The Goldman conference is always busy. I’ve already sent coverage of the PayPal, Amazon and SoFI interviews. I plan to cover several more for Saturday. There are 15 left that I’d like to get to (Alphabet, Cloudflare, Visa etc.), along with Oracle and Adobe earnings and the rest of this week’s news. Realistically, some of the 13 will be pushed to next week.

Shopify has consistently grown its take rate for the last few years through the successful introduction of more merchant solutions. These solutions (payments, cross-border, marketing, credit, tax etc.) round out the Shopify product suite to ensure customers have everything they need in one place. It then partners with important industry players (like Amazon on fulfillment) to plug any remaining product gaps.

The two largest opportunities for Shopify to grow attach rate from here are in global expansion and large enterprise adoption. For its more established small and medium business (SMB) segment, there is still plenty of room for take rate gains, just not quite as much.

Shopify’s aggressive enterprise push, with Shopify Plus upgrades, Commerce Components by Shopify (CCS) and headless (back and front end separation), is really only two years in. Sales cycles take about 9 months and onboarding cycles take another 9 months. Shopify’s new clients usually start with fewer products and add more from there. All of this means the growth contribution from larger enterprises should begin to ramp starting now. Very early here. On the international side, take rate gains will be driven by debuting more of its products in more geographies. Most of its current markets do not have the full suite available. Getting closer to that point will naturally propel take rate higher.

These two items, as well as future inevitable product launches, bode well for 3.03% being a positive step in a longer, beneficial journey. Why does this matter? It means revenue growth can continue to lead volume growth in the years ahead as its overall opportunity continues to mature; Shopify can extract more financial value from the existing business. Furthermore, considering take rate is tightly related to pace of cross-selling, as take rate rises, Shopify’s margin ceiling does too. It costs far more for it to onboard its first product with a new merchant than any incremental product thereafter.

Large companies generally work with SIs like Accenture and EY when embarking on large overhauls of their technology stacks. Shopify hadn’t done much with SIs for years. It relied on word-of-mouth and perhaps was a bit too proud to embrace this channel. That has since completely changed. SIs are now an instrumental part of its growth engine and a vital cog in its ability to grow awareness and purchase consideration as the largest companies in the world select vendors.

During the interview, Hoffmeister told us that when Shopify started its 2022 enterprise push, “the pace at which basically all big SIs pivoted to it was really impressive.” It works with all of these big players, with one having 2x the number of its own employees trained on Shopify than Shopify has on its own. These SIs, based on the data Hoffmeister continues to observe, are advocating for Shopify’s suite as the best tech platform out there.

Shopify is not currently interested in monetizing AI as a separate, standalone package. Instead, the company is fixated on ingraining AI as deeply into its current suite as possible.

This is generally the approach the company takes to new product offerings like Managed Markets (for cross-border) and others. It takes its time in driving product-market fit and value creation. Then and only then does it monetize via more merchant solutions or subscription price hikes. AI is following the same pattern for Shopify. For now, the value creation focus is on nudging merchants to best course of action for various tasks. It’s also about using world-class models to automate marketing content creation, customer service, custom pricing, web design and so much more. Shopify’s AI suite (called Shopify Magic) is not meant to run a business for a merchant. It’s meant to offer data-driven ideas on how to optimize every aspect of their business.

Shopify’s macro view has not changed since the Q2 earnings call. It’s not seeing consumer spending or the overall environment improve or worsen. Its continued financial outperformance is not a byproduct of an easier backdrop, but structural market share gains. Its product suite, go-to-market, geographic diversity and team are all facilitating relative resilience vs. everyone else in its space. More of the same for Shopify.

Shopify has gotten significantly more intentional, sophisticated and precise with marketing since Hoffmeister took over two years ago. It now can track key performance indicators (KPIs) in real-time to lean into or away from channels as performance evolves on a daily basis. That has unlocked significantly more opportunities to invest in high-return marketing channels and has made Shopify more efficient from a lifetime value to customer acquisition cost (CAC) perspective. This is a testament to the systems that Shopify created to support this capability, and is broadly expected to lead to accelerating merchant gains in 2024, with revenue contributions ramping in 2025.

Hoffmeister also said he’s seeing the pattern of adding more merchants continue, partially thanks to better marketing. This points to last quarter’s fantastic new merchant result not being a one-off event. That’s especially important and positive considering its decision to shorten free trials during Q2 propped up new merchant adds for that period.

This improved tech does mean Shopify will invest more than expected in marketing if returns warrant that decision from time to time. Still, there’s a reason why it guides to OpEx as a % of revenue (as Hoffmeister said). That’s Shopify communicating its commitment to ensure it maintains spend discipline.

According to new Nielsen data, Celsius saw 30% volume growth and a 19% price cut through the month ending August 24th. That’s its best result since mid-July on volume. On pricing, Celsius leadership has consistently spoken about more promotional programs and bundling as of late to drive market share growth. It does have some control over that even though it distributes through Pepsi. Still, while some declines are unsurprising, the degree of decline was somewhat notable.

This data doesn’t change much for me in terms of my thoughts on the investment. To avoid repeating myself, those thoughts can be found here.

New CEO Brian Niccol penned a letter to investors describing his priorities and vision for Starbucks. After several weeks of touring stores and chatting with team members, he sees what we all see: a beloved brand that isn’t executing. In his words, he called this “drifting from its core.” To address the company’s issues, he is implementing a “Back to Starbucks” campaign. Chipotle investors are probably laughing at that name. Niccol, who comes from Starbucks, is known very positively for his back to basics Chipotle campaign that proved so successful over the years. He flawlessly executed there and seems to be determined to run a similar playbook with this coffee giant.

There are 4 priorities that make up this playbook. First, Starbucks aims to “empower baristas to take care of customers. That means giving them the “tools and time to craft great drinks.” Part of this will be via motivating employees with better career opportunities to drive higher employee retention and effort. This sounds like him addressing both the throughput and product consistency issues that Starbucks deals with. Along the same fixing product consistency lines, the second item is getting “every morning right.” This means consistently high-quality goods in an acceptable amount of time.

Thirdly, it will invest in its stores to “re-establish Starbucks as the Community Coffeehouse.” The in-store experience has significantly worsened in some parts of the U.S., and Niccol wants to fix that. Part of this will be better organization of to-go and for-here lines to cut wait times. Finally, it wants to “tell the Starbucks story.” This is likely his aim to repair brand damage across parts of the world while reigniting growth in China, although the initial focus for all of this work will be in the U.S.

To fix all that is broken, there are more investments in technology, its supply chain and its mobile ordering platform coming.

Since starting the position in the high 70s in June and sending out the investment case, the stock has performed well on nothing other than excitement surrounding Niccol. Despite this price appreciation being hype-based, I actually think the hype is warranted. That’s why I haven’t trimmed into the recent multiple expansion. Niccol is a world-class CEO joining a world-class brand is in desperate need of… well… a world-class CEO. The same issues that he fixed at Chipotle are the issues needing fixing at Starbucks. I like the setup.

When Airbnb IPOed nearly 4 years ago, its S1 featured an “experiences” business segment showing how the firm could “expand beyond the core” offering. At the time, the segment was already 4 years old and was a very small piece of the overall business. That was in 2020. With its experiences business now being 8 years old, the same is still true. This is partially related to still searching for product-market fit and also focusing away from this segment during the pandemic to ensure its core business was in great shape.

During the interview, Airbnb CFO Ellie Mertz told investors that its next two product release events will feature “real movement towards offering more services and products to deliver incremental revenue growth streams in the years to come.” Airbnb is gearing up to finally lean back into this segment. It thinks it has made great progress in tweaking its product suite here to drive real traction and incremental volume.

It’s frustrating that this has taken 8 years to be potentially ready for meaningful, sustainable go-to-market… but it’s also exciting. As leadership will frequently tell you, Airbnb has built a massive book of business with only one product. Its massive base of existing customers is ripe for cross-selling other services; the company is ready to try taking its piece of that opportunity once more.

The most intuitive area for core product expansion is offering more booking services for other parts of the travel experience. This could mean more direct help with transportation, restaurant reservations, suggesting local experiences, etc. It has done some things in this area, like a partnership with Resy for in-app reservations, but it hasn’t been a focus.

According to Mertz, Airbnb’s small presence here over the years has been deeply valuable. Airbnb has gained significant data, insight and experience, which makes efforts here thus far anything but time wasted. They now know that hosts need more help and guidance on which services to offer and how to offer them. The team also learned that any ancillary bookings business needs to be more thoughtful about timing up customer promotions and nudges. These will be big parts of the incoming, re-vamped experiences offering.

“How do we get in front of the consumer at the right time? How do we have the right product experience so that we can easily encourage you to attach [more services] to an existing sale that you're probably not booking at the same time? And so that really informs the product strategy of how we, in the future, go about expanding that product to scale.”

CFO Ellie Mertz

Beyond more services for guests, “expanding beyond the core,” also means doing more for hosts. This could mean things like listing help or using its GenAI partners to help with customer service tools like auto translation. As an aside, customer service is where Airbnb expects its first tangible GenAI benefits to be realized.

Another service that Mertz teased was sponsored listings:

“When we talk about expanding beyond the core, it's not just what can the guest purchase from us. It's also how can we help the host be more successful? How can we help them with services? That could mean how do we allow hosts to pay to meet their earnings objectives. Some people presume that Airbnb is allergic to Promoter Listings. That is not the case. We just have not prioritized it yet.”

CFO Elli Mertz

For now, this upcoming re-launch means more fixed cost without more coinciding revenue. That’s why Airbnb guided to a 35%+ EBITDA margin this year vs. about 37% last year. It wanted to leave plenty of room to invest in this opportunity to potentially accelerate revenue growth in the years ahead. Mertz is “proud” of the fantastic margin expansion Airbnb has delivered over the last couple years but wants to “drive higher growth.” It guided to 9% revenue growth for next quarter.

Tough Y/Y comps for 2024 and the law of large numbers both help explain part of the revenue slowdown, but not all of it. Airbnb’s expected 9% growth next quarter is a byproduct of weakness in long lead time bookings. Last minute bookings trends remained healthy. To Airbnb, this means its customers are becoming more uncertain, and so less willing to book well in advance. It wanted to bake some prudence into its guidance for this reason.

They’re cautiously optimistic that long lead time bookers will eventually come back. Still, that hasn’t happened yet, and it didn’t want to rely on that happening to meet guidance.

It’s full speed ahead on expanding its under-tapped market playbook to more countries. Outside of North America and a couple other developed nations, Airbnb sees a massive opportunity to grow adoption, market share and revenues.

Non-urban travel trends remain better and more sustainable than most assumed for post-pandemic times. Airbnb attributes this to heightened awareness and a differentiated offering. In a lot of these places, there aren’t even hotel alternatives.

On GenAI, Mertz told investors that a “lot of the transformative applications are going to take longer than maybe we all expected.”

The main message from CFO Margin de Los Santos’s interview was that Meli’s growth runway remains miles long. For its core business, which delivered 30%+ growth across its markets, e-commerce penetration rates still materially lag the U.S. and Western Europe. That represents a direct opportunity to enjoy more structural growth. Beyond that core commerce offering, its fulfillment, financial services, advertising and other segments remain very early on in their growth trajectories. It sees itself becoming one of the largest digital advertisers on the planet, as well as the largest fintech in Latin America (2 horse race with Nu).

This is not breaking news… but I enjoy hearing about how untapped opportunities are whenever a leadership team wants to say it to me.

To capture more of its large markets, Meli is getting more aggressive on marketing partnerships. 60% of its traffic remains organic, which depicts how strong its brand is, but also that there’s probably more opportunity to spend productive marketing dollars. During this year’s Copa America soccer tournament, it racked up 300 million paid impressions and saw concretely positive impacts to its brand awareness. It’s confident that these additional marketing dollars were dollars well spent.

Last year, MELI rebranded its loyalty program to Meli+. This wasn’t just a name change. The news coincided with some changes to and augmentations of the product. First, as we’ve discussed in the past, it greatly built out its fulfillment network and laced exclusive network perks right into the program. One of these perks is MELI delivery day, which allows customers to set a day in the week to receive all shipments with a minimum free shipping order size of just $6. These benefits complement exclusive deals and promotions on its commerce platform and Disney+ content access quite nicely; there’s also more to the program. Most recently, MELI introduced financial service loyalty program perks. These include credit card cash back, higher deposit yields and more.

The rounding out of the Meli+ program has consistently delivered intended incremental retention and lifetime value (LTV) gains. Beyond this revenue creation lever, better packaging of the program has also helped with retention and LTV. The company moved from offering 6 tiers of the loyalty program before rebranding, to a single tier afterwards. Eventually, it realized that many didn’t care about the Disney+ content aspect of the offer and really only wanted the other things. As a result, it now offers all benefits excluding Disney+ for $2/month and everything together for $6/month. The $4 difference is what it pays to Disney for access. The $6/month plan subs offer ever higher LTV benefits, making this partnership worth it for Meli.

Meli doesn’t expect any major CapEx from investments in GenAI. It plans to use open source and partner models to train with its own data. It’s not playing the game of build the biggest model. Early on, GenAI has been quite helpful in credit underwriting, real-time customer service and merchant product listing.

There was a bit of concern stemming from an intense MELI CapEx cycle. Meli plans to build 9 new facilities (some already built) across its footprint this year. While that sounds like a lot, it’s just more of the same for this massive entity. The proportion of MELI-fulfilled orders still being at just 70% in its most developed market also directly shows us it has significantly more demand to meet this incoming supply. This is “business as usual” and the team has no plans to significantly raise CapEx as a percent of total revenue. It's decision to start disclosing free cash flow last quarter is another hint of this being the case.

CFO Martin de Los Santos was also asked about potential margin dilution stemming from all of these new facilities. They take time to ramp to full capacity. He reminded us that Meli has been growing its network (and free shipping rates) for several years. Based on “focus on efficiencies,” it has been able to grow margin throughout that period. He sees margin expansion continuing in the future. That’s news.

A lot of these fulfillment investments are in its local metro centers. These have worked “extremely well” to expedite delivery times and ease bottlenecks throughout the rest of its network. It’s opening two more of these facilities in Brazil.

Meli is now the 3rd largest Latin American advertising platform, after Alphabet and Meta. Almost all of its success to date has come from product ads. The opportunity to develop brand and display advertising relationships remains in inning 1. Santos candidly told investors that the firm “needs to do a better job” here in terms of communicating to them all of the value MELI can provide. It has all of the needed tools in place; this is about developing go-to-market. If history is any indication for Meli, just give it time.

It also sees a real opportunity to distribute its advertising demand across the content landscape. Its first dip into this area comes in the form of a Disney+ partnership that adds the content creator inventory to Meli’s ad inventory. It plans to do this with other content platforms down the road as well.

"This is the first time we’ve gone outside of our ecosystem on ads. We think it’s the first step of many to come.”

CFO Martin de Los Santos

Meli’s oldest credit card cohorts are now becoming directly profitable on schedule. This is a great achievement and proof of concept for its underwriting models. Furthermore, the milestone does not even include the financial benefits of cross-selling. Its credit card and financial service users place more marketplace orders and contribute more revenue vs. non-users.

Meli remains skewed towards growth mode and away from maximize profit mode. For long term shareholder value creation, considering how large its opportunities are, this is the correct approach. It does think it can continue delivering operating leverage in the years ahead. At the same time, per Santos, “in the short term MercadoLibre has many growth opportunities it doesn’t want to miss.”

It’s broadly known that Nvidia has the best GPUs in the market for accelerated compute and GenAI… by a wide margin. What’s less broadly known is how it got here. This wasn’t luck or chance, it was decades of preparation for this very moment. Nvidia invented the GPU 25 years ago, and it has been racing on innovation ever since.

It’s ability to lead this lucrative race in 2024 is a direct byproduct of the architecture that Nvidia has built over time. This architecture is what allows it to move so quickly (without any bottlenecks). It’s designed with consistency in mind and in a way that fixates on compatibility with older software and hardware. This means the new products built today can use older software and “accelerate it” (modernize it) to run on its latest and greatest platforms. Per Jensen, this commitment to consistency is what “protects the investment of software developers” to further motivate them to build with Nvidia’s tools.

Nvidia’s platform is also instrumental in making a customer’s accelerated compute transformation as easy as possible; it’s a powerful assistant in unlocking a firm’s ability to actually use its best-in-class hardware. That’s the only way they can actually build accelerated compute apps.

Cuda is its software platform built to facilitate this hardware adoption. It helps distribute work across the general compute CPUs and accelerated compute GPUs by deciding when cheaper CPUs can work and when they won’t.

The platform includes the industry’s first libraries of tools, like Cuda Deep Neural network (CuDNN) and Cuda Data Processing (CuDF). These are what allow general compute apps to be reconfigured and accelerated. These libraries routinely deliver explosive app and agent performance gains and were constructed algorithm by algorithm to ensure Nvidia had product offerings for all major industries.

“All these different libraries have to be invented to take the code and algorithms that run in the application and refactor them in a way that our accelerators can run… we have a rich library for self-driving cars… we have a fantastic library for robotics… And if you use those libraries, then you get 100x speed up… And the reason for that is usually 5%-10% of the code represents 99.999% of the run time. And so if you take that 5% and offload it on our accelerators/GPUs, then technically, you should be able to speed up the application 100x. That’s typical.”

Founder/CEO Jensen Huang

Unifying all of its software are Nvidia Inference Microservices (NIMs). While its library of tools offers guardrails for modernizing apps, NIMs simplify the commercial deployment of enterprise AI systems. NIMs pull from its hardware, Cuda and libraries like CuDNN to make inference capabilities more efficient than anyone else can (per Nvidia). Essentially, NIMs are out-of-the-box products to “bring AI factories and apps to life” in a full-service manner.

Nvidia isn’t just leading in hardware innovation. It’s also ensuring it has the software tools in place to expedite the pace of the new computing revolution. Huang thinks this end-to-end suite of products and services is a large differentiator.

“It makes sense for the chip designer and the supercluster designer and all the software that goes into it to be from the same company. They'll be more optimized, there will be more performant, more energy efficient, more cost effective.”

Founder/CEO Jensen Huang

There’s rising anti-trust scrutiny over Nvidia’s use of software to drive vendor lock and less open ecosystems. This quote above is probably how Huang would respond to those claims.

There has been rising investor concern that customers aren’t getting compelling ROI on GenAI spend. If true, that would likely mean spend will sharply slow. Jensen is not concerned about this for several reasons.

First, the hyperscalers that buy its hardware and rent out compute capacity to customers are enjoying great returns. Specifically, for every $1 spent on its products, they’re enjoying $5 worth of overall rentals. They’re also all capacity constrained, as Microsoft’s CTO again told investors this week (review of that interview coming Saturday).

Secondly, per Huang, the performance and savings gains are fantastic for accelerating general purpose infrastructure. He offered an example of a data processing application built with Spark enjoying 20x performance gains. After accounting for more expensive cost of compute, this leads to “10x” (so 90%) savings for running the same applications and agents. This savings doesn’t include the tangible benefits of 20x performance gains vastly expanding the menu of use cases and applications that are feasible and affordable to pursue.

In reality, general purpose CPUs are great for simplistic, linear instructions that don’t require massive amounts of data and information processing or inference. For these types of use cases, they’re cheaper than GPUs so more rational to use. Accelerated apps in the age of GenAI require rapid, massively scaled processing of data and information to train models and drive compelling model inference capabilities.

Through slowing performance gains over the last several years, CPUs have not caught up with the amount of data being introduced annually into our world. They’re too slow for GenAI and so wildly expensive to continue utilizing for modern apps and agents (what Jensen means when he says compute inflation). That’s where GPUs come in. They complement the CPU to greatly accelerate whatever task that CPU specializes in. That is what makes scraping the entire internet on a simple Gemini query or creating digital twins of the entire planet actually feasible to do. CPUs are not driving the performance and speed gains annually needed to keep up with the growing world; GPUs are.

Huang sees a massive runway for continued rapid performance gains for GPUs on an annual basis. There are 7 different chips that go into its current Blackwell platform and its planned Rubin platform. There’s significant room for improvement across all 7. And remember, the “software developed yesterday” will run on the hardware created tomorrow thanks to Nvidia’s architecture.

“In the event that we have to shift from one fab to another, we have the ability to do it. Maybe the process technology is not as great. Maybe we won't be able to get the same level of performance or cost, but we will be able to provide the supply.”

Founder/CEO Jensen Huang

The clearest sign of the GenAI chip boom still having legs is supply scarcity. If Nvidia still (somehow) has more demand than supply, that will drive pricing power and continued growth as it builds capacity. Jensen offered some comments that point to these constraints still being quite material (a good thing for bulls):

“We have a lot of people counting on us. Demand is so great that delivery of our products is somewhat emotional for people. It directly affects their revenues and competitiveness. We probably have more emotional customers today, and deservedly so. If we could fulfill everybody's needs, then the emotion would go away, but it's very emotional. It's really tense.”

Founder/CEO Jensen Huang