe? And how much of it can automation offset or maybe your thoughts on this topic, please?"
— Pedro Caravina, XP

A: "On the second change in the country, we are preparing for the next stages where the working week will be gradually reduced. I think it's a bit early on that regard. Clearly, we need the hours in order to process all the demand that is flowing through our network, which, in the end, will mean that we will need more employees in our logistics operation."
— Ariel Szarfsztejn, CEO

Assessment: The pre-emptive August reorganisation is good management and the neutrality claim on that half is credible. The hours reduction is a 9% cut in legal weekly hours with no offsetting pay adjustment, landing on the largest cost line in the market that produces 56% of revenue, and it is unsized, unmodelled by consensus and absent from every sell-side action we reviewed after the print. The offsets named were scheduling efficiency, workforce turnover allowing compensation experiments and automation, in a fulfillment network where management described automation scale as still relatively low. More detail was promised next quarter. This exchange is the most important thing on the call that the market has not yet priced.

What They're NOT Saying

1. The dollar impact of the government debt-renegotiation programme. Participation from late July was confirmed, a positive impact was confirmed, and sizing was declined twice. Renegotiated loans rose from 1.5% of the portfolio to 2.2% in the same quarter, which is $393M against $246M. Without the figure, the stability of the early-stage delinquency ratio cannot be separated from the programme.
2. How much of the capacity cost comes back in Q4. The only forward commitment on the call was that "part of that" reverses. The part was not quantified, and consensus needs 84 to 129 basis points of sequential operating margin expansion.
3. The size of the affiliate step-up in dollars, and its steady-state level. It is 44 basis points of revenue this quarter by arithmetic, roughly $47M, but management described the intensity as tactical and the channel as strategic without saying where the run-rate settles inside the stated 11% to 12% range.
4. Advertising revenue in dollars, for a third consecutive quarter. Growth decelerated from 62% FX-neutral to 51%, and this is the highest-margin line in the business and the most credible structural offset to shipping. It cannot be sized from outside, and the one quarter it decelerated meaningfully is the quarter it most needed to be visible.
5. The Brazilian pricing investments, now in the base for two full quarters. The buyer discounts and seller take-rate reductions introduced in April were named at Q2 as the largest single driver of a 276 basis point gross margin decline. They are not mentioned anywhere in the Q3 letter and have never been quantified.
6. The quantified cost of the Brazilian work-week reduction. Management confirmed it will require more logistics employees and said it is early. A 9% reduction in legal hours with no pay adjustment in the market that produces 56% of revenue is an input-cost change that will show up in 2027, and no number exists in any public document.
7. Off-balance-sheet credit card commitments, now $14,892M. Third consecutive quarter with no mention in the letter and no analyst question, on a balance 1.76 times the drawn card book. The news this quarter is favourable, since growth fell to 6.0% from 18.1%, which makes the continued silence harder to explain rather than easier.
8. Assets under management, AUM per user, active sellers, conversion, credit exposure per user, loyalty subscriber growth, cross-border GMV growth, the three-or-more-category share and insurtech policies per user. Every one of these was quantified at Q2 and none is quantified in the Q3 letter. The nine withdrawn metrics were the evidentiary base of last quarter's ecosystem argument.
9. The ecosystemic share of the user base, for a third consecutive quarter. Ecosystemic user growth decelerated from 37% to 33% and the uplift multiples were restated, but the share of the base they represent was again not given, which is the one input that would make the multiples modellable.
10. Any explanation for Other Countries. Direct contribution margin fell 508 basis points sequentially and has nearly halved since Q1, in the segment that houses the China fulfillment centre whose cross-border growth metric was also withdrawn. No commentary, no question.
11. Brazilian payroll loans. Flagged at Q1 as launching soon with government integration already complete. Second consecutive quarter with no mention in the letter, no mention on the call and no analyst question.
12. Any capital return. Treasury shares rose by 662 for the nine months, the same figure as for the six months, so nothing was repurchased in Q3. A $1bn bond was issued in September and net debt rose 25% in the quarter. Capital allocation was not raised by management or by any analyst in any of the three quarters we have covered.
13. The revenue-tax tailwind inside the gross margin. Revenue-based taxes fell from 5.8% of revenue to 5.5%, worth 30 basis points of gross margin and roughly a third of the sequential bridge. It appears only in the quarterly filing. Management named three gross-margin drivers and this was not one of them.

Market Reaction

• Pre-print setup: Closed at $1,739.88 on November 4 ahead of the after-market release. Down 13.6% year to date against an S&P 500 up 12.3%, down 24.5% over trailing twelve months and down 6.5% over trailing thirty days. The 52-week closing range entering the print was $1,546.81 to $2,316.40, so the shares came in 24.9% below the high and 12.5% above the low.
• The pre-print shock: On October 26, after the Brazilian incumbent won a surprise fourth term, the shares fell 7.8% from $1,810.15 to $1,668.72 on 2.4M shares, roughly four times normal volume, having gapped far lower at the open. Brazilian equities fell broadly in that session. The stock then recovered 4.3% over the seven sessions into the print.
• After-hours move: The release was filed at approximately 4:05 p.m. ET and the video and call began at 5:00 p.m. ET. The decline built through the evening and carried into a 7.5% opening gap the following morning.
• Reaction session: Opened at $1,609.99, which was also the session high, traded down to $1,469.01 and closed at $1,530.28, down 12.0% or $209.60. Market capitalisation fell from $88,196M to $77,580M, so $10.6bn of equity value was erased on the print.
• Volume: 2.1M shares against a 30-day average of 0.6M, roughly 3.7 times normal, against 3.1 times in Q2 and 6.0 times in Q1.
• New 52-week closing low: $1,530.28 is the lowest close of the trailing 252 sessions, $16.53 below the prior low of $1,546.81 set on May 15, and 33.9% below the pre-print 52-week closing high.
• Market context: The S&P 500 fell 4.4% in its worst session since April 2025 and the Nasdaq Composite fell 5.5%, its largest daily decline in nineteen months, on expectations of tighter United States regulation of artificial intelligence following the midterm results. The semiconductor index fell more than 12%.
• Peer reaction: The Brazil equity ETF fell 1.5%, Nu Holdings 2.1%, StoneCo 2.6%, PagSeguro 3.3%, dLocal 3.5%, Amazon 2.2%, Sea Limited 5.7% and Shopify 6.1%.

Separating the market from the print. A 12.0% decline on a day the index fell 4.4% invites a lazy conclusion in either direction. The peer table settles it. Latin American financial and payments names moved 1.5% to 3.5% and the Brazil ETF moved 1.5%, so this was not a Latin American session and not a Brazil-macro session. The selloff was specific to artificial intelligence and semiconductor regulation risk, a theme with no direct bearing on MercadoLibre, which propagated into the high-multiple growth complex. MELI underperformed the S&P 500 by 7.7 percentage points, the Brazil ETF by 10.5 and Nu Holdings by 10.0. The honest attribution is that roughly seven to eight points of the decline is the print and the remainder is multiple compression that would have hit a 46x stock on that tape with or without earnings.

Why the print itself cost seven or eight points. The quarter beat on revenue and missed on earnings by 5.1%, which on its own is a mid-single-digit event. What took it beyond that is the third consecutive quarter in which the margin went down for a different reason than the quarter before. In Q1 it was provisions, in Q2 it was gross margin, in Q3 it is marketing and recruitment. A market that has spent three quarters being told which line to watch, and has watched that line get fixed while the operating margin kept falling, is entitled to stop underwriting the next explanation. That is what a 12% session on 3.7 times volume into a 52-week low expresses, and it is not an unreasonable read of the pattern even though the composition of this particular decline is the most benign of the three.

The shape against the prior two quarters. Q1 produced a 12.7% decline on 6.0 times volume because the shareholder letter's "not changing materially in the near term" language reset the whole year during the call. Q2 produced a 4.8% decline on 3.1 times volume, with the after-hours print landing within 0.6% of the eventual close, because the market priced a known margin composition correctly on first read. Q3 sits between them on volume and close to Q1 on price, which does not fit either template. The after-hours move built through the evening and the opening gap widened further intraday to a 15.6% low, which is the signature of positioning unwinding rather than of a fundamental reassessment. On an ordinary tape we would expect this print to have cost five to seven points.

Street Perspective

Every published action on the reaction session was a price-target reduction with the rating left intact, and the cuts averaged roughly nine percent against a twelve percent decline in the shares. The aggregate target still implies close to forty percent upside, so the sell-side is treating the margin trajectory as a timing problem rather than a value problem. That is the central disagreement to resolve.

Debate: Did the Gross Margin Stabilise, or Did the Decline Just Move House?

Bull view: The bull case on the Street is that the quarter passed its own test. The gross margin held at 40.8% with the Brazilian pricing investments in the base for two full quarters and with no provision reversal left to spend, which is exactly the clean read the market asked for. Brazilian cost per shipment fell 5% in local currency and the lowest ASP band turned variable-contribution positive for the first time. The entire sequential operating margin decline is now attributable to a marketing channel management calibrates quarter by quarter and to recruitment costs management expects to partly reverse. Product development delivered 72 basis points of leverage while absorbing more AI spend. Strip the deliberate and the temporary and the underlying margin turned this quarter.

Bear view: The bear camp reads the same bridge and sees the explanation move house for the third consecutive quarter. Provisions in Q1, gross margin in Q2, marketing and recruitment in Q3: each quarter the named cause is fixed and the operating margin falls again. 40.8% is the lowest gross margin in eighteen quarters, and half the 60 basis points of relief in the sequential bridge came from a revenue-tax line management did not mention and does not control. First-party revenue, which the filing says carries a structurally lower margin, is growing 67% and is now 14.4% of the total. The operating margin has fallen for three straight quarters to its lowest level since Q1 2022, and management again refused to name a target or a timeline.

Our take: The bulls win the narrow argument and the bears win the general one. The specific test we set was whether the gross margin had found a level, and it had; that removes the downgrade scenario and is the single most important fact in the print. But the bear observation about the explanation rotating is not a debating point, it is the pattern, and it is the reason a 12% session was possible on a quarter with this composition. What decides it is Q4. Management has committed that part of the capacity cost reverses, consensus needs 84 to 129 basis points of sequential margin expansion to hold FY26 together, and the last two years delivered 304 and 38 basis points of seasonal lift. If the peak-season quarter cannot produce leverage with a named reversal behind it, the rotation of explanations stops being a communication issue and becomes the thesis.

Debate: Is the Credit Book Improving, or Is the Improvement Borrowed From a Government Programme?

Bull view: Every forward measure in the credit book improved and the most important one turned decisively. Cost of risk fell for a second consecutive quarter to 31.3%, from 33.0% and 36.8%, provisions fell to 55.5% of credit revenue, portfolio growth decelerated to 9.2% sequentially from 16.4% two quarters ago, and the gap between portfolio growth and revenue growth narrowed from 38 points to 17. Off-balance-sheet unused card commitments, the forward pipeline of the provision charge, grew 6.0% against 18.1% and 32.1%. The card cohort ladder was disclosed for the first time and shows more than half the Brazilian book already net-interest-margin positive. Management has tightened at the margin on the riskier Brazilian segments while still growing 62%.

Bear view: The bear camp points at the one thing that is not in the flow data. Renegotiated loans jumped from 1.5% of the portfolio to 2.2% in the quarter, about a 60% increase in dollars while the book grew 9.2%, and that is the quarter management confirmed participating in a government debt-renegotiation programme and a credit-bureau discount campaign and twice declined to size the benefit. Renegotiation moves balances out of delinquency, which is precisely why the 15 to 90 day ratio held at 7.0% and the 91 to 360 day bucket posted its smallest increase in three quarters. Behind that, the late-stage bucket still rose to 19.33%, as it has in every quarter of 2026, total past due reached 28.71%, allowance coverage reached 27.09%, nine-month write-offs are up 80.9%, and originations still grew 70% against a 62% portfolio.

Our take: The bulls have the stronger case this quarter and it is still not strong enough to close the argument, because the one disclosure that would close it was requested twice and withheld. The deceleration in unused card commitments is a genuine, unheralded positive that nobody on the call and nobody in the published sell-side reaction mentioned, and it is a better forward indicator than any ratio management chooses to cite. But the early-stage stability and the slower late-stage deterioration both arrived with an undisclosed assist in the same quarter. Treat the Q3 improvement in the early buckets as unverified. The Q4 ladder, printed with no programme running, is the clean test, and it arrives at the same time as the margin test.

Debate: Does 40x Forward Earnings Finally Pay for an Unresolved Margin?

Bull view: The valuation argument has flipped. In May the problem was a stale estimate; in August the problem was that the estimate got honest while the price rose 12%. Both are now resolved. FY26 consensus of $38.18 needs Q4 revenue growth below what the last two Q4s delivered and 84 to 129 basis points of sequential margin expansion inside a seasonal range of 38 to 304 basis points, so the number is credible rather than aspirational. The shares are at 41.0x trailing and 40.1x forward against 49.8x and 47.9x three months ago, 2.05x FY26 revenue against 2.39x, and 27.7x FY27 earnings. This is the first quarter of our coverage in which MELI is cheaper than at initiation on both trailing earnings and our own forward estimate, on a business compounding revenue at 45%, taking share in all three major markets and accelerating FX-neutral commerce revenue growth for a third consecutive quarter.

Bear view: A cheaper multiple on an unresolved margin is still an unresolved margin, and the earnings base has gone nowhere for three years. FY2024 diluted EPS was $37.69 on $20.8bn of revenue; FY2026 consensus is $38.18 on $41.8bn. Revenue doubles and earnings do not move, so the business has converted an entire doubling of scale into reinvestment with no demonstrated return in reported profit. Trailing EPS of $37.28 is still below FY2024's. At 40x, investors pay a premium growth multiple for a 4.2% net margin business whose net debt has risen 71.5% in nine months, whose capital intensity is rising again, which has returned no capital in this cycle, and whose management declines on principle to give any margin target. There is no multiple support until the earnings base moves.

Our take: The bear framing of the three-year earnings record is the single most useful sentence anyone can write about this stock and it is the reason the rating does not go to Outperform on a 12% drop. But the bull arithmetic is now correct in a way it has not been in either of our prior two notes. The specific reasons we held in May and August were a dishonest denominator and a rising price on a falling number, and both are discharged: the denominator is credible, the price has come to us, and the stock is below the initiation multiple on trailing earnings and on our own forward estimate for the first time. What is not discharged is that the operating margin has fallen for three consecutive quarters and the only forward commitment available is qualitative. The asymmetry has genuinely improved; it has not yet inverted. One quarter of evidence, with a named and falsifiable commitment attached, is worth more than one quarter of upside.

Model Update Needed

The comparison below is against post-print Street positioning and against the framework set out in our Q2 note.

Valuation impact: At the reaction close of $1,530.28 the shares carry a market capitalisation of $77,580M and an enterprise value of $85,608M against net debt of $8,028M. On our FY26 EPS range of $37 to $40 the stock trades at 38x to 41x, against 41.0x trailing twelve-month earnings of $37.28, 2.05x FY26 consensus revenue and 27.7x FY27 consensus earnings. Three months ago we wrote that the arithmetic producing our Hold was intact at a higher price on a lower number. That has reversed: the price is 16.4% lower than at the Q2 reaction close on an unchanged number, 12.0% of it in the reaction session, the forward multiple has compressed from 47.9x to 40.1x, and for the first time in our coverage MELI is cheaper than at initiation on both trailing earnings and our own forward estimate. The move also closes the performance gap we reported in August. Between the Q2 and Q3 reaction closes MELI fell 16.4% while the S&P 500 fell 4.6%, so MELI gave back 11.8 points of relative performance this quarter and now trails the index by 5.7 points since initiation, against a 7.9-point lead at the Q2 recap: MELI is down 6.3% from the initiation reaction close against an S&P 500 down 0.6%. We said the entry point we were waiting for never arrived. It has now mostly arrived, which is a different statement from saying it has arrived.

Thesis Scorecard Post-Earnings

Overall: Modestly stronger in aggregate for the first time in our coverage, on two moves in opposite directions that do not cancel. The valuation pillar genuinely improved: the estimate is credible, the multiple de-rated 12% in one session and 16% since August, and the stock is below its initiation multiple on trailing earnings and on our own forward estimate. Corporate and technology leverage strengthened, the credit flow data improved on every measure and the forward provision pipeline decelerated hard. Against that, the operating margin fell for a third consecutive quarter, the explanation for why moved house for a third consecutive quarter, nine quantified engagement metrics were withdrawn, advertising decelerated eleven points of FX-neutral growth in the quarter it most needed to be visible, and a government programme of undisclosed size sits inside the credit improvement. A new unquantified structural cost arrived in the market that drives 56% of revenue. The thesis has not changed; the balance of evidence under it is better than it was, and the price is materially better than it was.

Action: Maintain Hold, conviction to 7 from 6. The upgrade condition we published in August was explicit, a Q3 gross margin at or above 41.5% with the 91 to 360 day bucket flat, and neither half was met: gross margin printed 40.78% and the bucket rose 64 basis points. Honouring our own stated test is the discipline that makes this scorecard worth keeping, so the rating does not move. The downgrade condition, another leg down in gross margin alongside building late-stage delinquency, was also not met; the gross margin decline stopped. What changed is that both original reasons for the Hold have weakened: the denominator is now credible and the multiple has come to us. We are one quarter from a genuinely clean test, with a named and falsifiable management commitment attached to it, in a quarter that is seasonally the company's strongest. Revisit toward Outperform on a Q4 operating margin at or above 7.5% with the 91 to 360 day past-due bucket flat or lower, which together would confirm both that the capacity cost reversed as promised and that the 2026 Brazilian vintages have stopped seasoning adversely. Revisit toward Underperform if the Q4 operating margin fails to expand sequentially at all, since a peak-season quarter with a named cost reversal behind it that still cannot produce leverage would make the margin decline structural rather than elective, or if the renegotiated share of the portfolio rises again while the late-stage bucket accelerates.