We’ve sent a lot of earnings reviews this season. All of those and my current portfolio/performance can all be found here.

ServiceNow is an AI workflow automation leader. Their AI control tower offers a central dashboard for monitoring, maintaining and optimizing AI agents. It makes AI work completion & triaging intuitive to show enterprises how to maximize return on investment and boost AI efficiency “from project inception to retirement.”

And while this is nice, it’s missing most of a key ingredient. Accelerating pace of enterprise AI adoption will require ensuring those clients can embrace this exciting new technology without crippling security trade-offs. That’s a big source of friction right now, as companies are often required to choose between modernization of assets and protection of those assets. NOW dabbles in security, but it hasn’t been a main focus to date. Veza changes this. It’s meant to fix that trade-off with an identity security focus, and the timing is very good. Identity demand is enjoying a revival at the moment. That’s why CyberArk just set net new ARR records, why Palo Alto is spending a fortune to buy them and why CrowdStrike is increasingly focused on this growth opportunity as well. Agents represent a new form of identity to secure. And this machine-based form can scale exponentially faster than human-based identities while accessing information from around the digital ether more expeditiously than people can. That’s both exciting as a source of productivity and new identity demand… and also anxiety-provoking as a vast extension of the overall attack surface.

Enter Veza. Their asset graph can visually lay out and contemplate identity access points across human and machine-based assets. They make sense of the massive attack surface so it is manageable rather than chaotic. It handles permissions, access requests and powers “end-to-end visibility that legacy solutions can’t match.” With Veza, companies have full knowledge of who (or what) has access to which systems and an ability to fix improper access in real time. All of this happens from a central dashboard that provides end-to-end visibility and will perfectly complement AI Control Tower’s existing capabilities. This Control Tower will now be the orchestrator AND the protractor of proliferating agents, providing a clear source of incremental cross-selling, enhanced customer value, rising lifetime value and higher retention. This makes NOW an even more powerful point solution consolidator in the large world of workflows.

I like this move a lot. It’s not a tiny purchase, but it’s not massive either. The company was most recently valued at $800M by private markets this April and delivered 100% Y/Y ARR growth in 2024. Estimates for 2025 revenue widely range from $60M-$250M. Veza gives NOW significant incremental value to provide customers in ways that are highly relevant and related to its existing offering. This is a great source of total addressable market (TAM) expansion without veering too far from NOW’s established lane.

Another underwhelming conference from PayPal. There was nothing shocking in there, just more reiterations of 3 negatives. First, they continued talking about a weak consumer and macro backdrop weighing on branded checkout for Q4. Tougher Y/Y comps are also slowing growth, but macro was the main excuse. It’s hard to think challenges here are purely macro when we hear all big banks talk up macro resilience, observe Shopify delivering strong growth and witness apparel names from across that sector doing the same. Even Block’s macro commentary is more positive than PayPal’s right now. None of these are perfect comparisons for PayPal, but they’re all relevant and they’re healthier than the story this firm’s leadership is offering. If macro is worse for you than it is for your competition… it ain’t just macro.

Second, they reiterated an accelerated pace of investments next year. That will lead to slower profit growth as expected. It will also most likely postpone their investor day targets set earlier in the year. They again didn’t explicitly delay that timeline, but danced around it in a way that makes it pretty clear in my mind. If this item was solely based on agentic commerce and BNPL, I’d be a lot less irked. Those are exciting growth opportunities and it makes sense to pursue them intensely. But there’s another source. Their drive to build customer habits and frequency is taking “longer than hoped for.” Either they’re moving too slowly here or customers aren’t as interested in PayPal incentives as they assumed. Either way, that now means OpEx growth will mirror transaction margin dollar growth in 2026, which is meaningfully worse than the framework they offered less than a year ago. The incremental irritation here doesn’t come from the I-day target delays. We already knew that. It comes from their strategy to win back branded customers not going as well as expected.

The third item is the one that has me most annoyed. Like the first two, it’s not new or shocking… but it is convincingly negative. They’re moving much more slowly than expected in terms of releasing their modern checkout flow to customers. Only about 13% of transactions today are on their newest optimized flow. The other 87% of consumers still endure an objectively inferior experience and can easily switch to another option (out of many) for a better process. As long as this is true, I do not think branded checkout will find its footing or deliver the reacceleration they’ve previously laid out. And I think they’ll continue using macro as a scapegoat. I get it… 15 years of disparate and clunky integrations are hard to solve overnight. That’s 100% fair. Still, I can’t help but recall how vocal Chriss was about keeping his promises to build investor trust at the beginning of his tenure. This timeline, in my mind, was the most important promise he made. And it’s not being kept. I’m honestly frustrated with this company and team. Previous leadership’s tenure was plagued by not keeping their word. Every time I hear this new team talk… they sound more similar to their predecessors. I like Alex Chriss. I think he’s a nice person. I think he’s smart. I think he’s capable. And I think PayPal needs to execute better.

Branded checkout decay is why this company is so cheap. If that decay continues (like it will for Q4), this will remain extremely cheap and rightfully so. They can buy back all the shares they want to. They can deliver Venmo and Braintree inflections like they have. They can spend a fortune on Will Ferrell marketing campaigns. But? None of that matters nearly as much as core PayPal Branded Checkout still struggling. That will overshadow all of the other positives. I find the branded recovery to now be more speculative and lower probability than I have in the past. For this reason, I am seriously considering reducing PayPal exposure from 4.4% of holdings to 3.5%. I will keep you posted as always. No final decisions have been made. That cash would probably immediately go to ServiceNow (NOW).

Data Cloud Foundation:

Snowflake’s overarching platform is called the (AI-fueled) Data Cloud. It's a “single foundation to eliminate data silos” with a relational database makeup. Relational uses structured query language (SQL) and stores in rows and columns to make insight gleaning straightforward. This differs from document-oriented database vendors like MongoDB, which use Not Only SQL (NoSQL) and stores data in (as the name indicates) flexible documents. Snowflake is considered best suited for online analytical processing (OLAP), while document-oriented is considered best for online transactional processing (OLTP).

As a general rule of thumb, OLTP is considered better for simple, high-volume and real-time data operations. OLAP is (as the name indicates) more analytical and better for complex, strategic queries. Both have their different strengths for AI, and SNOW is now actively expanding into OLTP with its Crunchy Data acquisition. MongoDB would tell you SNOW's relational foundation isn't best suited for processing unstructured data for AI applications. SNOW would tell you the opposite.

Snowflake’s infrastructure unlocks affordable data storage, organization, querying and learning at gigantic scale. It offers these services with elastic compute capabilities, allowing for flexible scaling up and down of usage (just like MongoDB). The architecture naturally separates the functions of data storage and consumption, unlike legacy data warehouse solutions. In turn, this helps control costs and waste, handles diverse workloads and resolves potential scale bottlenecks. That’s why it’s highly capable when it comes to processing large volumes of complex queries with a diverse base of datatypes.

Under this framework, I can store as much data as I want to without the requirement of immediate processing. Data is utilized virtually, which removes the need for dedicated customer hardware. Snowflake does all of this for clients in a managed fashion, minimizing talent and infrastructure needs. All of this routinely delivers 50% total cost of ownership (TCO) reductions for its customers.

Snowflake splits its products into 4 categories:

Analytics Category:

This includes the Snowflake Data Warehouse, which is where structured data is stored and (on command) processed. Structured data is formatted data. It’s utilized for record keeping and report creation. Data can be easily fetched via a few different structured query language (SQL).

The Data Lake does what the warehouse does for unstructured data. Unstructured data is messy, unformatted and difficult for humans to quickly make sense of. AI, on the other hand, is fantastic at deriving meaning from this mess of information.

Data Engineering Category:

Key products here include Snowpark. This is its developer platform and data-equipped playground to build new things. It enables work in virtually any source code language. With it, customers can process and visualize data (through Snowpark functions) and build apps (through Snowpark Native Apps).

Next, GenAI models are voracious data consumers. Snowpark Container Services allows GenAI models to run closer to relevant data. This enhances performance, expedites model training and diminishes costs.

Dynamic Tables is a newer data streaming tool. It automates pieces of structured data querying within the data warehouse. It offers auto-updates to reports as new data is added to its cloud. 

Unistore is SNOW’s hybrid table product, which can ingest, store and organize both transactional and analytical workloads. Again... they're working hard on building out OLTP capabilities to expand beyond their analytical workload specialty.

Snowconvert is another product that helps with easier migrations, as it automates conversion and modernization of source code and data.

AI and Applications Category:

Cortex AI is what Snowflake calls its “AI layer." It includes a slew of GenAI-powered tools to (as Snowflake always says) bring AI, application-building and analytics “right to a customer’s data.” That conjoining routinely lowers data transfer and storage costs. Cortex AI offers unstructured text summary, sentiment analysis and helps beginners write SQL.

Considering SNOW's great access to data, they should be able to build valuable agents for customers. Speaking of which, Snowflake Intelligence is the name of its unified and highly capable AI agent that neatly ties all Cortex tools together.

Collaboration Category:

This is where the Snowflake app marketplace and its data connectors reside. Snowflake Data Sharing is its secure product for sharing data among the rest of Snowflake’s participating users. As more opt in, a compelling network effect of relevant data builds and the firm’s value proposition deepens.

Model:

Snowflake’s revenue model is consumption-based in nature. Customers book and pay in advance, with billings and revenue recognition coming as credits are utilized. This means visibility compared to SaaS business models is not as strong. It also means customers can more easily scale down (or up) usage when times are bad (or good).