Next week, I plan to publish a review of the Adobe earnings transcript and materials. Part one, which includes all quantitative charts and data, is in section 2 of this article. I will also publish a catch-up review on Lululemon next week. I was planning on pushing that to the end of the month to include it side-by-side with the Nike earnings review and I think I’ll have time to do it next week. Next week’s article will also include updated valuation comp sheets and the typical weekly newsflow.

MongoDB is a key player in data storage and analytics with a document-oriented setup. This differs from legacy relational-style databases and next-gen versions like Snowflake’s. How so? Legacy relational databases store data in rigid rows and columns linked by pre-set relationships. These databases look like giant Excel spreadsheets and use structured query language (SQL) to work. Legacy relational databases cannot seamlessly handle unstructured data like MongoDB’s data lake can. Legacy relational databases struggle to horizontally scale and unlock the most advanced, flexible and multi-stage querying. The datasets are fixed, with formatting and filtering more limited. This is a large limitation, considering how important unstructured data is for GenAI use cases. The inability to provide “not only SQL” (NOSQL) slows performance and diminishes value. MongoDB’s NoSQL database, called Atlas, and document-style storage fix these issues, which is why the pace of migration continues to ramp up.

“MongoDB’s platform enables developers to more easily represent the messiness of real-world data. This includes understanding relationships between structured and unstructured data, and managing data that is constantly evolving and changing. This fundamental architectural advantage provides customers greater flexibility, faster time to market, and the ability to scale without rearchitecting.”

CEO Dev Ittycheria

Atlas’s cloud-native database architecture uses a group of servers (or a cluster) to store data for app creation within its platform. The nature of MongoDB’s Atlas product allows clusters to be easily added to or subtracted from for seamless tweaking as needs fluctuate. It also offers MongoDB Realm as a mobile environment for app creation and MongoDB Search for data querying. Finally, MongoDB Data Lake is what specifically houses unstructured data for querying and analytics.

“MongoDB removes the constraints of legacy databases, enabling businesses to innovate at AI speed with our flexible document model and seamless scalability.”

CEO Dev Ittycheria

More Core Atlas Products:

Product Expansion:

MongoDB thinks it’s great at automating the cumbersome data modernization work needed for migrations. Its Relational Migrator product is a helpful tool to take client headache out of this. Now, it’s focused on leveraging Relational Migrator to actively help companies re-code their legacy applications and create a more end-to-end, full-service migration service. With this holistic strategy, MongoDB doesn’t just aid with data migration, but directs automation of software modernization to seamlessly run apps through its document-oriented foundation.

MDB also recently bought Voyage AI for another product expansion opportunity. Voyage is a key player in GenAI and agentic AI model trustworthiness. While models routinely create delightful, jaw-dropping experiences, they’re also wrong a lot. Hallucination rates (wrong answer rates) can often reach 25%. This makes models really only useful when you know what a right answer should look like and are double-checking. That limitation puts a tight ceiling on utility and is what Voyage aims to fix. It has two products. First is its set of Vector Embedding Models. As leadership puts it, these are the “bridge between models and a client's private data.” They facilitate meticulous information transferring into models. It organizes data as a number sequence to more closely group data points in a much more standardized manner that models can read. All of this helps with machine learning accuracy rates and uncovering patterns to reduce model mistakes. Secondly, it offers reranking models. These ingest and reorganize materials for a given model input based on data relevance. They filter model responses to help eliminate waste and inaccuracies. With Voyage, MDB gains a broader suite of GenAI services to round out a more cohesive platform. Now, it has Atlas, significant help for app and data migration/modernization, real-time data streaming, world-class search tools and highly-regarded embedding and reranking models under one roof.

“By combining these into one platform, we make it dramatically easier for developers to build intelligent, responsive apps without stitching together multiple systems.”

CEO Dev Ittycheria

And lastly, MongoDB also offers an environment for GenAI app creation. MongoDB AI Applications Program (MAAP) provides a slew of templates, guardrails, 3rd party integrations and configuration tools for fostering AI app creation.

Non-Atlas Business:

Enterprise Advanced (EA) refers to its on-premise (Atlas is cloud-based), database and app bundle. It allows companies to purchase licensing for subscription-based usage (rather than paying for consumption under Atlas).

Business Model:

Note that most of MongoDB’s growth is based on acquiring new customers and migrating workloads onto the platform, as well as consumption-based revenue from customers using its data products and growing workloads.

“Overall, we posted a strong Q1 despite a dynamic and fast-changing macro environment.”

CEO Dev Ittycheria

Guidance continues to assume that non-Atlas revenue comp headwinds will represent a $50M headwind for annual revenue. The raise was due to Atlas (a good thing).

Recent negative profit quarters make valuation charts far less valuable. That’s why I used my least favorite metric (sales multiples).

MDB trades for 70x forward EPS. EPS is expected to fall by 15% Y/Y as it laps an abnormally strong period of high-margin non-Atlas revenue. EPS is expected to grow by 20% the next year and by 32% the year after.

Demand:

MongoDB added its highest total number of new customers in 6 years. Retention rates were strong, with net revenue retention steady at 119%. For Atlas specifically, consumption was as expected. They did see a dip in April following trade war drama, but February and March were both very good, while May showed a strong M/M recovery. This all helped growth for Atlas accelerate – even when adjusting for more days in this calendar quarter.

Competition & Postgres:

The main theme of the Q&A was developers seemingly embracing Postgres, with both Snowflake and Databricks making purchases in the area to bolster their offerings. Does MongoDB see this as a threat to their NoSQL niche and a reason to think SQL will take share back from its offering? The short answer is no.

Leadership sees this M&A as “validating” the importance of its world-class capabilities within Online Transaction Processing (OTP). Competition is skating to where MongoDB already specializes.

“Architecturally, we are far better optimized for this new world of complex modern applications, especially in the world of AI. MongoDB was designed to address the needs of this modern world.”

CEO Dev Ittycheria

Vendors like Snowflake are fantastic at analytical workload processing, but are less mature in the transactional space. Their move to Postgres is to get better in OTP, which is vital for catering to GenAI and agentic AI needs. It’s what provides needed access to real-time information to power inference capabilities at scale. Dev reminded us that Databricks and Snowflake (through Unistore) had been trying to build out these capabilities internally; he thinks their decisions to buy Postgres vendors serve as evidence of how unique and defensible MongoDB’s architecture truly is. It’s hard to match. The “bolted-on” capabilities these two vendors purchase, per Dev, mean document sizes beyond 2 kilobytes cannot be ingested without conceding “very poor performance.” MDB scales far beyond that because of its natively JSON-fueled NoSQL foundation. 

“Some vendors are retrofitting products with JSON or vector support as brittle afterthoughts. This is a tacit admission that our approach is the best way to model real-world data. While their features may check a box, they fall apart in production.”

CEO Dev Ittycheria

Dev then reminded shareholders of how rich MDB’s product suite is beyond data storage & querying. Companies need several point solutions beyond Postgres for things like keyword search, vector search, embedding models and more. They can have all of that in one place via MDB – with superior NoSQL functionality.

Finally, Dev also pushed back on the idea of Postgres taking market share from NoSQL use cases. In his mind, the rising popularity of Postgres is because Oracle and other SQL offerings are ceding rapid share to it. Not because NoSQL demand is shifting back to SQL. Dev thinks Postgres will win more SQL workloads and MongoDB will keep thriving in NoSQL areas – such as AI.

Fixes Working & Self-Serve Momentum:

MDB attributes a “solid new business quarter” to two main things. First, its product innovation is resonating. For example, MongoDB 8.0 is enjoying its “fastest uptake of any major release,” with adoption pace doubling the previous one. Voyage debuted two new retrieval models with better model accuracy and 80% lower data storage costs. From there, the combined entity launched its own Model Context Protocol (MCP) service with Cursor, GitHub, Anthropic and other needed 3rd-party connections. These connections are what enable developers to create agentic workflows that can more freely tap into the information they need across companies to augment task completion potential.

Second, its revamped go-to-market approach is working. MDB shifted resources to a batch of large enterprise accounts with great up-sell potential. Sales productivity for this newer team is “high” and it’s pleased with momentum. 

Part of this go-to-market shift entailed letting its mid-market business become more self-serve. Many mid-market customers prefer initially consuming MDB products in this way, so removing some direct sales resources from this channel (in favor of large enterprises) has not cost them much at all. For review, self-serve means MDB products are procured in a way allowing developers to use them with more autonomy and less MDB support. MDB bases its platform off of what’s best for developers, and in the age of AI, self-serve is routinely the answer. Self-serve means easier on-boarding and building on top of the MDB ecosystem with less red tape. Considering how rapidly AI is evolving, that’s attractive. MDB also offers a semi-managed self-service format, which gives support for the backend maintenance work, thus giving developers that same degree of building freedom without requiring them to do tedious work. That’s really working. Finally, tight integrations with all the other products MDB offers mean this self-serve style comes with seamless access to all tools a developer needs in one place.

Finally, it thinks it’s doing a better job of educating customers on the value of Atlas. Whether that’s more thoughtfully equipping its team with certifications or being more present in investor events across the globe, they’re actively seeking out new business rather than relying on in-bound interest. They’ve also added onboarding documentation in four new languages to support global growth.

AI Positioning:

MDB envisions the FY 2026 financial contribution from AI as modest. It does have some companies experimenting with GenAI apps built with its platform, but “enterprises are still early in overall AI adoption.” Hesitation comes from finite knowledge and skills, as well as lack of trust in models. Its fully managed app-building ecosystem (and broad integrations) + Voyage helps with those things. It’s adamant that when infrastructure focus shifts to building custom GenAI apps, its platform is poised to capture a significant chunk of that demand.  It knows that its ability to help modernize data and apps, paired with Atlas and products like MAAP will be a winning recipe in AI. We hear from countless firms that an AI strategy is not possible to implement without a strong, non-archaic data foundation. And that’s really where MDB thrives. 

Leadership:

MongoDB hired Mike Berry as their new CFO. Berry doesn’t plan to change much beyond the previous go-to-market fixes. His main focus will be on advancing operating efficiency and shifting some focus to margin expansion (hallelujah). He has been the CFO at McAfee and NetApp, along with 5 other scaled companies as part of his 30 years of experience.

Much better quarter than last time around. The company’s new CFO seems determined to prioritize margin optimization more than his predecessor. That’s already showing up in results without sacrificing any kind of revenue generation. The real AI demand contribution hasn’t even begun yet for MDB, so this demand resilience hints at the cost cuts eliminating redundant, unneeded pieces of the business… rather than needed assets. This company got bloated and is now fixing things. That leads me to wonder “what was the old CFO doing,” but the more important thing is that Berry is focused on the right things. I think Atlas is an excellent business and Voyage + app migration and building tools give them more of an overarching platform to sell. That should allow them to more effectively bundle products together, juice cross-selling opportunities and augment customer retention.

At 72x forward EPS and 99x forward FCF, this name is very expensive. That’s especially true considering EPS will fall this year, before compounding at a roughly 25% clip over the next two years (for a 2.9x growth multiple even when ignoring this year). Its FCF growth multiple is also right around 2x. While I think this is a high-quality company that just delivered their best quarter in nearly a year, I find other deals in software to be more compelling.

Demand:

GitLab beat revenue estimates by 0.6% & beat guidance by 0.9%. Its 30% 2-year revenue CAGR compares to 31.2% Q/Q & 31.7% 2 quarters ago.

Profits & Margins:

Balance Sheet:

Guidance & Valuation:

GitLab trades for 56x forward EPS. EPS is expected to grow by 2% this year, following 270% EPS growth last year. EPS is then supposed to grow by 23% the following year and by 35% the year after that.

Demand:

Profits & Margins:

Balance Sheet:

Annual Guidance & Valuation:

Adobe trades for 19x forward EPS. EPS is expected to grow by 12% in each of the next two years.

Macro & Quarter-to-date:

Lapointe reiterated what we heard from the team last week. Its ultra-prime niche continues to deliver robust debit interchange growth and consistently strong credit quality. They’ve “never felt better about the business,” as they “continue to see really good momentum in Q2.”

Loan Platform Business (LPB):

SoFi has spoken a lot about expanding LPB originations to customers outside of its credit parameters. It’s currently servicing about $20B in loan demand and rejecting $80B. There’s significant capital market demand for part of that $80B, and it thinks it can safely capture it. That would also mean saying yes to far more customers, while it maintains that relationship and enjoys massive cross-selling potential to other products. Doing this through LPB means adding that incremental demand without adding balance sheet risk, although I will say lower credit quality borrowers always come with more capital market demand cyclicality. This would make the peaks and valleys higher and lower in my mind, but that still makes sense. I’d rather have $10 of profit during the peak and $4 of profit during the trough than $8 and $5. Still, something to keep in mind.

The new part of the LPB discussion centered on expanding from unsecured personal loans to including its home and student loan businesses in the product offering. That would greatly expand the market size while improving the overall level of credit quality for its average LPB paper. I candidly like this idea better than expanding to subprime. I’d like to see this happen first.

Credit Demand:

Loan demand on borrower and capital markets sides, for both its held paper and LPB, has never been better. And for some new evidence, beyond a banner Q1 for LPB and its personal loan sales, it just closed a 2nd $700M securitization deal for LPB loans. The first of these since 2021 came last quarter and now it has closed another. Both deals came with low credit spreads and strong SoFi yields, which should support even more LPB partner demand. If these investors see easy, liquid access to secondary markets, the risk of being stuck with unwanted credit diminishes and the eagerness to add more of these loans to their balance sheet grows.

Tech Platform:

We got more reiterations of previously bullish commentary. Candidly, I’ll be happy to hear these reiterations as many times as they want to give them. The segment has enjoyed a “really good uptick in demand” for a few quarters. The federal, Wyndham and Mercantil Banco deals will all ramp into 2026, while they “expect to announce a few more deals” this year. All in all, they have a swelling pipeline, boomeranging clients (churned customers that came back) and 10 deals that will start contributing to growth in 2026. The stars are aligning (finally).

Ingredients for the Turnaround:

I think PayPal’s work to improve archaic products and add new tools like FX as a Service and Fastlane are largely understood at this point. During this interview, Kereere got into a few other reasons for PayPal’s stabilized and improving results, along with the robust multi-year guide. Talent has been a key enabler of this better pace of product improvement. They didn’t really have a scaled base of quality people in AI, data science or other key categories. This meant they didn’t have the capabilities to take advantage of their vast data advantage. As a customer came into their ecosystem or a merchant’s site, they had the assets to know what products they probably wanted, what payment option they liked best and what level of promotion would get them to convert. It just wasn’t using any of this information to power an unmatched level experience personalization. Now it is.

Next, they also didn’t have appropriate levels of local talent in their key growth markets. Through office openings across the Middle East and more hiring of folks across major countries who “understand how business is done in those markets,” it has addressed that issue as well.

And finally, PYPL started… well… actually listening to its customers. Even the innovation PayPal was delivering wasn’t focused in the areas its customers wanted. You read that correctly. Their product suite was disjointed, confusing and sold separately. Their newness was irrelevant. There was very little correlation between product requests and product delivery and very little organization of the suite of products it did have. They’ve overhauled go-to-market into more clearly segmented, yet interoperable teams. They’ve also begun to model their product roadmap after what they’re asked for. What a concept.

Braintree:

Kereere told us that “Braintree has never been stronger.” That’s true across the USA, Latin America and newer markets like Australia. They’re on track to deliver a revenue reacceleration towards the end of this year. And? That reacceleration could come with even better margins than expected. Conversations and contractual changes with merchants to finally “price to value” are going better than planned. Companies are embracing the more holistic value proposition Braintree and its services together provide. That’s yielding pricing power and surprisingly modest losses in volume per merchant as it normalizes fees. Braintree can easily compound revenue at a 10%+ clip for a long time… it can positively benefit overall revenue and translation margin dollar growth. I expect that to happen once this strategic pivot concludes by the end of this year.

Selfbook Partnership for Agentic Commerce:

PayPal announced a new partnership with Selfbook. They provide “AI agents for travel experience search and discovery.” The more partnerships, the better. They all allow PayPal to tap into more data sources to enable its own AI agents to collect more context and provide better output tokens. They also allow PayPal to actionably insert itself into search tools like Perplexity to turn consumer interest into targeted, personalized promotions – guided by PayPal’s unmatched consumer dataset. Now, a consumer won’t just get a chatbot answer. They’ll get a chatbot answer tailored to their own interests, with actionable ways to purchase a good or service in the most affordable way possible. Pretty powerful. In turn, that boosts shopper conversation rates and is something PayPal can easily monetize. Many more partnerships to come.

Oracle provides a slew of software and hardware tools for on-premise and cloud environments… with an understandable focus on shifting towards cloud deployments. It has 3 main segments that tie very closely together. 

Oracle Cloud Infrastructure (OCI) is its fully managed business for infrastructure services (virtual machines, storage, managed high-performance compute data centers etc.). This segment also includes platform services to build apps in its safe, controlled environment (serverless and container-based).

Strategic software as a service (SaaS) includes Oracle NetSuite. This is a set of applications for enterprise resource planning (ERP), customer relationship management (CRM), human capital management (HCM), e-commerce and more. It’s hard at work on launching more industry-specific software apps across areas like healthcare. It has a more customizable, feature-rich version of this product suite called Oracle Fusion geared towards larger customers.

The last segment is its slew of database products. Creating valuable apps from GenAI infrastructure requires great models and rich, organized information access to properly season those models. That’s where its database capabilities come into play. It provides an SQL tool, but as we saw in the MDB piece, that’s somewhat dated at this point. Its more interesting product here is the not only structured query language (NoSQL) database for unstructured data,

Oracle closely integrates with the 3 big hyperscalers, allowing its database offerings to run anywhere. This also means that customers can migrate their on-premise databases to the cloud via OCI or through any of these hyperscalers, diminishing the friction associated with using it. Oracle believes that this data cloud interoperability provides innate data transferring cost advantages. Cost is estimated to be “several times cheaper” for model training than any competitive product, according to leadership. Oracle has re-emerged as a digital infrastructure titan. While the company did take longer to roll out its high-performance compute product suite, it has since achieved fantastic traction.

Oracle raised FY 2026 revenue guidance from $66B to more than $67B. This represents 16% Y/Y FXN growth and beat estimates by at least 1%. Cloud revenue growth is expected to accelerate from 24% Y/Y to 40% Y/Y. Within that, cloud infrastructure growth is expected to accelerate from 50% in FY 2025 to 70% in FY 2026. RPO growth is set to exceed 100% Y/Y in 2026, which is also giving them fantastic visibility to raise their previous 20% Y/Y FY 2027 growth guidance. They are increasingly confident in meeting or likely beating their FY 2029 targets as well. Pretty good!

Notably, Stargate (project with Oracle) is part of this guidance. Still, it sounded like it was a modest piece of the impressive forecasts. Furthermore, per Ellison, if Stargate scales as expected, their RPO guidance is “understated.”

For next quarter, $1.48 EPS guidance met estimates. Oracle trades for 30x forward EPS. EPS is expected to grow by 12% this year and by 22% the following year.

CapEx & Free Cash Flow:

Hefty CapEx to fund its aggressive infrastructure build outs led to negative FCF for the full year. Operating cash flow was $20.8B, but CapEx was $21.2B. That is assumed to be over $25B next fiscal year, which will continue to weigh on cash flow margins. Depreciation costs will also ramp from this, but they still see more operating leverage ahead. At the same time, these investments are connected to near-term revenue opportunities. They simply don’t have the capacity to service all of the wonderful demand they’re enjoying. And when customers are quite literally telling them “we will buy all of the capacity in any region you have available” that’s understandable. This has never happened to them before, and just goes to show how fantastic demand currently is. The correct decision is to spend and so they are.

“CapEx is going to go up because the demand right now seems almost insatiable.”

Founder/Chairman Larry Ellison

OCI:

OCI is absolutely killing it. Oracle Cloud@Customer data centers, which entail adding OCI to a customer’s own private data centers, rose by 104% Y/Y. It has plans to double that footprint again in FY 2026, as signals remain exceedingly strong and as demand comfortably outpaces supply. Importantly, this is for non-GPU server components and power. They are having no trouble getting needed GPUs at this point. 

“We actually currently are still waving off customers or scheduling them out into the future so that we have enough supply to meet demand. This is a situation we have not seen in our history.”

CEO Safra Catz

Total OCI consumption rose by 62% Y/Y in Q4 and they think that’s going to keep accelerating in FY 2026 to 70% Y/Y. Aforementioned thriving RPO growth, which is made up of non-cancelable bookings, provides a high degree of visibility to inform this optimism.

The same things that have resonated to date continue to do so for OCI. Its uniform data center layout drives simplicity and a single stock of inventory to easily add new server racks to. Customers can easily buy incremental compute at the exact size they need – no minimums. It pairs this with what it views as superior levels of automation that drive cost efficiencies and lower human error rates. From there, best-in-class server density allows it to pack in more compute per data center, while Ellison is adamant that Oracle’s ability to move data in and out of GPU clusters is more efficient than with anyone else. All of this contributes to lower total cost of ownership and these excellent results. The recipe allows Oracle to profitably service smaller batches of demand that AWS and Azure simply won’t touch. Oracle will and it will also happily deploy OCI within those competitor environments via what it calls multi-cloud data centers. It plans to triple its capacity here next year and is confident that its ability to provide all Oracle services and apps in every single cloud (not just subsections of its offering) will help it keep standing out.

Database Services:

“One of the reasons OCI is more secure and faster, is because we have an autonomous database that runs it.”

Founder/Chairman Larry Ellison

Its database products are winning. And they’re all perfect complements to OCI. Combining data and infrastructure vendors routinely lowers transfer and processing costs to augment the advantages already laid out in the OCI section. The flexibility of data product usage is also an edge. Oracle offers these products through OCI, every large hyperscaler, multi-cloud environments and deployed within a client’s private data centers as well. Interoperability also… you guessed it… drives incremental cost efficiencies. Noticing a theme? Ellison sees Oracle’s newest database (Oracle 23AI or its “AI Data Platform”) as the best at conjoining public data and customer data… from a wide array of sources… to make it seamlessly available for large language models (LLMs). The new offering handles theme-based or vector search and retrieval-augmented generation (RAG) to make sure search results can be added to large language models (LLMs). The introduction also entailed a large batch of new partner data integrations, like with Nvidia and IBM. Lastly, as Ellison puts it, companies want to use all of their data without making it available to large language models (LLMs) for public consumption. Oracle unlocks full access to their data, as well as secure integrations with all the models and apps they want to work with in one place.

Now for some more numbers. Multi-cloud database revenue rose 115% Y/Y to mark continued acceleration vs. 92% growth last quarter.  The tripling in mutli-cloud data center regions is a great sign as well. That should help this bucket maintain 100% Y/Y growth and should keep pushing 47% Y/Y growth in autonomous database consumption higher next year. 

“Other companies say they have all the data. So they can do AI really well. They can build all these AI agents on top of all of that data. The only problem with that statement is they don't have all the data we do. We have most of the world's valuable data. The vast majority of it is in an Oracle database.”

Founder/Chairman Larry Ellison

As an aside, for some context on how large the cloud database opportunity is, there’s a 5x revenue opportunity from migrating on-premise support revenue to the cloud. There’s far more component-and-networking-level revenue to enjoy and that’s where the market is heading.

Software Applications:

Oracle is aggressively infusing AI agents, or reasoning capabilities, into its core applications. This is already leading to higher renewal rates and strong incremental bookings and should lead to this segment also enjoying a growth acceleration next year. The SaaS segment overall delivered 11% Y/Y growth, while back-office applications rose 20% Y/Y to reach an annualized run rate of $9.3B. The team continues to expect this bucket to support demand for OCI and its database products. This goes back to the frequent theme of vendor consolidation driving lower costs, easier access to needed information and better outcomes.

Excellent quarter. While Oracle’s start in AI infrastructure was a tad slower than others, it has more than made up for lost time. Its efficiently scalable data center architecture is allowing it to help where others aren’t willing, while its apps and database offerings are perfect complements for that niche. The company is thriving in every sense and next year should look even better. I think leadership deserves a ton of credit here. This has morphed from somewhat of a dinosaur, to enjoying a large chunk of multiple secular growth stories for themselves. This is one of my favorite companies not currently in the portfolio and does feel like one that got away. Congratulations to shareholders. You’ve been so right.

On Co-Founder/CEO Marcos Galperin Retiring & Stepping Into Chairman Role:

Cuccioli sees very little strategic change stemming from the leadership transition. Commerce President Ariel Szarfsztejn has been attached at the hip to Galperin for years and has run the commerce business (the largest piece of MELI) for a long time. When Galperin passes the torch, Szarfsztejn will be determined to make things business as usual. Considering how historically successful Galperin has been with leading MELI, that’s good news.

Resource Allocation:

Commerce and fintech remain the focus areas for MELI – and rightfully so. Its 3 core markets are about a decade behind the USA in e-commerce penetration and places like Mexico still have 40% of their population without a bank account. It’s tempting to shift a lot of the focus to other markets like Paraguay and Colombia, but return on investment is simply too compelling throughout Brazil, Mexico and Argentina for that to make any sense. They have a proven playbook for growing their core business segments and doing so in established countries is both lower risk and higher reward. No-brainer.

Their 30%+ e-commerce shipment CAGR requires hefty investments in logistics footprint to keep scaling its high-quality service with demand. Its planned credit card launch in Argentina will consume considerable resources as front-loaded provisions come before cohorts mature. This is where dollars should be allocated. Focus on the core and the core will keep treating you well.

And a big part of consistent prioritization will remain on creating a best-in-class customer experience in its main markets. That’s why it again just slashed free shipping threshold in Brazil to a little over $3 per order. It has unique attributes like unmatched cross-selling, superior marketplace economies of scale and a better, more localized logistics footprint. That all inherently drives things like better buying frequency, lower marketing spend per product sold and the ability to collect profit dollars from low-margin business more reliably than the other guys. So? It has earned the ability to pass on all of these unique efficiencies via more value for its customers. Whether that’s more deposit yield, lower loan interest rates, more everyday essentials (which it can rationally and profitably provide), more affordable content access, or this news, it’s a winning formula and a formula it will keep leveraging. That’s how it will stay ahead of Amazon. That’s how it will stay ahead of new entrants from Asia.

Every single time MELI has boosted free shipping offers in Brazil, order frequency and engagement gains have made up for that decision on the bottom line. That has happened while MELI has extended its customer service lead vs. everyone else. Power of the platform; power of scale; power of Mercado Libre.

Grocery Business:

MELI continues to gain confidence in the grocery business being a scalable profit driver going forward. Notably, while virtually all of its non-fresh goods are first-party-sourced, its fresh business will continue to grow through partnerships. It does not have interest in building out those capabilities internally.

Advertising:

Off-platform is the future of MELI’s advertising business. It is now a powerful player in attracting sponsored listing and performance marketing dollars through its demand side platform and marketplace. It has built a strong business by connecting ad dollars to direct sales. Going forward, a lot more focus will be placed on display and video advertising. A partnership with Disney+ and its growing Mercado Play’s free streaming offering are two early pieces of this expansion project – and there will be many more.

Latin America’s ad penetration as a percent of GDP greatly lags North America and Western Europe. That market also still has 50% of its ad spend outside of digital channels vs. just 25% in the others mentioned. There’s a massive opportunity here, and MELI expects to build the largest media business in Latin America as it takes advantage. They’ll be very careful to avoid growing ad load to a point of diluting the customer experience. Still, there’s a boatload of work to do to better monetize its existing traffic without making that concession.

Argentina:

MELI is excited by the prospects of launching a credit card in Argentina later in the year. Financial service penetration rates in that nation resemble Haiti, per Cuccioli. Relatively high GDP per capita paired with lack of penetration is a great combination for growth. That’s especially true as macro in that country keeps brightening and inflation keeps slowing. This slowing inflation is also allowing them to invest more in fulfillment capacity, as seller costs become less prohibitive. Raising its self-fulfilled penetration rate from 22% to closer to Brazil and Mexico should have the same profoundly positive impact on customer service and growth.

The Next Large Product Expansion Category?

Healthcare will be the next product frontier for MELI. They’re testing some things in Brazil to try to create an offering with a sky-high net promoter score (NPS). From there, they’ll expand to the rest of Brazil, Mexico and probably Argentina. Primary care telehealth and drug discounts were cited as the two largest opportunities within this sector.

Overarching Announcements:

Apple told us that intelligent Siri still needs more work and will likely come out later this year. For Apple Intelligence, they added several more languages to tableset more global expansion. Considering iPhone sales in markets with this available lead markets without it, that should be positive for growth.

Next, they’re opening the on-device model that powers Apple Intelligence to 3rd-party developers. If Apple is perhaps finding trouble driving impactful AI software innovation, why not let its massive base of developers do the work for them? Meta certainly is embracing that philosophy. The company will still be handsomely compensated if any of these releases lead to app store revenue, so the financial risk of not directly controlling this innovation is low. This is a good decision, in my opinion. They teased demos such as Kahoot using the on-device model to turn pictures of class notes into quizzes.

Apple revealed liquid glass as its first new design in several years. This will be implemented across all of its devices, with a more aesthetically pleasing display that reacts to movement, touch and light. Rectangular displays have all been scrapped to more closely fit the shape of our screens, while translucence enables app and background colors to change as we switch apps.

Finally, operating systems will now be named after the year they come out. The new iPhone Operating System (iOS) is called iOS26, for example.

iOS 2026:

Apple showed off some interface tweaks like more pleasing swiping and phone unlocking, as well as dynamic wallpaper and 3D album art in Apple Music. For the camera, it streamlined the interface to only show most commonly used options as the default. This is cleaner and allows users wanting to explore other choices to seamlessly summon them on command.

CarPlay added the ability to field calls without directions going away and added “CarPlay Ultra” to extend the software to all vehicle screens. This also comes with new CarPlay apps like changing the temperature or radio station all from the same ecosystem.

For the phone app, like the camera, it updated the default display to show only the things we use most – call favorites, recents and voicemails. The keypad and contacts list can be easily accessed when prompted. They also added call screening for telemarketers. This answers calls for us and prompts salespeople to offer their name and reason for them bothering us. It will then ring our phone, show us who these people are and allow us to more confidently ignore calls. Good release here. For another materially positive launch, it added hold assist. This automatically detects hold music, asks if we want to wait, keeps us connected and rings our phone when the person on the other line is available. It tells that person that we’ll be on the call shortly to avoid them hanging up on us. That’s awesome.

For the messages app, it added another screener tool to sift spam out of our inbox. Those annoying messages will now go to a separate area where we can check all of them and choose what to allow into our main message feed. Group chats added typing indicators and Apple Cash settling right from within the app. Consumers can also now combine two emojis and prompt ChatGPT to generate entirely new ones. That tool is available to developers.

Like OpenAI, Google and Meta, it launched live translation. This offers real-time captions via Facetime, automated text translations and spoken translation via call. I didn’t find this as impressive as Alphabet’s demo, but this was a needed launch and it will only get better from here.

Apple Wallet added more auto partners for locking and starting a car from within that app. It will also launch digital IDs this fall. These won’t replace the need for passports when traveling abroad, but will work for domestic travel or for someone verifying their age at a bar, for example.

WatchOS:

The workout buddy product was the most interesting part of this section. This uses all of our workout data and fitness history to personalize insights. It takes the personality of a personal trainer to encourage us on our long runs and praise us for sticking to our health regimens. New tools also allow for custom workouts and the ability to race against our personal records. I’m going to love that. Finally, WatchOS added new wrist flick gestures to hide notifications, mute calls and more.

MacOS:

Upgraded Spotlight was an interesting launch here. This creates way more opportunity to tap into data and conduct multi-app tasks right from that bar. It’s no longer just a search bar, but an extremely actionable, interactive app usage bar. You can create events, start recordings, tap into editor apps from a work document and use quick keys to access all of these things even more easily. Apple added new 3rd-party integrations with Google and Dropbox to extend what this product can do.

VisionOS:

iPadOS:

Apple added flexible app sizing on iPad like we enjoy on Mac. This way, people can have more applications open at the same time.

Thoughts:

A lot of these announcements were pretty cool. They will subtly bolster the value proposition for Apple and create delightful experiences. At the same time, the emphasis is on the word “subtle.” None of this looks all that new besides a liquid glass display that I don’t really see as needle-moving. Apple needs to re-ignite the innovation engine. Whether that’s through purchasing leaders in the smartglasses field or something else, they can’t keep rolling out nearly identical products in perpetuity with some new bells and whistles. They can’t be complacent forever, regardless of how powerful their ecosystem moat is. And it is powerful. 

That’s why none of this bearishness will matter in the coming years. Their success has meant a giant cash pile is available to manufacture profit growth. North America (and many other parts of the globe) still lives on their devices – myself included. All of my next devices will be from Apple. My social circle would all say the same thing. And while I think that dynamic probably feels safe and comfortable to Apple’s leadership team, I don’t think it will last forever. Meta is ahead of them on next-gen wearables and is moving faster. See RayBan and Orion for evidence. Tim Cook needs to shake things up a little bit to ensure their last 20 years of domination leads to 20 more years of domination. Again… not relevant in the near term. But will be relevant eventually. And at 30x forward EPS and a roughly 7% EPS CAGR, they’re priced as if people assume they'll figure out the next wave of innovation. TBD.

Shopify added USDC support, through Base (owned by Coinbase), to its Shopify Payments and Shop Pay checkout. More payment flexibility naturally bolsters conversion rates, and stablecoins have real promise in areas like payment speed, lack of cross-border fees etc.

Anything Shopify has done in the past to create more choice has been uniformly positive for this business. I expect the same to be true here. This sounds like phase one of the partnership, with phase two entailing cash-back rewards for using this form of payment. That’s very easy to do, considering the more favorable fee structure of a stablecoin vs. a credit card.

Meta is purchasing a 49% stake in Scale AI for $14.3B. As part of the deal, founder/CEO Alexandr Wang will join Meta to lead its new Superintelligence lab. Love that. ScaleAI is a highly-regarded data labeling, curating and sorting machine. Essentially, they prepare data and optimize what is actually used in models to make sure their pace of LLM improvement is as affordable and effective as possible. 

The company can ingest massive amounts of unstructured data, understand what can actually help a model improve and remove the rest. ScaleAI leadership calls this fixing the “garbage in and garbage out” problem. All of this helps mightily with hallucination rates and model KPIs by giving ScaleAI a holistic view of model hygiene evaluation. The firm also greatly alleviates data ingestion scale bottlenecks by augmenting the amount of high-fidelity context that can be fed into LLMs. It boasts a large roster of 3rd-party data integrations to ensure these LLMs can access whatever they need. In the world of agentic AI, where algorithms race around the internet and pull info from dozens of sources, that interoperability is needed. Their business deploys a hybrid AI and human approach (called “human-in-the-loop”) to pair best-in-class levels of automation with elite research teams to help with error rates and fringe cases.

It’s easy to see how this is an extremely attractive company for Meta and its Llama models. 

Wang is also considered a world-class talent in this field and a visionary. He was a math and programming prodigy and was working as a capable engineer in Silicon Valley by age 17. Like Zuckerberg, he dropped out of school to create ScaleAI and is now the youngest self-made billionaire on the planet. The two have a lot in common, including now a vested stake in seeing Meta succeed. Wang’s company has bellwether clients such as the U.S. Department of Defense, Anthropic, Cohere, Microsoft, Nvidia (an investor), Amazon (another investor), Toyota, Uber, GM, Samsung, Accenture, Databricks and OpenAI. Nice list… I guess. And interestingly, Wang was Sam Altman’s roommate for a while during the pandemic.

I think news on Meta delaying its Llama 4 Behemoth model has been overblown. I also think model providers will constantly leapfrog each other over the coming years as we likely consolidate down to a few major players. I already thought Meta’s chance of being one of those players was quite strong, as Llama 3 was not a fluke. These chances just got a lot stronger with this best-in-class company.

Starbucks will cut prices by $0.70 for some iced drinks in China to become more competitive in that market. It is struggling there more than anywhere else as the landscape is bitterly difficult. And with that said… Starbucks is supposedly enjoying significant interest in its China business. While a potential deal is still in its early stages, I’d love to see them sell at least a minority stake. As I’ve said many times, the Chinese government does not want an American company to own 100% of the economic value they’re extracting from that economy. Starbucks does. By shedding majority ownership, I think they take the target off their backs, enjoy a large cash infusion to invest in their other markets (which are all working better than China) and can still benefit from the new investors maybe turning this around. I’ll keep saying it. That’s the correct decision. Make it happen.

Next, Starbucks will add one dedicated assistant store manager to most U.S. stores. They’re currently in 20% of them, and it thinks many more are needed. This should lead to better, more organized customer service and alleviation of crippling throughput issues. It will free managers to focus on cross-store operations and supervision, rather than tedious day-to-day work. The change also gives its employees a more desirable path for internal career mobility, which supports its goal of 90% of retail leadership roles being internal promotions. This will test in 6 markets this year and scale in 2026. I expect this to be successful, considering these people are already in ⅕ of their stores. Clearly, they’re leading to positive outcomes. Along similar lines, it hosted 14,000 coffeehouse leaders in Las Vegas last week to rally them behind Back to Starbucks initiatives. This also featured its first-ever Starbucks Global Barista Championship, which not only sounds fun, but should also make it crystal clear what a successful team member looks like at Starbucks.

“This isn’t just a reset—it’s a recommitment to who we are when we are at our best. This event is our moment to recommit to a culture of hospitality and excellence. We’re making progress, have real momentum with our 'Back to Starbucks' plan and are on the right track to turn the business around.”

COO Mike Grams

The company is testing a protein vanilla latte with banana cold foam and no sugar as it modernizes its menu to cater to health and wellness trends. Unlike the old team, which guessed and hoped at what would work, Niccol’s Starbucks will obsessively test new launches until they’re certain they will work.

In other Meta AI news, it’s trying to poach talent from other tech giants to build a new general artificial intelligence team.

More charts showing Google Search losing share to ChatGPT surfaced this week. I got a lot of questions so I wanted to discuss them. I expect Alphabet to keep losing market share in search. And? I also expect the search business to keep compounding at a steady clip near 10% for a long time. These two views are compatible in a growing market. The market share charts entirely ignore two things: First, Gemini is taking some of this share from Alphabet. It has 400M+ monthly active users (MAUs) and is quickly establishing itself as one of the largest chatbots on the market. That makes sense, considering this firm’s world-class models and its unmatched access to multi-modal data. Market share going from one part of a company to another part is not concerning.

At the same time, this chart also showed ChatGPT rising from 1% to 6% market share over the last year. And this leads me to my second point. GenAI and agentic AI are vastly expanding the overall search pie. The use cases are bountifully growing and the efficiency at which the products connect users to answers is rapidly leveling-up. I do not care if ChatGPT is taking market share from a pie that is greatly growing. And that’s what is happening. This is why I do not think share losses will materially harm the search growth engine in the near future. I still think the mega-cap has a great chance to win back some of that share as folks realize their offering is better than the others, but that isn’t needed for my thesis to come true.

Now onto some other data from Morgan Stanley. Data from that firm revealed that 16-24 year-olds are using the Search Giant for product research, price comparison, and product search (already know what they want) more frequently than last year. Research share rose from 51% to 56%; price comparison share rose from 36% to 40%; product search share rose from 20% to 30%. It’s vital to note that this includes Google, Gemini and YouTube. Still, YouTube’s durable popularity isn’t nearly enough to explain these large spikes. Core search resilience and Gemini proliferation is. This company stinks at “dying.” Spencer Walsh on X shared this chart and I wanted to thank him for it.

OpenAI signed a new cloud deal with this search titan. Interesting to see two competitors partnering up on infrastructure. There were rumors of OpenAI looking to diversify away from Azure, but I assumed they’d pick AWS – who they don’t really compete with that much.

Nvidia will fully exclude all potential China revenue from its forward guidance starting next quarter. If restrictions are lifted in the slightest, that would likely mean considerable upside. Nvidia is also collaborating with Novo Nordisk to accelerate drug discovery, partnered with Deutsche Telekom for German Sovereign AI and debuted new self-driving models. There were several press releases on bringing more of its hardware to Europe throughout the week.

AMD hosted an event in which it featured Sam Altman and leaders at Amazon, Meta and others. It was an impressive lineup of heavy hitters. They’re all using AMD’s new GPUs for some things as the company (some think) looks to close the performance gap vs. Nvidia Blackwell this year. It will be interesting to see if Nvidia’s next platform (Vera-Rubin) will be something they can also match. If so, the opportunity is large.

Amazon and Walmart are toying with the idea of adding stablecoins as checkout options. I think this is pretty inevitable and is just something payment processors will need to adapt to if they’d like to avoid being left behind. That’s why PayPal, Visa, Mastercard and others are so focused on this area of the payments markets.

Uber and Wayve are extending their partnership to deploy level 4 autonomous vehicle rides in London for testing. Uber also added Dick’s Sporting Goods as a new delivery partner.

Tesla delayed the rollout of its Austin robotaxi rides to June 22nd. They’re still currently testing. I hear anecdotes about how well things are going and how there’s no supervision or geofencing. And then… I hear anecdotes about the cars being remotely operated with supervisors close behind and more testing needed. I’ve seen footage of them working. I’ve seen footage of them drilling a test dummy on staged routes. There is so much noise surrounding this launch (especially on X) and it’s hard to know what’s actually true and what isn’t. Time will tell.

Eli Lilly will only partner with firms not selling personalized GLP-1 copycats. That probably makes a Hims arrangement pretty unlikely at this point. GLP-1 will continue to feature a rapid pace of changing headlines.

DraftKings added a $0.50 transaction fee for sports bets in Illinois to recoup some of the recent tax hike. This news passed with far less pushback than their surcharge idea (despite it being a nearly identical concept). I think that’s great news. It will embolden other vendors there to follow suit and should mean DKNG claws back all (or more than) the amount of the hike. For context, $0.50 is worth an incremental 1% in hold rate vs. their average bet size. This will inevitably push some to black market operators, which means modestly lower volume for DraftKings and everyone else. It also means zero tax revenue for Illinois for those gamblers. Why they didn’t contemplate this before passing another tax hike is something for which I don’t have a good answer. I quantified the impact of the change last week (section 6).

Inflation Data:

Consumer Data: