Some personal news! I’m getting married today. Thinking she’s a forever hold. Hopefully she feels the same way. Work is a bit too busy for me to be going on a honeymoon, so I’ll be back and ready to go Monday Morning. Have a great weekend!

CEO CJ Desai’s Perceived MongoDB Positioning One Year Into His Tenure:

MongoDB’s platform continues to resonate in the AI era. Its document-oriented architecture, run anywhere approach and online transaction processing (OLTP) bread-and-butter is well suited for harnessing vast sums of unstructured data. That’s obviously imperative in this world where models and agents are built with this data providing essential context and structure to create actual value. You can’t vibe code this information. You can’t create synthetic simulations without having the real thing in the first place. You cannot emulate MongoDB’s scale at the cost and performance it provides on your own. This is one of the most well insulated companies against any AI threat, with a platform that disruptors rely on to do their disrupting. That’s why they’re having so much success selling data, model optimization and developer tools to AI natives. It’s also why they’re now beginning to build traction with frontier labs just like Snowflake and Databricks have done with their competing businesses. MongoDB leadership, as they often do, cited Snowflake and Databricks pushing further into OLTP use cases as evidence of this being a great place to be.

Sources of ongoing Leverage:

R&D is going to keep growing at a rate slower than revenue, offering MongoDB clear margin expansion visibility over the coming years. While its margins can spike higher in a given quarter if consumption was especially robust, R&D and other sources of OpEx based improvement, there’s plenty of core OpEx-based leverage to enjoy.

Conversely, they’ll keep hiring aggressively for sales and marketing, so that won’t be a big source of EBIT margin expansion at least for now. They are notching some material savings from layering more automation and AI onto their own platform, but they’re reinvesting most of this into more growth rather than letting the incremental dollars immediately flow down the income statement.

Go-to-Market:

Under Desai, MongoDB has moved away from a sales philosophy that strictly prioritized direct pursuit over distribution partnerships. That is changing. Much like we saw that strategic reconsideration works wonders for Cloudflare and CrowdStrike, this is now contributing to the stronger growth patterns seen below. It’s not just AI. It’s going after this demand directly while also working more closely with the other enterprise software platforms its customers want to work with (diminishes friction). Companies like ServiceNow and Cloudflare are actively integrating and working with MongoDB to serve joint customers. While this doesn’t perfectly emulate the stickiness of just doing everything by itself, that’s not realistic for MongoDB (or anyone else). And integrating with complementary offerings is a great way to plug the gap in a lower risk way.

They’re also doing a better job of communicating value propositions directly to C-suites. While this company has been around for a long time, there are still massive enterprises that don’t really know much about them.  According to Desai, merely telling big prospective customers that they even have vector search or embedding models to improve output quality is often a pleasant surprise. Sometimes it really is as simple as being more vocal about the value that your products already provide.

Lastly, while they’ve made a lot of progress, MongoDB is still working on fixing some underperforming pieces of their overall sales engine. That’s why they brought in new sales leaders with tight focuses on the U.S. Federal segment and on better Asia Pacific execution. This included some restructuring to cut underperforming salespeople. MDB thinks this past quarter’s strength partially represents  early returns from these changes.

Atlas:

While the overall business has notched a material acceleration (seen above) over the last few quarters, that is largely the result of its on-premise business. Atlas (cloud-based delivery) hasn’t accelerated at all while competitors like Snowflake and Databricks have meaningfully done so. Why is this happening? 

MongoDB thinks the big growth catalysts haven’t manifested yet but soon will. First, auto embeddings will automate data conversion to formats for AI search to unify its embedding models (make outputs more accurate with better data availability) and vector search into one powerful offering. They also just launched a fully managed model context protocol (MCP), which standardizes agent and data connection and has already delivered a noticeable uplift in Atlas revenue since launching a month ago. You could argue that Snowflake and Databricks did a better job of innovating more quickly and positioning their suites for faster success. I would agree with that opinion, but MDB is fully capable of catching up and I think they’re going to do that.

I also think there’s another factor at play:  MongoDB has a more meaningful skew towards on-premise deployments than the other two, and on-premise remains highly important. As we often cover, voracious data consumption and massive transfer volumes lead to leakage risks. On-premise deployments eliminate much of that concern. It may be more appealing to the more highly regulated industries that favor more stringent asset guarding over more flexibility or potentially lower cost. While that does take some overall demand away from Atlas, average Atlas consumption per customer using its on-premise solutions is much higher than the overall base.

Mythos Moment:

As a reminder, the “Mythos Moment” is when Anthropic held back a model debut because they feared it would be used by adversaries in ways that current security systems couldn’t defend against. Whether that was a marketing ploy to promote tighter open-source regulation or a legitimate concern, it worked. It raised the sense of urgency for customers to have best-in-class security tools to protect against rising AI threats.

And while that shift in mindset undeniably occurred, the immediate uplift to security business growth wasn’t all that noticeable across the board. Palo Alto said little other than “we’re having more customer conversations,” without much material financial acceleration outside of a big inorganic CyberArk boost. SentinelOne and Zscaler didn’t show much of a pickup. CrowdStrike did. They reported a sharp acceleration in net new ARR growth to above 50% Y/Y and that is because they’re taking advantage of heightened AI security demand better than others. They view this evolution in AI security focus as durable and think they’ll continue to be a primary beneficiary. Agreed.

Security Information & Event Management (SIEM) – An Underappreciated Edge:

CrowdStrike doesn’t charge customers for first-party data ingestion. Before they offered a SIEM, customers complained about having to move data off Falcon to another platform that charged them for storage. Now, CrowdStrike offers this capability without charge and from the same interface that companies use for most of their security work. They don’t need to charge outside of third-party data ingestion. They can instead use this as a way to provide highly differentiated value against point solutions thanks to the added product utility, cost cutting and cross-selling capabilities.

Guardian & Sizing the Guardian (AI Detection & Response) Opportunity:

As laid out in the Fal.Con 2026 review, Guardian secures all AI-related assets, permissions and relationships with scaled detection & response capabilities that move at the speed of AI. It secures agents across the endpoint, SaaS applications, cloud workloads and the browser. It offers complete AI agent inventory (including Shadow AI). With Guardian, customers know what agents are doing and can seamlessly trace actions to understand issues faster and fix them. Interestingly, this also includes an AI Gateway (coming soon), which processes agent traffic and directly competes with Zscaler in doing so. The two are close partners and just recently deepened their relationship around AI Governance and adding new product integrations. Zscaler is part of CrowdStrike’s Project QuiltWorks too. At the same time, this application will be a crucial part of the overall agent security opportunity. CrowdStrike is pursuing it alone, rather than using network-centric partners to provide the service.

CrowdStrike thinks this business will ultimately be far larger than endpoint detection and response. Co-Founder/CEO George Kurtz talked about the general (admittedly rough) consensus envisioning a world where every human-based endpoint is accompanied by 90 agents that can work and process data at exponentially faster speeds. He also added that this in no way felt like a ceiling, and the true number could be significantly higher. There was a time when Bill Gates was mocked for saying that every household in the USA would eventually have a computer in it. Now, the norm is having far more than one. I think the same will be true for current estimates surrounding eventual agent traction. Nobody can really envision everything these machine-based assets will do for us. So? We can’t possibly know how many we’ll need. I’ll take the over.

AI Labs – Friends Not Foes:

Frontier Labs like Anthropic are excellent at uncovering vulnerabilities and writing code to patch them with unmatched automation and scaling vs. solutions used even just a few quarters ago. This is a major part of exposure management, which CrowdStrike also provides. At the same time, AI labs on their own lack the needed access and data to come close to fully understanding a customer’s digital estate. This makes CrowdStrike and partners like Anthropic perfect matches for vastly improving threat management in a complementary fashion. Beyond that, without decades of outcome data and governance experience, AI companies lack the ability to build any of the in-line breach protection that companies routinely rely on. These capabilities can’t be vibe coded without proprietary enterprise context that CrowdStrike will never give away.

Within these partnerships, Kurtz thinks frontier models will continue being required for the most complex and work-intensive tasks, but not for anything else. For everything else, cheaper models, including open-source options, its SafeMind model and others, can and will be used.

Updated Financial Model:

For 2026, ONON reiterated low-20%  constant currency (CC) growth. As a reminder, the highly promotional environment with intense price competition pushed them to hold off on fulfilling some wholesale demand. They weren’t willing to create a revenue sugar-high in exchange for heavier discounting and the associated brand degradation. This company is always taking the long view, and right now, that’s pushing them to make decisions that the public markets do not like. Probably the right thing to be doing, but that doesn’t mean a moody, short-sighted Mr. Market will think so. For Q3, this headwind will lead to 17% CC revenue growth. While they’re not willing to explicitly predict when this market weakness will begin to ease, they did indicate that Q3 should mark the worst of the obstacles. Wholesale conditions aren’t totally in their control, but there are still some levers for ONON to pull. They’ve gotten more active in helping partners like DICK’S Sporting Goods improve displays and storytelling, and they added that sell-out rates improved during Q3 as a sign of the strategy working.

For 2026-2029, ONON expects at least 5.6 billion Swiss francs, which represents high-teens revenue compounding. That revenue is 3% below consensus, but again, this was an “at least” guide. For some historical context, at ONON’s 2023 Investor Day, it guided to at least 26% CC growth through 2026 and ended up a little more than 4 points higher than that. There is good reason to think they were highly prudent in modeling, and they even explicitly said they were with newer sports like golf and soccer. More later.

Regarding margins, ONON thinks GPM will stay around 65% during this period. Ideally that would still be expanding, but they need to balance margin optimization with pursuit of expansion. Furthermore, 65% is already excellent for apparel and testament to their full-price discipline. GPM tailwinds like direct-to-consumer mix shift, manufacturing economies of scale and inventory management improvements will all remain in place. They’ll just use that potential added margin to reinvest in the business. Over time, they don’t see 65% as a GPM ceiling. Finally, they do not expect to get aggressive with price increases to support GPM, and continue to push for accessible price points on the performance side, with more expensive offerings in lifestyle.

Despite the lack of presumed GPM expansion , the company plans to push EBITDA margin from a little under 20% this year to 22%+ by 2029. That leverage will come from distribution leverage and then G&A leverage as ONON grows beyond heavy fixed cost investments to support scaling.

Other guidance notes:

Product Roadmap:

ONON is embarking on one of the more aggressive footwear refreshes in company history. They’re “refreshing every important running line” in expedited fashion, each optimized for something different. Cloudrunner will fixate on support, Cloudsurfer on comfort, Cloudmonster on bounce and Cloudboom on speed. They’re augmenting the lineup with these goals in mind. Two things are true: First, ONON has a founder-led leadership team with a fantastic culture and track records. Second, I always get nervous when apparel companies talk about accelerating the pace of product newness. That creates heightened execution & existing inventory risk, as there’s never any guarantee of launches building traction. With ONON, the chances are as good as they are for anyone else. But still… far from certain.

Golf and Soccer are the next two expansion sports for ONON. Golf feels like the perfect activity to add to its overall assortment. As leadership rightfully pointed out, spend per average player is higher than any other sport, the audience is getting younger and more into fashion and the lifestyle/sport intersection is as natural. They think competitors are innovating slowly and know their wholesale partners have been yearning for ONON to do a lot more in this category. Now they will. The organization aims to have a “toe-to-head” offering during the first half of 2027.

On soccer, their launch with Kylian Mbappé and Sydney Schertenleib was hugely successful from a social media perspective. This launch campaign collected 50.5% of all football-related media attention at the peak, with the audience skewing very young too. As with other categories, ONON prioritized working with famous athletes who truly felt connected to the brand and wanted to participate in shaping product design. This is where they landed. They’ll begin scaling product offerings and marketing next year, with the meaningful ramp likely coming during 2028.

Interestingly, while the company will offer soccer products for men and women, it looks like they’ll focus a bit more on the women’s segment. They see a larger shoe and apparel quality gap in this part of the market, with most female designs merely retrofitted versions of men’s offerings. They aim to create more assortment purpose-built for the 50% of the market where there’s less intense competition. Makes sense.

The same female focus will extend to apparel, as ONON pushes to “win with her” as a main goal. They will still provide plenty of shoes and shirts for men too, but they think there’s more runway on the women’s side with higher growth potential and an easier ability to stand out.

Quick Take:

No change to my opinion here. If I had to own a name in the space, it would be this one, but I have chosen zero exposure. This is the nicest house in a neighborhood that I don’t want to live in. I think the team is fantastic and they’ve executed as well as anyone else for a few years now. While the accelerated product refresh cadence makes me nervous, they certainly have a much better chance of threading the needle than Lululemon did a couple years ago.

There’s still room to grow brand awareness and greatly expand store count; I just prefer to focus elsewhere. You can’t own them all and it is nearly impossible to be a long-term winner in consumer retail. Your brand momentum can be as fleeting as a popular teenage influencer deciding they like a slightly different shape or color or style than you offer in any given season. I don’t pretend to know how to model that abstract reality, which pushes all of these names into my “too hard” pile.

SoFi and Mastercard have taken their SoFiUSD settlement partnership live across SoFi Bank's card programs on Mastercard's global payments network. This was first announced in March. SoFi says its full card program will migrate to blockchain-based settlement and they’re proud of getting this live in 6 months. The progress is expected to scale to $25B in annualized volume post migration, and they’re in ongoing negotiations with large merchants to drive even more acceptance and adoption. 

Merchants don’t need to maintain separate accounts for holding stablecoins or create new payments infrastructure to support it. They can instantly settle in fiat through one SoFi Bank Account and access 24/7 withdrawal availability as well. This is why it’s expected to be so popular. Cheaper… more convenient… more access… fewer accounts. It should easily be a good value-add to their Big Business Banking offering alongside a single regulated banking platform for fiat and crypto. And through SoFi Tech Solutions, 3rd party issuers can use these same systems, which will hopefully help improve overall tech solution adoption. That’s their expectation as we head into 2027.

I also published a lengthy SoFi podcast episode during the week.

6. Headlines & Macro

Edgewater came out with a new note calling out fewer opportunities for larger market share gains (For who?) and tougher price competition from other solutions. To them, the company entering a “show me” phase. They cut Q4 and 2027 revenue and profit estimates. Following the underwhelming Q3 and blaming it on model update cadence, I don’t love hearing this. Need to see a clean quarter from them to keep the Trade Desk comparisons from getting very frequent.

Oracle issued something called a “force majeure” notice for its New Mexico data center project. This “frees parties from their legal obligations when an extraordinary, uncontrollable event prevents them from fulfilling a contract.” Their justification is to maintain needed financial flexibility and liquidity in case things turn sour, but despite this news, the project remains on schedule.

In Macro-land, manufacturing data was much better than expected alongside hot services data and resilient jobs data. There’s currently at 68% chance of an October rate hike vs. 71% earlier in the week following news about potentially easing tensions in the Middle East.

Vibe Check:

Meta Founder Mark Zuckerberg and Meta Superintelligence Lab founder Alexandr Wang took to the stage to present Meta’s latest innovation. The Muse launch from a few weeks ago kind of stole the thunder from this event, but there was still plenty to share. And throughout that sharing, I couldn’t help but notice a softer, friendlier, more personable tone than we’ve heard from the company in recent years. It wasn’t just the lovably quirky Alexandr Wang referring to Muse as a “banger, scientifically speaking.” It wasn’t just Zuck wearing a “building is my love language shirt.” The vibes were fantastic all-around. Jokes were flowing, smiles were large and, if we look at all of the successful recent product innovation, who can blame them? I certainly won’t. 

Meta’s models are rapidly catching the AI frontier at prices that vastly undercut. Their upcoming “Watermelon” launch is expected to match, if not exceed, the best models currently on the market. 

Muse:

On the agent side of things, according to Appfigures, they outpaced apples-to-apples ChatGPT downloads by 40% through the first 12 days. I realize Meta enjoyed the ubiquitous brand and global distribution that ChatGPT hadn’t yet built, but this is still notable. It's not like they piggybacked on Instagram sign-ups the way they did with Threads and, furthermore, ChatGPT was arguably the most successful consumer launch ever. Even more encouraging is that Muse reached 450K daily active users in 10 days compared to 7 weeks for ChatGPT (and the chart below shows that momentum has accelerated. That metric relies on consumers finding enough value from the product to build habit and routine. It’s happening with Muse at a best-in-class pace.

The argument that Meta couldn’t monetize AI has been silly from the start. They’ve shown more ability to use this tech to augment their core business than any other mega-cap peer. On the other hand, they hadn’t yet shown signs of monetizing AI outside of their existing business. Muse will accomplish that in a highly disruptive way. 

Meta is offering much more free compute to users than competitors. For context, I canceled my paid subscription with Claude this past week because Muse is giving me everything and more for $0. The idea here is very similar to Meta’s legacy apps business. Build gigantic scale without charging customers and figure out other ways to monetize once you’ve captured their loyal attention. I don’t think this is the only reason why Anthropic and OpenAI have more recently prioritized cheaper model launches (open source competition is another big one), but I do think it’s contributing. Muse monetization will happen via transaction enablement. Meta has rapidly added bellwether merchants like Walmart, Best Buy, Sephora, Expedia, DICK’s and so many more. It has also added important partnerships with three payments and commerce infrastructure titans including PayPal, Shopify and Stripe. Amazon is holding off on integrating for now, but it seems highly likely that the two titans will sort things out and partner here too.

Hardware:

Meta announced new VR Glasses that will come out in Spring 2027 for $1299.99 per pair. It’s 80% lighter than Quest 3 at 100 grams, doesn’t have a clunky head strap and features a nice upgrade to battery and display quality, with eye tracking and hand gesture controls too. Hearteningly, Oculus founder Palmer Luckey made an appearance to promote the new product, as Anduril and Meta collaborate on U.S. military glasses. The relationship between these two world-class founders seems to be repaired, which is great news for both companies. 

Other interesting features like dual screens for working and hologram-based calling were included. Importantly, it also supports direct, USB-C-based wired connections to computers and gaming systems, which customers have been yearning for. I think two things are true here:  First, Meta is making rapid progress with a product that looks increasingly normal, comfortable and reasonably priced. The $1299 price tag is about a third of the Apple Vision Pro and $900 cheaper than the most similar Snapchat alternative. Notably, Meta is doing this while moving along what it sees as a clear path to positive hardware profit.  I think this product offers the best blend of performance, cost and look in the market… by a wide margin

Second, Meta still has a ways to go to drive this to massive consumer scaling. They need more time to miniaturize, iterate and build applications (like Muse) that can be brought to life by this form factor. They’re getting there faster than everyone else, but they’re not at the finish line just yet.

In other smartglasses news, Meta is launching a cheaper camera-free version of the Ray-Ban product for $349. This should help with privacy concerns and offer a much lighter alternative for those not wanting the camera feature. They also refreshed their Ray-Ban Meta glasses with a 3rd generation lineup that, shockingly, comes with better battery, camera quality, microphone performance and an action button for Meta AI and Muse access.

They also introduced new Meta-branded glasses at a $249 price point. In my mind, the Ray-Ban and other partnerships were great ways to build faster adoption with brands that consumers already knew as sunglasses makers. Now that consumer awareness has grown, it would make sense for Meta to shift some of the focus to its own glasses, which don’t require them to profit- share. That means they can do things like economically charge just $249. The margin on that sale may not be incredible, but it gives Meta a lot more access to its consumers with a gateway product and allows it to up-sell software add-ons and monetize Muse usage in more ways. That’s the real profit driver. 

Generally speaking, there were several new versions announced and collaborations with global influencers like Kylie Jenner too.

One more thing here. Meta came out with “Muse Charm” as a hand-held gadget built to facilitate Muse usage. It can  be activated with a simple tap and no need to unlock your phone as another way of potentially making customers less dependent on these mobile devices. It features cameras, microphones, a speaker and 5G connectivity and will be ready for the holiday season. They’re still finalizing the design, but it sounds like everything is on track.

Consumer Ecosystem:

I think the Meta vision is becoming a lot clearer. They aim to offer best-in-class consumer hardware at a small fraction of the price of others and build giant scale. They’re ok with accepting a thin margin there, as they know more Meta glasses in circulation means more usage of its models, agents and other profit-rich software add-ons. The plan to generate profit via outcomes and transactions from a quickly growing list of partners, rather than charging an arm and a leg for the compute directly.

So, to summarize, they’ll have high-quality offerings across every part of the consumer AI opportunity. They’ll undercut everyone to build a more rapid scale. They’ll take advantage of global distribution and 3 billion daily active users to cross-sell products with higher margins to still make a fortune.

I think Meta is positioning itself as the consumer AI leader, and I actually think they can say “thank you, Apple” for this increasingly compelling market power. Why? I think Apple restricting Meta data access a few years ago lit a fire under Zuck’s behind. Plugging the signal gap was a painful process, but it became loud & clear that Meta could not rely on Apple for anything and that owning more of their ecosystem was imperative. Fast forward to today? They're leading in consumer agents. They are leading in next-gen wearables. They are rapidly catching up to model leaders. They are taking control of the entire ecosystem and their obsession with doing so was born from a friend surprisingly turning foe.  Thank you Apple.