I sent two portfolio updates (including an exit), Fed presser highlights & a short piece on rate cut beneficiary themes during the week. The SoFi investment case is quickly progressing. 

To new readers: While I do closely cover news for my holdings, most of the coverage at this point is non-holdings. Talking about a stock does not mean I own shares.

Adobe is a software giant that invented the .pdf file (co-founder John Warnock specifically). It provides programs to create and imagine, handle customer interactions and process documents. Revenue is split into two main buckets: Digital Media and Digital Experiences. Digital Media is made up of its “Creative Cloud” and “Document Cloud.” The Creative Cloud includes Photoshop and Illustrator. It’s what empowers creation, iteration and perfection of digital design. The Document Cloud, including the ubiquitous Adobe Acrobat, allows for secure PDF management and collaboration – among other things.

Finally, its Experience Cloud includes Adobe Analytics and other products like “Campaign.” Campaign is its (intuitively-named) marketing campaign tool. Experience Cloud covers end-to-end customer interactions with a real-time customer data platform (CDP), ensuring that those interactions are optimized. It also publishes some greatly appreciated macro data on overall commerce spend.

The snapshot of this report was posted last week. This is the more detailed version.

Beat revenue estimates by 0.7% & beat its guidance by 0.8%.

Adobe trades for 27x forward earnings. Earnings are expected to grow by 12% this year and by 13% next year.

Adobe has an unparalleled amount of data within the unstructured PDF format. With the newer AI assistant in both Adobe Acrobat and Adobe Reader, it’s now actually able to allow data scientists to query conversationally; soon it will enable the creation of complete summaries or even slide decks from these PDFs. Pretty cool. Usage of the AI assistant in document cloud rose 70% Q/Q off of a small base. Its series of GenAI models (called Firefly) powers the AI assistant. For example, the image generation model is what will be used to form structured presentations out of seemingly jumbled PDFs. It’s also working on deeper integrations between its document cloud AI suite and Adobe Express (AI-infused app for creating digital assets) and, simply put, GenAI-inspired upgrades to its entire document cloud suite.

For the segment overall, net new ARR (NNARR) set new records and rose 24% Y/Y. Usage and user growth were positive across Reader and Acrobat, with the help of strong growth from added Chrome and Microsoft Edge extensions.

Adobe (shockingly) added more tools for its Firefly models to Photoshop to “accelerate core creative workflows” while automating tedious tasks. That was the focus of this product section. In Illustrator, it enhanced text-to-pattern capabilities, debuted Generative Recoloring and added Generative Shape Fill (automates the conversational creation of vector graphics). It also added Generative Fill for Photoshop. Generative Remove is another AI-powered tool for Adobe Lightroom (photo editing); its strong adoption was explicitly called out by leadership. It also added generative expand functionality for Premiere Pro video editing.

All three Adobe clouds are becoming more cohesive and interoperable with the help of GenAI. Document cloud is being used to vastly enhance the pace of creative cloud materials. Those materials are being readily used to drive better customer interactions in the Experience Cloud. It’s all better together. Adobe Express is really where all of this comes together, with go-to-market for this product ramping now and 1,500 enterprise customers signed during the quarter.

“We are amplifying creativity and productivity by enabling the convergence of products like Photoshop, Express and Acrobat as knowledge workers and creatives seek to make content more compelling and engaging. We're bringing together content creation and production, workflow and collaboration and campaign activation and insights across Creative Cloud, Express and Experience Cloud. New offerings, including Adobe GenStudio and Firefly Services, empower companies to address personalized content creation at scale with agility and enable them to address their content supply chain challenges.”

CEO Shantanu Narayen

In GenAI, Adobe sees its massive data as a key ingredient to compete. At the same time, per the team, the “greatest differentiation” comes at the user interface layer. It sees an unparalleled ability to drive value through many GenAI applications across its broad suite of tools. It also believes this broad suite of tools (again, massive dataset) better trains AI models to uplift utility.

“Given the early adoption of AI Assistant, we intend to actively promote subscription plans that include generative AI capabilities over legacy perpetual plans that do not.”

CFO David Wadhwani

For another example of products converging around GenAI innovation, Tata Consultancy Services (massive company in India) is using Adobe Premiere Pro (video editing) to transcribe videos and the AI Assistant in Acrobat to create written summaries of these videos.

The company launched the Adobe Content Hub during the quarter. It’s a central destination for cross-department collaboration and content editing within the Adobe Experience Manager (AEM) (for optimizing interactions with potential customers). Broader collaboration also means more open access to AEM for 3rd party agencies that its clients routinely will work with.

Since debuting in March 2024, the Adobe GenStudio is already making an impact for some large clients. This product taps into GenAI to automate content creation, turbocharge the content supply chain and vastly bolster the amount of marketing materials a company can produce. It also comes with deep performance analytics indicating what’s actually working. Vanguard is using this tool to raise quality engagement rates with investors by 176% via more personalization. That personalization is a direct effect of GenAI making it rational to create marketing materials for smaller customer cohorts and interests.

I thought this quarter was rock-solid. GenAI adoption was a key risk for this firm a few quarters ago, but it has absolutely figured things out. While it’s not moving the financial needle like it is for infrastructure players, that should change as inevitable app monetization begins. The guidance miss isn’t concerning, as it is related to deal closure timing and this firm just continues to wonderfully compound top and bottom line at its large scale.

Oracle provides a slew of software and hardware tools for on-premise and cloud environments… with an understandable focus on a continued shift 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 (server-less and container-based).

Strategic software as a service (SaaS) offers apps for human resources, enterprise resource planning (Oracle NetSuite) etc. It’s hard at work on launching more industry-specific software apps across areas like Healthcare. The full suite of these apps is called Oracle Fusion.

The last segment is Oracle Database (OD). Creating valuable apps from GenAI infrastructure requires great models and great data products to properly season those models. That’s where its Oracle Database (OD) product comes in. It provides a no standard query language (NoSQL) database for unstructured data which is highly important in the age of GenAI. Oracle closely integrates with the 3 big hyperscalers (Azure, Alphabet and most recently Amazon) to allow its OD database products 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 OD. Oracle believes that this data cloud interoperability provides inherent cost advantages with data transferring. Cost benefits are estimated to be  “several times cheaper” for model training than any competitive product, according to leadership. Oracle is also building 23 new OCI data centers for Azure and 12 for Google Cloud to support their GenAI developments (including OpenAI’s). These multifaceted partnerships are expected to be several year drivers of OCI and database growth.

Oracle is (re)-emerging as a digital infrastructure titan. While the company did take longer to roll out its high-performance compute product suite, that has since achieved fantastic traction. The results you see below are the byproduct of it taking its fair share of this massive high-performance computing infrastructure boom.

The snapshot of this report was posted last week. This is the more detailed version.

At an analyst event right after the earnings call (not sure why they didn’t just do it then), Oracle set new fiscal year 2026 and 2029 targets. For 2026, it raised revenue guidance from $65 billion to at least $66 billion. Sell-side estimates at the time of the change called for $64.5 billion. Strong… but 2029 targets are exceedingly strong. Oracle sees more than $104 billion in fiscal year 2029 revenue. That compares to estimates looking for $89 billion for a 17%+ beat. The stock exploded higher on this news. Finally, it sees EPS growth ramping to 20% Y/Y by 2029 as well. From its current FY 2025 to FY 2029, sell-siders called for EPS growth ranging from 13% to 16%... meaning this is also better than expected. Oracle is back. I get that this is very far off, but its large backlog and multi-year subscription business does provide incremental visibility here.

Oracle trades for 26x forward EPS. EPS is expected to grow by 13% Y/Y this year and by 14% Y/Y next year.

Demand continues to comfortably outstrip supply, as Oracle gears up to double CapEx in 2025 vs. 2024 to support this phenomenon. There is massive demand to train LLMs and OCI + its databases (it thinks) provide the most flexible and efficient way to do this.

Its CapEx will continue to directly track bookings activity, and that activity remains very strong. For evidence, OCI consumption rose by 56% Y/Y and cloud RPO rose by a whopping 52% Y/Y despite tough comps; it also reversed a normal seasonal pattern of Q/Q RPO declines for this quarter. That’s the fantastic large contract momentum in action.

Oracle now has 85 live cloud regions with 77 more projects coming soon. The largest of this is an 800 megawatt data center that supports “acres” of Nvidia GPU clusters. It will soon build a datacenter with a full gigawatt of capacity. To account for these massive energy needs, it plans to build small nuclear reactors. We’ve also just learned about Azure using nuclear power partnerships for its own infrastructure.

“That's what's required to stay competitive in the race to build one, just one of the most powerful artificial neural networks in the world. The stakes are high and the race goes on.”

Founder/Chairman Larry Ellison

Ellison spoke on Oracle’s Exadata database cloud services running on its own RDMA networks. Exadata is a specialized computer system that is blazing fast and can handle massive swaths of data. RDMA stands for Remote Direct Memory Access and allows for cheaper, quicker, more efficient communication between computing systems. He also thinks this setup provides an “order of magnitude” improvement in performance, scalability, reliability and security vs. other databases, while being the first fully autonomous database for training models.  The end result of this foundation is Oracle being a compelling database vendor for countless clients. Apparently, the 3 hyperscaler partners believe that as well.

Oracle did get a bit of help on operating expenses from extending the useful life of some servers from 5 years to 6 years. This is it trying to milk utility from general compute infrastructure and avoid spending more there as it embraces the high performance compute boom. We’ve seen many, many other companies do the same… Meta, Amazon, Alphabet, Zscaler, Cloudflare etc. Nothing shady here, although EBIT would have been a small miss without this help.

Cloud gross margin on the app and infrastructure sides continue to rise as it finds efficiencies through improving economies of scale. Really good to hear.

The quarter was really good… the 2029 guidance that came after it was elite. It clearly shows all of us that Oracle is far from a dinosaur. It is still a go-to database vendor, with cutting edge apps and infrastructure to round out the suite. Growth is accelerating, RPO growth points to that continuing and margins are not at all suffering during a time when many are spending like crazy to keep up with competition. Really impressed with this company. Congratulations to shareholders.

Snowflake CEO Mike Scarpelli was asked about guidance and inter-quarter trends, but declined to comment as he always does. Here’s the vague commentary he gave:

“I would just say as we gave guidance, we like the trends we were seeing and our guidance reflects what we're seeing at that point in time.” –

Snowflake CFO Mike Scarpelli

Zooming out a bit, a hot topic of debate lately has been how quickly and how meaningfully new Snowflake products can impact overall financial results. Snowpark is an environment for developing apps (with pre-built templates called Snowflake native apps) and is the only new product baked into guidance this year and represents a 3% contribution to revenue. So when will Cortex AI (suite of AI tools like Document AI unstructured data processing) and all of its other products make their own impact? Scarpelli is encouraged by early trends, but “needs to see a few quarters of history” before he’s ready to put it in the guidance. He essentially committed to these products being a material part of next year’s forecasts.

He also reminded us that the core data warehousing and engineering business is quite healthy. That’s true, but it’s also slowing more quickly than most expected, with these new projects not ready to pick up the slack. Scarpelli wants to see a revenue growth re-acceleration next year, which will be powered by core business stability and new product acceleration. We shall see!

As a reminder, Iceberg Tables are open-sourced data storage offerings, often with lower storage costs, open-source integration flexibility and more data control. They were actually originated by Netflix. The risk is that this will lead to data storage and duplication revenue headwinds, with the impact expected to be the worst in Q4 and storage making up about 10% of its business. But there are positives here too. By supporting this form of open source storage, Snowflake is making it easier for customers to use more data sources within Snowflake for analytics and consumption. The other 90% of revenue potentially stands to benefit from this broader interoperability. Early on, it’s already seeing incremental workloads being moved into the Snowflake ecosystem for querying, insight gleaning and usage thanks to Iceberg. And? Here’s what Scarpelli had to say about how the storage risk is playing out:

“I have not seen any customers move data out of Snowflake that was being stored in Snowflake as a result of Iceberg. Even for some of our big customers who have said they want to adopt Iceberg, I'm still seeing them do big migrations to Snowflake.”

Snowflake CFO Mike Scarpelli

Again, the biggest impact is expected in Q4, but it is still undeniably encouraging that it hasn’t seen any negative impact to date. Perhaps it was too pessimistic here.

Well, well, well… finally a company pushing back on endless GPU spending. Scarpelli told investors that Snowflake is done buying GPUs (aside from in a few markets where supply is not yet available) until he sees real revenue to back up this spend. That stands in stark contrast to what we hear from everyone else. To all of the other players, the risk has uniformly been called “under investing” into this opportunity. That puts them in real danger of falling behind in our rapidly changing world. Snowflake doesn’t see it that way. It thinks it has the GPUs that it needs and is done making these investments until it needs more. It’s not interested in building out capacity ahead of demand like the hyperscalers and many others seem to be. We shall see what the correct approach was over time. But for a company that’s reputation has moved from disruptor darling to potentially losing ground against Databricks and others… for a company whose new team has consistently talked up accelerating innovation… this surprised me.

For this year, the impact could actually be quite positive. It slashed its profit forecasts earlier in the year due to an incremental $50 million in GPU investments. Perhaps it didn’t need to allocate that much, which would foster EBITDA upside. That’s sorely needed for this company right now as margins remain challenged.

Like many others, Snowflake still sees GenAI apps in experimentation mode. Most companies have not figured out how to effectively use the software made possible from high-performance compute data centers and most projects remain in beta testing. It has deployed some apps internally for faster data cleansing and workload migration help and is now focused on using GenAI to automate support functions, which could diminish G&A needs over time.

Snowflake sentiment has soured around its ability to unleash and organize unstructured data to train GenAI models and build apps.

Snowflake Unistore is Snow’s hybrid table product, which can ingest and organize transactional and analytical workloads. Snowflake has primarily been an analytical workload specialist  until now. This will be fully released later in the year, but isn’t expected to drive revenue until fiscal year (FY) 2027. It will unlock many new types of applications that rely on transactional workloads in an efficient manner. So why isn’t it already ramping? Per Scarpelli, it’s because “nobody has ever done this before.” SNOW sees itself as blazing this trail and creating this new product-market fit at a margin profile that makes sense for the firm. It won’t rush and these things take time; it thinks things are tracking well.

Fal.con 2024 was as interesting as ever. The company walked us through product innovation, go-to-market tweaks and an updated view of the July outage impact. We’ll dive into all of the highlights right here.

CrowdStrike debuted a new user experience called Project Kestrel. This more holistically merges the natural usage and workflows of more CrowdStrike modules to “remove silos and unify data” more effectively across Falcon. This will mean more context fed into every event and decision to uplift the overall value and security of the platform. It’s also meant to make other products not being used by clients more visible to drive more module cross-selling. It comes with highly malleable access controls and a “single view of all assets, vulnerability and misconfigurations.” CrowdStrike has always strived to deliver that single view, and this is an upgrade on that constant journey.

While endpoint security is CRWD’s most mature product category, it remains in aggressive innovation mode here. Complacency in a rapidly evolving world is how companies die, and this team is determined to keep its R&D pedal to the metal… despite the July 19th outage and coinciding impacts (much more later).

CrowdStrike Signal is its new “family of AI powered engines” to scrape similar events and vulnerabilities into less anecdotal, more meaningful insights. This improves its ability to prioritize the right issues, enhance coverage and minimize false positives (a real cost and time burden for customers).

In other endpoint news, CrowdStrike added coverage for older Windows operating systems, including Windows XP and more. Many customers still run on older Windows operating systems, and now Falcon can work with them.

CrowdStrike added AI Security Posture Management (a more specific version of cloud security posture management (CSPM)). This continuously monitors the hygiene and configuration of large language models (LLMs) and AI apps presiding in cloud environments. Another tool to cross-sell within this rapidly growing, yet untapped segment.

Data Security Posture Management is also now fully integrated into Falcon. This tags, organizes, and protects unstructured and structured data in motion or at rest. With this, CrowdStrike is now the only cloud security provider to offer both DSPM and App Security Posture Management (ASPM) within its overarching Cloud Native Application Protection Platform (CNAPP). I know, I know… these guys are in love with acronyms. With Falcon, this data never needs to be stored on an endpoint. It is streamed to cloud infrastructure for more efficient storage in real time. That’s unique and allows CrowdStrike to avoid limiting what data is actually collected (doesn’t have to worry about overwhelming endpoints) while also lowering latency for querying and processing. This is one of the key differentiating Falcon platform factors Kurtz discusses constantly.

CRWD President Adam Sentonas spent a lot of time on cloud security. He believes its platform is more complete than any other and that continued hyperscaler growth will support more success for years to come. It’s currently a $515 million annualized business (growing north of 80% Y/Y), and CrowdStrike sees that reaching about $2.75 billion in 6 years (at most) for a robust 32%+ compounded annual growth rate. The company thinks visibility and posture management are the easy parts of the cloud opportunity. A key competitor (SentinelOne) agrees, which is why it has even talked about offering these products as free add-ons.

The real value CrowdStrike envisions is in actually protecting cloud runtime security. It offers what it views as best-in-class agent and agentless security tools to combine a broad understanding of the cloud environment with cloud protection.

Whether it’s cloud detection and response (CDR), cloud workload protection (CWP) or all of the other agentless acronyms we’ve already mentioned, it has everything a customer could possibly need right in Falcon. Nobody else does. Simply put, the first wave of the cloud opportunity was visibility and posture management. The second, larger wave will be protection. CrowdStrike thinks wave one is very crowded and wave two is a greenfield opportunity.

CrowdStrike  is pushing to make identity protection (not an identity brokerage like Okta) part of its core offering. During the event, it announced Falcon Privileged Access. This makes the principle of least privilege (PoLP) more of the de facto mode for the product. PoLP simply means giving your employees and vendors only the access they need… nothing more, nothing less. The new launch also features just-in-time access, so access to needed tools is only granted when it’s actually needed. This eliminates the threat of dormant permissions being breached and abused by adversaries. All of that helps to keep a tighter lid on infrastructure and allows companies to customize individual access.

Interestingly, Microsoft CEO Satya Nadella made a remote appearance on the stage with George Kurtz. Microsoft and CrowdStrike are fierce cybersecurity competitors and CrowdStrike leadership routinely throws shade at them. Like at the recent Microsoft security event, Nadella spoke on the importance of an open ecosystem and kernel (core software infrastructure) access for shared Windows customers. Some were concerned that Microsoft would try to eliminate CRWD’s kernel access following the event. Instead, they’re just adding more configuration and security controls, while increasing software package deployment prerequisites for 3rd parties like CrowdStrike. Good news for the entire industry, in my opinion.

In other Fal.con Microsoft news, CrowdStrike added real time threat protection support for Microsoft Entra ID. This got an extremely loud applause from the attendees. Entra ID is a cloud native identity and access management (IAM) tool. It offers single sign-on, multi-factor authentication, identity protection and granular access controls. Developers and IT professionals have been clamoring for this for a while.

SIEM collects, gleans insight from, organizes and unleashes the power of 1st and 3rd party data. It allows for log management (data storage, collection, monitoring & analysis) at great scale, providing companies with an overarching view of their data. At the same time, this scalable data ingestion is a key part of its endpoint product called extended detection and response (XDR). XDR is EDR with more access to 3rd party data sources. Accessing all of that data relies on the SIEM suite. CrowdStrike bought Humio in 2021 for $400 million to provide the foundation for this SIEM offering. It’s now doing $220 million in high-margin, triple-digit growth annual recurring revenue. Good purchase… I guess.

The firm sees SIEM as ripe for the taking. The industry is cloaked in entrenched incumbents failing to deliver needed innovation. Especially in the age of GenAI, data is exploding in size and importance. Incumbents have failed to deliver the changes needed to rapidly and affordably process all of this data, and are delivering cost inflation to their customers without acceptable performance. Enter CrowdStrike. Because so much of a company’s data is already in Falcon, it makes sense to move the rest into it as well. That’s what leadership will readily argue. 

CrowdStrike debuted AI-generated parsers for data log management. AI parsers pull from GenAI models to automate data extraction and pattern creation from raw logs as data is ingested. Insights lead to new parser creation and parser enrichment from previous findings. This process means the process of data ingestion, processing and analysis is constantly improving and allows companies to find extremely subtle issues missed by analysts. In turn, that lowers false positive rates and helps mightily in threat detection, software performance optimization etc. This uplevels a customer’s ability to seamlessly use their data in the Falcon environment, or, as Kurtz puts it, just makes ingestion “easier and better.”

This recipe is extremely powerful. It means customers have ALL of the data and ALL of the security and observability tools needed under one roof. It makes CrowdStrike the truest platform in cybersecurity.

Aside from this announcement, CrowdStrike debuted detection posture management. This works in tandem with MITRE ATT&CK to guide customers on techniques to uncover protection gaps with “prescriptive recommendations to strengthen security posture.” Finally, it announced new workflow automations with a new library of 300 workflow templates for clients.

I’ve covered both of these concepts in detail in recent articles. For a quick review, Falcon Flex is its malleable subscription offering. It’s not consumption based, but it does allow customers to draw down on pre-set, multi-year commitments at their leisure. No annual minimums. At the same time, customers can easily mix and match modules and add more to spend commitments if needed. In the short period that Flex has ramped to $700 million in total contract value (TCV), customers are routinely consuming more than expected. 

As another quick review, Customer Commitment Packages (CCPs) are its concessions given to customers in response to July 19th. They’re temporary incentives. This can mean extended subscriptions, longer free trials, comped professional services, discounts and Falcon Flex credits. CrowdStrike wants every single customer to receive the credits. Why? Because its products are excellent and its platform is extremely sticky. This, as CFO Burt Podbere describes, “seeds the world” with more of its product. And it’s extremely confident that when Flex credits expire, customers will continue using the comped modules and grow overall spend.

Perhaps as another customer concession (although it wasn’t called that), the company announced CrowdStrike Financial Services (CFS). This is a high yield savings accou… just kidding. This is CrowdStrike’s new internal financing division to help clients map out payments over their contract terms. It sees this as highly “channel and go-to-market friendly.”  All of this is designed to make the CrowdStrike selection as easy as possible. As its total addressable market explodes from $116 billion in 2025 to $250 billion in 2029, products like this give it a better shot to capture even more of the opportunity. This is a wholly-owned CrowdStrike subsidiary and it is self-financing.

CrowdStrike maintained all fiscal year pre-incident 2029 margin targets. Customer commitment packages will weigh on gross margin and cash flow for about a year, but the firm expects no material impact thereafter. CFO Burt Podbere hinted at getting there sooner.

On the path to $10 billion in annual recurring revenue (ARR), it maintained the “by fiscal year 2031” target, compared to “2029-2031” previously. It continued to hint at the potential of still reaching $10 billion by 2029. $10 billion by 2031 includes:

Their TAM remains just 6% penetrated.

“Pipeline activity levels have returned to pre-July 19th levels.”

Founder/CEO George Kurtz

Podbere also walked us through a few CCP customer examples and impact on overall TCV and ARR. For customers electing extended trials, there’s no TCV hit, but there is an ARR hit. When customers instead pick more Flex credits for more modules, the actual impact on ARR is quite positive. In one case, Burt talked about offering $600,000 in flex credits to motivate more module uptake and immediately more ARR from $1 million to $3 million ($3.6 million when the credits wear off). I don’t think this should be overlooked. CrowdStrike turned a historic blunder into a real opportunity for more platform adoption and ARR per customer. They made the absolute most out of an awful situation. Props to them.

I’ve owned CrowdStrike since the IPO. I’ve loved this team since the IPO. Still, as Max subs know, I trimmed the stock when it was extended near $383 and then trimmed another large piece on July 21st following the outage. I haven’t added since.

Based on the immense uncertainty we faced at the time, that was appropriate in my view. At the same time, their response has exceeded my expectations and solidified their bright future in my mind. I’m waiting on another Mr. Market fit or maybe some litigation headline risk to resume adding to this existing position. Sell-siders continue to line up to defend this name following the event, but the headline risk of the outage is not gone, the multiple is again very lofty (80x forward earnings and a 4x PEG using consensus 3-year EPS CAGR) and I think patience is warranted. I’m happy to own what I own today.

Cloudflare sees its biggest GenAI opportunity in inference. Models aren’t constantly trained. They’re seasoned with data and periodically upgraded with new context to enrich capabilities. Inference is what constantly happens after initial training. Models get the knowledge they need to begin driving new insight, new ideas and new patterns from a seemingly unstructured mess.

And while it’s too early to expect a material revenue uplift, signs so far are good. 80% of the top 50 GenAI companies are on its network to help with inference (and also training). Furthermore, Workers Platform is its server-less (so fully managed by Cloudflare) product suite for developers to build, maintain, secure and deploy applications. This allows for caching of content and apps across Cloudflare’s global network for faster delivery (shorter distance). This includes its newer Workers AI product, which allows developers to access models and infuse GenAI tools (like sentiment analysis) into Cloudflare-hosted apps and networks. Overall adoption of Workers is robust, with developers on it rising from 2.0 million to 2.4 million sequentially. 

Per Seifert, the Cloudflare network’s global reach, which allows customers to be 50 milliseconds from any connection, is a differentiator. This allows customers to do more inferencing “close to the interface,” but not directly on the device.” This saves time and money. It aims to emulate connectivity performance levels enjoyed within a data center. It has also gotten very good at optimizing hardware needs to power its inference business. As Seifert puts it, “not every job needs an H200.” Some, less complex tasks can get away with less powerful and cheaper chips. Cloudflare is excellent at uncovering that minimum compute need, while using its global network moat to “ensure GenAI investments” are as efficient as they can be. To them, it’s “less so about investing a crazy amount in GPUs and models.” All of this means CapEx can stay relatively low, its margin profile can stay sky-high and it can pass savings on to customers to drive lower total cost of ownership.

Cloudflare is not yet focused on driving GenAI monetization. It simply wants more traction for its Workers AI platform and other GenAI products and also wants to help customers optimize GPU utilization to make new projects run through Cloudflare actually make economic sense. All 50 of its products also can run on “any server, in any location” in a completely “hardware agnostic” manner, which makes it a great universal partner for GenAI players. It has no conflict of interests with regard to hardware or software systems. It’s whatever works best for the customer. With this approach, it is confident that monetization will follow and is in no hurry to force that issue.

Secure Access Service Edge (SASE) is a term for how Cloudflare conjoins web performance products like Magic Wide Area Network (for secure, blazing fast connections) and its security use cases like Zero Trust Network Access (ZTNA), which seamlessly verifies requests at every turn, eliminating lateral threat movement (picking on the weakest part of a network and getting free, perpetual access thereafter). SASE drives vendor consolidation, controls costs and augments performance. Cloudflare One is its overarching product bundle subscription combining its suite. While NET got a later start here than others in the field like Zscaler, it now sees its offering as at parity with other offerings out there. SASE is the “second wave of Cloudflare products.” Wave one was its network connectivity and DDoS mitigation products, which are world-class. And according to this commentary, wave 2, which includes SASE and ZTNA, is now getting there (wave 3 is GenAI).

This is another tool in the tool kit for Cloudflare driving better outcomes and lower overall TCO via vendor consolidation and point solution displacement. Specifically, this broad bundling of network optimization and network security routinely cuts TCO by 30%-50%.

“I think the demand environment, unfortunately, is unchanged. Every deal is a fight. It’s currently about grinding through processes to get deals across the finishing line. There's no improvement in sight, to be very honest there.” – Cloudflare CFO Thomas Seifert

There wasn’t a lot of new information from its last earnings review. Here were the brief highlights.

Longer Term Contracts:

Datadog is enjoying incremental interest in clients signing larger, longer term deals. That’s a byproduct of its successful product suite bolstering and also some discounting that it's doing to incentivize bundling and platform adoption (like everyone else does).

Like other software firms, Datadog doesn’t benefit from GPU sales or reselling from Nvidia and hyperscalers. It needs models to be created, apps to be incubated and then finally deployed for its observability and app monitoring business to actually benefit. Per Obstler, companies don’t even have a grasp on return on investment yet to understand what kind of runtime software opportunity there will be. Its infrastructure monitoring business does benefit from the explosion in high performance compute data centers, but none of its other segments do.

Still, GenAI will lead to an acceleration in app modernization demand. General compute powered apps are expeditiously becoming more and more antiquated and companies can evolve or be left behind. That will eventually support Datadog’s business, as its bread and butter is monitoring and perfecting the performance of all of these assets. While Datadog says 4% of its ARR now comes from GenAI, that’s really just the tools created by GenAI native enterprise software firms for everyone else. The “everyone else” part of that equation has not begun deploying apps in mass.

“If LLMs are put in models and cause an acceleration of the investment cycle in re-platforming or acceleration in the ability to develop new applications that would complement Datadog.”

Datadog CFO David Obstler

RUM: Tracking actual user actions to optimize and flag suspicious flows rather than simulating users.

Synthetics: Datadog Synthetics is its way of testing interactions and processes in a zero stakes environment. It allows companies to predict and simulate usage patterns to test interface and product resilience.

Flex logs: This is its cost effective means to store and retain large batches of logs. They’re priced at $0.60 per 1 million annually and allow for separation of storage and query costs. This makes it ideal for long term data storage and regulatory compliance. With Flex Logs, storage and computation can scale in a parallel, independent manner. This separation for Datadog’s clients unleashes far more data scalability, customization and cost optimization. Conversely, querying from a flex log is slower than for Datadog’s standard log tier. That makes Flex Logs better suited for lower priority data.

Obstler wouldn’t commit to Datadog’s newer products getting to $500 million in ARR like its core offerings have. He is optimistic and sees a lengthy runway, but didn’t seem confident in the opportunity for these products being quite as large. This could just be a savvy CFO not wanting to offer too much information in a random investor conference.

PayPal is now a checkout option for Amazon Buy with Prime merchants. As a reminder, Buy with Prime allows merchants to offer Amazon’s world-class fulfillment to Prime members through their own sites. This means better control over the shopper experience and the coinciding, lucrative data without having to manage the fulfillment process. This will be done with a seamless, temporary link back to PayPal, with minimized clicks and pages to juice conversion (as always).

Over the last month, PayPal has announced the following partnerships and contracts:

That’s what happens when you re-shape the company to drive faster innovation, such as Fastlane for guest checkout or more software add-ons for Braintree utility-building. That’s what happens when you focus on things like authorization rates and fixing your grossly antiquated mobile checkout business… rather than irrationally undercutting competition to prop up volume growth at the expense of respectable margins. PayPal was never a dinosaur… although its old CEO may have been (sorry if that’s harsh). This company was yearning for a leader to get this train back on the tracks and he has done so with admirable speed. I expect all of these partnerships, new products, Braintree’s margin trough, better Venmo monetization to lead to a much stronger multi-year period for PayPal than we’ve had in nearly a decade.

I think the pictures below embody what I’m getting at. It shows you how fixated this company is on taking care of every customer. It was once known for awful customer service and just being too big to care about individual inquiries. That’s no longer true.

SentinelOne announced a new partnership with the world’s largest PC maker, Lenovo. As part of it, SentinelOne will integrate its Singularity Platform into “millions” of Lenovo devices. This also includes Purple AI – its GenAI assistant – as well as optional upgrades for existing users. Wells Fargo immediately came out with a note that this will be material to next 12-month results.

Between this news and the recent Alphabet partnership, SentinelOne is quickly rounding out a needed roster of powerful go-to-market partners. This is how you extend reach and, especially with Alphabet, how you better cater to larger enterprises. These largest enterprises routinely buy software through system integrators and enterprise app stores. SentinelOne must do a better job of forcing its foot in the door for all of these channels to give it a better shot at gaining more visibility. It has an elite technological reputation, and (as we’ve written about many times) needs to match that with better go-to-market. This is a great step. I’d love to see more system integrator announcements in the near future. 

Between this and leadership’s decision to forgo baking in the positive expected impact from the CrowdStrike outage, I see a good chance of them outperforming quarterly expectations in the months ahead. They explicitly told us that they’ll be ready to model the CrowdStrike boost next quarter, and I would assume this deal wasn’t included in guidance as it hadn’t been closed as of its last quarter. That demand outperformance should flow directly to the bottom line and keep feeding the currently explosive profit inflection. When pairing all of this with a gross profit multiple that’s roughly half of CrowdStrike’s and the quickly broadening product suite that feeds cross-selling, I continue to like this setup.

Amazon announced a series of updates for Buy with Prime (including the PayPal note above) at its Accelerate conference for sellers.

First, it updated us on how well this product is doing. Overall fulfillment from Buy with Prime (BwP) is up 45% Y/Y with a 16% merchant boost to revenue per shopper. This is Amazon better using its excess capacity to trim deadweight loss and juice the margin ceiling. Merchants are enjoying 13% lower out-of-stock rates with this and 24% better inventory turnover. But wait, there’s more. Buy with Prime traffic on merchant sites is up 50% Y/Y (+300% Y/Y during Prime Day 2024) and the overall merchant count is up 25% Y/Y. That works.

Relatedly, Multi-Channel Fulfillment (MCF) orders fulfilled by this rose 70% Y/Y. This is routinely used in tandem with BwP and lets companies store products in Amazon’s fulfillment footprint to supply orders across all merchant channels (100+ channel integrations). MCF is speeding up merchant delivery by 40% and raising conversion (with MCF badges to boost shopper confidence). It’s pushing harder to move more MCF orders from 2-day fulfillment to same-day, which boosts conversion by another 20%.

Amazon also extended its demand-side platform (DSP) for programmatic advertising campaigns to BwP merchants. Google Shopping Ads and TikTok Ads were the two named integration partners.

Merchants can now tap into Amazon’s fortress datasets to precisely target eligible shoppers with offers to drive traffic. 80%+ of campaigns so far are reporting return on ad spend at or above expectations. Would love that to be 100%. It also added delivery estimate displays to BwP TikTok ads. Delivery promises directly prop up conversion rates.

Finally, aside from the PayPal partnership news, it deepened its integration with Shopify (SHOP). Now merchants can access BwP right from Shopify checkout. Sticking with Shopify, that is why I thought its decision to sell the fulfillment business was so spot on. In addition to eliminating a massive margin drag, it turns Amazon into slightly less of an enemy and opens the door for partnerships like this. 

Amazon debuted a new chatbot for sellers called Amelia. It offers an “all-in-one” selling companion to guide best practices for customer service, conversion and growth. It was built on Bedrock. 

Amazon is making employees come back into the office 5 days per week. It says this is due to productivity benefits, which is probably accurate. I also think this will push some voluntary employee exits as this company continues to optimize its headcount. Not to sound insensitive… but you don’t need to pay severance when workers quit. In Amazon’s obsessive pursuit of margin expansion, this is another unfortunate lever to pull. I expect this to become a clear trend over the coming quarters as excess labor supply dissipates, unemployment ticks up a bit and the employer/employee power balance shifts back to employers.

The Search Giant reportedly offered to sell its Ad Exchange (AdX) product to end the anti-trust probe in Europe. AdX is its marketplace for sell-side publishers to offer impressions to buyers. Because YouTube and other company assets are also buyers, there’s actually a material conflict of interest in this case. This is the first time it has offered to sell anything amid regulatory concerns, and the offer was rejected. They want the company to sell DoubleClick for Publishers (DFP), which is a campaign creation and management tool for advertisers (similar conflict of interest).

The sale concession does point to rising anti-trust pressure on this company, but as a new shareholder, I am ok with that. As I’ve said before, I think it’s very easy to see how these divested businesses would be valued well in excess of 17x GAAP operating income. The sum of the parts is very likely worth more than the whole. Still, the ecosystem is extremely powerful here. It’s nice to be able to own a massive publisher campaign platform, a massive ad marketplace and lots of inventory to sell as well. This ecosystem weakens with every sold business unit. I would have to see what the forced sales actually were here, but if drastic enough, I’d consider lightening up on the holding. Most recently, the firm actually won an appeal to overturn a $1.66 billion EU fine for abusive advertising practices. Still, most of the findings were upheld and the probe is ongoing.

Powell Presser and Fed Statement highlights sent earlier in the week.