Then I wondered about the big tech companies who are NOT in the list of top model developers but somewhat tentatively building models: Apple, Microsoft, and Amazon. Can you spot the commonality of these companies? We cannot run the counterfactuals, but I did wonder if Jobs, Gates, and Bezos were still at the helm of their respective companies, would they make a different choice? This isn’t necessarily criticism of the “manager class” and yet another ode to “founder mode” mantra. The reality is building, and (perhaps more importantly) staying at the frontier will be incredibly capital intensive endeavor with very uncertain payoff. The reason founders can pursue such endeavor is due to myriad of factors such as voting control, and cultural buy-in from broader stakeholders (e.g. investors and employees) due to their past successes. Managers, by definition, do not have access to such reservoir at their disposal, at least not to the extent founders typically do.

I do not, however, believe the pursuit of the frontier by these founders comes down to mere whims, adventure, or voting control. I sense the primary motivation to be entirely different: relevance! While I have had somewhat skeptical tone in the past about the economics of frontier models, I have been gradually updating my view that owning the frontier will likely prove to be very, very useful even if the precise knowledge of long-term economics is somewhat unknowable today. I carefully chose the words “long term economics” because it is not enough to generate attractive economics today, rather the ability to sustain such economics for a very long period is what will drive valuation for these companies. My skepticism of the frontier primarily stemmed from my inability to spot a source of long-term competitive advantage for the frontier labs such as Anthropic and OpenAI. However, I have updated my view here as I have started to internalize the likely source of their long-term differentiation: data. It is quite revealing how Meta is willing to essentially give away its latest coding model “Muse Spark 1.2” for “free” if you are fine with sharing the data with them.

Going back to my point about relevance, I suspect frontier models will unlock newer products and revenue streams in the future which can only be accessed if you are at the frontier. These founders intuitively understand the long-term arc of this technology and are willing to swing for the fences to stay relevant despite the uncertain payoff. They also seem to understand that the fate of these foundational technologies can potentially be decided early in their lifecycle. Indeed, Android was released just one year after iPhone was launched.

In some sense, owning the frontier may mean you are at a new branch of technology, and thanks to potentially recursive self improving (RSI) nature of these models, it may be lot easier for such company to jump to the next branch. Almost nothing may be off limits for these models in the long term (which admittedly may not be “long”) if the bottleneck to your problem is intelligence. From chip designing to Riemann hypothesis, everything may be a fair game once you’re at the frontier.

The reality is investors probably care less about Riemann hypothesis and want to know more about how much these companies will make money. Unless your estimates are going up and you can articulate clear ROI for your capex buildout, public market investors do not seem to share these founders’ fascination around pursuing the frontier. If anything, given the uncertain payoff of such pursuit, investors are much more willing to sell compute capacity to the lagging model developers. It is not hard to understand the appeal for selling capacity. Given that almost every buyer is repeating ad nauseum “demand is outstripping supply”, it just seems lot easier to underwrite a data center buildout for third-party customers. On the other hand, it is a lot more difficult exercise to underwrite the curve of the pay-off structure from the most leading frontier model company to the second to the third and so on.

Given that context, it is perhaps no surprise that “managers” feel a lot more comfortable in supplying the capacity to companies pursuing the frontier than chasing it themselves. Managers understandably do not feel they have the license to pursue the speculative frontier as vigorously as founders might. I was a bit hesitant to even include Apple in this list as they clearly are not in frontier model race. Amazon, on the other hand, still does not exhibit such clarity. After articulating why Amazon doesn’t need a frontier model to succeed, Andy Jassy went onto then explain why Amazon is still choosing to build “frontier” models:

“…we are pursuing our own Frontier model. And we’re doing it for a few reasons. First of which is it just gives us additional control over cost, cost for our own consumer applications, but also we’re trying to drive costs down for customers. And having a player like ourselves, it’s always focused on trying to take the price performance and the cost down for customers all the time, we think will help keep the models more cost effective for customers.

I think also it allows us to have more control over prioritization of what models focus on. And we have, both from our own external customers as well as our internal customers inside the company, certain priorities that matter that we want the models trained especially well for. And then it gives us some control on speed.”

Like Ben Thompson, I also suspect Jassy may have a very different definition of “frontier” than what is typically understood. My best guess is what he’s essentially saying is Amazon will have a model on their own and for some reason, he’s just labeling it as one of the many “frontier” models the world will have. As I have mentioned before, the frontier has not been as dynamic this year as it was in 2025 when Gemini was the best model for a brief period.

Oh Gemini! Last year, I have written quite a few pieces (see here, or here) articulating why I believe Alphabet has a deep, structural advantage in pursuing the frontier model race. Unfortunately, they have been mostly missing from the race this year. While Alphabet remains founder controlled company, it is led by a manager. I do have a more positive perception about Pichai than most Alphabet skeptics out there, but as I have alluded earlier, managers inherently likely will find it very challenging to navigate the race to frontier given the speculative pay-off vs more amenable spreadsheet math of building capacity for serving 3P customers. Brin only seems to show up intermittently, but given the competitive intensity here, I’m less confident than ever before that such intermittent appearances will be sufficient to lead the frontier. Larry Page is allegedly willing to go rather bankrupt to pursue the elixir of AI, but I cannot actually recall the last time we heard from him and I do not have a good sense how deeply he is still involved in the company’s strategy.

While Alphabet’s deteriorating position in the model race is surprising to me, I am actually somewhat similarly surprised by the absence of any Microsoft model in the aforementioned Artificial Analysis Index. If someone asked me three years ago to estimate the probability of Microsoft staying in the vicinity of frontier models in 2026, I would actually give them a decent chance given they have access to all the IP of OpenAI. The fact that they are still nowhere to be found in that index should be a source of at least some annoyance from their shareholders. Given their near term estimates are going up, I guess such annoyances can be kept at bay for now. I haven’t followed Microsoft as closely as I have followed Meta, Amazon, or Alphabet over the years, but after covering Microsoft’s earnings for the last several quarters, I have noticed that management has a vexing propensity to decide on a narrative that they want investors to believe about Microsoft and then ONLY disclose objective-looking numbers to fit such narrative. As I have covered before, Microsoft’s goalposts seem to change every quarter depending on whatever the “current” narrative they believe in. It was particularly amusing when Microsoft management mentioned that 90% their “cloud revenue” came outside of frontier model companies. Such disclosure can assuage unsuspecting shareholders about lack of concentration risk of frontier model companies for Microsoft’s incremental growth, but I do want to highlight what consists of “cloud revenue”: Microsoft 365 Commercial cloud, Azure and other cloud services, the commercial portion of LinkedIn, Dynamics 365, and other commercial cloud properties. Presumably, most Microsoft shareholders actually want to know how much of Azure’s revenue came from OpenAI. This clever disclosure (or lack of) actually prompted me to make this meme which I then sent to a group chat I am in!

Thankfully, even though Microsoft management is coy about certain disclosures during the call, OpenAI is an equity method investee which qualifies as a related party under ASC 850. That forced Microsoft to disclose in their SEC filing how much of their revenue came from OpenAI and that number is $24.1 Billion in FY26. Nearly all of these revenue likely flow through Azure which makes Azure revenue the more appropriate denominator to understand the concentration risk of the capex buildout. Microsoft, of course, doesn’t even disclose Azure revenue consistently (only the growth), but they did mention Azure revenue exceeded $100 Billion in FY’26. I guess it doesn’t quite fit with the Microsoft’s current preferred narrative that nearly one-quarter of Azure’s revenue is coming from one customer! I do suspect that as long as OpenAI and Anthropic continues their meteoric and frankly incredible growth trajectory, shareholders need not worry about such concentration. But we will get much more clarity where the skeletons are hidden in the seemingly “easy” business of selling capacity to frontier labs if that trajectory ever stops.

To be fair, as of today, there is nothing in the horizon that will make you think such ever increasing annual revenue run-rate trajectory is about to slow down. But these things are inherently unforecastable. SpaceX is talking about building up to 10 GW capacity by 2027, and both Microsoft and Amazon promised to double their capacity by 2027. On top of this, Jensen Huang seems to be on a mission to teach hyperscalers a lesson for building their ASICs and want to further fragment the broader compute market. There is an incredible amount of supply of compute that’s coming in the next 18-24 months and if there is any hangover at all from such spike in supply, this business won’t seem as “easy” as it seems today.

One company that seems to be making a very difficult choice today is Meta Platforms. Like Amazon, they too want investors to believe that their core business can succeed without owning a frontier model which I do find to be credible. Unlike Amazon, Meta’s financials are visibly saddled with the massive talent and compute requirement to stay at the frontier. In the latest call, Zuckerberg tried a scattershot approach to assuage investors that such massive investments will lead to new revenue streams in the coming years. From the call:

“The opportunity in front of us is massive. First, we are now at a point where our investments in AI are accelerating every major part of our core business. They’re improving the experience for people using our apps, driving better performance for advertisers and helping our teams build new experiences and ship faster.

Second, we are developing new personal agents that will be the foundation for our next wave of products and revenue lines in the months and years ahead. And third, we see a large enterprise opportunity to sell to businesses, including APIs, business agents, potentially selling compute directly and other services that we’re building for large customers.

…we also expect to grow a large business serving large customers as well.

We’ve built our API. We’re rolling out business agents. We’re getting a lot of offers for compute at a significant premium over what we paid for it. And we have more coding and productivity tools on our road map as well.”

The only problem is apart from the first one, the rest of these potential revenue streams are at best nascent and at worst mere ideation stage. In the most recent Sharp Tech podcast, Ben Thompson criticized such scattershot approach which he thinks ended up confusing (or worse, alarming) investors. However, he was actually quite constructive about all of the monetization opportunities mentioned by Meta except the large enterprise opportunity. He pointed out how long it took for Google Cloud to be where they are today and Meta doesn’t have any credible “go-to-market” strategy to make a dent. I believe Google Cloud is the wrong analogy here. A more appropriate analogy might be OpenAI or Anthropic. Why are OpenAI and Anthropic running circles in enterprise market today? The OpenAI case study is perhaps much more instructive. To borrow Thompson’s own words, OpenAI was the “accidental consumer company” company and despite being late to the enterprise opportunity, they are clearly putting up a fight against Anthropic. Observing OpenAI’s success closely, I believe the limiting factor here is actually having a frontier model, and IF you have it, you can take care rest of the details much more quickly than ever before.

Imagine for a moment that Meta manages to build a credible frontier model with equally impressive harness. I suspect all the hyperscalers will be tripping over each other to host Meta’s closed models and sell the API to their customers. As much revenue as these hyperscalers are making today from the frontier labs, they also know they don’t want to live in a world where the labs have monopsony power over their suppliers. In this scenario, I don’t think Meta needs to build a GCP-sized platform to monetize their model appropriately; they just need to make their models available everywhere to at least recoup the training cost. They make plenty of money from other ways anyway.

But…this is a hypothetical scenario! And as mentioned earlier, the pay-off is lot more speculative today than supplying capacity to frontier labs. But when I imagine myself what would I do if I owned 100% of these companies, it becomes more clear to me that indeed I too would pursue the frontier despite such uncertain pay-off. Zuckerberg is doing exactly that and I suspect every Meta shareholder who would make a different choice may want to leave the boat before it gets potentially messy.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Current Portfolio:

Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.

I made a slight change to my portfolio yesterday.

Read more

Of course, part of the story here is terminal value concerns related to AI which is definitionally unfalsifiable regardless of near-term operating performance. But more importantly, the stock reacted really poorly after its earnings last week because of 3Q’26 guide which implies an anemic sequential revenue growth QoQ.

For context, in February they guided to 30% growth (against a ~24% sell-side consensus for 2026), raised to 35% in May, and are now guiding for 39% revenue growth in 2026. Back in my 4Q’25 update I wrote it “won’t surprise me at all” if actual 2026 growth landed close to mid-30s; the guide itself has now blown past that. Despite such outlook, the sequential growth of ~1% next quarter has likely spooked investors. Moreover, while Figma raised the revenue outlook for the year, the non-GAAP operating income guide was held flat at $125-135 million despite the $40 million revenue raise. So investors may be concerned about the incremental revenue as “empty calories”.

I will discuss other key metrics and some key takeaways from the quarter as well as some additional thoughts on Figma’s AI risk behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

First of all, digital advertising industry is defined as Alphabet’s advertising revenue+ Meta’s advertising revenue+ Snap’s ad revenue+ Pinterest revenue+ Microsoft Search advertising revenue+ Amazon advertising revenue+ Trade Desk revenue+ AppLovin revenue+ Reddit’s ad revenue. Since Snap started reporting ad revenue separately since 1Q’24, I started including only ad revenues now. However, before 1Q’24, I assumed Snap’s total revenue as their ad revenue which is perhaps closer to reality anyway. Similarly, Reddit’s number has only been included since 1Q’23 since we don’t have their ad revenue numbers in earlier years. Keep these in mind while looking at YoY comparisons in the past. Also note that for visual convenience, I have colored all the Q2 of the last five years.

Obviously, the actual digital advertising industry is larger than this, but this snapshot helps me gauge the broader digital advertising industry’s big picture. TikTok and Walmart advertising are notably missing here. Well, TikTok is a private company and Walmart doesn’t disclose their advertising revenue consistently.

Some key takeaways from 2Q’26:

Digital advertising industry decelerated slightly from 22.2% in 1Q’26 to 20.3% in 2Q’26. We will likely see further deceleration for the next couple of quarters given the tough comps (both 3Q’25 and 4Q’25 grew by 18.5% YoY which were the highest growth quarters last year).

While Meta’s share gain continued, the pace of share gain has moderated a bit in 2Q’26. Given its incremental share of 44.5% is still comfortably ahead of current share of 34.8%, Meta will likely to continue to gain share in the foreseeable future. One challenge with inferring too much based on quarterly snapshot is FX changes likely play a material role in the rate of changes here. Among these companies, Meta likely had the highest mix of revenue coming outside the US and any FX tailwind or headwind would affect Meta’s market share more than others. See more on how FX likely affected Meta’s 2Q’26 ad revenues here.

Over the last five years, Google Network business’ market share has halved from ~8% in 2022 to ~4% in 2026. While Google Search too steadily loses market share, incremental market share has actually improved in 2Q’26 YoY. I am curious to see if search can continue to stabilize around ~30% incremental share or drifts lower to mid to low 20s over the next 24 months. Search seems fine for now, but as OpenAI is likely to ramp up their ad efforts and Meta is under tremendous pressure to grow their ad revenue in order to gain permission from the investors for further capex increases in 2027 (and beyond), there is likely a decent probability that ~30% incremental market share for Google Search may not quite prove to be “steady-state” over the medium term. No strong opinion here, but gun to my head, I suspect we may see mid-20s incremental market share for Search in 2027-28.

Among the sub-scaled ad players, Reddit and Applovin are punching above their weight; look at their current share vs incremental share. TradeDesk has been an unmitigated disaster as growth has fallen off the cliff. Given that the company guided for ~12% revenue decline in 3Q’26, the worst is not behind them. I guess it’s not a surprise that the stock is down ~90% from its December 2024 peak. I haven’t studied Trade Desk closely, so I have no opinion about their future.

I should, however, note that some of the definitions of advertising here are not quite apple to apple as some companies report advertising dollar on a gross and some report on a net basis which is why I like to focus more on incremental share data.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Current Portfolio:

Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.

My current portfolio is disclosed below:

Read more

Given such sustained GMV growth, Shopify's GMV now stands at ~44% of Amazon's estimated retail GMV, up from ~10% back in 2017. In 1Q'26 update, I suggested Shopify could reach half of Amazon's GMV by the end of next year; after this quarter, that trajectory not only looks firmly intact, they may actually get there slightly ahead of that timeline.

On an incremental basis, the story moderated slightly but remains extraordinary. Shopify added ~80% as much GMV YoY as Amazon retail likely did in the quarter, down from the ~95-130% range of the prior few quarters. For context, Shopify CFO cited eMarketer data on the call highlighting Shopify merchants have captured nearly half of all incremental US e-commerce dollars since the start of 2025 while being only ~14% of US e-commerce.

The reason Shopify can match Amazon almost dollar-for-dollar remains the same one flagged in the 3Q’25 update: Shopify’s platform model taps growth reservoirs that are structurally harder for Amazon’s marketplace to reach. In 2Q’26, international GMV grew 37% (Europe +34% FXN vs. North America +28%), offline/point-of-sale GMV grew 32%, and B2B GMV grew 76%. B2B is decelerating as the base scales; it was doubling annually through 3Q’25, then +84% in 4Q’25, +80% in 1Q’26, and now +76% but it remains a meaningful tailwind. Payments penetration hit 68% of GMV (+3 points YoY) with European penetration up more than 350bps YoY.

There were some interesting details around why Shopify’s growth has sustained over such a long period. I will discuss that and more about their margins as well as key takeaways from the earnings call behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

Since Booking reported its own 2Q’26 a couple of days earlier, I will first go through Airbnb’s quarter through the usual KPIs, then compare and contrast against Booking, and finally discuss some broader takeaways from the calls.

Airbnb’s performance exhibits a lot of what I have been expecting and writing over the last few months, but let me start with something that has been somewhat diverging from my expectations: take rates! LTM take rates actually went down by ~31 bps YoY. Moreover, even though during 1Q’26, management expected full-year take rate to be higher in 2026, they have now guided it to be flat. This is despite the fact that their recent single service fee model is a tailwind to overall take rates AND significant growth in insurance related revenues which was up +45% and +60% in 1Q’26 and 2Q’26 respectively.

So, what explains this divergence? Management cited two primary reasons: a) Reserve-Now-Pay-Later (RNPL) which was ~20% of GBV, and b) higher customer incentives. Airbnb is actually expanding RNPL given the strong results they have seen and they like the fact that RNPL lets host lock in earlier calendar share (vs other OTAs). As a result, RNPL is likely to be a continued headwind to the reported take rates even though it is net positive for the overall business. Similarly, to promote new verticals such as experiences and services as well as hotels, Airbnb is providing “credit” to guests that they can use in the future on Airbnb. So, Airbnb is sacrificing some or lot of the take rates in such GBV to grow the overall marketplace. Both decisions are highly likely to be net positive for the business over the long term even if it lowers the reported take rates in the near term. Once RNPL is more widely rolled out and early investment phase in new verticals are behind us, I still think take rate will go up over the medium to long term. But the pace of such improvement may be more sluggish than I initially expected, especially given Airbnb’s ambition to launch new verticals almost every year which may lead to more sustained investment in customer incentives.

Geographically, the reported revenue growth was quite balanced: North America +15.8%, EMEA +15.6%, LatAm +26.0%, APAC +16.9%. As usual, FX impacts these reported numbers a bit too much to derive useful comparisons, so N&S growth and FXN ADR remain my preferred lens.

The standout on this quarter is North America (NA): after six quarters stuck in LSD-MSD nights growth, NA reached HSD for the second consecutive quarter, with FXN ADR up 7%. EMEA also moved up from MSD to HSD. In other words, the core markets are accelerating. CFO attributed part of the ADR strength to the continued disproportionate popularity of larger homes; bedroom nights grew faster than nights booked and accelerated more on a YoY basis, which frames the ADR appreciation as incremental value delivered rather than pure price inflation. Meanwhile, expansion markets continue to grow about twice as fast as core, first-time bookers accelerated to +11% (the best in four years, with Gen Z the fastest-growing cohort), and app nights grew 23% YoY to 64% of total nights (vs 59% a year ago) which is quite encouraging to see.

I will benchmark Airbnb’s quarter against Booking as well some other key takeaways from the call which will be behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

Speaking of DashPass, management shared some interesting details about its trajectory. I will discuss more granular aspects of the quarter, including the per order economics, as well as a portfolio change behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

While premium subscriber number was above management’ guide, Monthly Active Users (MAU) came out 1 million below the guide. As I have shown below, Spotify has experienced a noticeable acceleration in 2022-23 period in incremental MAU growth even though Premium subscriber trend was more steady which led to a decline in free to paying subscriber conversion. Spotify management indicated that now is the time to focus on conversion than merely growing the top of the funnel. From the call (emphasis mine):

“…We’ve been working on turning our outperformance in MAU into revenue growth. So to capitalize on this opportunity, we are adjusting elements like product optimization and ad load, among other things in select emerging markets. Now this strategy carefully increases friction in our free service with a goal of driving higher user conversion and revenue growth down the line.

Yes, this will show itself in our Q3 MAU, but we believe it’s well worth it. And as we’ve shared previously, the free-to-paid conversion cycle in emerging markets grows differently than our established markets. So while a move like this one will take time to play out, the opportunity is vast. This will be additive to our potential over time.

I will discuss segment wise breakdown, product development as well as margins and valuation behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

Let’s start with the numerator in ROIC: Net Operating Profit After Tax or NOPAT. To make this number more comparable over a longer time series, I have added back all the one-off expenses over the years e.g. FTC fine in 2019, layoff and restructuring related expenses in 2022, and the most recent legal and severance related costs in 2Q’26. Then I used a normalized tax rate of 21% (instead of actual effective tax) to net out tax variability over time.

The denominator is the Invested Capital. Before I explain how I calculated invested capital, let me first mention that invested capital should include capital that's ACTUALLY deployed AND generating the NOPAT in your numerator. Anything you add to the denominator needs a corresponding adjustment to the numerator, or you're just mechanically depressing ROIC without economic meaning.

So, I started with total assets and subtracted cash and marketable securities, equity investments (this has no impact on NOPAT but it went from zero in 2Q’20 to ~$30 Billion now which would depress ROIC in recent quarters following ScaleAI stake last year if you don’t adjust for it), and non-interest bearing operating liabilities. I used to not net out operating lease liabilities which was an error on my part; if I leave the lease liabilities unsubtracted (keeping ~$24 Billion of Right-Of-Use assets in the capital base) while NOPAT already bears the full lease expense including its embedded financing cost, I would have understated ROIC. Similarly, I didn’t net out “long-term income taxes” before, but since I am already using normalized tax rate than the typically lower effective tax rate for Meta, my numerator already captures such impact and it should be netted out from my invested capital calculation. Perhaps most importantly, I have subtracted “Construction in Progress” or CIP from my invested capital base since this non-depreciable asset by definition is currently generating zero return for the business and hence have no impact on NOPAT. Once this asset becomes “active”, it will transition from CIP to depreciable asset and will have associated NOPAT impact.

After making all these adjustments, I get to the below LTM ROIC chart for Meta Platforms. While 2Q’26 did see slight downtick in ROIC, it is quite remarkable that companies of this size are actually reporting mid-50s ROIC. But this chart is also why investors are perhaps wary about the capex spree. For context, Meta’s average invested capital in my methodology in 2Q’26 was just ~$127 Billion. CIP in 2Q’26 was a whopping $80 Billion, almost all of which is presumably going to come online in the next 18-24 months. So, the invested capital base is going to go up significantly in the coming years (and the depreciation expenses). As a result, Meta’s LTM ROIC in this methodology will almost certainly keep going down as well. Generally speaking, investors don’t really like it whenever the incremental ROIC is noticeably worse than the overall ROIC. While that is understandable, given the scale of capital deployment investors should be okay with ~20-30% incremental ROIC which would still be lot lower than current “true” ROIC but obviously much, much higher than cost of capital. You can quip that given the incremental ROIC is likely going to be lot worse than current ROIC, shouldn’t the multiple for companies such as Meta go down? That makes intuitive sense, but important to remember the broader context here. One of the big drivers for multiples is reinvestment runway. Even if your ROIC is sky-high and has limited reinvestment runway, your multiple might be lower than a company with respectable ROIC but an immense reinvestment opportunities. This is essentially why Amazon has consistently traded at a materially more premium multiples in the past than either Alphabet or Meta despite having much lower ROIC. Investors were just much more comfortable about Amazon’s reinvestment runway than Google Search or Meta’s FOA business both of which were already outright dominant companies in their lane and investors perennially worried about their limited reinvestment opportunities. Google Cloud was an early indication why such concern may have been premature, but AI is a potential lifeline for a massive reinvestment runway for both Meta and Alphabet, albeit with likely an inferior incremental ROIC than their current “monopolies”. Of course, there is a vigorous and legitimate debate around long-term incremental ROIC of the capex build out for these companies. You can be skeptical about the actual return on the capacity that is about to come online in the next 2-3 years and still acknowledge that so far, the reporting numbers have shown very little deterioration of these businesses’ ROIC (I am just showing Meta’s ROIC here, but the broader message/trend is also very much applicable for Alphabet. Since Alphabet didn’t report “asset not yet in service” line item before 2024, we cannot build a longer term time series chart for their ROIC).

Beyond the persistent ROIC debate, I also want to share some thoughts about Meta and Alphabet after going through their 10-Q behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

For the first time in more than a year, Floor & Decor (FND) reported a quarter and did not cut its annual guidance. FY25 was guided down at every revision i.e. three consecutive cuts across the four reports in 2025, and 2026 started with yet another cut in 1Q, so much so that I wrote last quarter I was a bit surprised they hadn’t given up on providing annual guidance altogether. Well, this time they got to keep it. The comp outlook was maintained at flat to -4%, adjusted EBITDA was nudged up to $550-$585 million from $545-$580 million, and adjusted EPS was raised to $1.88-$2.13 on the back of the 2Q beat and a greater expected contribution from buybacks. Management went out of their way to say they have “high confidence” in hitting the midpoint of the comp guide. Holding a guide is admittedly a low bar to celebrate, but given the last two years, it’s not nothing.

The headline comp itself was hardly inspiring: -2.1%, which makes it 13 negative quarters out of the last 14. As I mentioned in my 1Q recap, this one was near certain to be negative since April was already tracking -4.5% when they reported. What was much more encouraging is the cadence within the quarter: April ended at -5.1%, May improved to -1.3%, and June was almost flat at -0.3%. Transactions remain the primary culprit, but even that improved to -2.9% from -5.5% in 1Q, and average ticket grew +0.8% despite lapping last year’s strongest quarterly ticket growth of +3.8%.

Just as I was wondering things may finally be turning a bit, it turns out July was a bit of a head-scratcher. Management mentioned “some pretty ugly days” around the July 4th holiday that they hadn’t seen in a while, before the business returned to the May/June run rate in late July and into August. 3Q-to-date comp stood at -2.2%. Importantly, none of this sequential improvement can be attributed to macro since US existing home sales (EHS) were still hovering around ~4 million annualized units in June, near historically low levels. Despite such subdued macro, two of FND’s three regions (West and now East) posted positive comps excluding cannibalization, and 8 of their 16 districts were positive on that basis.

Within different customer segments, the Pro market is the more attractive one, and thankfully, the Pro mix shift continued its steady march in 2Q’26. Pro sales grew ~4% YoY against a company that grew 3% and now account for ~55% of sales, up from half last quarter and only 35% in 2021. The new pro app launching next year, which will pull purchasing, loyalty rewards, pricing, and project management into one place, suggests the mix shift may not have peaked. Alongside that, online penetration reached 20.3% of sales (from 18.6% a year ago), and the company kicked off an 18-24 month digital transformation. None of this will move the near-term numbers much, but these are the kinds of investments that will be far more visible in the P&L once EHS normalizes.

The GAAP optics this quarter were almost comically good: 48.2% gross margin (+430 bps YoY) and $0.89 diluted EPS (+53%), but almost all of that is the tariff refunds. I do want to note that these tariff refunds will keep affecting the P&L through the back half of this year. Adjusting for the refund, the underlying quarter was okay rather than spectacular: adjusted gross margin of 43.7% was down 20 bps YoY.

What I found most interesting from this call is what management intends to do with the tariff refunds. They laid out three buckets: offset the inflationary pressures they were otherwise bracing for (oil earlier in the year, and now rising domestic trucking rates), selectively invest in price where they see elasticity to drive share gains, and then the usual capital allocation waterfall. The competitive kicker here is that most independent flooring retailers source through two-step distribution and are not the importer of record, which means they will likely never see a tariff refund check. From the call:

“…when you think about tariff refunds, because they’re not the importer of record, unlikely that they’re going to get those tariff refunds and be able to reinvest into their business, which because of that, I think would prevent them from getting too aggressive around price or promotion in the second half of the year. Everything that we have seen, generally speaking, has been prices going up and certainly not going down.”

As I mentioned in the past, one of the core parts of my FND thesis was that “FND may even look at this period a decade from now as a blessing in disguise since this severe softness in existing home sales may be contributing much more to eliminate competition than investors are appreciating”. The tariff episode looks like yet another asymmetric blow: FND absorbed the tariffs with scale, got the money back with interest, and can now reinvest it in price especially where independents cannot follow. The longer this cycle drags, the more lopsided the eventual competitive landscape becomes.

Like gross margin, adjusted EBITDA margin also went down slightly. Please note while management’s adjusted EBITDA definition adds back SBC, I do not. In some sense, I do find it rather impressive that despite posting 13 of the last 14 quarters negative comp, the company has been largely able to maintain double digit EBITDA margin. If FND can navigate one of the worst EHS trends in decades while maintaining double digit EBITDA margin, it bodes well for the future margin once the cycle turns.

Inventory was up just 0.7% vs year-end. More importantly, in 1H’26, FND generated $278 million of operating cash flow (vs $155 million in 1H’25), and that was despite the ~$90 million tariff refund receivable that only converted to cash after quarter end. So 2H cash flow will get that boost too. FND also began executing the $400 million buyback announced last quarter: in 2Q’26, they repurchased ~1.3 million shares (~1.2% of shares outstanding) at $49.36/share. I believe buyback at these prices will prove to be excellent capital allocation decision over the long-term. The full-year capex guide was also trimmed to $240-$275 million.

Stepping back, my thesis for FND never required 2026 to be a good year, but it does require FND to keep taking share while the cycle grinds down its competitors and to be positioned for the other side whenever EHS mean-reverts, the timing of which remains way above my paygrade. On that note, this quarter was a bit encouraging: positive comps ex-cannibalization in two of three regions with EHS still stuck at ~4 million units, and a tariff windfall that landed asymmetrically in FND’s favor.

When an analyst pressed on whether this is finally the inflection, the new CEO Bradley Paulsen’s answer was, “I’m going to be really careful here because I don’t want to be the initiator of a false start.” Indeed, I have no idea whether the June exit rate marks the trough or just another head-fake in a cycle full of them, but the improving competitive positioning, Pro mix, and respectable margins even at cyclical lows keeps me comfortable that the downside is limited and the eventual upside skew is intact.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Current Portfolio:

Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.

My current portfolio is disclosed below:

Read more

I will start with a clarification. When I said Google Cloud’s growth is more "broad based" than AWS, it was only an observation of 2Q’26 and not a prospective one. What the mix looks like going forward is a different question, and given GCP's smaller base plus upfront recognition on TPU hardware sales, “Southern” may well be right that Anthropic becomes a larger percentage of GCP's revenue than AWS's over the next year. But even if that proves to be true, I'd argue the risk remains more asymmetric for AWS. If lab revenue disappoints, Google can slow capex and absorb capacity into its massive 1P workloads i.e. Search, YouTube, Gemini, DeepMind can all soak up TPUs. AWS has no comparable internal sink; its infrastructure is built to be rented. Higher revenue exposure with a shock absorber can be a better position than lower exposure without one.

On the question of labs being still rather insignificant part of overall cloud revenue, I think a snapshot of today's mix understates where this is heading. Every hyperscaler is reporting hundreds of billions of backlog and that backlog is indeed tomorrow’s revenue. So even if labs are ~10% of AWS run-rate today, the composition of what's contracted to convert into revenue over the next few years is far more lab-heavy than the current mix. Ultimately, a lot hinges on what AI labs’ revenue will be in 3-4 years. If its hundreds of billions, the incremental revenue will have much higher concentration from AI labs. Sothern himself mentioned “Trn/GPU clusters serving labs are basically the lowest rev per GW part of MSFT/AWS business, being long-term wholesale contracts. Clouds charge much higher prices selling to the enterprise (where they also cross sell high margin cloud services), and serving inference tokens.” Indeed, that’s also my concern. The incremental revenue from the capacity coming online in the next few years will be inferior to current and traditional cloud economics. And every year the concentration increases, the labs bargaining power over hyperscalers should increase. There are multiple moving pieces here though. If compute demand remains comfortably higher than supply for years, this increasing bargaining power simply may not materialize in hyperscalers’ financials anytime soon since AI labs will be more busy focusing to get the capacity they want/need. As an investor who cares about moats which plays a critical role in what multiple we should be paying for the earnings streams, I am still alarmed by the evolution of AI labs bargaining power here even if they do not necessarily materialize in the next 2-3 years in current compute constrained environment.

Southern also says “Ultimately the labs will become profitable and want to build their own GPU training and inference infrastructure, so I don’t expect them to get better terms from clouds, just to build their own infra in time”. I’m not sure I understand how that is actually bullish for hyperscalers in the long-term. The possibility of insourcing IS labs’ bargaining power. Once a lab matures to build its own capacity, it will know its all-in self-build cost per token with precision. That number then becomes the ceiling on what it will pay any landlord, plus perhaps a convenience premium for speed, and flexibility. The landlord's margin will become that convenience premium and the premium should shrink as the lab matures, and capital markets get comfortable financing compute against lab offtakes directly. The sweetener of hyperscalers’ circular deals will no longer be relevant and labs will act like a “full adult” in the negotiating table. Admittedly, there are some real barriers for such in-sourcing dreams even if labs are fully capable of financing and operating such data centers. There are potentially multiple bottlenecks in the AI value chain. Power could be a real constraint for transitioning away from 3P hyperscalers to 1P insourcing. So the “convenience premium” could end up being more attractive than it sounds if the incumbent hyperscalers simply have all the pieces of the puzzle. Moreover, we may very well see more concerted political backlash for more and more data centers as the buildout continues.

As you can see, I am much more ambivalent than offering any certainty how this long-term potential economic tussle between hyperscalers and frontier labs may settle in 5-10 years. Given the ambivalence in my mind, especially in current multiples for hyperscalers, I think it makes sense for me to largely be observant for a few quarters or years. I still do have some exposure to this debate via GOOGL, so I’m not quite entirely a distant observer yet.

Frankly speaking, if you listen hyperscaler CEOs closely during this earnings call, they are sort of telling you they are not as nonchalant about labs increasing bargaining power. For example, during the prepared remarks, Satya Nadella started his commentary by mentioning two of the goals Microsoft has. One of those goals is “empowering every organization to build their own continuous learning loop and ensuring that they don’t outsource their core IP.” He later explained during Q&A:

“…at the end of the day, every firm is going to evaluate who are the providers who are helping them with their outcomes and their knowledge creation. I think that, that is now fairly clear, and it’s going to become clearer by the day. This is not going to be about come in and take all my knowledge and benefit yourself, whereas I am not getting anything out of it.

So given that direction of travel, we are very, very clear about the architectural sort of design of the platform, which is you’ve got to keep your harness separate from the model. When the harness will ensure that your memory, your context, all of that is external. That means any given model at any given time is swappable. You should and you can use frontier models. There’s no reason not to. But you also can use multiple of them, right? So if you look at some of the stats I gave. It’s a great example of how to use the frontier models for what they deliver, how to use low-cost models for what they deliver, and in fact, train your own model when you don’t want to use any external model itself because after all, you have all the output, you have all the traces, you have all the context. That’s really the enterprise design architecture that we are going to evangelize.”

I am confident that Microsoft will certainly use “swappable” models in their 1P software products/solutions, but not nearly as confident that enterprise customers are nimble enough to follow Microsoft’s footsteps.

In my last Microsoft earnings update, I mentioned while Microsoft now evangelizes a multi-model world, their own disclosures suggested ~90% of Foundry customers were still using models from just one company. This quarter, the disclosure goalposts moved again. Microsoft now says Foundry has 100k customers (up from the 80k disclosed in FY 1Q’26) with revenue more than doubling YoY, and that customers building with models from multiple providers are up 5x since the start of the year. Notice they did not repeat the “used more than one model” count from last quarter, nor the OpenAI+ Anthropic customer count from FY 2Q’26. Each quarter we get a new numerator on a new base, which makes penetration very hard to track; my best triangulation is that the large majority of Foundry customers are still effectively single-model, even if the multi-model cohort is growing quickly off a small base. None of the disclosures (or omission of certain data) in these earnings call is accidental. If it were conducive to the narrative of multi-model or “swappable” model world Microsoft is trying to propagate, we would have consistent data about it. Maybe it’s just too early though; the reality is Anthropic and OpenAI have been on a different gear in 2026 than the rest of the pack. The fact that Gemini and other closed models are increasingly missing from the frontier conversation is a huge impediment to the future Nadella and Jassy hope to see. Of course, Chinese models made material progress, and the idea that Chinese open weight models are closing the gap only got momentum recently. So we may get more adoption of multi-models in a couple of quarters. However, OpenAI has already made it clear that they are very much willing to price aggressively to compete vigorously against open weight models. Ultimately, it seems the best way to hurt frontier models is to actually be at the frontier and make the frontier more crowded space than it is today.

Unfortunately, the frontier keeps moving. Andy Jassy mentioned the models keep leapfrogging each other at the frontier, but in 2026 such leapfrogging has been largely confined to Anthropic and OpenAI. While Jassy was talking about leapfrogging in earnings call, Amazon actually completed their $50 Billion investment on OpenAI yesterday that had some interesting initial conditions:

Indeed, the very moment it appeared Chinese open weight models closed the gap, we may be on the cusp of again further divergence between frontier and open weight models.

OpenAI and Anthropic both seem to have much more capable models internally that may be released soon. Just today, OpenAI mentioned an internal version of Astra (OpenAI’s next major model family), “solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.” If that didn’t raise your eyebrows, get this: “The total number of tokens needed to find solutions to these problems would cost roughly $2,000 at Sol API rates.” We really are likely entering “an era of proof abundance”.

As impressive as it, I don’t know whether we are on the verge of finding multiple new capabilities unlocking hundreds of billions of revenue for the labs, or the fact that even such complex mathematical problems require $2,000 to solve may mean we are perhaps already in the “intelligence is too cheap to meter” era. So, selling raw intelligence to generate hundreds of billions of revenue every year may prove to be a high bar. I know, I know…”Jevon’s Paradox”!

Let me be very clear. The whole AI space moves so quickly these days that it honestly becomes difficult to have a rigid thought about the future for weeks, let alone months or years. The unfortunate reality for investors is that stocks do get priced for their long-term future and market demands a lot of certainty for the future that I still do not quite share for much of the AI value chain. I am hoping once the labs IPO in the coming months or year, we will have slightly more clarity.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Current Portfolio:

Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.

My current portfolio is disclosed below:

Read more

I will talk more about AWS later, but as I typically do every quarter, I will first start with non-AWS segments of Amazon.

Ex-AWS, Amazon’s North America and International segments margin trajectory were a bit underwhelming. International segment’s operating margin was flat YoY and even though it may seem North America’s operating margin expanded YoY, it includes $600 million tariff related refund. Excluding such impact, North America’s operating margin declined by 18 bps YoY.

If you look at worldwide paid unit growth vs shipping+ fulfillment cost growth, you would notice that the latter used to consistently outpace the former pretty much all the time since 2015 until 3Q’22. Since then, unit growth has largely been faster than shipping+ fulfillment costs, indicating operating leverage in their logistics footprint. In 2Q’26, however, these lines converged at 17%. However, during the call management mentioned excluding the impact of higher fuel and linehaul rates, shipping cost would grow more slowly than unit growth “at a pace that is relatively consistent with last quarter.” That likely means the operating leverage theme still has further legs here once or if fuel costs come down.

A big driver for retail profitability is advertising. Given Amazon ads are perhaps more of a competitor to Google than Meta, I pay close attention to Amazon’s incremental share in advertising compared to Google Advertising. Last quarter, I wondered if Amazon is giving some share back to Google. That concern has evaporated by 2Q’26 numbers as Amazon’s incremental revenue share as a percentage of Google Advertising share went back to the long-term trend of ~40% following a dip in 1Q’26.

Okay, enough about non-AWS business. Let’s get into AWS related discussion which will be behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Read more

I will begin with the comparison of incremental revenue YoY between Meta’s Family of Apps (FOA) and Google Search. Just like in 1Q’26, FOA’s incremental revenue was ~41% higher than Google Search revenue. By 2Q’27, I think it’s likely that the LTM ad revenue of Meta’s FOA will exceed Google Search revenue. For context, back in 2Q’22 Meta’s FOA ad business was just ~72% of Google Search revenue. So to go from there to exceed Google Search revenue by 2Q’27 would be quite remarkable indeed. Of course, Google Search hasn’t been a sloth either which only highlights the rapid pace of growth in FOA despite the ever increasing larger base.

Daily Active People (DAP) on Meta increased by 40 million QoQ and 120 million YoY to reach 3.6 Billion. Instagram now has 2 billion Daily Active Users (DAU) and Threads has 500 million Monthly Active Users (MAU).

The geographical revenue trend was quite interesting in 2Q’26. I would specifically highlight that revenue in North America accelerated and was above 30% YoY for the first time since 3Q’21. Of course, 3Q’21 was benefitted substantially due to pandemic and ZIRP era quirks; so it’s notable that Meta North America ads grew >30% despite revenue being double that of 2Q’21. Europe and APAC were the reasons why Meta’s overall ad revenue growth decelerated by ~544 bps despite North America’s acceleration. Ad impression growth was actually fairly consistent across the regions, so it was mostly the price per ad that drove the deceleration in Europe. A good chunk of that is likely explained by FX trends, but I suspect less personalized ads in Europe was also the reason that explains relatively soft ad prices in Europe. In fact, in the follow-up call Meta highlighted the latter aspect to explain somewhat softer (again, relative to expectations) guide for Q3 too (emphasis mine):

First, we are lapping a quarter of accelerated impression growth, which benefited from engagement-related ranking improvements, mainly on Instagram feed and reels, as well as some ad load optimization on Instagram feed and stories. Second, we’re seeing the impact of less personalized ads offering in Europe now that it’s fully rolled out, and that may be an additional headwind. And we’re also expecting an impact from our continued integrity enforcement efforts.

Worldwide average price per ad was flat QoQ while impression growth decelerated by 500 bps. However, Meta shared data points to substantiate that their ranking and algorithm improvements are still throwing decent payoffs in impression growth.

Instagram time spent grew double digits globally, Facebook video time grew 9% globally (10%+ in the US & Canada) on top of Q1's gain, and over half of recommended content on Instagram Feed is now less than a day old, more than double a year ago. Meta also shipped what it called its largest single ranking release ever on Reels, worth 15 bps of incremental Instagram sessions.

During the call, Meta shared more granular details to explain why they expect further improvements in their ranking and recommendation infrastructure. Some key excerpts from both the earnings call as well as the follow-up call (emphasis mine):

“This quarter, we introduced Meta Generative Recommender, a paradigm shift in how our ad system works. Rather than scoring every possible ad individually, we are now using LLM to reason about ad content and user preferences together and predict the best ad for each person. This makes our ad matching more intelligent and more precise, which compounds performance gains for advertisers.

We deployed the first generative model into our ads retrieval system and saw notable improvements in ads performance. Early pilots using LLM to better understand user preferences drove a 1% increase in app event conversions on Instagram.

In Q2, we also advanced our user understanding models to analyze ads and organic activity and simultaneously improve both user experience and advertiser performance. Combined with our GEM model for ads ranking and sequence learning, these advancements generated an 8.3% increase in ad clicks and a 15.7% uplift in conversions on Facebook.

“we certainly see further headroom to continue improving recommendations over the rest of the year and into 2027. And we expect that will help us drive additional gains on both engagement on Facebook and Instagram.”

eventually, we hope to get to a place where we can collapse our current multi-stage recommendation model to a simpler unified model that, similar to how LLMs do next-token prediction, can directly produce a set of tokens to show users, which we can then match to either organic or ads content within our inventory. So we’ve already begun early validation work on LLM-native recommendation systems, and we’ll look to ship components of that to complement parts of the current stack.

While FOA’s ad business understandably gets all the attention, its “other revenue” is making some noticeable progress. For context, Meta’s other revenue doubled between 1Q’23 and 1Q’25, and then nearly doubled again from 1Q’25 to 2Q’26. It wouldn’t surprise me if it doubles again by 2Q’27. While other revenue is still below 2% of FOA’s total revenue and hence doesn’t get much attention, that may start to change in coming years. They just launched Meta Business Agent platform which does have the potential to become a more material contributor over time. Meta management actually sounded more excited about this opportunity during the follow-up call (emphasis mine):

“we also launched the Business Agent Platform earlier this month, which includes a suite of APIs and tools that allows enterprises to customize their business agent, integrate it through their existing systems, so they can connect to systems like catalog, CRMs, inventory management, giving their business agents the ability to take action on behalf of the business and to deploy those conversations on WhatsApp. So we are just getting started here.

So we’re very excited about this. We think the product is getting ready to scale. We’re investing in driving greater awareness of it, while also continuing to improve the discoverability and onboarding and user experiences. And we’ll be introducing more capabilities into the business agent over the rest of this year and into ‘27. And we’ll also keep improving our models, which we think will drive increased performance and demand for Meta Business Agent going forward. So overall, I think this is an opportunity that we think is pretty unique. We are building a turnkey solution that’s going to leverage existing social media posts, ad campaigns, web presence, making it really easy for businesses to set up. And we want Meta Business Agents to work day one for each business. And we think there will be a big opportunity here again, given the size and scale of advertisers on our platform.

Additionally, earlier this month, we announced a volume-based pricing model for messages sent using Meta Business Agent. And effective August 1st, we’re planning to charge on a per token basis for Meta Business Agent messages with one token base charge that will encompass both AI agent processing and message delivery”

It’s probably a bit too early to get excited about this, but I would expect this may become a source for potential upside in estimates in 2027 (and beyond).

In terms of margins, overall reported EBIT declined by 6%, but adjusted for one-off expenses mentioned earlier, EBIT would have grown by 9% YoY. Of course, in an era when capital is becoming more of a constraint, the ever persistent $4-5 Billion losses in Reality Labs is increasingly even a more pronounced source of embarrassment for Meta. LTM losses here may have peaked, but that is hardly much of a solace as Meta management should start to promise a more concrete path to much lower losses in the medium term. Eric Seufert characterized “Meta’s metaverse initiative from a few years ago as an albatross around the company’s neck that has left a trust deficit with public markets related to large-scale infrastructure investments.” Well put!

Meta Compute and AI Ambitions

One encouraging data point that Meta shared is that once they integrated Muse Spark to their AI assistant, number of people interacting with the assistant daily increased by 60% and continues to grow week-over-week.

Predictably, the call had a lot of tidbits around what Meta plans to do with the massive capacity they are building despite not having a 3P cloud business. Zuckerberg mentioned Meta is getting “a lot of offers for compute at a significant premium” over what Meta paid for it. In the follow-up call, Susan Li also reiterated that the offer they’re receiving is “multiples” of what they paid for it. However, they’re still focused much more on internal uses first. Let me quote the detailed explanation (long excerpt, but important to understand) Meta provided during the call to explain their philosophy here (emphasis mine):

“Our approach to building capacity is strongly influenced by several key elements. First, the broad environment for building infrastructure is dynamic and uncertain in both near-term and longer-term time horizons. The industry has underbuilt historically for the wave of AI adoption, making existing capacity, including our own extremely valuable. Longer term, the supply chains need to be built out to support the capacity that we anticipate we and others will need for AI-powered experiences. Second, we have high confidence in our ability to utilize capacity to scale and build on top of our existing experiences as well as continue to invest in foundational models that will create substantial new opportunities.

Consequently, our current plans are geared towards maximizing 2026 and 2027 capacity. When we have had incremental capacity in the past, it has proven extremely valuable in scaling experiences like Reels. And we are confident that this will be true in this time frame as well.

Longer term, it’s harder to predict the exact usage scaling curves, but we believe that our distribution advantages will give us the opportunity to serve AI products that are valuable for everyone, both our 3.6 billion users and millions of businesses. This should be true regardless of whether our models are on the frontier, but we believe that being on the frontier will unlock new markets and opportunities for which we may need additional compute. Therefore, our longer-term capacity strategy aims to give us the flexibility to continue growing compute in 2028 and beyond by laying down data center and network foundations to accommodate future server decisions.

The long-lived nature of these assets inherently provides the flexibility that will make it possible to adjust our investment to the pace of AI adoption. In addition, we have been making strategic investments in areas like our internal custom silicon effort, which will provide long-term strategic flexibility and supply chain leverage. This will be helpful in driving better returns on those long-term investments.

Finally, we believe that overall industry capacity is going to remain tight for the foreseeable future. As we’ve said earlier, we strongly believe that the models, consumer experiences and enterprise offerings that we are building will be the best and highest ROI use of our infrastructure. Those enterprise offerings have the potential to take multiple forms…agentic tools, our API or monetizing compute directly given outsized market demand. We expect that remaining nimble about these opportunities will help us fund our build-out more efficiently while preserving our strategic flexibility to have the compute when we need it and provide us multiple pathways to generate returns on invested capital.

…we believe that there will continue to be a significantly higher margin on selling intelligence rather than selling compute directly. But we think that there is a big opportunity, obviously, to sell compute as well.

Now in terms of running the business, obviously, a common trade-off that we need to make is around how much do you monetize something today versus develop future assets for the future? And I think that it’s always a portfolio, right? It’s not like you don’t want to only do long-term things and you know like and not kind of prove the markets out that exist in the near term. But I also think it would be foolish to basically just sell all of the compute and take a short-term profit. But when you have the opportunity to build intelligence on top of it, which will be a kind of a multiple and that compounds the value of the compute on top of that.

So I think the answer is what we’re doing, which is to basically use a lot of our capital to build out compute, having confidence that we have the ability to monetize the compute directly when that makes sense, but also knowing that we have quite a number of different use cases to monetize the intelligence on top of the compute, including the enterprise cases that we talked about and including some of the consumer and cases that we talked about, and including just the core business, which is not even necessarily new products that we haven’t talked about, but just in terms of using that to be able to further add intelligence and improve the ranking and recommendations and ads in the core services.”

My read from this long excerpt is that Meta will mostly sell compute when the offer is so good that they cannot quite refuse. If some of the frontier labs underestimate compute demand and need some capacity soon to serve the demand, they might want to pay an arm and a leg for a short period of time. It can make a lot of sense for Meta to serve that market.

The other possibility is market may simply be too impatient to see Meta’s non-advertising monetization and force management’s hand to sell some capacity. In either of the cases, I don’t think consensus estimates still embed much revenue from such deals. This perhaps explains why the stock rallied so hard when market thought such deals may happen lot sooner than they realized. Now that Meta management somewhat downplayed the immediacy of such deal coming online, the stock gave up all such gains. Nonetheless it’s perhaps a useful message to Meta’s management what can reverse the flailing stock price which can be a competitive disadvantage in a hot AI talent market and potential capital infusion requirement from secondary equity offerings.

While investors remain fixated on Meta’s ROIC, I am yet to be concerned. I will expand on my relative nonchalance about such concerns behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Read more

You can listen to the Deep Dive here

The very first Deep Dive I ever wrote for MBI Deep Dives back in 2020 was on Uber. While discussing Uber Eats in that Deep Dive, I made the cardinal mistake of doubting the American eaters! I thought the pandemic likely pulled forward consumer propensity of ordering food online and once things normalize, people would likely go back to dine in restaurants and food delivery companies would face noticeable headwind afterwards. Six years later, as I was studying DoorDash over the last few weeks, I can now confirm that I was, in no uncertain terms, wrong!

My mistake was viewing food delivery purely through the lens of pandemic-era consumer convenience, rather than as a fundamental infrastructure bottleneck for local commerce which was unleashed during Covid to become increasingly integral part of everyday life for millions of Americans. To understand why the habit stuck, let’s go back to when that bottleneck was first identified.

In the fall of 2012, four Stanford students: Tony Xu, Stanley Tang, Andy Fang, and Evan Moore were working on a class project in the GSB’s “Startup Garage” course, interviewing small business owners around Palo Alto to figure out what technology they could build for them. One of those interviews was with a one-person macaron shop. During the interview, the shop owner pulled out a thick booklet pages and pages of delivery orders she had been forced to refuse because she had nobody to deliver them. As Xu recalled in Sequoia’s “Crucible Moments” podcast, small businesses running on 17-18 days of cash don’t have the luxury of turning away business and yet, the owner was forced to do it since there was not quite any viable solution.

The four of them spent the following weeks interviewing another ~150-200 small businesses and kept hearing the same complaint: everyone wanted delivery, but nobody could afford to staff it. So, they thought they might be onto something here. In an afternoon, they stood up a landing page called PaloAltoDelivery.com with PDF menus of local restaurants and a Google Voice number that rang their own phones. DoorDash’s S-1 mentioned what happened next: on January 12, 2013, “…the first DoorDash consumer ordered prawn pad thai and spring rolls”, and one of the founders drove it over himself. For months, the founders were the dispatchers, the customer support, and the delivery fleet. Xu himself was among the company’s very first Dashers, a tradition that survives today in the form of “WeDash,” a program under which every salaried US employee must do delivery shifts each year. The company went through Y-Combinator in the summer of 2013 and renamed itself DoorDash.

In some ways, restaurant food is perhaps the worst thing to deliver as they can be perishable within minutes, ordered in unforgiving lunch and dinner spikes, with customers who expect the whole thing done in under 45 minutes for a minimal fee. Indeed, Xu wrote for Sequoia that when they started, they had exactly three questions to answer: “Would customers be willing to pay $5 for this service?” “Would restaurants be willing to pay us a certain percentage?” “Would drivers be willing to work for this wage?”

If you can build a logistics network that can reliably move hot scrambled eggs across a suburb in 30 minutes, you can probably move basically anything else in local commerce later i.e. groceries, convenience, flowers, the tube of toothpaste...you name it! Restaurants also happened to be the largest category of local commerce with essentially no delivery infrastructure of its own; outside of pizza and Chinese food, most American restaurants simply didn’t deliver. Today, we take food delivery for granted, but it’s actually not that long ago it was far from the norm. Given that context, even though DoorDash’s ambition is to build last-mile logistics for every merchant on Main Street, restaurants were probably the hardest, AND the largest beachhead which means it made a lot of sense to solve that problem first!

While Grubhub, Seamless, and Postmates fought over Manhattan and San Francisco, DoorDash deliberately went to the suburbs. When DoorDash finally launched in San Francisco in 2015 (two years after founding), Xu told TechCrunch that they had wanted to first learn in cities that looked like the average American city rather than the largest market next door. Suburbs had less competition, higher car ownership, easier parking, larger family order sizes, and more importantly, households for whom delivery was a genuinely new capability rather than a substitute for walking a block. Even though suburbs may have looked like a niche strategy at the time, it turned out to be one of the most important pieces of the puzzle to solve the broader problem.

However, for a long time investors were highly skeptical. When Xu and his co-founders were raising money, more than 100 investors said no since on-demand delivery still carried the stench of the dot-com graveyard (Webvan, Kozmo etc), and the margins looked impossibly thin. Sequoia’s Alfred Lin, who would later lead a round and take a board seat, initially called Xu just to pass on the seed. Xu later admitted that solvency worried him for three years as he recalled in this podcast:

I always was nervous, and I think a lot of people involved with DoorDash were very nervous about, you know, our solvency. We were always worried about running out of cash and it was three years of being worried.

…Besides just the constant financial stress, we had people lose confidence in the company, right? Even internally—about maybe a quarter or 20 to 25 percent of the company—voluntarily left over those three years.

Xu and DoorDash went through those dog years to eventually become the dominant player in food delivery market.

On December 9, 2020, DoorDash IPO’d at $102 per share. Then the pandemic-era market did exactly what it did to every 2020 IPOs: shares popped ~80% on the first day of trading, briefly pushing the market cap toward $60 billion. Enterprise Value (EV) reached ~$80 Billion during the 2021 peak and then crashed to just ~$15 Billion in 2022. Since then, the company has recovered in value but EV is still hovering around $80 Billion.

In 2025, DoorDash processed ~$102 Billion of Marketplace Gross Order Value (GOV) across 3.2 billion orders, generated $13.7 Billion of revenue, posted its first meaningfully profitable GAAP year, and after acquiring Wolt in 2022 and Deliveroo in 2025, it now operates across more than 40 countries with over 56 million monthly active users. Not bad for a company that started with a PDF menu website with a Google Voice number.

In the rest of this Deep Dive, I will first walk through how the three-sided machine actually makes money and dissect the unit economics of an order. I will then discuss the competitive dynamics: whether food delivery is recession proof, whether the model is actually net beneficial to all three sides, why DoorDash beat Grubhub and the competition it faces against Uber, its chances in grocery against Amazon, Walmart, and Instacart, and whether AI is an opportunity or a disintermediation threat. Finally, I will cover capital allocation and management incentives before getting into valuation.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Business Overview

Read more

Read more

The divorce has been sort of telegraphed for months. Waymo hasn’t added a new Uber city since Atlanta in June 2025, and the two sides have traded accusations over safety incidents and “unsustainable” economics.

Uber has actually been lobbying for “hybrid network” rules which means if you call an Uber in a market with AVs, you can get matched with either an AV or a human driver. If such a regulation were enacted, this all but guarantees Uber’s primacy in ridesharing for years (decades?) to come. TechCrunch laid out plainly:

If Uber is successful and its hybrid network idea is adopted in D.C. — or elsewhere — it would leave AV developers like Waymo with two choices: put their robotaxis on ride-hailing apps like Uber’s, or employ human drivers who provide ride-hailing services alongside the robot cars that have taken years and hundreds of millions of dollars to develop.

Uber’s “AV transition white paper” is very eager to make the case that they are on the same side as labor which is a bit ironic given the company has for years on the opposite side of this labor and tech platform owner debate. Some interesting excerpts from their white paper (emphasis mine):

We owe people the truth. Autonomous vehicles may expand mobility overall, but they could also mean less work for many drivers over the long run, including people who rely on platform work as a flexible safety net. We believe a phased transition to a hybrid model where drivers and autonomous vehicles work side-by-side offers the most promise for workers and greater reliability for consumers and cities. A too-rapid shift to driverless fleets risks exposing gaps in today’s labor and benefit systems, pushing the burden onto the workers and communities to absorb the disruption. If we want an autonomous future that holds together socially, we have to plan for any workforce disruption with the same seriousness we bring to the underlying hardware and software.

In San Francisco and Los Angeles, where drivers compete against AV-only networks, driver utilization and hourly earnings declined last year. Given that the vast majority of drivers work part time, and an AV is online most of the day, we see that in California, one AV does the work of about four drivers.

The right path is phased and hybrid, moving fast but not all at once. This reduces real risks: worker displacement, uneven access, and pressure on city infrastructure and safety systems.

While Uber is making the argument for hybrid network to the regulators, it’s simultaneously making the case to investors that AV-only competitor is simply bad business. For example, during 4Q’25 call, Uber pointed out that any AV-only competitor must either hold significant underutilized supply to match Uber’s reliability and prices, or deliver a worse consumer experience and leave demand on the table.

There’s an interesting academic paper titled “Sharing economy in the era of full automation,” that was published earlier this month. The paper never mentions Uber vs Waymo debate and actually touches mostly on Elon Musk’s dream of your privately owned autonomous car moonlighting as a robotaxi while you sit in meetings at work. But if you look past that framing, it is a general model of the exact problem underneath the Uber-Waymo debate: how a platform facing spiky, spatially lopsided demand should combine always-on, capital-heavy base-load supply with flexible third-party supply. In the paper, the flexible layer is private AVs rented by the hour instead of being owned by the platform itself.

The paper sides with Uber’s case that a rational fleet owner sizes to base load, which structurally leaves the peaks, the bad-weather surges, and the tail geographies to a flexible layer. Whoever aggregates that layer pays it marginal opportunity cost and keeps the spread.

However, the paper models exactly one platform, and its result shows that whoever owns demand and dispatch captures essentially all the economics, while flexible suppliers (human or robotic) get paid opportunity cost. That is wonderful for Uber only so long as Uber is the one doing the dispatching, but as the news flow suggests, Waymo obviously understands the reality of the long-term economics here if they don’t own the demand directly.

In fact, an analyst actually asked Elon Musk during last week’s Tesla’s 2Q’26 earnings call if Tesla would like to consider third-party distribution partnerships, such as with rideshare providers, to increase utilization. Elon Musk doesn’t appear to be interested as he said: “We expect to be vertically integrated with robotaxi as we are in the rest of our business.”

The truth is both Elon Musk and Waymo deeply understand the underwhelming economics of being at the mercy of the demand aggregator. It would make long-term strategic sense only if they see no path at all to control the demand aggregation layer. So, I would expect both Tesla and Waymo to give their best shot at being at the demand aggregation and only succumb to partnership with incumbent ridesharing companies if they come to the conclusion that there is no viable path to aggregate demand themselves. We may not have to wait years to get our answers how the debates will be settled here. Driverless Digest Substack pointed out that while AVs are being launched in more and more cities, in the next 12 months Waymo will likely transition from mere footprints to building density in many of these cities. From “Driverless Digest”:

Waymo has already entered or announced many of the largest markets. Based on my demand-weighted view of U.S. ride-hail, it is currently live in metros representing roughly 27% of estimated demand and has announced or signaled markets representing another 43%. In other words, Waymo now has a presence—or a stated intention to establish one—across close to 70% of estimated U.S. ride-hail demand.

That does not mean Waymo serves 70% of the market today. Nor does it mean its service areas cover entire metros. This is a metro-level demand proxy, not a measure of actual coverage. I count Los Angeles as a live market because Waymo operates there, for example, even though its service remains concentrated within a limited portion of the broader metro area.

But that still matters. Waymo is already in, or heading to, many of the markets that drive U.S. rideshare demand. Now it needs to make the service meaningfully available in each one: more vehicles, shorter wait times, broader coverage, airport access, and more consistent reliability.

In robotaxis, availability is the product. A service that works great but comes with long wait times or is not available in your area is still not a true substitute for Uber or Lyft.

The key point is simple: Waymo has established the market footprint. The next 12 months are about building density within it.

That is both the opportunity and the risk. Waymo’s next chapter is less about proving the Driver and more about industrializing the fleet.

Ultimately, the only question that may matter is whether Uber will continue to be the aggregator with dominant market share in ridesharing. If regulators demand hybrid network in all platforms, it pretty much ensures Uber’s prime position in ridesharing. Without such regulatory blessing, it appears the intensity to be at the demand aggregation layer will intensify in the near term. Uber competed tooth and nail over the last decade to dominate ridesharing today, and this decade is supposed to be the harvest decade for all the cash it deployed last decade or so. Such harvesting period may need to wait a bit if Waymo and Tesla start a new war in the demand aggregation layer in the next couple of years. I don’t have any strong point of view yet how such battle will be settled, but I intend to follow it more closely going forward to see if the picture is getting any clearer over time.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Current Portfolio

Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.

My current portfolio is disclosed below:

Read more

Back in 2012, I visited Harvard for the first time to attend a conference. During that conference, I attended a social get together where I met a bunch of undergrads at Harvard (I myself was a Junior in college at that time back in Bangladesh). It was still socially acceptable to ask for each other’s Facebook account and stay connected through Facebook. So, after the event, I became “friends” with a bunch of Harvard undergrads.

A couple of weeks ago, one such friend (who once won Gold at International Mathematical Olympiad) posted the below status update on Facebook:

“As International Mathematical Olympiad is being held in Shanghai right now, I’m having strange feelings while experimenting with AI model performance at the most well known global academic competition in the world. Our conclusion is that AI models are now practically better than almost all humans, even at IMO. Day 1 perfect score is confirmed, and likely to be the same on day 2 as well. This surpasses our previous expectation which was merely a gold medal level performance.

If I was 15 year old right now, I know I wouldn’t be able to motivate myself to get better at math as I did back then, this makes me sad. At that age, I felt real romantic love towards mathematics, and AI is like a much stronger bully stealing my love. It may sound absurd, but that’s how I feel!”

Then a couple days later, he confirmed the AI models’ supremacy in another status update on Facebook:

“GPT 5.6 Pro solved all 6 problems from IMO 2026 on the first attempt without any human help or steering…The problems are considered incredibly hard, usually a performance at this level is only accomplished by < 5 contestants from the whole world.”

While reading these status updates, I was reminded of another interaction I had with a girl when I was visiting Bangladesh a couple of months ago. She was just in her 9th grade, and it was pretty evident to me that she was perhaps one of the smartest 9th graders I have ever interacted with. Her father used to teach my wife Mathematics when my wife was in High School, and he wanted to visit us with his daughter when he learned that we are in town.

The daughter clearly got the Mathematics bug from her father and she was telling me how she spends much of her day solving Math problems and rummaging through different problem sets and discussions in different internet forums. it was a good reminder that despite being born in a second-tier city of a third world country, a gifted kid can still keep pace with the world through the blessings of the internet. She wanted to represent Bangladesh at IMO and was curious to learn more about college admission process in the US. The reality is most US colleges would be lucky to have her in their class, so I encouraged her to not fret over such mundane things. However, I did have one suggestion for her: “keep your identity small…don’t make mathematical proficiency your primary identity, rather just internalize that it is one of your gifts that the world may or may not value in a decade or two. Nurture the inherent joy of Mathematics that got you hooked in the first place so that even if the world becomes less appreciative of your gifts, you still have your reservoir of curiosity and joy to remain infatuated with Mathematics.”

AI’s rapid progress may make these questions more urgent than I may have internalized even a couple of months ago, but while this discomforting reality may be a relatively new phenomenon for today’s mathematically gifted kids, many other gifted kids indeed grappled with such reality for a long time. Imagine a kid in India who is the national Champion in some random Olympic sport but fail to make the Olympic podium. Can you imagine how challenging it is to become national champion in anything in a country of a billion and half population? And yet, there are perhaps numerous such gifted kids who do not find much of an economic value for their gifts.

Of course, Mathematics is no equivalent of a random Olympic sport as it is far more instrumental and fundamental in our understanding of the world. But I do often wonder Ilya Sutskever’s October 2023 tweet: “if you value intelligence above all other human qualities, you’re gonna have a bad time”

Indeed, if intelligence becomes too cheap to meter in not-so-distant future, that may fundamentally reshape how we approach mathematics (and mathematicians) too. Terence Tao, one of the greatest living Mathematicians, actually gave a profound (and very accessible) presentation titled “Mathematics in the age of AI” just two days ago at “International Congress of Mathematicians” (ICM). He didn’t mince the soul searching that may be required in the field of mathematics sooner rather than later (emphasis mine):

“…in the early twentieth century, discoveries such as Russell’s paradox (1901) or the Gödel incompleteness theorems (1931) forced practicing mathematicians to critically re-examine their implicit assumptions about the foundations of mathematics.

This crisis in foundations (∼ 1900–1930) was a turbulent period for mathematics. But the end product was extremely valuable: an explicit, rigorous, and standardized foundational framework. There is scope for further improvement. But our current foundations have survived strenuous testing and are a trusted environment for mathematics.

I believe we are entering a similarly turbulent period —a crisis in the foundations of mathematical values and practices.”

Since there are plenty of people out there who remain deeply skeptical of AI’s actual capability (especially given how jagged it can be), Tao was careful in explaining his “AI Capability Conjecture”:

At some point in the near future, some AI tools will, at some expense, and with some level of human supervision, be able to correctly accomplish some research-level mathematical tasks in some fields of mathematics, with some non-trivial success rate, and at some level of correctness and quality.

He then cited some independent assessment that led him to formulate his “working hypothesis” for AI’s capability:

AI tools will, reasonably soon, become capable of performing a reasonable fraction of research-level mathematical tasks, with reasonable levels of success, quality, supervision, and cost.

In such a world, what should be the “Goals” of today’s and tomorrow’s Mathematicians? Tao goes through different iterations of the goals in his presentation and decided on the following in his fifth attempt:

Solve unsolved problems, verify them to be correct, ensure they are clearly communicated, and have them digested, accepted, and incorporated into the definitive theory of the field.

The last bit may not seem super prestigious today, but Tao reminded a wonderful quote by William Thurston in ““On proof and progress in mathematics”:

"We are not trying to meet some abstract production quota of definitions, theorems and proofs. The measure of our success is whether what we do enables people to understand and think more clearly and effectively about math."

Of course, digestion and acceptance may be the actual bottleneck relative to the generation of novel proofs, but that’s exactly where humans can leave their mark. From Tao:

Community acceptance of a result, by its nature, is slow and human. It can be encouraged with good exposition and careful writing. But it is ultimately an external process that cannot be optimized purely by the authors and their AI tools

Our current publication infrastructure relies on human editors and referees to voluntarily provide this community acceptance as a service. This work is often regarded as less prestigious than that of generating proofs in the first place. But it is an essential component of our profession. It is also how we convert the individual achievements of mathematicians into collective progress and understanding.

Near the end of the presentation, Tao acknowledged as we are likely moving from an “era of proof scarcity to an era of proof abundance”, the mathematics community needs to evolve with it:

In short, we will transition from an era of proof scarcity to an era of proof abundance

we need to decrease the emphasis on proof generation, and of being the “first” to solve a problem; and increase the emphasis on “proof digestion”: exposition, publication, and canonicalization.

my suggested rule of thumb: if the authors cannot convincingly demonstrate that they can give a clear, expert-level talk on their results, that is correct and properly attributed, then the result should not be published.

The whole presentation is worth going through; you can sense the internal struggle due to the rapidly progressing AI models, and the wisdom that came out of it from one of the greatest living Mathematicians. As the foundations of mathematical values shift from generation to digestion, the best path forward might just be the one I suggested to that 9th grader in Bangladesh. When the novelty of simply finding the answer fades, it is the inherent joy of the process and our ability to share that clarity with others that will truly endure. Ultimately, our greatest enduring moat may simply be our curiosity and our shared human experience of understanding.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Disclaimer: All posts on “MBI Deep Dives” are for informational purposes only. This is NOT a recommendation to buy or sell securities discussed. Please do your own work before investing your money.

Let me dig into 2Q’26 earnings of TI behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more

Elon Musk

as you can tell from the many collaborations on so many fronts with SpaceX and there’s a lot -- there’s more and more overlap, especially with Terafab, that’s really going to be a gigantic project.

So -- but obviously, we can’t talk about combining companies and that kind of thing on an earnings call. It’s got to be done with the appropriate process. And with that, I’ll turn it over to Brandon, our General Counsel.

Brandon Ehrhart

that’s exactly right. We continue to benefit from our relationship with SpaceX, and we’ve -- they’ve been a great partner, and we have numerous beneficial transactions with them. And earlier this year, we deepened our relationship through an investment and a framework agreement. This will allow us to continue to work with them on projects that Elon mentioned like Terafab and Digital Optimus.

Elon Musk

Yes. And there’s many other things, too. Obviously, you’ve got Grok in the car. So -- and Grok helping drive Digital Optimus. You also got Starlink being integrated into the Cybercab and Starlink will be integrated into all of our vehicles, at least for markets that Starlink is active. Because for a robotaxi situation, you need to have coverage everywhere. And there are many places even in Silicon Valley where the cellular coverage is terrible or sometimes nonexistent, which is surprising for Silicon Valley. But I know when I want to drive to work with the first 10, 15 minutes, I can’t actually do any calls because the cellular connectivity is so bad.

So we can’t have robotaxis getting stuck in these like Bermuda triangles of lack of cellular connectivity. So Starlink with its ability to do connectivity anywhere is actually quite important. So we don’t have robotaxis missing in action. And then obviously, if people are sitting in the car that they’re going to want to do high productivity stuff or entertainment. And with Starlink, you can watch 4K live sports in the car and with very low cost per gigabyte of data that’s really not feasible via the cellular system. And there’s many other situations.

That's perhaps about as loud as a CEO can wink without formally announcing a merger. If I were either SpaceX or Tesla shareholder, I think I would actually welcome the merger.

The reality is Neither of these companies is best understood or valued on their current core operating businesses. Both companies are essentially a portfolios of deep out-of-the-money call options attached to a cash-generating core. Given the cash generating core today doesn’t really come anywhere close to explain their current valuation, investors are likely paying hefty premium for the deep-out-of-the-money call options.

To realize the full potential of such call options, both Tesla and SpaceX may need to deploy gargantuan amount of capex in the coming years. Tesla’s unsupervised FSD or robotaxis, Optimus, SpaceX’s orbital data centers, and frontier model ambition, or their joint Terrafab project…all of these call options seem pretty capital intensive. And it’s really hard to know which of them will take off at what timeframe. Given such uncertainty, it may indeed make sense to have all the call options under the same roof and then add fuel to the ones that actually start to pan out. I have no idea about the probability of success of any one of such call options. In some ways, Elon Musk’s companies increasingly seem to be “Berkshire inverted”. Buffett assembled a portfolio of short-vol cash streams and used float to buy more of them; following the very likely merger of Tesla and SpaceX, MuskCo mostly appears be a portfolio of deep-out-of-the-money calls funded partly by an industrial float but perhaps increasingly more and more be the generosity of the strangers by their willingness to pay such a hefty premium for a portfolio of deep-out-of-the-money call options. I wouldn’t touch it anytime soon, but I would also not bet against the possibility that if a couple of the call options come to fruition, that may “justify” the valuation of MuskCo in the coming years.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Current Portfolio

Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.

After going through Alphabet’s earnings closely, I have decided to add another ~120 bps to my existing Meta position. I will expand on my rationale below.

Read more

Here’s a more granular segment wise growth trajectory since 1Q’23.

And these revenues weren’t really empty calories either. Operating income grew ~25% to $40.8 Billion for a 34% consolidated margin. Note that net income and EPS grew much faster than that, but it was primarily unrealized gains in the equity portfolio (SpaceX and Anthropic).

As mentioned earlier, the showstopper was Google Cloud: operating income more than tripled YoY to $8.8 Billion, with margin expanding from 20.7% to 35.6% while still absorbing the Wiz integration drag i.e. margins would be even higher if we adjust for the Wiz integration.

However, let me repeat the caveat I mentioned last quarter:

In both these segments, I should highlight that there is a slight wrinkle in comparing current segment margins to historical segment margins. In 2Q’24, Alphabet disclosed that “AI model development teams previously under Google Research in our Google Services segment are included as part of Google DeepMind, reported within Alphabet-level activities, prospectively beginning in the second quarter of 2024”.

As you can see below, there has been a noticeable step up in “Corporate costs” which includes the “Alphabet-level activities”.

Given that context, it is better to focus on consolidated operating margins than any segment level margins.

So revenues are accelerating and profits are growing at healthy margins. What are investors worried about then? I don’t want to engage in the day-to-day price action in current market environment, but there are indeed some legitimate concerns despite such operating performance which I will discuss more behind the paywall.

Subscribers get the daily journal and five+ years of Deep Dives, i.e. full-length analyses with financial models on 65+ companies. The daily is just how I think out loud between the Deep Dives!

Subscribe now

Read more