During the week, I published Shopify, Airbnb, MercadoLibre and Nvidia investor conference reviews along with Celsius and Starbucks news. I also published PayPal, Amazon and SoFi investor conference reviews along with an Oracle earnings snapshot.

I plan to cover the Oracle and Adobe quarters in more detail and a few more interviews from this past week’s Goldman Sachs conference next week. Some of that coverage will include AMD, ServiceNow, Datadog and MongoDB. I continue to work on the upcoming SoFI Investment Case refresher.

Kurian spent the bulk of this event offering investors an insight-packed overview of the cloud business. The presentation was split into GOOGL’s 5 GenAI offerings, which all sit on the Vertex AI platform.

Starting with the bottom infrastructure layer, Alphabet offers a series of chips, switches and memory bandwidth optimized for its cloud environment. Kurian was also quick to remind investors that this cloud business offers 70x the liquid cooling capacity (new new data center technology replacing/supplementing air-cooled) vs. any other hyperscaler. 

The company takes a similar approach to AWS on chips. It’s investing heavily in the category, with its latest 6th generation Tensor Processing Units (TPUs) offering 3x performance and 2.5x inference cost lead vs. competitors for some specific workloads. At the same time, the search giant openly partners with best-in-class GPU players like Nvidia and AMD to let customers use whatever hardware they want within the cloud environment. There’s some real evidence of the company beginning to flex its muscles in GenAI hardware. 90% of all AI unicorns use GCP for model training and inference, while 60% of all AI startups and large incumbents like Ford do as well.

The next category is developer tools. The firm aims to give developers all of the support they need to connect AI models (such as Gemini or models from partners such as Anthropic) to enterprise systems for eventual app and agent creation. Ultimate model choice lets developers and enterprises standardize on Google Cloud Platform (GCP) regardless of which tool works best for them. Again, client choice and avoiding vendor lock is a powerful way to create more open, valuable and sticky cloud contracts. If developers don’t have what they want, they’ll just go somewhere else. 

GCP also supports a wide range of model services. For example, its model grounding tool vets query accuracy, with adaptive grounding to prioritize which responses actually need double checking. That’s a big cost saver for companies. It also features model distillation capabilities to shrink massive parameter models down to smaller, task-specific models for clients like Samsung. All in all, 2 million developers are now building with its cloud-native, AI platform. 45% of these projects, with the help of system integrators like Accenture, are now live and ramping usage.

The next category is data analytics, which is an imperative complement to GenAI models and apps. As I frequently say, competing models can stand out from the size and relevance of the datasets they’re trained on. Knowing this, GCP (like both AWS and Azure with its Vertex debut) takes a fully open approach here. Customers can connect any data source (structured, semi-structured and unstructured) to GCP with “super low latency.” From there, it offers dynamic data agents/assistants for conversational querying, optimal data migration, visualization, insights and spreadsheet building. As a result of all of these tools, machine learning query volume for Big Query (how it monetizes) is up 80% year-to-date. The agents should also, per Kurian, drive more subscription seat growth, as it lowers the barrier to conducting complex, multi-modal data analytics.

The next piece of the opportunity is in security. This category is not only a separate growth vector, but a place where GCP can showcase some self-proclaimed advantages to drive customer trust and broader GCP adoption. For example, it sees 75% lower downtime rates vs. competitors and 50% lower vulnerability rates vs. “other clouds.” 

Mandiant, its threat hunting arm, helps collect threat intelligence from Google’s gigantic databases. Its traffic is rapidly growing, per Kurian. Mandiant also prioritizes threats, tests existing system configurations, flags issues and tells you exactly where you will probably be attacked while even writing the audit file for the case. This greatly helps in expediting remediation. 

Finally, we have GenAI applications. The company is actively lacing in GenAI tools like content creation, text summaries, translation and customer service automation to uplift the overall value of Google Workspace. One hospital system is using these tools to manage bed throughput and save $250 million per year in costs; Techniker Krankenkasse (largest health insurer in Germany) is using GenAI Workspace tools to cut claim handling from about 25 minutes to 3 seconds. 

While GenAI is deeply integrated into its established Workspace suite, Alphabet is also building new GenAI apps in areas like customer service. Many companies are doing the same thing, but Kurian sees their offering as best-in-class in a few areas. First, it can handle traffic from web, mobile, call centers and brick-and-mortar environments. Next, Kurian sees this product as world-class for answer accuracy rates, which is especially important for regulated sectors like healthcare and banking.

Kurian is determined to create the best environment out there for developers to build their own GenAI apps. For evidence of this objective being in reach, 75% of independent software vendors (ISVs) work with GCP for AI.

Kurian got the same batch of questions that AWS’s Garman got earlier in the week. It’s notable how frequently their answers were nearly identical. On go-to-market changes for GenAI, GCP is acting identically to AWS. It is focused more on “selling outside of the IT organization” to customer experience, marketing, financial and other teams based on the specific use case. It’s focused on training its teams to specialize by industry and stress the points of tangible value creation and short payback periods. Secondly, and also like AWS, it’s dedicated to building the exact solutions that customers want to diminish onboarding friction and pace of GenAI adoption.

Since the CrowdStrike outage, conversation surrounding technological architecture for security platforms has been front & center. CrowdStrike prides itself on the ultra-light weight agent, real-time data streaming and fantastic 3rd party review scores. But SentinelOne’s scores are also excellent, and it thinks its approach is actually better. CEO Tomer Weingarten got an immediate question about SentinelOne’s somewhat heavier, more algorithm-packed security agent compared to CrowdStrike.

In his mind, it’s not about the weight of the agent, but how lightweight SentinelOne actually is in terms of touching a client’s kernel (the most sensitive, core part of its operations). SentinelOne touches the core far less frequently and intensely than CrowdStrike and requires fewer software updates too. To the company (and intuitively), not touching the kernel materially diminishes the risk of its own software or algorithm updates leading to breaches of these all-important cores. It has created a work-around in Mac and Linux (not yet a Windows option) to emulate the data and insight produced by a kernel to learn from it and shield it without getting in its way. Bright people in the industry will subjectively argue about which approach is better than the other.

As Weingarten candidly said during the call, no platform (in any part of security) is perfect for every need. It’s about which batch of strengths and weaknesses are the most compelling for an individual deployment. SentinelOne finds itself being the most compelling more times than not; CrowdStrike would invariably say the same thing. In reality, they both should continue to find profitable growth within the massive, growing sector.

SentinelOne, like Zscaler and so many others, has been hard at work on revamping its go-to-market approach. It’s pushing to tighten relationships with channel partners and bringing in proven talent to usher in the next several years of growth. As part of this, it recently made another c-suite change by bringing in Barbara Larson as its new CFO (recently hired a new CMO and CRO too). Larson previously served as Workday’s CFO and the Senior Director of Field Financial Planning at VMWare.

This change is a response to SentinelOne’s ramping size in markets. It wants to “switch gears for how it grows scale” to push its revenue base across $1 billion and well beyond. It thinks it needs a “tremendous financial leader” to do that, and sees Barabara as the person. The praise for Larson can also be taken as a bit of shade thrown at outgoing CFO David Bernhardt. Considering this and SentinelOne’s issues with accounting in previous years, it’s good to see Bernhardt staying on with the company for another year to ensure a smooth transition. Furthermore, during the interview, Tomer told investors that this decision has been in the works for a long time. This wasn’t sudden or a response to any internal issues.

“Dave has done remarkably well for us. But he’s never seen the type of growth we envision… we're nearing that point that we just want to make sure that we're doing the most efficient way possible.”

Co-founder/CEO Tomer Weingarten

“Obviously, there have been some quarters where we didn't really get to what we believed that we could have. And I think now things with this team are looking just very, very different than any other point in time for the company.”

Co-founder/CEO Tomer Weingarten

There was a bit of investor disappointment from not getting a large upward guidance revision last quarter amid CrowdStrike’s issues. We have to remember that this only happened 8 weeks ago and the situation is highly fluid.  SentinelOne was not ready to assume what that benefit would be last quarter, so elected to forgo doing so. This could lead to compelling outperformance in quarters to come. Tomer told investors that they should be ready to quantify the financial impact next quarter. SentinelOne does expect a concrete benefit from this outage, and is already seeing pipeline favorability along with rising win and close rates. 

The firm sees many customers forgoing more workload growth with Falcon to invest more resources in SentinelOne. This joint coverage is routine in network security, and should become more popular following this summer’s outage.

Somewhat surprisingly, SentinelOne isn’t getting that aggressive on pushing customers to displace CrowdStrike. It’s simply trying to be there for customers and help in any way it can. This, it thinks, is what will maximize the positive impact of its competition’s mistakes.

Tomer talked about the idea that SentinelOne has led in MITRE evaluations for four years, yet still doesn’t see other companies or the market fully embracing that reality. Why? Because SentinelOne’s marketing budget is a small fraction of Microsoft's, CrowdStrike’s, Palo Alto’s, etc. It sees the most discerning buyers picking its product over the others, but also sees a lot of market confusion driven by (it thinks) false claims from competing products. 

Step one for SentinelOne accelerating its own marketing engine was inflecting to profitability. That’s now in the rearview mirror. The more important step two will be reaching annualized profitability. This should be in Q2 or Q3 of next year and is when it plans to really lean in. It doesn’t struggle with great tech; it struggles with go-to-market. A lot of that struggle is due to lacking resources to drive awareness and it’s now building the war chest needed to alleviate this growth bottleneck.

It has grown from 1% market share to 5% in a few years with a small fraction of the cash to drive awareness. Imagine what it can do as it becomes capable of more meaningfully spending budget here.

SentinelOne remains primarily focused on new customer wins. It is readily launching products across cloud and data security, and it has rounded out its platform quite a bit over the last couple years – organically and via M&A. Still, the opportunity for new customers remains the most promising. And? It merely gives SentinelOne more opportunity to drive more cross-selling growth down the road with what it views as best-in-class products.

SentinelOne’s main focus in cloud security is on runtime security, rather than Cloud Security Posture Management (CSPM). It sees CSPM as largely commoditized and has a product that it’s even considering offering for free in a bundle with other tools. Runtime security (like cloud workload protection (CWP)) is really where it thinks it can stand out. The market is highly fragmented and ready for the taking. 

The team was again asked about the changing seasonality in its Fiscal Year 2025. As a reminder, its billings guidance was ever-so-slightly below expectations, but reliance on a Q3 & Q4 acceleration to meet the number left some concerned. CFO Remo Canessa gave us the same explanation as on the call and as I’ve already covered. Zscaler signs three year deals with contracted billings being collected on an annual basis. The first half of 2022 and 2023 were weaker periods for macro and billings growth. That past softness is now flowing into contracted billings growth for Q1 and Q2 of this year. Specifically, contracted billings is expected to grow by 7% Y/Y through Q1 and Q2 (13% billings growth Y/Y overall) and accelerate to 23% for Q3 and Q4.

Like several other software companies in recent quarters, Zscaler has embarked on a large go-to-market overhaul. For now, the new salespeople in place as a result of this change are still ramping and won’t reach full productivity until the second half of next year. Elevated levels of employee attrition amid all of the change have also led to some go-to-market disruption. This dynamic is another, smaller factor for the enhanced billings seasonality. For strong evidence of things moving in the right direction, Zscaler has had a “strong 6 months” of hiring and now thinks it’s in a position to slightly moderate the pace of go-to-market headcount growth in 2025. For more evidence, pipeline quality is improving, sales productivity is boosting and close rates are rising.

Through all of this noise, leadership talked over and over again about how strong the underlying, structural trends of the business are. It thinks it’s in a “much better place” than last year on go-to-market and hinted at expecting to see revenue growth accelerate after the transition is fully finished. 

“After the 2025 transition, I'm personally here to grow at a much faster pace. That's what Mike Rich is here for, that's what we are here for. The opportunity is there, the platform is there and we're going through the right set of changes and transitions. We are one of the very few companies that can do actual cost takeout by removing a bunch of firewalls, VPNs and more products to show unreal savings. … in my view we’re in a great position to do well.”

Co-Founder/CEo Jay Chaudhry

Finally, Zscaler is always prudent on guidance, as the team strongly suggested during this past week’s interview interview that this fiscal year guide takes the same approach.

“If you look at our history, we've been prudent with our projections. And we're also prudent going forward.”

CFO Remo Canessa

Chaudhry sounded a tad annoyed with yet another analyst question on Zscaler competition. There have been several of these lately, which stem from many other security companies entering various aspects of the cloud security space. Zscaler isn’t seeing the competitive concern play out in conversations with customers:

“So many questions on competition. Why is that? Because I don’t see that when I talk to customers… You guys are only talking to the vendors. If you talk to more customers, you will have a different perspective.”

Co-Founder/CEO Jay Chaudhry

Chaudhry reminded all of us that Symantec’s Blue Coat had 85% of the Fortune 500 as customers at the time of its IPO. Cisco, Juniper and several others have competed with Zscaler across its product pillars. There have been entrenched incumbents since the company’s inception. Palo Alto and Fortinet both still have sizable firewall businesses and while Chaudhry continues to be adamant that firewalls will eventually all disappear, they’ve also entered Zero Trust and next generation cloud and network security businesses (to subjective degrees). Zscaler leadership is highly confident in its product suite being superior to anything else out there. Other vendors would argue, but its years of elite profitable growth at scale and sky-high retention speak for themselves.

There have also been countless new entrants. These will only continue to flow in. considering how large and compelling ZS’s markets are. Zscaler has been able to overcome this competition for years and sees itself as poised to keep doing just that.

“A lot of class B vendors tried to come in and say they compete with Zscaler. They’re all gone. Sold out. One of them is still trying to figure out who to sell to.”

Co-founder/CEO Jay Chaudhry

Great products across network and cloud security need massive amounts of traffic/data to train algorithms and enhance their usefulness. Zscaler sees 500 billion transactions per day, which it calls “an order of magnitude ahead of any other security vendor.” That means its best-in-class suite of cloud security products continues to improve faster than the pack. Chaudhry pointed out that there’s a reason why other vendors who claim to be entering its network space don’t disclose traffic metrics.

While macro is materially improving, cybersecurity spend remains resilient. It has been the least discretionary spend bucket in enterprise software for a long time, and this year is more of the same. Zscaler’s product suite eliminates seamless lateral threat movement. Lateral threat movement refers to an adversary being able to hack the weakest part of an ecosystem & then gain free, unconditional access to everything else thereafter. That’s what firewalls enable. Zscaler’s zero trust approach doesn’t, and routinely leads to better coverage and lower cost vs. alternatives. Its product efficacy and savings-fostering point solution displacement are popular in any environment.

Leadership was asked about its work with SMBs. They candidly told us that most of their work remains on large enterprises. They have a decently sized SMB business, but it hasn’t been a focus for them at all. I think this represents a long term opportunity when they’re finally ready to allocate meaningful attention here. The potential reminds me of what CrowdStrike successfully did on the endpoint security side with Falcon Go. It created smaller, more flexible product bundles (and then marketed the heck out of them) with more flexible contractual terms (under Falcon Flex) to give them more control over budget allocation. The result has been the SMB segment turning into a real contributor for it. I think Zscaler, which dominates with giant brands in network security like CrowdStrike does in the endpoint, can pull something similar off with its business too.

Just like during the earnings call a few weeks ago, Dara again told investors that consumer trends are “quite positive.” No changes here. Uber’s business is somewhat discretionary (especially delivery). Considering this, how is it faring so well? It could be because its customers are also younger and more affluent on average than the country. 40% of its base is between 18-34 and makes over $100,000 a year. While that makes some sense, its lower income users are also quite healthy and are actually growing faster than its richer consumers. Every cohort and demographic is spending just as liberally as they were a year ago.

While several smaller factors could be at play here, I think this is simply a byproduct of Uber’s unmatched business. Its driver scale is best-in-class, which means lower surcharge rates and better availability. Its product suite is broader than anyone else’s, and includes several low cost products to cater to sensitive customers. Lyft tried to do something similar on shared rides, but had to back away from the offering. It doesn’t have the scale, balance sheet, cost of capital or margin profile to have the luxury of offering these lower-value products. Uber does… which also means it gets all of the demand from this use case, grows engagement and juices cross-selling to its higher-value offerings. Uber then uses this product breadth to augment the value of its subscription and pass even more savings onto customers… and uses its merchant scale to create more delivery promotions for customers. Its massive market presence and focus on affordability and access all make its product offering uniquely value and its macro resilience more understandable.

“We have been growing faster than the category while improving margins pretty significantly on a year-on-year basis. And at this point, there's no reason to think that that won't be true going forward.”

Uber CEO Dara Khosrowshahi

Uber continues to push to be the partner of choice for AV fleets. As I’ve written about frequently, it wants to bring its strengths of demand aggregation and scaled supply to the equation, rather than building out its own fleets. It thinks its world-class data scale will be highly useful for optimizing route matching, customer service and throughput. It also thinks it will be needed for all AV fleets to supplement their supply through the multi-year transition. Why? Because if these fleets were just built to meet peak demand today, most of those cars would sit unused for most of the day and would burn too much cash. That’s where Uber’s driver supplement comes into play, which gives them a great chance to be a key part of this evolution. It also simply provides more demand, which has been shown to already raise utilization rates for fleets.

This is probably why Waymo, the most advanced player in the AV taxi space today, just extended its Uber partnership to more cities while now using it for more services like car cleaning. This partnership continues to tighten, but I question whether Wayma (or Amazon or Tesla) will need from Uber when they finally have their desired fleet sizes. That’s when things get a bit more dicey for this investment case. Sure, Uber can be the transition partner and help on managing supply/demand dynamics and service levels. But can these giants not figure that out on their own over time?

We are still years away from this being an issue, but not decades. Dara sees the software side of this equation (which will unlock hardware growth) being ready in the next 3-5 years. Amazon Prime and Google Maps provide elite network effects just like Uber does and I don’t think solely providing demand will be something these other players can’t do on their own. Waymo is already racking up solid SanFran market share with very finite capacity in 2024. 

All of this tees up why I’ve sold a majority of my Uber stake over the last few weeks. I wanted to take triple digit profits and respect the reality that Uber is moving from category domination, to something far less certain as the AV wave arrives. It can absolutely figure things out in this equation and it can absolutely sustainably grow for many more years and decades. I think that’s likely enough to still warrant a smaller holding. The probability of that happening is simply lower than I think it was in early 2023 when I bought the stake.

Uber sees a real opportunity to add more subscription value on the mobility side. It has done a lot on the delivery side, but thinks more focus needs to be paid to mobility. Whether that’s cheaper rides or awards, skipping wait lines etc. there’s a lot more coming here. Uber One penetration is still much higher in its most developed markets like the USA and U.K., with a massive runway to up-sell existing consumers across all other markets.

“Other players have programs. We believe ours is growing faster because we offer programming no one else has, so to speak. We’re like Netflix competing against the other streamers because we have exclusive perks.”

Uber CEO Dara Khosrowshahi

CTO Kevin Scott was asked about the pace of model improvement and if builders are seeing any diminishing marginal returns. According to him, we are “demonstrably not at the point of diminishing returns on how capable AI systems can get.” Rapid progress should continue in the years ahead and new use cases from this progress will slowly be unlocked to usher in the era of app-and-agent-level GenAI monetization. They are not close to running out of ideas on how to make models or copilots far more powerful. Part of this is being driven by a new phenomenon in GenAI. It inspires developers to go “chase new ideas” on model construction and usage. At the same time, it also gives them the tools to exponentially expedite that chase. This positive innovation flywheel means quicker idea discovery… leading to more tools to drive more of that quicker idea discovery.

This will be music to Nvidia shareholder ears. Despite massive investments in GenAI infrastructure, demand continues to outstrip supply. This should keep the large cloud service provider orders humming.

“Things are moving so quickly that it’s hard to imagine a world where you do the infrastructure build out and wait for it to be done to make decisions on the next (app) layer… you have to move fast because the tech is just evolving at such an incredible pace.”

CTO Kevin Smith

The OpenAI Partnership:

Smith called this partnership the “most consequential” in the company’s history. That’s not surprising, considering OpenAI is powering its copilots and new Bing search. Per Smith, OpenAI and Microsoft have a perfectly complementary set of skills. OpenAI has world-class research teams building arguably the best models on the planet. Microsoft has a gigantic dataset to season these models and the global scale to maximize adoption. It also has the cloud infrastructure to ensure OpenAI can avoid technological or compute-related bottlenecks. The two companies continue to work independently to drive their own AI innovation, but also very closely together to accelerate the pace of progress. There's a reason why Amazon and Alphabet are now pouring billions into Anthropic (One of OpenAI’s best competitors).

CEO Micahel Miebach interviewed with Goldman Sachs this week. He spoke on the health of the consumer, which is highly valuable insight, considering MA’s massive global spend volume. Mastercard does not assume credit risk, so comments here are only a great read through for volume (not repayment or delinquency rates).

During Mastercard’s Q2 earnings call, it called consumer spend healthy and resilient. That was being supported by continued employment market health. Since then, we’ve seen slowing but still decent job gains and an unemployment rate that just ticked back down from 4.3% to 4.2% (aided by government hiring, but still). That backdrop continues to enable “consumer spending remaining healthy.” CEO Micahel Miebach shared the company’s August spending growth data, which is nearly identical to July. No changes here.

Both high-and low-income consumers are demonstrating this resilience. High end consumers continue to spend on things like travel, concerts and other events, which continued into this week. Low end consumers are proving quite savvy in their ability to hunt for deals and stretch their budgets further.

Visa spooked investors a bit on its last earnings call when it talked about softer July spending trends. It cited several factors including weather, shopping event timing, Crowdstrike’s outage and more, but did not cite any macro deterioration. Some assumed this was macro-related anyway, and that its customer base (representing a great synopsis of the global customer) was growing more hesitant. Not so fast, my friend.

Quarter-to-date trends recovered and are now looking very consistent with Q3 in the USA. Payment volume is up 5% through August and stable Q/Q, while transaction growth recovered and cross-border volume is back to exactly in line with its rather robust forecast. He continues to feel good about the firm’s quarterly and annual guidance, while also feeling good about maintaining 10%+ Y/Y growth in the years to come.

“Trends have been stable. The consumer has held up remarkably well.”

This is for the USA, and trends in most of the world remain equally strong. The exception is Asia Pacific (APAC). Total volumes and cross-border trends have underwhelmed and remain below 2019 levels.

JP Morgan sees net interest income expectations from analysts of $90 billion next year as too high. This is solely related to more expected rate cuts.

JP Morgan sees no changes in credit health assumptions. They “haven’t seen any deterioration” and continue to expect a 3.4% loss rate this year and 3.6% for next year. It sees a U.S. economy that is having “some slowdown but still doing ok.” The consumer is driving that via stabilizing discretionary spend and slower-but-still-strong non-discretionary spend. Excess savings levels have come down, but are “still elevated” thanks to resilient employment and wage growth.

“From auto financing to mortgages to small business to middle market to large corporations, we haven’t seen any deterioration. Clearly, there are challenges in commercial real-estate. We haven’t seen the bottom of it. We are very well reserved for our exposure.” – Daniel E. Pinto

Ally’s commercial auto and corporate financing businesses continue to perform very well as expected. The credit card portfolio, which it pulled back from last year amid ramping losses, is “stable and in line with expectations.” The mortgage business is also performing in line with expectations.

So what isn’t performing? Retail auto financing. Net charge-off rates are above expectations and above what was assumed in its guidance. It sees its pool of struggling borrowers rising and these borrowers less able to repay debts than anyone else I listened to this week. It does do a material amount of retail auto business with lower credit quality borrowers, and when macro sours, this is the byproduct of that decision.

I think Visa, Mastercard and JP Morgan’s comments matter a lot more for markets than Ally. Most of the credit companies we cover don’t even touch retail auto and most also cater to stronger borrowers. Ally’s problems seem to be more unique to its own model and consumer auto financing specifically. The economy is clearly slowing, but still growing at a respectable clip and still comfortably maintaining full employment.

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