Max Subs — there have been zero changes to the portfolio since the mid-week update detailing my new holding. I’m also going to publish the Starbucks article on Monday. I reached the maximum article length for this newsletter platform and was not able to include it in this piece.

Broadcom creates & manufactures a slew of semiconductor-related equipment within data center, networking and industry-specific use cases. It also offers a range of software tools, which significantly broadened out with its VMWare acquisition. This company does not compete with Nvidia in terms of designing GPUs. Instead, it focuses on networking and connectivity, which competes with Nvidia’s switches and its SpectrumX networking product.

Broadcom’s VMWare acquisition is impacting several company metrics. It’s greatly benefiting overall revenue growth, as well as infrastructure software revenue growth. Even for the Q/Q comparison, a full quarter of VMWare contribution impacted growth materially. Finally, the acquisition is greatly hurting GAAP margins due to the M&A-fueled stock compensation, restructuring and integration costs.

Broadcom comfortably beat revenue estimates by 4.1%. Organic revenue growth was 12% Y/Y. AI-related revenue rose 280% Y/Y to reach $3.1 billion.

Broadcom trades for 35x this year’s earnings. Earnings are expected to grow by 14% Y/Y this year, and 23% Y/Y next year.

Infrastructure Software Solutions and the VMWare Integration:

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.

The integration is “going very well.” Broadcom has re-grouped about 8,000 different niche products into 4 foundational pillars to simplify go-to-market and cross-department communication. It’s also making “good progress” in moving all VMware revenue recognition to annual subscription models.

So far, 3,000 of Broadcom’s 10,000 largest clients have opted for these new products to better-leverage on-premise hardware. Most are signing multi-year contracts, which is leading to VMWare annualized bookings value (ABV) rising from $1.2 billion to $1.9 billion sequentially. While VMWare did about $3.3 billion in quarterly revenue during its last report as a standalone company, Broadcom sees cross-selling momentum pushing this quarterly revenue rate to $4.0 billion.

But VMWare is not just another revenue growth driver for Broadcom… it’s also a real profit driver as well. SG&A functions have been consolidated to eliminate monthly costs, and VMWare’s quarterly expense rate fell from $2.3 billion to $1.6 billion Q/Q. Broadcom sees this getting down to $1.3 billion by the end of its fiscal year. Growth and margin accretion are what M&A is supposed to deliver. This was a massive undertaking and has gone quite smoothly so far. Props to the team.

Semiconductor Solutions for AI:

Broadcom’s GenAI niche is predominately in networking revenue. Superchips 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, which is why the explosion of AI infrastructure (or when Nvidia’s Jensen Huang calls “AI factories”) is leading to its networking revenue rising to 53% of total for this bucket. All in all, it expects this rapid build-out of HPC infrastructure to power 40% Y/Y networking growth next year. This compares favorably to previous 35% Y/Y growth expectations.

Broadcom doubled the number of switches sold Y/Y as its collaboration with Arisa, Dell, Juniper and Supermicro on the new Tomahawk 5 and Jericho 3 switches went well. Switches allow individual GPUs connect to one another. Its Network Interface Cards (NICs) are the actual high-speed connectors to AI factories that enable vastly accelerated pace of data processing, model training and more. They extract as much efficiency out of switches and servers as humanly (or machinely) possible. While it currently offers robust 800-gigabyte bandwidth for AI factory connectivity, this will be doubled when it rolls out the next generation of its hardware. It needs to keep innovating here quickly, as Nvidia has made switches and networking a more prominent part of its own roadmap.

AI accelerators are another important area within next-gen data center connectivity and networking. These are designed as separate machines used to augment and bolster AI workload and data processing. Per the team, “networking these AI accelerators is hard, but Broadcom has the deepest understanding of what this takes.” That’s why 7 of the 8 largest AI clusters on the planet use Broadcom Ethernet connectivity.

“We are the AI accelerator of choice.”

CEO Hock Tan

Semiconductor Solutions – Other:

This was a great quarter. Despite real headwinds in its non-AI-related businesses, Broadcom’s networking positioning within this wave is quite compelling. Its VMWare acquisition has been flawlessly integrated and this company just continues to deliver strong execution. There are no companies in the space besides Nvidia currently enjoying a more noticeable uplift in AI-related revenue.

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) to ensure those interactions are optimized. It also publishes some greatly appreciated macro data on overall commerce spend.

GAAP margins & cash flow margins last Q hit by $1B in M&A charges. EPS growth was helped by expense discipline.

Guidance assumes stable macro headwinds, continued product traction momentum and a strong Q4.

Adobe Annual Guidance (first update since FY Q4 2023):

Adobe Second quarter guidance:

AI Momentum:

The theme of this call was AI momentum. Adobe has been cast aside as an afterthought in the GenAI race. To me, that’s a byproduct of where we are in the current monetization cycle for GenAI. Foundational infrastructure is being laid today to support the value-creating software and apps of tomorrow. 

While Adobe is firmly entrenched in the software and app side of things, it’s still finding early success with its GenAI features. Its Firefly GenAI model series trains on a massive sum of unstructured and structured data from documents, campaigns and more. And while GenAI certainly is bringing a wave of rapid change, it will still favor the incumbents that have the data and assets in place to actually make models and apps valuable. That’s Adobe’s strength. 

Examples of products to focus on here include Adobe Express (within mainly the creative cloud), which is a full-service tool that creates visuals, improve marketing content and augment campaign design. It comes with Firefly, which infuses GenAI into these processes for more automation and rapid iteration. MAUs for this product doubled Q/Q off of a very small base.

On the Document Cloud side of things, Acrobat AI Assistant is the centerpiece. This morphs the process of reading static PDFs into one where a user “talks to the documents.” It creates document summaries and can even automate the creation of presentations from these PDFs.

For the Experiences Cloud, its AI assistant within the Adobe Experiences Platform (AEP) is helping to improve its journey optimizer and experience manager products to make every consumer touchpoint more relevant and profitable for clients. Many more GenAI AEP apps are in the works.

“We're driving strong usage, value and demand for our AI solutions across all customer segments and seeing early success monetizing new AI technologies across our Digital Media and Digital Experience businesses.”

Adobe CEO Shantanu Narayen

More on the Document Cloud:

Acrobat link sharing allows for PDFs to be exchanged with a simple URL instead of attaching a PDF file. This allows for easier collaboration and more permission control over who can do what with a document. It’s a material positive for productivity. This quarter, link sharing rose 100%+ Y/Y, drove wonderful top-of-funnel traction, enjoyed 60% free monthly active user (MAU) growth and delivered strong free-to-paid conversion. Adobe Acrobat overall saw strong MAU growth.

Wins for this segment included AstraZeneca, Chevron, Florida and Illinois State Governments and Wells Fargo. Net new annual recurring revenue (NNARR) rose 24% Y/Y to reach $165 million.

More on the Creative Cloud:

Adobe released the latest series of its Firefly Image foundational model (FM) for automated image generation. It’s beta testing generative fill and reference images. Generative fill allows for image creation based on conversation. Reference images allow users to place images next to requests to better guide the token created from the query.

Adobe Lightroom is its product for photographers to easily edit and manage photos. Generative remove is a new GenAI tool for this product to eliminate unwanted lighting or objects from a photo to help capture that perfect shot. Adobe Premiere Pro is a suite of services for optimizing audio within videos. It’s working on several new GenAI augmentations of this product as we speak. This should help make its Frame.io offering (which is its platform for video collaboration) all the more useful.

Digital Experiences:

This was a rock-solid quarter for a company where sentiment was quickly souring. This should change the narrative at least for the time being. It shows how capable Adobe is when it comes to mining its massive database and creating compelling, unique GenAI tools from it. For a firm that was thought to be falling behind in this race, this quarter makes it clear that it’s one of the most aggressive software players when it comes to early GenAI monetization and direct subscription up-sells. The revenue contribution remains small, but is already ramping. Strong showing from a wonderfully boring, profitable, compounding machine.

CFO Chris Lapointe sat down for an interesting interview this past week. In it, he updated us on credit trends, how SoFi is progressing towards its 2026 targets and a lot more. I’ll cover that here, with another section on my updated thoughts towards the overall investment case.

SoFi remains committed to every single piece of the 2026 guidance it offered a few years ago. This calls for a mid-20% tech platform CAGR, 50%+ financial services compounding and reaching $0.55 to $0.80 in GAAP EPS.

A few thoughts on this. First, I find the financial services CAGR expectation to be both bold and encouraging. A big piece of financial service growth today is simply bolstering net interest income thanks to the explosion in its deposit base. To extend this runway, it will need to give these depositors compelling ways to use their funds. The multi-year CAGR tells you management sees that playing out.

Secondly, 2026 guidance assumes zero product launches and no contribution from anything that it doesn’t already offer. SoFi continues to get more comfortable with its underwriting credit card models, which should allow it to accelerate go-to-market and product innovation there in the coming quarters. That could easily be one of many sources of more upside. Wall Street doesn’t expect SoFi to get to $0.67 in 2026 like it thinks it will. I have no reason to think of this team’s expectation-setting as anything but candid. And with this layer of intentional pessimism baked in, I think $0.67 could even prove to be conservative. We’ll see.

“We are really confident in $0.55-$0.80 in 2026 GAAP EPS.”

CFO Chris Lapointe

Lapointe again told us about all of the Requests for Proposals with top U.S. and Latin American banks. He again said the conversations are going well. Sales cycles with these types of clients are long, but we’ve been hearing similar anecdotes from the team for well over a year. I’d love to see some real announcements made in the near future. 

At the same time, I’m not sure if Citi or another client would want to publicize offloading their tech stack to a competitor that is trying to steal market share from them. SoFi’s reiterated mid-20% growth guide likely relies on some of these deals actually closing. So? Don’t be surprised if we continue to see a mysterious acceleration here based on large client wins that just weren’t publicized. SoFi has to be careful to make future prospects comfortable with using competition for their banking core modernization.

More and more customers are using SoFi for their Primary Banking Account (PBA). This directly supports cross-selling, lifetime value, retention and overall financials. In a new disclosure, 30% of new members are now taking out a second product within 30 days of onboarding.

SoFi Invest momentum was also called “great.” Growth here has been masked by it being forced to exit the crypto business due to securing its banking charter. This could be another fantastic customer engagement and retention lever to pull if it continues to build momentum. 

SoFi continues to reject a whopping 80% of personal loan applicants (it is instead selling those lower credit worthy opportunities to others via SoFi Lantern). It’s leaning conservative amid volatile macro uncertainty. Variable-to-fixed refinancing is a big piece of SoFi’s personal loan bucket. As rates fall, there’s less demand for fixing floating interest costs, as those costs are likely falling. That would tell you rate cuts would be a headwind for personal loans. But more context is needed. The aggressive pickiness that SoFi is practicing today is leaving it with some pretty meaningful pent-up demand. If it got more confident in the forward rate path (not even confident in a lot of cuts coming), that would lead it to getting more liberal on approvals, originations and using its excess balance sheet capacity. I see that potential as more than offsetting the diminishing fixed credit demand. I also see rate stability and cuts as a somewhat counterintuitive boost to this business for this specific cycle.

Beyond personal loans, rate stability and/or cuts would directly bolster demand for its new home loan business and its primary originator student loan business. And as an aside, SoFi’s Galileo collects a large portion of its revenue from payment processing volume. As rates fall, velocity of money rises and that business should be helped (along with basically every financial service that it offers). I say this not to get anyone overly excited. I say it to offer more context on why I believe the stock’s weakness is macro-driven.

A key bear/bull debate lately has centered on how attainable SoFi’s 7%-8% life of loan loss rate peak truly is. Skeptics say they’ll breach that level, which will pressure fair values and its overall results. While that’s entirely possible, I don’t think it’s probable. We got more great context this week to hammer the idea further home.

First, Chris Lapointe feels “very comfortable and increasingly confident” in the life of loan loss rate targets. That’s nice to hear, but what about some data to back it up? Losses for its Q1-Q3 2023 cohorts (which are now seasoned enough for data to be meaningful) are delivering losses 20%-40% better than Q1-Q3 2022. They’re 40%-50% better than its previous loss rate peak of 8% in 2017. Between stringent underwriting, higher loss rate vintages maturing and an ability to sell delinquent loans if need be (like many financial institutions routinely do), this target feels quite safe.

Capital market demand levels are well above what it’s willing to fulfill. These loan buyers know rates are going to soon fall and want to lock in large agreements with assets currently offering higher yields. If this is the case, why not accelerate personal loan originations right now? If your capital market partners are eagerly awaiting more originations, that makes it sound like SoFi can collect more lending-related revenue without more balance sheet risk.

This is likely a matter of SoFi not being confident enough in the rejected loans to want to stick them on their capital market counterparties. It has spent more than a decade building a reputation for delivering quality loan pools at scale to buyers. It’s really not worth sacrificing that reputation based on wanting more revenue for the next couple quarters.

I’ve gotten a lot of questions on how my point of view towards SoFi has evolved over the last few weeks. With the stock now around $6.50 and the Qatar Investment Authority liquidating their investment (2% of SoFi shares) this week, I wanted to address those questions head-on.

First, after doing a bit of digging on the Qatar fund, the sale makes sense. The fund invested pre-IPO and held most of its stake in preferred shares that are now redeemed. It had a board seat, but was required to retain a certain amount of equity to keep that seat. I don’t believe its remaining common stock position (following preferred shares being redeemed this spring) was enough to make that happen. They would have had to go buy more shares on the open market. They’d already locked in strong profits on the preferred stake and were likely going to lose their board seat next year without a major incremental investment. Considering this, I don’t find this news alarming.

In my mind, nothing about the investment case has changed. The only difference is stock price and loud, usually uninformed hecklers making staying the course more challenging. SoFi today is weathering the current macroeconomic turbulence as well as anyone could expect. It has gotten significantly more conservative on lending amid violently fluctuating rate expectations, and who can blame them? Leaning into originations today is gambling with your balance sheet.

Nobody on the planet knows where rates will go over the next 6-9 months, and originating like things are certain and the backdrop is incredible is just not prudent. I’d much rather have SoFi be overly prudent, delay revving the origination engine for a few quarters, and leave the downside risk to lower revenue generation for 3-6 months. What’s the alternative today? Gambling. It’s hoping that things get better from here, using your precious balance sheet, and crossing your fingers that macro doesn’t deteriorate. A delayed revenue downside risk is far easier to digest than a balance sheet blow-up risk.

SoFi has done a great job taking that potential blow-up off the table. It has used its charter to drive considerable interest savings and net interest income thanks to lower balance sheet turnover requirements. It has begun to roll-off fair value premiums from some of its loans; it has found robust capital market demand at hefty gain on sale premiums to add more origination capacity without more balance sheet risk; it has remained steadfast in its commitment to 7%-8% life of loan loss rates.

And despite this conservatism, it’s still finding roughly 15% overall growth this year thanks to the breadth of its product offering. Financial services growth is rapid as balance sheet optimization unfolds, interchange revenue exponentially builds and other products gain more traction. A reacceleration for its tech segment has commenced as expected. It continues to explosively expand margins; it continues to manufacture more balance sheet capacity to be poised to pounce during the next fun part of a cycle; it continues to find robust lender and capital market demand for its credit; it continues to rapidly grow membership, products and deposits; it continues to execute.

Am I supposed to get angry with them because the stock is falling? Am I supposed to criticize a team because the macro backdrop is forcing them to play defense? Am I supposed to wave the white flag despite the multi-year compounding engine remaining quite compelling in my view? No, no I am not. I will harshly criticize this company if/when it gives me a fundamental reason to do so. Not when its sector is in the toilet, macro headwinds are raging, it’s faring as well as it possibly can and social media keyboard warriors are saying I told you so. They’ll disappear when the macro brightens, don’t worry. The negative sentiment will reverse. This company will be rewarded for strong execution. Growth will accelerate when lending becomes less volatile. That is my opinion and nothing I see in this firm’s data is leading me to want to do anything but stick to it. A falling stock is not evidence of a souring investment case. For thriving companies at fair valuations, it is evidence of a coiling spring.

Bulls and bears will continue to incessantly bicker about every fine detail pertaining to this investment case. I will continue to fixate on the body of evidence pointing to their 2026 target of $0.67 in EPS being reachable… or even conservative. If that’s the case, then nothing else matters besides my willingness to remain patient through the frustrating price action. In a diversified portfolio, there will always be stragglers like SoFi and standouts like CrowdStrike. SoFi the stock, not the company, is one of those stragglers today.

Finally, CEO Anthony Noto made a small open market purchase of shares on Friday.

Okta splits its business into three subcategories. Access management is by far its largest. The other two are governance and privileged access management (PAM). Access management serves as a gatekeeper for which identities and credentials are allowed to enter a certain environment. Governance gives clients a birds-eye-view of identities and access across various apps to optimize hygiene and observe any potential vulnerabilities. This ties very closely to access management and has been its most successful product cross-sell to date, as it is essentially an extension of the access management for corporate workforces. PAM is Okta’s zero trust approach to identity. It offers access only as needed, doesn’t offer consistent privileges to any devices, flags unfamiliar usage patterns and demands verification at every turn. Like Zscaler in network security, this prevents free, identity-based access to an entire software stack after penetrating the most vulnerable piece of it.

This isn’t an exhaustive list of its products. It also offers posture management to observe any misconfigurations and proactively flag issues among some other tools. Still, the products already mentioned encompass all of the revenue drivers today. And for PAM, as well as all other new products in the works, it’s still in product market fit mode. It wants these all to be best-in-class before getting aggressive on selling. It isn’t there yet.

Within access management, Okta further splits its product buckets into workforce and customer management. Workforce is the access management broker for Okta’s clients; customer is its access management broker for the customers of Okta’s clients.

Macro remains a headwind for Okta. Things aren’t getting better or worse, and the guidance assumes that remains the case for the rest of the year. It is “still a heavy headwind” and is (per the team) materially holding back its current growth rates.

The biggest area of weakness for the company has been within new logo acquisition. That makes sense in an uncertain world, as up-selling more products to existing customers is simply easier than signing and onboarding a brand new customer. Still, we’ve heard the exact opposite from CrowdStrike, SentinelOne and Zscaler within other complementary areas of cybersecurity. Just a weird backdrop impacting everyone differently. The majority of Okta’s new business is coming from cross-selling more products like governance and PAM.

Similarly to everyone else, the most pronounced weakness is within the small and medium business segment.

Okta has endured some high profile security breaches. As a cybersecurity company, that’s not really great for business. These things happen to Microsoft pretty frequently, but Okta isn’t Microsoft. Okta doesn’t have a world-class enterprise bundle to make second-rate security products good enough. It must be better than Microsoft to overcome that fortress of a moat.

This year, it is investing heavily in its own infrastructure and revamping its security controls. Fortunately, it hasn’t seen any real impact in terms of customer win rates or pipeline activity. Still, that doesn’t mean prospective, pre-pipeline customers aren’t a tad more hesitant to approach Okta. That’s not necessarily happening, but it’s possible and is why this area of focus is so important today. The foundation must be strong.

Another priority is “reigniting growth.” That will need to wait (at least partially) for the macro backdrop to brighten. Still, there are levers it can pull to help momentum in the short term. Customer access management, governance and PAM are all in the very early stages of broad adoption.

To nurture revenue growth, Okta wants to collaborate more with its partner ecosystem. It thinks there’s a lot to do with cloud service providers like AWS, where it enjoyed 130% Y/Y growth in contract value last quarter, reaching $175 million. Its Google Cloud partnership is brand new, and it likely won’t be partnering with Microsoft Azure in the same capacity due to the direct competition.

It also thinks it can do a lot more with Global System Integrators (GSIs). Okta’s access management product as a core offering is a gift and a curse. It is extremely easy to turn on and start to use. That’s amazing for customers, but GSIs ironically want this onboarding work to be tougher. Difficulty here for customers means more professional service revenue for them. Okta isn’t going to intentionally add complexity to onboarding, but broadening its product suite will have that intended outcome without annoying its paying customers.

Finally, Okta is now embracing a “hunter-farmer” model for its American SMB go-to-market. It is splitting the roles and incentives tied to winning new customers and up-selling already won customers. This will allow salespeople to focus more strictly on products that power its top-of-funnel, while others can specialize in PAM, seat expansion and other products motivating cross-selling. This added focus should mean more capable salespeople and more relevant customer touchpoints. They’re a quarter into this change and are “optimistic that it will be accretive to logo count and growth.” It could eventually expand this evolved model to the rest of its business. It thinks it has the sales capacity in place needed to implement this change and power go-to-market for now. It could potentially begin to accelerate hiring later in the year.

CFO Brett Tighe was quite candid in his self-critique of their customer identity access management go-to-market. 40% of its access management business comes on the customer side, and it sees a real opportunity to push that to 50%. The main competition here isn’t Ping and Microsoft like for workforce access management. Instead, it’s home-grown solutions. Okta must make the cost and value benefits from using it vs. building internally more apparent. It thinks those benefits are already material, but communication needs to get a bit better per the team.

I view the identity access broker space as the most mature piece of next-generation cybersecurity. The opportunity is smaller and more tapped than network and endpoint while the obvious avenues for cross-selling are less numerous.

With that said, I’m impressed by how Okta has executed over the last year. It has rationalized its cost base, delivered explosive and dearly needed operating leverage and is still generating respectable growth against this backdrop. It’s in the same conversation as Block, Airbnb, Shopify and Uber in this regard. At 36x this year’s earnings and a forward 2-year earnings CAGR of 30%, this does seem like a somewhat compelling growth/value hybrid in a space that surely still has growth left (even if the runway is shorter than in network and identity).

If there’s one thing the semiconductor industry loves, it’s constantly changing the names of products with an alphabet soup of acronyms for us all to juggle. Fun, fun. Those acronyms all fall into neat categories: chips, networking and connectivity, and software. It’s these ideas and AMD’s positioning within them that matter to investors. Not that they’ve memorized what an MI325 HBM3E chip stands for. That’s how we’ll frame this coverage. I’ll use Nvidia as the measuring stick based on its current technological prowess and considering that every question in this interview included Nvidia.

AMD leadership is adamant about it continuing to lead the neutral processing unit (NPU) race for AI-infused personal computing. “TOPs” stands for Tera Operations Per Second. This measures chip performance with more TOPs being better. AMD’s new Ryzen AI PC has 20% more TOPs than Microsoft’s own best unit. The PC includes a superchip with its latest CPU, GPU and NPU all in one, To them, TOPs superiority is imperative for running copilots and GenAI apps on personal computers with optimal latency, hallucination rates and performance. It’s how the firm claims to be a “leader in AI inference on the PC side.” The semiconductor space is a lot like the cybersecurity space in that vendors all think their technology is better than everyone else’s. In reality, they’re probably all right for specific use cases.

Within AI data centers, its new CPUs “extend leadership in performance per watt and dollar significantly.” GPUs are going to be the main workhorse for high-performance compute workloads. But CPUs still work just fine for static, step-series, instruction-based tasks. And when a CPU can be used instead of a GPU, cost optimization routinely favors the CPU. For this reason, it continues to focus on boosting core counts (or adding more processing power to CPUs) rather than unit volumes here. Chip unit volumes are challenged, but CPU core unit volumes look a lot better.

GPU (and software) is where Nvidia is thought to have the largest, most defensible lead. Like Nvidia, AMD is also shrinking the timing of new platform delivery from a few years to 12 months. Later this year, its newest MI series chip will feature “better memory and bandwidth than the competition.” Its 2025 release will boost those metrics by 35x and will “compete with Blackwell 200.” This is great to hear, but the issue is that Nvidia will have another chip after Blackwell that delivers more exponential efficiency gains next year. It’s very hard to catch Nvidia here, but AMD thinks it is making some headway. That Blackwell successor will be called Rubin, and AMD thinks it will get to these Rubin levels of performance in 2026. That is likely when Rubin revenue will start to ramp.

“We believe we are very competitive on the GPU side.”

CFO Jean Hu

Hybrid bonding and chip-on-wafer-on-substrate constraints continue to ease. There are still bottlenecks, which will likely last into the second half of the year.

Nvidia has made rapid progress with its NVLink switches and SpectrumX technology to blaze deeper, larger connections between GPU clusters. It’s working with Microsoft, Meta, Google, AWS, Broadcom and Cisco on a new open standard framework for linking up to 1,000 GPUs. SpectrumX can connect 10,000 GPUs.

AMD doesn’t want to develop all of this internally. It will lean on partners for networking (including Ethernet) to emulate the full service suite that Nvidia offers. As we covered in the Broadcom section, there are other very formidable players here that can help AMD close the gap without doing so on its own.

Its new software stack release comes with a broad suite of models, tools and MI chip integration help. Nvidia’s Cuda software suite has become immensely popular. Utilizing Cuda and AMD’s NPUs at the same time requires a lot of manual work and hinders adoption of AMD’s chipsets. AMD is pushing hard to lower the friction associated with porting Cuda applications to AMD’s chip framework. There is vendor lock when it comes to Nvidia’s Cuda, and AMD needs to proactively overcome that vendor lock for more intuitive back-end integrations.

The slow growth economy that AXP saw through Q1 has been stable so far in Q2. CFO Christophe Le Caillec reiterated the billings and 9%-11% revenue guidance offered on the last call. It’s seeing strength within its consumer niche (which skews very affluent) and a bit of weakness on the small business side of things.

American Express now has a 25% market share of this segment vs. 20% a few years ago. There’s a lot of competition, but it keeps taking more market share. It continues to expect to accelerate volume growth here through the back half of the year.

A big piece of this confidence stems from the accelerated cadence of product refreshes since the new team took over a year ago. These product refreshes continue to work and continue to deliver more pricing power on annual subscription fees. Encouragingly, price elasticity of demand remains strong. Per the team, its only regret about previous price hikes was not hiking more aggressively. Retention has been sky high in response to these changes.

Comments here are a solid read-through for the ultra-prime portion of credit health.

Its 750 average FICO for GenZ and Millennial customers continues to rival other ultra-prime credit issuers like SoFi. Its average FICO for these younger customers is on par with the average FICO for GenX clients with other major card networks.

American Express has also gotten very good at nudging customers with loan offers when they make big ticket purchases with its cards. They know customers better than anyone else, and so can more confidently offer financing. Based on its wonderfully robust 1.3% loan write-off rate (and a strong 2.1% write-off rate for credit cards) as of last quarter, this is working. AmEx sees these rates worsening over the course of the year, but remaining at or near best-in-class.

Progyny’s team again took to the stage to update investors on how things are going. We learned nothing new vs. the last chat. Still, based on utilization rate and medication mix volatility recently impacting results, I thought it would be valuable to reiterate that same update now a handful of weeks further into the quarter.

The utilization trough that it saw in March continued to recover through April and May. The Jefferies analyst cited his channel checks from April, which CEO Peter Anevski confirmed were directionally correct.

Utilization is not back to previous peaks, but it is steadily recovering. That should continue. The most rational explanation for the March dip was Alabama ruling that discarding embryos was wrongful death. This technically made IVF treatment illegal for a few weeks and led to utilization softness that was most pronounced in socially conservative states. Encouragingly, Alabama needed just three weeks to exclude IVF from this new law, and key members from both parties have vocally supported fertility treatment. The way to reverse falling birth rates is not to prevent the larger portion of women waiting longer to have children from seeking needed treatment.

Aside from this, Anevski again offered optimism on how favorably every key metric is tracking vs. last year for the selling season. It teased a few more health system channel partnerships that should soon be announced to join others like CVS and Blue Cross Blue Shield.

Venmo Checkout was added to eBay. PayPal needs Venmo to monetize more effectively and branded checkout volume is a great way to accomplish this.

KeyBanc released some Netflix research this week pointing to strong membership growth and engagement trends. It sees 15%, 13% and 11% growth for 2024, 2025, and 2026 respectively. This is slightly ahead of consensus for 2024 and 2026 and about 1 point ahead for 2025.

DA Davidson initiated Duolingo with a buy rating. It sees the name as best-in-class in the space, with the best positioning to capture GenAI monetization down the road. 

JP Morgan sees Amazon continuing to take e-commerce market share gains in its State of U.S. E-Commerce report. Amazon is gearing up to invest “billions” in Taiwan cloud infrastructure. Finally, Amazon and Vrio agreed to a deal where Amazon will provide satellite internet across South America as part of Project Kuiper. This project is becoming a real potential revenue driver. Vrio joins Vodafone and Japanese telcos as early wins for Kuiper.

JP Morgan also issued a bullish initiation on Shopify during the week. It sees deepening competitive leads and ample room for margin accretion. Evercore ISI upgraded Shopify to match uniformly bullish channel checks from sell-siders recently. The lone bearish note on Shopify came from Moffett Nathanson. It sees risks to cohort retention this year, although remains overall bullish on the company.

Uber signed an interesting deal with RedCap by Solera to offer courtesy rides and parts delivery to 5,500 more dealerships across the nation.

DraftKings re-appointed Erik Bradbury as its Chief Accounting Officer. Erik was most recently the Chief Accounting Officer for IAC. Needham issued a very bullish DraftKings note this week calling DraftKings the leader of the $35 billion North American online gaming industry.

During the meeting, the May consumer price index (CPI) data came in below expectations. M/M Core CPI rose by 0.2% vs. 0.3% expected; M/M CPI rose by 0% vs. 0.1% expected; Y/Y core rose by 3.4% vs. 3.5% expected and 3.6% last month. After the meeting, the May producer price index (PPI) came in at -0.2% vs. 0.1% expected.