The full Amazon review is in section 2 of this article and is for paid readers. Upgrade below to read.

Upgrade

MELI is the e-commerce and logistics king of most of Latin America. It has a thriving marketplace and fulfillment business, with support for 3rd-party merchants. It also features a rapidly growing financial services suite and payments platform, entertainment offerings through partnerships, a quickly expanding ads business and a loyalty program called MELI+. That's where it laces a plethora of product utility into one unique consumer bundle. The business model resembles Amazon without cloud computing and with financial services. Here are the names of its various products:

By market, Brazil revenue beat estimates by 7.8% and grew by 38% Y/Y. That’s a meaningful 13-point acceleration vs. last quarter. FX helped, but I was still encouraged to see its largest source of outperformance come from Brazil. That’s the market where competitive concerns from players like Amazon and Shopee have been the most pronounced lately. MELI's initiatives (covered later in the piece) are working to drive demand. Argentina revenue missed estimates by 10% and Mexico revenue beat estimates by 6.4%.

The sequential acceleration seen below was helped by easier Y/Y comps, but that comp also includes lower shipping and merchant fees compared to last year. That makes the acceleration a lot more notable.

“In Argentina, growth of GMV, buyers, and TPV remained resilient in Q3. Still, trends slowed through the quarter due to the challenging macro backdrop… looking ahead, we think that hopefully some of the volatility will go away after the results of the elections.” – CEO Martin de los Santos

So much to discuss for margins. There were a ton of moving pieces that we covered throughout the quarter and had material impacts. It’s accelerating operating expenses (OpEx) across a few categories to extend its competitive edge and capture what is still a massive future opportunity. Examples include its lowered free shipping minimums in Brazil, which had a full quarter impact for Q3 compared to just a 1 month impact last Q. Other examples are its reduced Brazilian merchant fees, investments to grow 1st-party assortment in categories like grocery, more marketing and higher credit provisions stemming from overall portfolio expansion. That last factor is why FCF missed by such a wide margin. The credit book was nearly $1B larger than expected and originations count towards cash burn. That credit item is easing in terms of Y/Y margin headwind, but was expected to ease a lot more than it did.

The impact on contribution margin (especially in Brazil) is a lot more pronounced than the EBIT impact. And that’s actually very encouraging. The costs described above are all contribution profit drains right now. But? If they can deliver material boosts to revenue generation, it means added fixed cost leverage in areas like G&A that helps buffer the EBIT margin contraction. The fact that this is playing out is direct evidence that the growth investments are yielding real revenue gains. Immediately. It’s also how EBIT still managed to grow 30% Y/Y despite all of these large incremental costs. Not only are the investments a continuation of proven things that have worked for long-term profitability, but they’re occurring while MELI still delivers great near-term profit growth. That makes the strategy very easy for me to support. Time and time again, they’ve shown an ability to allocate capital in a way that fosters rapid market share gains. And they've turned those improvements into great returns. This should be more of the same, with plenty of data later in this piece supporting that opinion.

For Argentinian contribution margin, macro pressures led to rising funding costs and a 450 bps Y/Y headwind for that metric. It would have been flat without this factor. For net interest margin after losses (NIMAL), the note on higher Argentinian funding costs led to that decline and miss. Credit quality is stable or improving.

Net income growth of 6% Y/Y vs. 30% EBIT growth was related to two things. First, its tax bill rose 75% Y/Y to 34% of EBIT vs. 24% of EBIT Y/Y. Second, foreign exchange losses more than doubled Y/Y to $102M. Because Argentina is in hyperinflation according to GAAP accounting, negatively fluctuating asset valuations held in Argentine pesos count directly against net income. With a stable Y/Y tax rate and FX losses, net income would have grown by about 37% Y/Y.

And finally, there’s an added challenge for 2-year margin comparisons. MELI made a series of reporting charges during 2024. Mercado Pago’s interest income/expense was moved above the EBIT line. Mercado Envios changed to reporting gross revenue, rather than deducting shipping costs and reporting net revenue. This represents about a roughly 2-point margin headwind vs. 2023 comparisons.

“We believe we are still in the early stages of realizing our full potential, with e-commerce penetration of just mid-teens in Latin America, and single-digit market shares – in some cases, low-single digits.” – Shareholder Letter

“In the long term, we continue to be very optimistic about the margin profile of our business as we continue to scale the business.” – Co-Founder/CEO Martin de Los Santos

It doesn't offer guidance aside from saying "the best is yet to come" every quarter.

MELI trades for 45x forward EPS. EPS is expected to compound at a 46% clip in 2026 and 2027 after just 9% growth this year.

Credit Health:

Credit growth is so elevated because credit health is so strong. Their ability to use extensive customer commerce data to know borrowers better and more effectively underwrite is working well. Again, the NIMAL decline was related to Argentina macro, with a mix shift from personal loans to credit cards providing the rest of the decline. I repeat... credit quality is stable or improving. First time payment defaults set yet another record low in Brazil and improved in Mexico. These cohorts are clearly showing signs of future profitability, with around half of its Brazilian credit book now delivering a positive NIMAL. Mexico is earlier on in its journey but is showing the same pattern. As the book scales, the front-loaded provisioning will become a less intense part of the overall business because new credit will shrink as a portion of the book. That means the margin drag will keep easing.

Commerce Segment:

This category is showing the most obvious signs of aforementioned growth investments working. The lowered free shipping threshold in Brazil led to its fastest rate of buyer growth in 4+ years. The rate was faster than the pandemic peak. Items sold growth in Brazil and overall both improved accordingly. Their “market share gains accelerated” while customer delight reached new heights. Businesses also responded positively. Merchants with sales between R$19-79 (lowered free shipping threshold from R$79 to R$19) rose by more than 10% Y/Y; new listings within that price range rose by 200% Y/Y. All of this assortment delivered record conversion and retention rates. Lower fees… fueling more customer value... which powers more customer engagement… which motivates more assortment… which generates even more engagement. 

“The impact on the marketplace is really enormous.” – Co-Founder/CEO Martin de los Santos

While MELI makes this formula look easy, it isn’t. It’s actually quite hard to execute. The firm's ability is the byproduct of its world-class fulfillment network and cost efficiency work. Along these lines, unit shipping costs fell 8% Q/Q in Brazil thanks to better usage of its slow shipping network. They're getting better at batching orders, using excess capacity and shrinking deadweight loss within this part of the logistics footprint. As a result, Brazilian shipments rose by 28% Q/Q with zero impact on customer service... and again... while Brazilian unit costs fell.

The trend in Mexico is similar. The same cost fell 12% Y/Y while on-time deliveries set a new record.

Marketing:

Part of the added growth spending includes efforts in social media marketing. As leadership told us, Latin Americans spend more time on those services than a typical human. As a result, they’ve increased focus and grown affiliates by 4x Y/Y. Beyond this, brand campaigns across its core markets and Chile led to “record brand preference scores.”

Advertising Business:

The newer Mercado Ads platform is performing well. Off-platform relationships with companies like Google are allowing MELI to provide far more ad inventory to its customers. Partnerships with Roku and HBO will build on that trend and should help display & video ad revenue maintain nearly 100% growth. For now, overall FXN ad growth accelerated from 59% Y/Y last quarter to  63% Y/Y this quarter.

Loyalty Program:

MELI added a new shipping benefit for MELI+ members in Brazil. Those members used to be the only people able to access free shipping below R$79. That went away with the lowered minimum, so MELI added a different perk. Now, it lets members tap into this cheaper free shipping option with the same or next day MELI network, compared to the slower network for non-members. In other news, it launched a new Meli+ “super bundle,” with Disney+, Netflix, HBO Max, Apple TV and all of its existing perks in one package. Finally, it announced partnerships with McDonald’s, Petrobras and others. Now MELI+ members using a MELI credit card with those merchants will get unique cash-back rewards. They’re confident in the consumer subscription remaining the most compelling in the market and have plenty of ideas on how to further boost value.

Fintech:

Fintech net Promoter score (NPS) in Brazil set new highs. Leadership believes they have the highest NPS in the nation, but that’s based on “internal data” so it’s hard to verify. Regardless, the experience they’re delivering is resonating. Looking ahead, creating more primary banking relationships (PBRs) is key. It raises retention, credit quality and overall engagement. The push is going well, as the proportion of members using MELI for primary banking in Brazil rose by 11 points Y/Y. The credit card is a big piece of this strategy. It drives usage frequency and lifetime value gains in Brazil, and they want to match that in Argentina as well. Mercado Pago is already ubiquitous there thanks to QR payments and its savings account. They have more pre-card launch traction there than they did in Brazil or Mexico and I'm optimistic that Argentina will look just like those other two (or better).

Acquiring TPV:

Acquiring TPV refers to the piece of MercadoPago that lets merchants seamlessly tap into these easier payment options, augmenting conversion rates (through software and point of sale hardware). Acquiring TPV happens whenever MercadoPago is used as the main payment facilitator/processor on the marketplace or on a merchant’s own site (“off-platform” acquiring TPV). They’re taking market share “across all markets” thanks to popular credit, bank service and solutions like point of sale software and sales data analytics. 

Other Countries:

Great quarter. If they’re going to show this much immediate progress from growth investments, I say go for it. We don’t even need to speculate. We can see how these projects are already yielding the kind of user and engagement gains that will lead to more profitable growth down the road. And furthermore, I see a lot of the Argentina FX margin weakness as very temporary. There are so many things weighing on its margin profile right now that will go away. As that happens, this will turn back into a rapid grower with expanding margins. And I think that will occur in 2026. MELI’s positioning in arguably the most compelling geography out there is simply ideal. They’re the clear service and share leader, with economies of scale that enable them to deliver value others can’t rationally match. Their markets will stay competitive. Sea Limited and Amazon will keep taking some market share. And I think MELI will continue to dominate regardless.

I’ve leaned into this name pretty aggressively amid the recent decline and am glad that I did. I’m so comfortable owning a large stake in this company and would look to add again if there’s any drama. Two things are certain… there will always be some level of macro noise pertaining to at least one nation in Latin America… and MELI will continue to overcome those headwinds and thrive across cycles.

“We're encouraged by the start of the peak season, and we are ready to serve customers in the coming months.” – CFO Brian Olsavsky

Amazon trades for 32x forward EPS. Excluding the Anthropic help, non-GAAP estimates won't move much. EPS is expected to grow by 20% this year, 15% next year and 22% in 2027.

Fulfillment:

Amazon continues to work hard on optimizing every part of the fulfillment chain. As we talk about a lot, doing so allows them to improve delivery speed and costs, creating a compelling win-win and fortifying the market leader’s dominance. They’re localizing shipments, getting better at combining packages, improving demand forecasting, precisely placing inventory near the final destination and implementing robotics in ways that enable them to consistently get better with timeliness and efficiency. There’s so much more improvement left to enjoy in every single one of these proven areas.

This progress is how they’re profitably unlocking same-day perishable grocery delivery in 2,300 U.S. cities by 2026 and generating everyday essentials category growth 2x faster than the overall business. Their unmatched network enables an ability to do things like service high-density urban orders in a few hours; that’s so important for these high-growth areas. On the other end of the spectrum, elite and always improving efficiency is also how they’re rationally entering more rural communities around the nation. Subscale competitors are being forced to back away from these communities as they struggle to profitably serve them. Amazon is able to lean in because of how much more scaled and optimized its fulfillment business is.

In other news, Amazon’s white-labeled fulfillment service, Amazon Multi-Channel Fulfillment (MCF) inked new partnerships with Walmart, Shopify and SHEIN. The three large commerce companies will now offer their merchants world-class Amazon logistics right on their websites.

AWS – Capacity, Hardware & Models:

Last quarter, I argued that AWS wasn’t falling behind in AI. I talked about AWS’s superior model choice and security, as well as an emerging slate of compelling AI hardware and app-based products. And I cited AWS backlog growth of 25% Y/Y as evidence that they weren’t struggling to find demand. What were they struggling to do? Place capacity. That is what Azure has been better at. They’ve more quickly turned OpenAI compute deals into cloud revenue than AWS has with Anthropic. And so Azure accelerated while AWS didn’t, making it look like they were falling behind in AI capabilities. In reality, Amazon feels like they’re second to none in terms of tool breadth. And while that’s subjective, I do think it’s pretty clear they’re on par at worst.

It’s now finally catching up in terms of compute volume gains, with Project Rainier launched and an industry-leading 3.8 gigawatts of power added to their footprint this year. And so, as you may expect, growth is now accelerating to its fastest clip in 3 years, with reasons to believe Anthropic will keep that trend going well into 2026. What evidence? Backlog grew by 22% Y/Y to $200B and deal volume in October (not reflected in this) already surpassed all of Q3. It sounds like there are a few cloud win announcements it will soon make. Furthermore, they’re on track to add more power during Q4 than the quarterly rate so far this year. They’re immediately monetizing this capacity as soon as it comes online. So? If they’re accelerating pace of power additions, faster revenue growth should follow suit. Jassy echoed this sentiment by stating he thinks AWS can maintain this current momentum, but he wouldn’t explicitly offer forward growth guidance.

“Because of its advantaged capabilities, security, operational performance, and customer focus, AWS continues to earn most of the big enterprise and government transformations to the cloud.” – CEO Andy Jassy

On the hardware side of things, its Trainium2 custom AI chip is “fully subscribed,” generating billions in annual revenue and growing 150% Q/Q. Project Rainier was a big reason for this, as it boasts a massive 500,000 Trainium 2 chip cluster to power Anthropic's frontier model work. That will soon scale to 1 million chips. While Trainium is only used by a small number of customers (mainly Amazon and Anthropic), they see that changing with Trainium 3. The team expects to greatly broaden the customer roster, thanks to 40% price performance gains vs. the already highly efficient Trainium 2 chip. Rainier has been great proof of concept for this product line, and Amazon is ready to now take the training wheels off.

For models, it added new OpenAI, DeepSeek and Anthropic models to Bedrock. Furthermore, it debuted Multimodal Embeddings for its Nova model, which should enable lower hallucination rates for chat, image, video and audio-based token generation.

One more note here. Google is the best full stack AI player, but Amazon is no slouch at connecting various pieces of the AI opportunity:

AWS – AI Application Layer

Amazon split this conversation into two sections: agents for customers not wanting to build from scratch, and tools for those that do. For customers wanting to use existing agents, there were a few updates to note. Kiro is its integrated developer environment (IDE) and coding agent. After racking up 100,000 users in one week, it has doubled sequentially. Customers are flocking to its conversational app-building capabilities and simple templates. Transform is helping customers automate the once-cumbersome cloud migration process and 4x the speed for customers like Thomson Reuters. Overall, Transform is up to 700,000 hours of manual work saved less than a year after launching. And finally, its contact center agent called Connect is up to $1B in annual revenue. 

To encourage faster, easier adoption of agents, Amazon debuted Agentcore this quarter. It provides all needed foundational building blocks, including memory, compute, security and everything else a customer needs to safely embrace agentic AI. We covered it in more detail here (section 8). 

“When we talk to our customers, Agentcore really resonates. There is not anything else like it. It's changing their timeframe and their receptivity to building agents, and it's very compelling for them.” – CEO Andy Jassy

Quick Suite, which was also released recently and is covered in section 3 of this article, is a slew of enterprise-ready agents to deploy for various use cases. Jassy talked up the product set’s ability to drive 80%+ time savings and 90% cost savings for certain workflows and products.

For customers wanting to build on their own, its Strands agentic development kit (ADK) makes building agents a lot less intimidating, with a full suite of Bedrock models and tools to pull from within a single, secure interface.

Marketplace:

AI is helping its marketplace too. Just like with Meta, Amazon is using this tech to benefit its core business today. Rufus is delivering a 60% boost to shopper conversion probability and has been used by 250M customers. That’s eventually expected to deliver $10B in incremental annual revenue. Its “Help Me Decide” tool is also infusing AI into the product discovery process, more granularly matching customers with interests based on data-driven insight to improve the experience.

On the merchant side, 1.3M sellers are now using its AI tools for listings and other use cases.

Interestingly, Amazon sees agentic commerce speeding up e-commerce adoption. They rightfully pointed out that Amazon is best when a customer knows what they want. If they’re looking for product testing, comparison or education, brick and mortar is still better. With capable agents answering all of your questions and still connecting you to a lot more assortment, that could change in the years to come.

In non-AI news, selection grew 14% Q/Q while features such as “add-to-delivery” (add another item to a cart post purchase) have already been used 80 million times over a very short period of time.

Advertising:

Prime:

Prime Video delivered 65% growth in viewership for their popular title “The Summer I Turned Pretty.” They continue to notch strong engagement trends for hit titles and grow more confident in this being a great business over the long haul. Thursday Night Football on Prime is enjoying 16% Y/Y viewership growth, while Prime just added Peacock and Fox One as new subscription options to keep momentum going.

New Bets:

30,000 Employee Layoff:

Amazon didn’t make this announcement because of AI, robotics or to enhance margins. That’s what Jassy told investors. Instead, this was made to improve company culture. They’ve gotten too bloated, with too many layers of middle management slowing the builders down. They hired way too aggressively and indiscriminately during the pandemic, and that led to these issues. They’re now trying to fix the problem, get leaner and move faster. Alphabet and Meta both made similar moves in recent years. While this sucks for those impacted, it’s good to hear this isn’t AI taking everyone’s job and this is also probably going to make Amazon a better-run company.

Strong quarter. To be expanding North American EBIT margin on a Y/Y basis (excluding one-off charges) despite tariffs is so impressive to me. But I suppose everything about Amazon is impressive. Their marketplace’s combination of speed, assortment and pricing is unmatched in many markets, while their category expansion is going very well. Kuiper is now obviously going to become a successful business and start turning large front-loaded costs into scaled revenue. For cloud, sentiment for AWS felt divorced from reality. Folks interpreted capacity growth timing delays as evidence that AWS was falling behind in AI. As I said last quarter, they weren’t. They just needed to move faster in turning backlog into revenue. And now that’s happening. I also think Jassy did a better job this quarter of actually explaining all the exciting things going on within the segment. Between the Trainium disclosures, rapid AI tool adoption and October deal commentary, they gave investors a lot to be excited about. I am happy to pay 30x forward earnings (ex-equity investment help) for the leader in the two categories that arguably boast the best combination of rapid and structural growth (e-commerce and cloud). I have absolutely no interest in selling any shares.