Table of Contents:

Access 40+ earnings reviews from this past season & my updated portfolio/performance here.

Next week's content will include updated valuation multiple comp sheets, a review of MongoDB's Investor Day, a Rubrik Earnings Review & so much more. I have been traveling this past week and am excited to get back to the office and in the swing of things.

The two companies struck a multi-year partnership announced this week. PayPal will use Alphabet’s AI services to bolster its operations and will “partner with Google Cloud Platform (GCP) to reimagine its technology foundations and infrastructures. That sounds like a contract is coming for GCP. PayPal will get branded checkout and Hyperwallet Payouts distribution gains across the half dozen products Alphabet provides to 2+ billion people. PayPal Enterprise Payments (Braintree-based) will become a “key processing provider for GCP, Google Ads and Google Play” as well. 

For Alphabet, this is a fine partnership. They probably get a new cloud contract and incremental traffic for their AI services to improve more quickly. For PayPal, this is great news. All they had to offer was using Alphabet’s world-class AI services and cloud infrastructure. They should be doing those two things anyway. And now, PayPal has a real opportunity to incrementally boost volume and overall market share thanks to what should be a nice boost to visibility and distribution. More distribution alone does not mean more volume. There are a lot of checkout options. But this is why CEO Alex Chriss and PayPal have worked so hard to make their most known checkout option also the most rewarding. It now can more meaningfully stack rewards thanks to a larger partner network and strong customer lifetime value and its budding ads platform (now expanding to Germany and the U.K.). And it can do so in omni-channel settings thanks to its aggressive “PayPal Everywhere” push. 

The only annoying piece of this news is how vague the added distribution language was in the press release. That makes it hard to quantify how large this partnership truly can be. It should be very meaningful, but we’ll have to wait and see.

Smartglasses Traction:

Zuck confidently proclaimed that the company’s Ray-Ban smartglasses are on a “sales trajectory similar to the most popular consumer electronics of all time.” That probably means the company will sell 5-10+ million units this year, considering he thinks eclipsing that benchmark for 3rd-generation hardware is a clear signal of future ubiquity. He’s not declaring this the next iPhone or guaranteeing omnipresence, but he is saying the product is off to a fantastic start.

New Ray-Ban Glasses:

Meta announced an updated Ray-Ban model with double the battery life  as well as 3K video resolution. It comes with a cool new feature called Conversation Focus. This helps users amplify the voices of people who they’re talking to in a crowded room and tune out everyone else. They also tried to tease Live AI in a real-time demo, but it did not work as planned. That can happen when you’re introducing new features in a truly live fashion. The company blamed the wifi connection, and the CTO Andrew Bosworth (Boz) was visibly annoyed with himself and his team. I’m sure it will be rectified.

New Oakley Vanguard Glasses:

Zuck unveiled a new Oakley model with up to 9 hours in battery life. It offers 3K video recording with a 122-degree field of view and stabilization technology to maintain quality footage – even on rocky terrain. Its open-ear speaker is 6 decibels louder than the Meta Oakley HSTN.

Vanguard model comes with a Garmin integration, enabling a seamless ability to pre-set triggers for video recording (speed,  heartbeat etc.). Meta preps an aggregated video of the footage with stats to share.

Meta Ray-Ban Display:

Zuck also excitedly debuted a version of the Ray-Ban glasses with a new, Meta-built display. They’re fittingly called Meta Ray-Ban Display. It’s large enough to read a text, and is placed off to the side, avoiding blocking views. The technology is “crisp on the sunniest of days” and is "sharper than any headset” on a pixel-per-degree basis. In the image below, you can see what this looks like from Zuck’s point of view on the right:

This more advanced hardware will also come with Meta’s first neural band, unlocking hand movement-based hardware control. It has 18 hours of battery life and is water resistant. In a live demo that actually worked, Zuckerberg texted Boz, but the video call he tried to start with it did not go as planned. They blamed the wifi again. 

Another cool feature for these glasses is live subtitles. Customers can have conversations they’re having displayed for them with minimal latency, which should be quite valuable for hearing-impaired individuals. The starting price is $799 and this will start shipping in a couple of weeks (I’m sure after they fix some bugs).

Meta Horizon Studio:

Meta believes they need to improve support for creators building metaverse worlds and immersive new experiences. While many think consumer interest and demand is the primary bottleneck holding this part of the business back, Meta thinks turbocharging quality supply growth will be a massive help for the Metaverse. Along those lines, Meta Horizon Studio combines all of the AI tools Meta has been working on for Metaverse world creation. It greatly automates and expedites the process of creating new games and other experiences, with an agentic AI assistant coming soon to make things even more efficient.

Meta Horizon Studio will be powered by the new “Meta Horizon Engine.” This will replace the Unity Software product they were using and will provide the foundation and backbone for future app-building in the Metaverse. It allows Meta to support 5x the number of people per world. As part of this product announcement, Zuck also introduced a product called “Hyperscape Capture.” This allows people to use their Quest headsets to scan a room and quickly generate a full digital twin. Whether it’s revisiting cherished memories, remotely attending a concert or upgrading real estate touring processes, there are many use cases for technology like this.

“Together, these two products will enable immersive and interactive worlds across all of our products… starting with VR and eventually your glasses and social media.” – Meta Founder/CEO Mark Zuckerberg

Entertainment Hub:

Meta is launching an entertainment hub as a new video content viewing destination on Quest. Like Roku, this will provide a centralized dashboard that organizes various channels and applications. Universal Pictures, Blumhouse and Disney+ were the content partnerships Zuck announced today. I’m sure there will be many more.

Take:

Regardless of live demo hiccups, Meta is clearly the leader in smartglasses and it’s easy to see why this is the perfect hardware for the age of AI. As Zuck so frequently says, glasses can easily see and hear what we do – without distracting us. They can seamlessly collect relevant data to help users more effectively than something like a phone. I do think these are going to become more popular as the products advance and Meta should be the primary beneficiary. I’m still not all that bullish on their metaverse focus, but I am quite excited about smartglasses.

For 2025, GDP is now expected to rise by 1.6% Y/Y vs. 1.4% as of the June SEP. 2026 estimates also rose from 1.6% to 1.8% since the June release. Long-run GDP growth expectations remain 1.8%.

For 2025, there have been no changes to the 4.5% unemployment rate projection since June. The actual unemployment rate is currently 4.3%. For 2026 and 2027, projections were 4.4% and 4.3%, respectively. Both improved 10 basis points from the last SEP release. There are also no changes to its longer-run projection of 4.2% unemployment.

For inflation, the Core PCE for 2025 is still expected to be 3.1%. For 2026, estimates rose from 2.4% to 2.6% since the last meeting. For 2027, they stayed at 2.1%. No changes to its long-run 2% inflation target.

The Fed funds rate median projection for 2025 moved from 3.9% to 3.6%, which points to 2 more cuts this year. Notably, there’s actually an FOMC member with an end-of-year rate projection implying 5 more cuts this year. The interest rate projection range also widened from 80 bps to 150 bps since June, pointing to a lot more board member disagreement on the path of forward policy. For 2026, the rate target moved from 3.6% to 3.4%; for 2027, it moved from 3.3% to 3.1%. No change to the Fed’s 3% long-run target.

Employment:

Monthly job gains have averaged 29,000 over the last 3 months. That slower pace of hiring is partially related to modestly softer labor demand, which is why they see unemployment moving from 4.3% to 4.5% by January. This is what Powell was referring to when speaking about greater downside risks in employment metrics. At the same time, a lot of this slowing job growth is not demand related. The greatly reduced pace of immigration has materially impacted the labor supply available to employers for certain jobs. That’s having a large impact on slowing payroll growth, but doesn’t put upward pressure on the unemployment rate. This creates what Powell called an “unusual” dynamic in which payroll and employment metrics aren’t as tightly correlated as they typically are.

Inflation:

Powell rightfully pointed out that the modest pickup in recent inflation readings is goods-related. Importantly, services disinflation has continued. This supports the idea that inflation is currently a one-time, tariff-driven adjustment event, which means the new policy should work its way through the inflation data in the coming months and metrics should again improve. If inflation was still driven heavily by services or something else more structural, that would be more concerning. Powell conveyed willingness to look past what he probably expects will be a short-term dose of inflation in order to support labor markets.

The Fed is clearly not declaring victory on inflation, but they do think it has moderated enough from its peak to allow this cut (and probably a couple more this year). That is because incremental labor market fragility is presently a greater risk than inflation.

Cuts are very good for liquidity, risk asset valuations and (on a 6-12 month delay) consumer confidence. At the same time, the consumer confidence piece is really only true when cuts aren’t in response to a sharply deteriorating labor market. When that’s the case, employment weakness is a materially larger economic headwind than modestly easier policy is a tailwind. The worsening in the labor market has been quite gradual and orderly to this point while the unemployment rate still looks great. To me, the most important thing of note is that hiring weakness is tied to the reduced immigration rather than a structural decay in hiring appetite. If that’s true (and I think it is), consumers should keep spending and the economy should be poised for steady growth.

Unofficial Monthly Active User (MAU) Note:

Leadership disclosed having more than 130M MAUs during the presentation. This wasn’t an official disclosure, but more of a casual update without much additional context. If we unrealistically assume it’s 130M total and that they won’t add more during the final two weeks of the quarter, that still represents 15% Y/Y MAU growth. And growth will end up being much higher than that.

The Duolingo Score:

A key theme of the product event was making the Duolingo Score a bigger company priority. They want this to be the “global standard for language proficiency.” To get there, for example, the goal is for everyone to describe themselves as a “Duolingo French 60,” rather than offering vague, less standardized explanations on language proficiency.

Scores range from 0 to 160. Levels aren’t randomly assigned, they’re based on the Common European Framework of Reference for Languages (CEFR). Duolingo gets speakers up to 130 for English, Spanish and French, with 6 languages set to join that roster in the coming months. This should be a nice user growth unlock. 0-29 is the beginner score range. This is where people learn basic words and the present tense; 30-59 is pre-intermediate and is when sentence structure gets more complex and vocab expands. 60-99 is intermediate. Several different verb tenses are introduced and lessons continue getting more complex.

Score 100-129 is where it gets more interesting. They call this level “independent.” 130, which takes 554 hours to reach, is the highest level one can become on Duolingo. That’s where someone is talented and proficient enough to confidently be hired in that language. The speaker probably still has an accent and doesn’t quite get all of the subtle nuances of a native speaker, but they’re in the ballpark and fluent enough. 160 is considered perfectly fluent.

Von Ahn was quick to point out that people with 1,000-day streaks can get there by spending just 50 hours learning a language. To quickly progress towards fluent speaking, you need to invest more time. That’s why they work so hard on making that time as fun and competitive and possible.

Because the rankings are tied to the standard CEFR foundation, it’s easy to test how good Duolingo is at teaching quickly and effectively. As you can see in the included image, for people who started learning a language on Duolingo, aren’t using other tools and are a level 60, its learners are getting more advanced. And importantly, that’s without them spending more of their finite time.

I’m a big fan of pushing the scoring system to learners. Not only does it allow users to compete against their friends in a new, quantified manner, but it could also be great for Duolingo’s competitive moat. If they realize their goals here, scores will be included across every hiring website, dating app and social media network. Its massive base of users will enjoy an ability to convey their skills to employers and potential romantic suitors better than everyone else can. I think that will be a powerful domino effect for attracting many new users.

English learners (I’d think eventually more languages) can easily get these scores verified with the Duolingo English Test (DET). These tests have been accepted by 99/100 top U.S. universities and 6,000 schools across the globe. And now? LinkedIn too. The hiring giant was announced as the first integration, with an easy 3-step process to upload your official score onto your profile. Many more are coming.

Video Call with Lily Updates:

I would have liked to hear more about this product’s advancement. It’s the single-most important feature for driving Duolingo Max subscription conversions. It’s a great concept for testing speaking skills, without stressing over talking to a native speaker. They briefly talked about conversations more dynamically changing with observed learner performance, expanded Lily memory (to make conversations more personal and engaging) and new themes.

Other notes:

Management reviewed their split-test-powered approach to product iterating. In essence, more data powers faster split testing, which leads to better product enhancements and an overall product. They’re borderline-obsessive in a way that means updates aren’t based on what they think users want, but what they know users want.

We didn’t learn a lot about other subjects. Chess did expand to 6 languages, and the team walked through some product features like bite-sized lessons, which make learning less intimidating. Matches vs. Oscar (one of their characters) were showcased; he comments on moves and talks trash to make the process more fun.

New Data:

Yipit data for Duolingo bookings does not look great. Specifically, August bookings for Duolingo fell by 8% Y/Y in the USA. Globally, growth was 32% Y/Y (social media issue mainly a USA issue vs. 38% Y/Y growth in July. Consensus expectations call for 33% Y/Y growth for the quarter, but September comps get 13 points tougher, which puts expectations at risk. I continue to think this is entirely related to their pause in social media virality and will resolve itself as that normalization process wraps up. For now, it could lead to Q3 weakness and risk to 2025 guidance. The company loves to sandbag when they offer expectations, so there could be enough wiggle room to still be ok. Still, this data release was not ideal.

Take & Reiterated Investment Thoughts:

This was a fine product event. Not overly exciting but certainly not bad either. I do find the scoring system to be compelling for potential product stickiness.

As I’ve been talking about a lot, this is a battleground stock. Doubters think AI chatbots and translation replace the need for this product. I’d ask them if they think an applicant has as good of a chance at a job if they’re talking to a potential boss through an algorithm rather than fluently. The answer is obviously “no” in my opinion – and I’d ask skeptics similar questions about social circles and romantic relationships.

Bears also do not think user growth can re-accelerate as Duolingo’s social media engine normalizes post-AI backlash. A talented social media leader leaving the company shouldn’t be remotely enough to jeopardize its future; that’s a ridiculous argument to me. I think Zaria Parvez is extremely talented, and fully replaceable. I think bears are wrong here too.

Social media virality powers top-of-funnel growth and it still doesn’t look great. Normalization has not happened yet. They seem to still be in fix negative sentiment mode (from the AI-first customer backlash), which prevents edgy content that routinely goes viral. That means Q3 results could potentially underwhelm. 

For me with this firm at 28x forward FCF, the basis of my investment patience is very simple. If they show us clear signs of user growth acceleration coming as expected, I’d stay very patient. If that viral content returns and user growth keeps slowing beyond expectations, I won’t. This has been one of my biggest investment wins ever, as my cost base is just under $120 and I’ve sold a material chunk of the holding for several hundred percent gains this year. But… There’s no room for unconditional loyalty. They need to keep earning my time and attention like I expect they will.

Compelling Partner Study:

Canalys conducted independent research that bodes very well for the health of CRWD’s partner ecosystem. In it, they found that $1 of Falcon Platform revenue generated by a partner leads to $7 in services revenue for that partner. Leadership called this the “strongest cybersecurity partner ecosystem multiplier on record.” CrowdStrike has done very well with embracing managed service providers (MSPs), system integrators (SIs), resellers, integrations and other partnership types. And they’re responding, as 60% of CRWD’s new business last quarter came from these 3rd-party channels. Studies like these bode very well for that momentum continuing, as partners focus most on Falcon, considering the sky-high returns. That’s especially true considering ROI positively compounds with more product adoption and with time, rather than dwindling down as issues and maintenance lead to cost growth surpassing value creation. 

CrowdStrike is extending certain Customer Commitment Package (CCP) rebates that it offered as part of its July 2024 outage concessions. Not customer rebates… just partner rebates. This is happening because CrowdStrike saw how effective incentives were to accelerate adoption of its platform and lead to positive return on investment. So it’s maintaining the most lucrative pieces of these packages.

More on a Deepening Partnership Ecosystem & Broadening Reach:

The tight partnership with AWS continues to broaden. Falcon added native integrations for important AI products including SageMaker and BedRock. The existing Nvidia relationship is moving in the same direction. CrowdStrike added Charlotte AI AgentWorks (its model building environment we’ll discuss later) with Nvidia’s Nemotron large language models (LLMs). This includes models built by Nvidia customers on the NeMo Agent Toolkit. The integration allows shared customers using those models to tap into the diverse array of data CrowdStrike has across its dozens of modules and operational history right from Nvidia’s platform. 

CrowdStrike also announced a new Falcon Shield integration with the Salesforce Security Center. As a reminder, Falcon Shield offers Agentic AI-infused software as a service (SaaS) posture management and security, meaning security teams using CRM’s security center will now have full visibility and control over this software. That control ranges from human-based interactions to agent-based for a complete sense of security. Whether it’s for better posture or configurations or agents with unwanted access or suspicious activity, Salesforce Security Center customers will now have access to Falcon’s world-class offerings. One more announcement here. CrowdStrike is also integrating Charlotte AI with Agentforce, giving shared customers an ability to more confidently deploy and scale agents across their enterprise. This is made possible by Charlotte’s ability to automate a wide array of security tasks, making protection of this new asset class seamless, rather than daunting. Charlotte AI can now be deployed in Slack, helping with security work prioritization, coordination and completion. Both of these integrations will be completed this calendar year.

CrowdStrike and Meta announced a new, tangible framework for grading LLM performance for security-specific use cases. It combines Meta’s CyberSecEval foundation with CrowdStrike’s threat intelligence to score various security workflows. While this may not be a direct revenue driver for CrowdStrike, it should help accelerate enterprise AI adoption, which would mean more demand for its products down the road. Why? Because enterprises struggle with understanding what high-quality AI security looks like. There isn’t a standardized scoring system for this specific category and the two companies want this to create that. This testing and ranking process will empower enterprises to understand when AI is deployment ready in a low-stakes experimental environment – rather than in runtime. 

“By embedding protection with the world’s AI leaders, CrowdStrike is uniting the ecosystem to innovate with AI, secure AI, and lead cybersecurity into the AI era with the CrowdStrike Falcon platform.” – Investor Materials

The “Industry’s Richest AI-Ready Data Layer:”

CrowdStrike’s “enterprise graph” combines the vast data from its threat and asset graphs to combine adversary & asset-level context into one. This pulls from its data lake, consolidating, consuming and organizing data from all security use cases for insight gleaning. This world-class data scale and architecture is accessible with an AI-ready query language to make usage intuitive. It’s what feeds all of Falcon’s modules and agents to actually create differentiated value vs. less experienced, less prevalent and less capable competitors.

Agents, models and all other AI assets can’t just ingest a mess of unstructured data and easily turn it into automation and value. It needs software middlemen like CrowdStrike’s Enterprise Graph to organize, understand and make data available in a way that is comprehensible to these non-human assets.

Here, I think it’s worth reminding everyone about their Onum purchase. For review, that firm’s real-time data streaming product is able to vet and evaluate the value of data to know what to block before it’s used by products like the Enterprise Graph. That means a smaller, more efficient and more affordable flow of data into these products for CRWD’s customers.

If you missed the explanation of Onum and that M&A decision, I would highly recommend reading about it in this article. If you search for “SIEM M&A” in the article, it will take you right to that commentary.

Release the AI Agents:

Two more big announcements in the realm of AI agents. First is Charlotte AI AgentWorks. Like many other enterprise software firms including Salesforce and Microsoft, CrowdStrike is offering no-code tools and guardrails for customers to build their very own AI agents. They can tap into all of the data and algorithms CRWD has to offer, while infusing their own data into the mix, helping them build more granular and enterprise-specific agents. So… they get CRWD’s vast data layer built on 15 years of scaled security data to season agents, as well as their own data to enrich these agents for more specific use cases. Pretty powerful unlock, in my opinion.

All of this will mean better product market fit for its client base, as well as more traffic and demand for the Falcon platform. It’s worth noting here that this capability requires massively scalable access to 1st & 3rd-party data. The combination of its security data lake, SIEM to organize and unleash the value of this data, as well as Onum to enable real-time streaming and redundant data storage eliminations is highly compelling. And with no-code functionality, customers don’t need to be tech savvy to build these agents. They can simply query from a long list of LLM integrations with natural language to set workflows and rules.

As part of its agent push, CRWD also introduced a slew of its own security agents that are “mission ready to deliver machine-speed capabilities and accelerate outcomes.” The initial agents are as follows:

Importantly, partnerships with Rubrix, Salesforce, ServiceNow, Zscaler and many more broaden the data access these agents enjoy, allowing them to complete work even more effectively. All of these products add more value and time savings to CRWD’s overarching Falcon Platform. It’s already fantastic at allowing customers to do more with less. This will amplify that reality.

Pangea Security  M&A – Entering the “Interaction Layer” of AI Security:

CrowdStrike has worked hard to build products for protecting key AI assets such as models, next-gen apps and agents. Now, it’s buying Pangea Security to expand into AI asset interaction security. Whether it’s human-to-AI (visibility into how employees are using this technology in good or bad ways) or AI-to-AI interaction, Pangea Security has companies covered. And that is vital in the agentic AI era. These agents race around the web collecting eligible information from a broad range of resources. Making sure these interactions are done in a kosher and responsible manner is simply imperative. Enterprises cannot embrace AI without knowing darn well sure it won’t jeopardize the stability of their core operations. With Pangea Security, CrowdStrike becomes much better at helping customers eliminate the perceived trade-off between AI innovation and security risk. This inspects prompts (model inputs) as well as responses (model outputs) to block bad actors from carrying out malicious prompt injections or other model manipulation. In essence, it’s a virtual security guard approving every single interaction based on a dynamic, data-driven set of rules.

Looking ahead, Pangea will facilitate a brand new AI Detection and Response (AIDR) module as part of Falcon. This will create holistic security for AI data, models, agents, identities, infrastructure… and now interactions. So… companies are free to open their arms to AI innovation… while still sleeping very well at night. Between this purchase and the Meta news already covered, CRWD is becoming a key accelerant for enterprises fully adopting this cutting-edge technology as they move from experimentation to deployment.

CrowdStrike M&A will likely continue to involve small, tuck-in acquisitions like this one and Onum. 

“Agentic AI is Cybersecurity’s largest opportunity yet… 100x the identities, 100x the complexity and 100x the opportunity… this is a $90B opportunity” – CFO Burt Podbere

Leaning into Identity Security:

A lot of the AI-infused hacks taking place recently (and just most attacks overall) are identity-based. Knowing this, CrowdStrike is no longer entering bits and pieces of this market; it’s competing in the entire thing. All I can say is watch out, Okta and company. CrowdStrike has far more room for bundled discounting. Furthermore, there’s reason to believe CRWD is capable of building products as effective as other leaders in this arena. Identity is a very close adjacency to CRWD’s other product pillars. And these other product pillars mean superior data telemetry (aggregated streaming from various sources) vs. companies with smaller suites. It’s this better context – across endpoint, cloud, identity, IT and so much more – that makes leadership confident in their identity products “closing critical gaps” left behind by other scaled players in the space. Shots fired.

The updates at this event included a multi-factor authentication (MFA) tool that comes with phishing protections. Furthermore, leadership talked about “modern” privileged access management, which includes Microsoft Teams and identity directory integrations to improve permission allocation. Finally, it spoke on continued identity threat detection and response (ITDR) maturation, as well as strong Falcon Shield adoption. CrowdStrike is clearly determined to project agentic identities as that new asset class exponentially explodes over the coming years.

“Traditional IAM and PAM were designed to manage access, not stop adversaries. CrowdStrike closes the gaps adversaries exploit with these fragmented solutions, securing every identity across every stage of the attack and environment.” – Investor Materials

AI Data Protection:

CRWD unveiled a batch of Falcon Data Protection updates in the spirit of AI. This pushes AI data security beyond just the browser, without support for applications or cloud environments. This product update effectively combines endpoint-based data loss prevention and cloud-based data security posture management into a unified offering. And rather than offering “snapshots” from each module, CrowdStrike offers a continuous view of data movement within them. It helps tag and categorize GenAI data, which helps power full visibility of AI assets. It reduces alert false positives and leads to better work. With this view, enterprises can understand when shadow or malicious AI is operating and to know how to remediate. And finally, it means 10X detection coverage thanks to cross-domain visibility offering an interoperable view across the entire Falcon platform.

IT Market Expander:

CrowdStrike introduced Risk-Based Patching for its Falcon for IT module that features a Falcon Exposure Management integration. This effectively combines IT and security teamwork under one console, interface and platform. It means IT teams and security teams can actually communicate on bug addressing and resolution, enabling concurrent, collaborative work. This should accelerate and improve patching workflows.

Misc. Notes:

Tying Product Innovation Together:

Between CRWD’s vast set of 31 modules across data, endpoint, cloud and more, as well as all of the work it’s doing with agents and Charlotte AI, it’s confident in being perfectly positioned to be the Security Operations Center for the agentic AI era. They’re adamant they can be the company to deliver “better outcomes at lower costs” across the entire security stack. That takes being a true platform, with a single, lightweight agent and a unified interface across diverse product types. That’s Falcon. CRWD President Mike Sentonas and CEO George Kurtz poked fun at other platforms that resemble “digital taxidermy” or platforms that “look alive on the outside but resemble Frankenstein underneath.” While leadership’s conviction is subjective, it’s hard to argue when digging into the pace of innovation (22 updates announced at Fal.Con in total) and its brightening financial results. Speaking of which…

Financials:

For this fiscal year (FY) 2026, CFO Burt Podbere reiterated 17+ net new ARR (NNARR) growth based on a strong pipeline and retention trends. From there, the company guided to at least 20% Y/Y NNARR growth in FY 2027. That would get it to a minimum of $6.32B for that year, meaning a 22% ARR CAGR from FY 2025-2027. It’s hard to say what 2027 growth will be, considering the 2026 NNARR guidance is also a minimum. At the same time, FY 2027 ARR growth expectations were 20% Y/Y and this forecast gives them a great chance of being 2-3 points higher than that. CrowdStrike is always prudent with forecasts and has never given next year ARR commentary this early on. I’m confident they’re being conservative in these targets. But wait… there’s more. The company set $20B ARR targets for FY 2036. They have a great track record in terms of hitting several year targets, which makes me slightly more confident in these numbers than I would be for any 10-year goal. This goal includes $10B in ARR by FY 2031 and 15% compounded over the next 5 years to reach $20B.

CrowdStrike now sees $1.75B in SIEM ARR by FY 2031 vs. $1.25B previously. Onum has a material part in this rising optimism, as CRWD 3Xed its SIEM total addressable market estimate from $16B 2 years ago to $48B today. Podbere maintained $2.75B cloud and $1.25B identity ARR expectations for that same year.

In terms of margins, CRWD sees its 21.4% EBIT margin in 2026 rising to at least 24% in FY 2027 (new) and 30% by FY 2029. They also think FCF margin will move from 30% next year (reiterated) to 36% by FY 2029 (reiterated). It also remains confident in an 82%-85% subscription gross margin by FY 2029, which greatly helps with conviction in the other margin expectations. It has a ton of work left to do on optimizing cloud and data costs over time.

Let’s do some extremely speculative, probably inaccurate, mainly for fun modeling for a moment. Assume it does $20B in recurring revenue and about $800M in non-recurring professional services revenue for FY 2036. Then assume it can extract four incremental points of FCF margin leverage from FY 2031 to 2036, as pace of improvement slows. That would give it $7.9B in 2036 FCF. A world-class company with world-class visibility like this one should continue to fetch a premium FCF multiple similar to a company like ServiceNow at that scale. Call it 45x. That gives us a $355B company in the year 2036 and a 12% return CAGR from now to then. Based on inevitable future M&A, market expansion and typical conservatism in their forecasts, I’d take the over.

Flex is helping drive larger deals and, as Podbere reminded us, is commitment-based, not consumption-based, meaning customers pre-set spending levels at the beginning of finite terms, rather than consuming at their leisure. The “flex”ibility is in a customer’s ability to consume whatever modules they want within that period. All of this means flex has no impact on revenue visibility. And as we talk about a lot, customers are welcome to consume more than they commit to and are routinely doing just that. 

Take:

I still have a little more to listen to from this week-long event. I’m confident this is the lion’s share of the news from this long event. If there’s anything else to talk about from the remaining presentations, I will include it in next weekend’s article (and in Discord early next week).

This event was packed with exciting innovation and encouraging financial disclosures. CrowdStrike has clearly moved past its 2024 mistake gracefully and admirably. The elite team, the quality of its technology and the stickiness of the platform all deserve a lot of credit for that. They turned a large negative into the most positive thing they could by “seeding” its customer base with more modules and temporary discounts. The discounts are now fading and the module retention remains excellent. This all paves the way for continued growth acceleration and resumed margin expansion in the years ahead. It’s back to business as usual.

This is one of the only companies on the planet that I’d own at its current 2.2x free cash flow growth multiple (FCF multiple / forward 2-year FCF CAGR) and its 2.9x EBIT growth multiple. It’s a special enterprise with a giant opportunity and a best-in-class ability to capture it. The event provided more evidence.

There was a bearish report published on DraftKings that got into the competitive risks of prediction markets. I extensively covered that topic in recent articles, and I wanted to share that information again here. 

For review, CEO Jason Robbins does not feel pressure to be the first mover in prediction markets and has chosen a wait-and-see approach. When/if they do offer this, initial focus would be on introducing this in the non-legal sports betting states. That’s because he thinks sports betting will take the vast majority of overall market share in states with both. Robbins pointed to the UK, where both exist and where prediction markets are a single-digit % of overall bookmaker revenue. When sports betting is available, that’s where people flock. When it isn’t, they’re more likely to try prediction markets. The patience is also to avoid ruining regulatory relationships they’ve worked so hard to build.

The DraftKings CEO also thinks that online sports betting will be a much better product than peer-to-peer prediction markets over the long haul. He doesn’t envision these companies ever having the full feature set that DKNG has because of risk management issues when you’re an exchange instead of the market-maker. DKNG can block and limit world-class betters to manage risk whenever they feel like it, which means they can offer a lot more bet assortment to players while still protecting the company. For prediction markets, it’s much more of a free-for-all, without control over who takes the action and with a requirement to allocate funding for every single possible bet outcome that their uncontrolled user base makes. 

Interestingly, Robbins eventually sees DKNG entering other financial services to compete with vendors like Robinhood. I’d like to see them focus on sports, iGaming and prediction markets. If they do enter the prediction market space however, that will require licensing that would allow them to offer other financial services. He doesn’t see downside to experimenting and that’s fine as long as  it isn’t a distraction to core operations.

I continue to view potential clear prediction market legality as a large net positive for DraftKings. It would mean it can enter every single state at lower tax rates than it’s currently paying in existing states. It would probably not make much of a difference in its existing states, because the sportsbook is such a better product. But it would be extremely incremental in the 11 states where sports betting is currently not offered. Nothing in that report changes anything about my point of view towards the company. 

Lyft and Waymo are partnering in Nashville. Uber didn’t love this news, but it was almost inevitable. Waymo is going to continue to experiment with many different business models and partners. They’ve explicitly said this and Lyft is a natural option for testing. The partners that prove they can most effectively maximize utilization is how they’ll choose where to lean in. Right now, they’re still in testing mode. There’s every reason to believe Uber will be better than any other potential partner at optimizing their key performance indicators and helping these cars make the most money. But it’s understandable to see this headline creating a little short-term anxiety for Mr. Market. I am not fretting.

Maybe TTD CEO Jeff Green offended someone at AdWeek? Did he call someone fat? Sure seems that way. The coverage from that publication recently has been uniformly negative – and in strange ways. One article this week criticized them for tweaking Kokai’s user interface after user feedback. Then? They shared (I think confidential) information from junior employees that led the outlet criticizing TTD for using those employees to set their roadmap. So… AdWeek has been critical of TTD listening to customers and doing what they request this week. And? It has also been critical via claiming TTD doesn’t do what customers want. So which is it? These articles also claimed the interface tweak was a feature tweak instead of changing how those features were presented. It almost tried to make it sound like there were more issues with core targeting and reporting products, which is just not the case.

They topped this off with a report that Yahoo’s competing offering is winning budgets from TTD and Google due to lower fees. The article totally ignored the large brands I’m certain TTD has inevitably won from Yahoo in recent quarters and sounded like a promotion for Yahoo’s struggling offering more than anything. Budgets move constantly. AdWeek hunted for a needle in the haystack to find advertisers claiming they moved some dollars to Yahoo. Selective, cherry-picked reporting. I lost some respect for the publication this week.

Mercado Libre is buying its Brazilian brokerage partner (Nikos DTVM) after a successful multi-year relationship. Owning this company will enhance the breadth and flexibility of the brokerage product MELI provides in the country. MELI has shown a clear ability to out-innovate basically everyone in the nation (except maybe Nu) in banking and financial services. Getting full control of this asset should simply add to its ability to differentiate and extend its value proposition in the region. That, in turn, should mean more marketplace cross-selling, more adoption of other financial service offerings like credit cards or savings accounts, and, generally speaking, more growth.

Consumer & Employment Data:

Inflation Data:

Output Data: