As a reminder, I liquidated my stake in April. The time-stamped sale and reasoning for the decision can be found via that link. No part of the reasoning changed from this report. I plan to finish reviewing this quarter alongside Nike when they report earnings later this month.

Lulu called the U.S. consumer “cautious” in its prepared remarks. The team thinks they’re doing reasonably well with improving product newness, which is perhaps why market share rose in both men’s and women’s premium athletic wear. Still, weak USA macro and trade wars were blamed for the negative revisions.

Lulu trades for about 17x its new EPS guidance and probably somewhere around there based on sell-side estimates once they fall to reflect this downward revision. EPS growth will be roughly 0% Y/Y based on new guidance this year. EPS is currently expected to grow by 10% next year, but I wouldn’t be surprised to see those estimates fall a bit.

I spoke about SoFi a lot in my weekly chat with Evan and Tannor on Wednesday. That conversation can be found here.

This event was unfortunately 25 minutes of Lapointe reviewing the business model and only 5 minutes of Q&A. My SoFi Deep Dive gets into all of the repeated ideas he covered. I’ll focus on the pieces of newness from the chat and offer some personal thoughts on the tech segment.

Current Trends:

Its ultra-prime customer base continues to spend at healthy levels, with him hinting at interchange volume growth accelerating into Q2. Borrower health remains strong and capital market demand for its loans remains equally strong. These things were all mentioned on their earnings call 6 weeks ago, but the backdrop is highly fluid and I take comfort in their important trends remaining healthy. I will gladly listen to them reiterate this as many times as they’d like to. Whether it’s unmatched breadth of person-to-person payments, strong SoFi Plus momentum, excellent cross-selling cadence, robust engagement, ramping student loan demand as payments resume or brightening tech platform momentum, things are going well.

More on Student Loans:

SoFi unsurprisingly expects the current acceleration in student loan volume to continue. It expects to be the primary beneficiary of that, considering its 60% refi market share and potential rate cuts fueling the fire (not needed but would certainly be welcomed). They see $280B in potential demand to serve, and again, thriving capital market demand for its paper to ensure it can capture these borrowers without the balance sheet ballooning in size.

Liquidity:

SoFi has an “internal stress buffer” beyond mandated capital ratio levels that gives it the required comfort to pursue all of its growth opportunities. Their 15.5% total risk-based capital ratio is well beyond the 10.5% minimum and that buffer. The balance sheet is in excellent shape.

Tech Platform:

There wasn’t anything new here, but I wanted to talk about this piece of the business a bit. Some folks think the Galileo and Technisys purchases were mistakes because revenue growth has underwhelmed for about a year now. I could not disagree more. We’ve well documented the four simultaneous headwinds that have raged for this segment: A cumbersome go-to-market change, a regional banking crisis, 5 points of rate hikes and diverting product innovation assets to fully integrate Technisys. But I wanted to add another item here.

Cost advantages are everything in banking and financial services. SoFi’s branch-less model matches the advantages of other disruptors. It then pairs that model with a bank charter to match the advantages of incumbent lenders. That combination is rare in its sector, and the tech platform deepens that rarity significantly. When the Technisys integration project is soon complete, SoFi will shed reliance on third-party vendors and the expensive contracts needed to run its front and back ends. It will own its entire tech stack, thus side-stepping hefty costs and creating an incremental revenue opportunity that I view as icing on the cake. That icing is beginning to taste much better, with recent commentary on pipeline and several late-stage deals coming (per Noto), but the main ingredient is the cost lead that SoFi enjoys over everyone in its sector. Nobody has the charter… the lack of branches… the digitally native operations… and the vertically integrated tech stack. 

This means SoFi can continue to pay more savings interest than most can. It means it can profitably offer more credit card rewards over time. It means it can offer significant discounts on loan interest rates. It means it can offer more travel and event perks under SoFi Plus. Generally speaking, it means it can rationally provide more value than the incumbents – which is where all of its market share gains are coming from. Owning Galileo is what makes SoFi unique. While I expect the tech segment to accelerate meaningfully next year, that’s not required for this to be a vital contributor. This is a bank and a tech company. Being both is what makes it special and is why it has compounded at such a robust clip across cycles and through loan moratoriums.

I think the Zscaler 101 section I wrote for last quarter is a great read to table-set everything I’m about to talk about in this piece. It offers product definitions and a 30,000 ft. view of ZS’s business. That can be found here. For folks familiar with the company, that will all be review and can be skipped.

AI Product News:

Zscaler introduced a slew of new AI products focused on seamless deployment, frictionless protection of next-gen assets and new layers of automation. Considering that AI agents are simply new assets needing protection and how viciously they consume data from various parts of the web, that means far more for ZS to secure. This makes it clear why they’re so focused on AI.

“AI agents are somewhat like users. We are the best party to provide secure access to AI agents.”

Founder/CEO Jay Chaudhry

The first tool is AI Data Security Classification. This adds what it calls “human-like intuition” to identifying sensitive data across 200 types of information. This accelerates the process of uncovering “unexpected sensitive data” and vulnerabilities. Rather than static if/then statements across a few categories determining if data posture is healthy, this makes that process far more dense and nuanced, without adding hefty cost of complexity. Simply put, this makes it much better at hunting and resolving data issues that can lead to harmful breaches.

Next, it announced enhanced GenAI security with more prompt visibility. This upgrades its LLM prompting “classification and inspection” to prevent malicious prompt injections. It allows customers to easily set rules and policies, with a powerful ability to guard against model prompting that violates them. This pulls from its data loss prevention capabilities and shields proprietary assets – without the need to deploy archaic firewalls that constantly fail.

Third, it announced AI Segmentation as an augmentation of its existing tools. Per the team, this is the “first purpose-built user-to-app segmentation product” with broad AI automation. User-to-app segmentation essentially means shrinking the network connection down to a single user and a single app, rather than one user gaining access to an entire network. The entire network entry invites lateral threat movement and massive security vulnerabilities. Zscaler’s default zero trust architecture + segmentation prevent that pressing issue.

Finally, it unveiled ZDX Network Intelligence. This provides a bird’s-eye view of internet service provider (ISP) performance. It uses a highly malleable connection network and outage data to refine Zscaler’s zero trust exchange efficacy. This helps performance and up-time, while access to all of this outage data naturally makes the product a perfect ally for “analyzing disruptive ISP issue trends and enabling faster remediation, as well as cost savings.”

AI automation is a key ingredient for SOC-level traction. Companies need to protect assets across the board… but they need to do so without costs exploding as a result. Zscaler’s (and other companies like CrowdStrike) AI innovation is enabling them to thread that elusive needle.

Zero Trust Everywhere – Expansion of Zscaler for Users to Zero Trust Branch and Cloud:

ZS debuted a unified appliance for all types of asset branches (campuses, factories etc.) to cut the need for expensive, dated hardware such as firewalls and virtual desktop infrastructure (VDI). This tool runs through a single console, thus extending coverage without adding complexity or cost of deployment headache. With a unified appliance, customers don’t need firewalls for every single branch. That’s a game-changer for many customers struggling with the costs and failure rates of that aging tool. This new product also unlocks secure, branch-level connections for contractors, giving them access to whatever they need. Nothing more… nothing less. 

Next, Zscaler announced its Zero Trust Gateway for Cloud Workloads. Notably, this pulls from its cloud connector product, which connects cloud apps and servers with a given network. This is specifically built for AWS, which has become a large selling channel for Zscaler. The product enables safe and responsible workload-to-network connections “in under 10 minutes without deploying dedicated virtual machines previously needed for this function. Like other products already mentioned, this shrinks the attack surface and risk levels by minimizing entry points for a potential bad actor and only granting access to specific, needed tools instead of an entire network. Legacy firewalls grant access to everything within it; this grants access in a far more granular, personalized and safe manner. To offer even more segmented, minimal access within cloud workload use cases, it added “microsegmentation.” This treats them all as individual assets.

“We have now done petabyte scale deployments for customer cloud workloads.”

Co-Founder/CEO) Jay Chaudhry

Finally, for Zero Trust Everywhere, it extended its Zero Trust exchange for business-to-business use cases. This has been a highly requested product from clients, and now it’s here. With this product, organizations shed the potential harm from connecting to other enterprises and leave it to ZS to make sure that is done safely.

More on Zero Trust for Branches:

This will be added to Zscaler 101 in the next earnings review. Zscaler’s Zero Trust for branches eliminates much of the need for software-defined wide area networks (SD-WANS). SD-WANS are sort of like a network control center that optimizes traffic based on best available connectivity option. It also often comes with security tools that serve as access gateways (using things like firewalls). That functionality approves traffic to enter an entire network. Once you’re in, you’re in. Meaning? A threat actor can freely move throughout everything within the network. This is what these companies always call “lateral threat movement.” Zscaler for Branches eliminates significant need for SD-WAN by no longer granting access to an entire network. Instead, it grants access for a user within a branch to a single piece, rather than the entire thing. And with microsegmentation (thanks to Airgap M&A) it can grant access down to an individual cloud workload while further grouping types of traffic and users within a branch.

This works for both on-premise and public cloud deployments. Just like ZS’s other Zero Trust products, it vastly shrinks the branch-level attack surface by granting access to only what is needed, rather than everything. This product, through “branch connectors,” can also easily plug into ZIA and ZPA to give that functionality to individual branches.

“Our customers understand that the future is not about firewalls and SD WANs. It’s Zero Trust.”

Co-Founder/CEO Jay Chaudhry

Red Canary M&A:

We got more information from Red Canary leadership on what makes this company successful and how it can help Zscaler pursue Security Operations Center (SOC) status more effectively. As a reminder, SOC is essentially an aggregated hub of security and information technology coverage for all needed assets in one place. It relies heavily on 3 things:.

Zscaler sees SOC as needing 3 things. The first two are: great security software and, access to scalable, organized, actionable data ingestion (which its data lake and data fabric enable). IT operations (ZDX Copilot) paired with SecOps (Risk360, Business Insights and UVM, exposure management and threat management) is the third ingredient. ZScaler already had some offerings here, but it was missing MDR. Red Canary’s MDR product is the final piece of the SecOps puzzle (massive piece of SOC), as it removes the talent bottleneck that countless Zscaler clients face when trying to fully cover their enterprises.

But what made ZS attracted to this MDR product specifically? Why not buy one of the many other players? Per Red Canary leadership, they’re just very good at their jobs. They ingest petabytes of data per day, complete 1 million investigations per quarter and uncover 60,000 threats per year. According to them, an enterprise’s SOC has a very hard time figuring out which of the 60,000 threats are actually there and how to prioritize them. Red Canary unleashes its purpose-built platform, its established investigation processes and its years of experience to gain a better understanding of the risk. They pull from all of their previous encounters and uncover clear patterns to more effectively and quickly fix issues. While agentic AI can be a great ally for effective MDR, Red Canary thinks of it as a “genius IQ-level toddler.” Sometimes they absolutely baffle you and fully handle investigations. But? Most of the time they just have a lot more to learn and need the help of its data and team.

Candidly, I think there are several capable players in the MDR space with similar capabilities. Still, Zscaler didn’t have anything here, and so was missing an important SecOps and SOC tool. Red Canary works well to plug that gap and will accelerate ZS leadership’s schedule to having a fully-functioning, end-to-end SOC offering by more than a year.

Per Adam Geller (Chief Product Officer), Zscaler’s positioning at the heart of network connections means it has a wonderful view of traffic and identities. They have the right to cross-sell many more tools with this lucrative role and the right to eventually be a client’s holistic SOC. It thrives with real-time security via Zero Trust Everywhere and boasts a constant, accurate understanding of security posture because of its central network position. It’s great at flagging and prioritizing risks in this regard. It’s great with configuration and hygiene analysis via vulnerability management, attack surface management and specific apps like Risk 360 to help proactively fix fragile infrastructure before a breach occurs. It has the needed data ingestion tools needed via its lake and data fabric to ensure needed, affordable access to 1st and 3rd-party data connections. It has data security tools like DLP and DSPM (already defined) to power its data security everywhere ambition. And now, with the help of Red Canary, it’s bolstering its agentic operations (SecOps products like UVM, Risk360 & more + ITOps products like ZDX Copilot) to accelerate its path to full SOC capabilities.

Quick Thoughts:

This was an encouraging event in my opinion. Zscaler is debuting the steady stream of innovation needed to allow it to compound revenue at a 20%+ clip for a very long time. This is not just a network security company… it does that, cloud, data and vulnerability management, now with identity tools too. The rapid traction for many recently-introduced products tells me their innovation is resonating and their expedient go-to-market fixes are all working.

Competition:

Co-Founder/CEO Jay Chaudhry did not mince words when asked about the competition. He rightfully said that competition has been claiming they can match its product suite for the better part of a decade. The issue? They’re really only securing user-to-network connections, leaving several gaps in coverage and not even doing that single thing all that well. They’re winning because their network-based suite and exchange blaze secure connections between everything… from users, to cloud workloads, to devices, branches, businesses, apps and agents. They’re also doing this while cutting customer costs by more than 50% post migration.

“These companies are kind of competition, but not really… they’re spinning up firewall technology on a virtual machine and calling it competitive. It’s not zero trust. They're trying to confuse the customer because they want to stay relevant. They don't want to get displaced.”

Co-Founder/CEO Jay Chaudhry

On Being More Expensive Than Secure Access Service Edge Tools Like Microsoft Entra:(SASE)

While Microsoft’s listed price doesn’t include license and directory fees, the price for its product vs. Zscaler’s Zero Trust Everywhere (ZTE) offering is significant. As a reminder, Secure Access Service Edge (SASE) provides an overarching, bundled suite of network security and performance tools. Zscaler will readily tell you that competitors call their offering SASE rather than Zero Trust because they do not emulate the granular, attack surface-minimizing traits of ZTE. He also pushed back on the idea that they’re more expensive. While price tags for ZTE may be higher than some offerings, the superior value proposition makes the true price far more compelling. Like in medicine, preventative care is cheaper than reactive care. Per Chaudhry, they rarely lose deals because of the relative price of their products.

Data Security and Vendor Consolidation:

Customers are increasingly drawn to ZS’s data products, which are another powerful lever for vendor consolidation, SOC pursuit, cross-selling and growth. These companies “don’t want 5 vendors to do data security.” As Chaudhry reminded us, data hygiene and posture management are already hard with 1 vendor – let alone 5. Their positioning at the heart of the network equips them with unique, scaled and high-fidelity data that customers are yearning for. 

Why Don’t They Have an Endpoint Detection and Response (EDR) Offering?

Chaudhry thinks ZS “does what needs to be done in the endpoint.” He doesn't feel a need to offer an EDR product to compete with CrowdStrike, as they effectively manage endpoint performance, data leakage and other core tasks. They seem to be satisfied with that product menu, although I’d still love to see them offer some kind of EDR product. Even if it’s bare-bones.

Why are Traffic Trends so Good?

There’s no secret ingredient here. The value proposition is simply resonating and creating robust traffic while its sector struggles to stay above 0% Y/Y growth. Its commitment to generous portions, hiking prices below the rate of inflation and offering reasonably healthy food is working. Its fixation on using data, intricate experimentation and evidence to guide its menu, throughput progress, store layouts, loyalty program and everything else is too. They don’t guess. They test… test… and test some more until it is painfully obvious if something will work or not. They’re too modest to say it, but Cava is winning because their team is world-class and executing better than anyone else in the space. They’re superstars in the highest-growth quick-service cuisine in the USA. Structural tailwinds paired with elite execution is a powerful recipe.

GLP1:

Early data from GLP1 medicine actually points to the drugs being net neutral for Cava demand. They’re pushing folks to seek out lighter options, which its 38-ingredient menu addresses.

Lots of Ingredients:

Cava utilizes nearly double the ingredients of a Chipotle, while the two structure ordering in nearly identical ways. This led many to believe throughput would eventually become an issue for Cava, but that’s not happening. Its top stores are approaching $6 million per year in sales and are still seeing throughput progress. This is related to targeted investments made in labor allocation optimization, as well as AI tools telling employees how much of an ingredient to make based on depletion rate and expected sales. This makes everyone more efficient and more capable of serving more people. They have a large pipeline of ideas left in the tank for juicing already healthy throughput levels even more. It’s impressive to see such a young company offer such a wide array of ingredient options and still execute at a surgical level. This unlocks the ability to service more tastes and feed more palates.

New Stores Outperforming:

As a reminder, Cava recently raised new store volume targets from $2.1M in year one and $2.3M in year two by $200,000 each. Flourishing new openings across all of its geographies prompted that change, as well as boosts to cash return and margin goals. Well? They’re probably not done raising this guidance. According to CFO Tricia Tolivar, she “imagines they’ll give updated information when it’s appropriate,” as all new openings are performing beyond the already raised forecasts. This just goes to show how strong the momentum is for this brand. It also means fringe markets that weren’t quite compelling enough to pursue now are, considering every single store is more efficient than they’re supposed to be.

Food: 

Chicken shawarma continues to test very well and will probably debut later in the year. They see a multi-year roadmap of food-level innovation to boost order values and appeal to more people. Some areas will probably include doing more with dessert and beverage, but even within proteins and bases, they have a ton of stuff left to introduce. As always, they’ll do so in a slow and calculated way, reducing the risk of poor debuts. They’ll also focus on adding more things without adding more complexity, like they did with the new pita chip flavor that simply requires using different, pre-blended seasoning on an existing item.

Tariffs:

Cava has great supply chain flexibility to shift some international sourcing to domestic providers. This should keep any tariff impacts quite modest. Not non-existent, considering they’re dedicated to sourcing some ingredients like olive oil from Greece… but still modest.

Some Thoughts:

The business is killing it and the mid-50s EBITDA multiple is beginning to look more attractive. I’m sticking to my desired mid-40s EBITDA multiple target to begin more meaningfully building out this stake. The valuation is the only part of the enterprise that I don’t love.

Broadcom creates & manufactures a slew of semiconductor-related equipment within data center, networking and industry-specific use cases. Chips and high-performance compute (HPC) can’t all be packed into the same corner of a data center. GPUs must be able to connect to one another to drive better bandwidth and performance, with faster, more efficient model training and inference to cut costs. This is where Broadcom thrives. 

It also offers a range of software tools, which significantly broadened out with its VMWare acquisition. VMware offers virtual, localized layers of software that sit on top of hardware. This allows the centralized hardware to run several different operating systems from the same place. The company, which is now a Broadcom unit, calls these “virtual machines” or virtual private clouds. By reducing hardware requirements, VMWare saves its clients money. 

This company does not compete with Nvidia in terms of designing GPUs. It does, however, create application-specific integrated circuits (ASICs) for more specialized workloads. It also makes accelerated processing units (XPUs) which are the high-performance accelerator subsection of ASICs. All XPUs are ASICs but not all ASICs are XPUs. These are often used to optimize data center, networking and GPU performance. In some cases, this can replace various needs for more generalized chips like GPUs. Furthermore, its core niche focuses on networking and connectivity, which competes with Nvidia’s switches and its SpectrumX networking product. Jericho and Tomahawk are the names of its Ethernet products.Tomahawk switches are purpose-built with bandwidth in mind and are better for data centers, AI and the cloud. Jericho routers are better for sectors like telecom and edge networks, although newer models also have AI use cases as well. Finally, it offers network interface cards (NICs) that bridge connections between endpoints and a network.

Broadcom trades for 38x forward EPS. That might fall by a turn following slightly outperforming profit and guidance. EPS is expected to grow by 36% this year (excluding VMWare M&A noise) and by 19% next year.

Semiconductor Solutions – AI Networking:

AI networking demand rose by 170% Y/Y as its Ethernet products are thriving. Customers are eagerly embracing this solution for the powerful inter-data center (DC) chip linking and DC-to-DC linking too. And based on strong AI networking demand (as well as XPU momentum discussed below), it expects FY 2026 AI semiconductor solutions growth to be similar to FY 2025 (so around 60% Y/Y). Much tougher comps? No problem.

To keep the momentum humming, it just announced its newest Tomahawk 6 product. This doubles AI networking speed vs. its predecessor and allows 100,000 AI chips to connect to fewer network layers to boost speed and performance of AI training. This is purpose-built for both adding more compute density (“scale-up”) to servers and linking more chips together (“scale-out”)

Semiconductor Solutions – XPUs:

XPU growth was above 10% Y/Y. As a reminder, it has 3 large hyperscaler clients with a serviceable addressable market (SAM) of $75B. The company has an existing 70% market share of that specific SAM, so it is confident in taking a large chunk of the opportunity. It also has 2 other hyperscaler “partners” that it has secured design wins that it expects to turn into tape outs (later-stage chip designs with higher probability of translating into real revenue) this year. And furthermore, it has 2 other new hyperscaler clients it’s “deeply engaging” with as of last quarter. While design wins are often easy to secure and casually allotted, AVGO is highly picky in what designs it’s actually willing to work on. They will only partner with high-volume vendors with strong reason to believe designs will lead to financial success. Per Hock Tan, they “continue to make excellent progress” with all 7 of these companies. He sees 3 or more of them deploying 1 million AI clusters in 2027 for frontier models, and they’re confident a lot of that will mean demand for its XPUs. They’re doing $60B in annualized revenue (using this quarter). The $75B opportunity on its own can significantly bolster that base, without considering the potential impact from the other 4.

Despite what it called a somewhat fragile macro, its customers are “doubling down” on their Broadcom commitments. And interestingly, that’s now being supported by a boom in inference demand. Last quarter, Broadcom told us that most of its demand was for pre-training models. That was different than many other players in the space, so it makes a lot of sense to see Broadcom seeing this transition. That inference acceleration could likely lead to an XPU demand acceleration later in 2026. And demand for training remains quite strong despite much more interest in inference and agentic reasoning models. It sees its XPUs as perfectly positioned to cater to both needs.

“I think there's no differentiation between training and inference in using merchant accelerators (like GPUs) versus custom accelerators… the fundamental value in creating your own hardware versus using a third-party merchant silicon that you are able to optimize your software to the hardware and eventually achieve way higher performance than you otherwise could.”

CEO Hock Tan

“Networking is hot, but that doesn't mean XPU is softening. It's very much along the trajectory we expect it to be. And there's no lumpiness. There's no softening.”

CEO Hock Tan

Semiconductor Solutions – Non-AI:

Tan called this segment’s revenue “close to the bottom” for the cycle. While the bounce-back has been slower than what it wanted, broadband, enterprise networking and server storage subsections all returned to positive Q/Q growth. Industrial and wireless are the main drags for this segment. It expects revenue growth next quarter to again be 0% Q/Q.

Infrastructure Software:

The company continues to effectively move large clients from license-based to subscription-based revenue. They also keep flawlessly up-selling customers to VMWare Cloud Foundation (VCF). This “enables the entire data center to be virtualized and customers to create their own private cloud environment on-premise.” It’s no longer predominately virtualizing CPUs in isolation. It is virtualizing CPUs, GPUs and every other piece of the data center. The product for GPU virtualization is called VMWare Private AI foundation (released in tandem with Nvidia). CPU booking activity is healthy, but that’s the more exciting growth vector in my mind. While 45% of AVGO’s largest customers had already adopted VCF as of 6 months ago, that is now up to 87%.

Just like last quarter, Broadcom poured some cold water on the idea that everything is moving to public clouds. In the age of AI, when sensitive data is moving at a torrid pace, some companies are prioritizing data privacy and slowing some public cloud workload migrations in favor of on-premise usage. They’re even seeing some customers “repatriate workloads from public clouds” in favor of this deployment option.

The effective integration is driving 10%+ ARR growth for its infrastructure software business overall, and that’s happening while AVGO surgically removes redundant costs from VMWare’s model. For context, quarterly OpEx is now $1.1B vs. $1.6B a few quarters ago and EBIT margin for the segment rose from 60% to 76% Y/Y. That’s how you integrate two large companies.

This company continues to admirably execute and take a large portion of the ethernet and XPU opportunities that AI is creating. I view this as the highest-quality hardware name in the Gen/Agentic AI space not named Nvidia. And just like for that elite company, this elite company will continue to post great growth and results as long as the AI hardware boom persists. Commentary from Jensen discussing miles to go in training and inference law improvement, as well as agentic models consuming 100x-1000x the compute of first-gen models points at the runway remaining long. Tan’s commentary on unwavering and rapid demand growth for its AI semi solutions does too. Great quarter from a great company.

I got some questions about whether or not DKNG was falling because Kalshi is rolling out sports betting to more markets. As I’ve said many times, prediction market makers being allowed to offer sports options is a large net positive for DKNG. It would mean it gains access to the other half of the USA and pays far lower tax rates with double the addressable market. And if ESPN can’t even take 10% of the sports gambling market in the USA in legal states, these companies probably won’t either. Kalshi is not why DKNG fell earlier in the week… Illinois is.

Illinois raised its progressive tax rate from 20% on the first $20M in revenue and 40% thereafter to 25% and 50%, respectively (again… prediction market legality would be great for DraftKings). The most disappointing move out of Illinois came when they first embraced that progressive tax format. This structure directly penalizes successful scaling and DraftKings and FanDuel specifically. Higher taxes in isolation are actually not that terrible for DraftKings. The company has demonstrated a seamless ability to recoup much of the profit hit via promotional and marketing cuts, and higher taxes on their own actually favor the big boys. They drive natural market inefficiency, which is easier to overcome with size. Progressive tax rates are a different story, and make Illinois a less attractive market for DKNG. They’ve been able to offset a lot of the headwind from the initial change, and they’ll have to keep offsetting things with the rates now raised again. This will surely push more gamblers to the black market, but it’s simply something DKNG will need to deal with.

Tax rates are probably going to keep rising in many states, but the progressive nature of Illinois policy fortunately looks like an anomaly at this point. There’s no real movement from other states to match it.

I think this news will likely prevent DraftKings from raising financial guidance when it reports Q2 results, but I don’t really expect this to lead to a downward revision. Illinois is 15% of the legal sports gambling market, and another 5% of incremental tax should have a roughly 0.7% impact on overall profitability for that segment. Illinois has no iGaming, so the overall impact will be even smaller.

With all of this said, I think stressing over rising state taxes is missing the forest for the trees. DraftKings is a market leader that is taking incremental share in a structural growth industry. It trades for 18x forward FCF, while FCF is expected to compound at an 82% clip for the next two years. And that growth doesn’t even include potential iGaming legalization for 86% of the population. DKNG has the top two brands in iGaming. Even if you cut profit estimates in half due to tax headwinds, which is wildly unrealistic, this would still be a growth company trading at a multiple I find more compelling than pretty much anything else out there. Taxes will rise. DraftKings will win regardless. That’s how I see things.

Adweek published what seemed like a recycled article about Amazon’s demand-service provider ad platform taking some budget away from Trade Desk. They cited an $80M migration that I’ve seen elsewhere and other clients shifting some, not all, of their spend to the e-commerce giant. All of this evidence is from before the end of Q1 and before TTD posted a great quarter with upbeat guidance. But? They struggle with the same conflict of interests that Alphabet has for years. They compete with pretty much all of their advertisers in some capacity and own a lot of supply too. This inherently motivates them to route demand to that supply, while TTD serves the entire internet, without owning supply and without that conflict of interest. Amazon will surely continue to win its share of advertising revenue as it aggressively pursues growth for its competing product. I just think the market is so giant, and the need for a non-conflicted vendor (which is only Trade Desk) is so massive that TTD will continue to steadily compound as far as the eye can see… Amazon or no Amazon. And furthermore, while Amazon is becoming a stronger competitor, Google is becoming a weaker competitor through de-prioritization and antitrust lawsuits. TTD competed admirably against a stronger Google for years and years. They’ll do the same against Amazon, in my opinion.

MELI is applying for a banking license in Argentina. This would unlock significantly more product flexibility. It would also give it freedom to run a deposit business on its own, without any third parties, and with more control over interest rates. In turn, this deposit base would provide more affordable access to capital to fund its thriving credit businesses.

MELI is lowering its free shipping threshold in Brazil from a $14.15 order minimum to $3.40. This is in response to increasingly localized, automated and efficient fulfillment capabilities and ramping competition from Asian competition. MELI has always had to fend off fierce competitors (like Amazon) and has done so admirably to date. I expect that to continue — thanks to things like this. But how can it afford such generous offerings? Its product ecosystem provides immense opportunity for cross-selling, which means more efficient, margin-rich operations with more opportunity to pass on value to customers. Doing so generates heightened engagement, loyalty and growth. That all pairs perfectly with its unique logistics footprint that allows it to delight customers with fast and affordable shipping. This delight is what has made MELI so wildly successful, and I think its commitment to leading its field in this regard is well-placed.

Starbucks named Mike Grams as their new COO. Grams, like CEO Brian Niccol, comes from Taco Bell. He spent 29 total years there climbing the ladder before taking over as their COO in 2020. Grams was named the EVP and Chief Store Officer of Starbucks this February, and now will add this role to his responsibilities. This is in response to Niccol wanting more progress at Starbucks. He sees good progress being made but they “need to move faster.” Other changes include combining its Global Coffee and Sustainability team with its Global Brand team. This is meant to communicate loudly and clearly that coffee is the Starbucks brand. It’s the main character. There were several other organizational structure and reporting changes at the senior management level.

Just like Niccol did with Taco Bell and Chipotle, I expect him to right this ship faster than pretty much anyone else could. The issue is that things were an absolute mess when he came in and Starbucks will take time to be fixed. As long as they keep making the right changes, which I think they are, I’m happy to be patient and slowly build out the stake.

Amazon is experimenting with humanoid robots to do deliveries.

Alphabet ended its integration with PayPal’s digital wallet (not its debit card). This hasn’t even been an option for users since April 11th. Considering that, PayPal knew this was coming when it offered guidance and when Chriss spoke at an investment conference a week ago. There was no mention of this having any impact on guidance, while Chriss spoke about the quarter going well.

In an interview with Bank of America this week, Palo Alto CEO Nikesh Arora told investors that the initial “shock” from tariffs has gone away and things are now back to “business as usual.” There wasn’t anything else new in that conversation.

Output data:

Consumer & Employment Data:

Inflation Data: