Q2 losses totaling $16.64 billion and massive layoffs of more than 15,000 employees worldwide have stained his tenure. After such a traumatic year, it's hard for a board that only cares about stock price and public perception to find a thick enough silver lining to keep Gelsinger around.

Pat Gelsinger took the reins after Brian Krzanicih’s horrendous tenure, which was mired in a lack of flexibility and open-mindedness. Brian’s team fell complacent with revenue growth, while net income was flat around $10 billion. Additionally, Intel's failure to double down on mobile and GPU computing left It grossly behind other giants (spoiler: it does not bode well when AI becomes a big thing).

Pat was applauded as the right choice for Intel. Previously the CTO, his deep engineering background was perfect for taking risky bets in R&D to set up Intel to recover their gap in competitiveness. However, 3 years have passed, Gelsinger is out, and Intel is again stuck in a rut. So, how did we get here?

Losing Their Edge

Ask any engineer who worked in California in the 80s and 90s and they’ll tell you that Intel was a Silicon Valley darling. Intel was the largest semiconductor manufacturer by revenue, had a majority market share, great financial prowess, and was recognized as a household driven by the “Wintel” boom.

Intel’s glory days were fueled by three consecutive CEOs - Robert Noyce, Gordon Moore, and Andy Grove. The mistakes started racking up when Intel brought in Paul Otellini, the first non-engineer as CEO. Otellini wasn’t too shabby while he was around. In 2012 under Otellini, Intel’s profits exceeded the total of key competitors Broadcom, Nvidia, Qualcomm, and Texas Instruments.

Where Otellini fell short was failing to bet on the right long-term investments. 2007 was a pivotal time for Intel. The iPhone came out and the company was met with a decision. Either they could reinforce their dominance in the computer chip market or bet on the burgeoning mobile chip market.

What if I told you they chose against mobile chip production simply due to a financial estimation error? Otellini, the supposedly “business-oriented” CEO, decided not to work with Apple to produce mobile chips because Apple’s ideal price was less than Intel’s forecasted cost. In hindsight, Otellini notes that the forecasted cost was wrong, and the volume driven by the mobile market was 100 times anyone's expectations.

Missing the mobile market came with a massive opportunity cost. Today, mobile accounts for over 33% of TSMC’s product line - even after the boom in high-performance computing chips. Part of missing out on mobile involved differences between Intel’s x86 architecture and the ARM-based architecture that were optimized for memory and power usage, and well suited for mobile devices. Rather than recognizing x86’s shortfalls when it came to smaller devices, Intel doubled down and believed their identity as a chipmaker and designer (IDM) would allow them to excel in mobile computing.

While Otellini dropped the ball on mobile, he did try to salvage it by presenting a wild idea to the board in 2005: acquiring Nvidia for $20 billion. The board hesitated and obviously, it did not go through. The indecisiveness marked the start of another failed opportunity. Intel would double down on CPU instead of GPU and ditch attempts to enter the graphics space. Unsurprisingly, Nvidia has emerged as a clear leader in both the consumer (gaming) and enterprise space with a commanding 88% market share and growing.

Source: Techspot

Splitting Attention

“I do not expect to join any company which is simply a manufacturer of semiconductors. I would rather try to find some small company which is trying to develop some product or technology which no one has yet done.”

In his resignation letter to Fairchild Semiconductor Company, Robert Noyce noted that he would rather work on an interesting product at a small company over a simple but large manufacturer. However, Intel is unique compared to other semiconductor companies because it is both a chip factory (fab) and a chip designer. This would be like if Nvidia and TSMC were rolled up into one company.

This worked for the first 30 years but has now led to concerns about their ability to move nimbly with changes in the market (ie producing mobile or cutting-edge graphic computing chips). It’s why Krzanich fell complacent and why AMD, Nvidia, and TSMC have overtaken Intel. As an Integrated Device Manufacturer (IDM), Intel splits its budget between two business lines, limiting the individual growth of each line. Hence, Intel moves slower than foundries like Samsung and TSMC and chip designers like Nvidia and AMD.

But Intel’s situation is not new. It turns out that operating a foundry costs billions to maintain and grow - AMD found this out the hard way. In 2008, AMD was in a similar boat. The company was bleeding cash and decided to spin off its Foundry division for $700 million. The results speak for themselves.

Source: Tim Culpan Substack

AMD’s spinoff went on to become GlobalFoundries - a $25 billion chip manufacturer. Maybe Intel can turn this around as an IDM, but the company seems better off divesting and doubling down on one or the other. Given Intel is close to production on their groundbreaking new 18A Panther Lake process, it’s more than likely their brand as a chip manufacturer will continue for the time being. So, what do you think Intel should do?

That’s all for today folks! If you enjoyed this piece covering Pat Gelsinger’s surprising retirement and Intel’s astronomical fall, hit the like button or drop a comment with your thoughts. Contact us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Thumbnail creds: https://www.hatchwise.com/resources/the-history-of-the-intel-logo

Inspired by McLovin from Superbad, Applovin is likely one of the most valuable tech companies you’ve never heard of. And there’s a pretty simple explanation for why; the company was worth half of what it is today a mere one month ago. That’s right - they were irrelevant in the tech space until recently. A year ago, they were hovering at a modest $10B valuation driven by two main products: AppDiscovery and MAX. 

So what changed? A surprise earnings report showed $1.2 billion in Q3 revenue and massive margin expansion with a near 36% net income margin. Financial success combined with their new e-commerce product line has seeded investor confidence in Applovin’s ability to expand its advertising service past mobile gaming and e-commerce to other niches.

Subscribe now

Mechanics Of Mobile Advertising

Within the overarching advertising ecosystem lies a smaller ecosystem for mobile network ads. These ads could appear as videos or pop-ups between sections of a game, as display ads at the top or bottom during your app experience, or as embedded rich media while scrolling through sections of an app. 

Mobile advertising may seem niche but spending is expected to surpass $400 billion in 2024. Part of AppLovin’s meteoric rise is that they’ve been able to hit extraordinary levels of profit facilitating only $10 billion in volume for a subsection of the economy, mobile gaming clients.

While in a niche subset of the ecosystem, Applovin still faces considerable competition from ad networks offered by the world's largest companies. AdMob by Google, Audience Network by Meta, and Amazon Publisher Services. 

Each platform has quirks, but advertisers can use one or all to promote their brand through mobile networks. Google, Meta, and Amazon are all quite similar. Each platform can be thought of as an exchange in financial markets. A network’s essential function is to present a marketplace where advertisers compete to bid for ad space on publisher apps. Several pricing models exist such as cost-per-impression, cost-per-click, or cost-per-action, based on the conversion event that matters for the advertiser.

Real-time bidding allows dynamic adjustments in the market for ad space, saving massive headaches for app developers who may have had to create ways to accept bids and publish ad content on their apps. These advancements in the ad network space have led to a new reality: the commoditization of AdTech. Unlike financial exchanges, there is no barrier to being traded on multiple networks, and it is in the best interest of app publishers to optimize multiple ad networks. This is known as ad mediation: Applovin’s plan to win the industry.

Winning In Ads

Applovin S1

This graphic from Applovin’s IPO prospectus laid out their plan of action. First, help app developers market their apps with AppDiscovery. Adam Foroughi started AppDiscovery after failing to build awareness for app projects over the years. AppDiscovery is built on AXON, the machine-learning algorithm that aims to improve the likeliness that your app is downloaded. Hint: this is where their use of AI impacts the business's long-term prospects.

While that’s cool, it does not change the fact that Applovin competes in a playing field among the largest data aggregators. Google, Amazon, and Facebook have built trillion-dollar companies scouring user data and optimizing the ad conversion experience for companies that use their ad networks. Given that, what must Applovin do to win in this cutthroat industry?

Applovin MAX. The company’s edge comes from its ability to succeed in mediation. MAX can tap into Applovin’s native ad network, which they testify to be 5-6x greater than many other ad networks (probably except Google or Meta). If that’s not as effective for you, MAX can optimize a combination of other networks to deliver the best eCPM (effective cost per mille) to deliver the best ad for the publisher. Each ad has to deliver the greatest revenue, with the caveat that they are quality: relevant, easy to skip, and appropriate for the audience.

To drive the greatest return, you want to find the best mix of ad units, fill as much space as possible, and optimize revenue. All while growing and retaining your audience.

Ultimately, customers are partial to the platform that drives the greatest amount of money. As such, MAX works to optimize ARPDAU (Average Revenue Per Daily Active User) while AppDiscovery tackles UA (user acquisition). The two give a mobile app developer a full growth flywheel, which serves as a growth engine for an indefinite time.

Applovin - Why Mediation is Critical

MAX has been winning since 2021, proving that mediation is not commoditized. Adam wrote a letter to the public in late 2021 explaining the importance of mediation as well as its impact on a few mobile gaming apps that utilized MAX.

That’s all for today folks! If you enjoyed this piece covering Applovin’s surprise takeover of the AdTech industry, hit the like button or drop a comment with your thoughts. Simply contact us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Moore hypothesized that this pace of growth up until the mid-1960s would continue for at least 10 years. And he couldn’t have been more right. This growth rate has stayed true even till today, with the most sophisticated TSMC fabs pushing out 3 nm chips. Apple’s M3 Max is a prime example of a 3nm chipset and has nearly 100 billion transistors packed into one chip.

The bottleneck in maintaining Moore’s law was not even in the hands of TSMC, Samsung, or Intel. In reality, to pack in more transistors in a single chip, there would have to be an evolving technique that made it possible to etch at an increasingly nanoscale. That’s where ASML comes in.

Similar to how newspapers used lithography to create print plates, photolithography was used to etch designs in chips for memory or computing. ASML perfected photolithography, which quickly became the method that offered the highest, throughput, lowest cost, and greatest accuracy.

Photolithography was discovered to be an effective way to etch circuit designs onto a wafer way back in 1950. Silicon wafers were coated with a photoresist, which a light source would then etch to create an integrated circuit. These processes initially used natural light sources visible to the human eye, but challenges arose as the need for smaller chips increased.

ASML had to innovate. By the 1990s, deep ultraviolet (DUV) light allowed chipsets to reduce to 193 nm processes. But DUV was just the start. Extreme ultraviolet (EUV) light, previously thought to be a theoretical concept, was manufactured into reality to enable the single-digit nanometer processes as seen on the Apple M3 Max for example.

However, we aren’t here to discuss the engineering feat at play. I’m surely not the one to be doing that. Moreso answering - is photolithography a defensible business model?

Subscribe now

ASML’s Volatile Monopoly

ASML smashed it with photolithography. By the 1990s, the incumbents Nikon and Canon were pushed into obsolescence due to their inability to keep up with the high R&D costs associated with cutting-edge ultraviolet technology. While they still existed in the cheaper areas of the lithography markets, the clear leader was ASML, which controlled 82% of the lithography market. However, their control stretched far past market share.

ASML is still the only manufacturer of these massive photolithography machines made up of over 100,000 components. Unlike the subject of our previous piece, Rivian, ASML orchestrates an intricate horizontal supply chain with over 5,000 suppliers. This leads to them outsourcing over 80% of the finished product. 

Given there’s not clear competitor to their EUV product, its safe to say that ASML should continue to maintain its monopoly through the next few decades. However, recent times haven’t been as favorable for ASML. Following an earnings blunder in which the company mistakenly released its guidance a day early, the stock tumbled more than 20% between 2 trading sessions. 

The fact is that semiconductors are tricky. While you see the crazy highs that Nvidia or SMCI put up, the industry often sees periods of silence. How often does a company order new $150m semiconductor equipment or update to the latest GPUs? Infrastructure is expensive, and cycles of low sales make it difficult for a company such as ASML to be properly valued in a market that isn’t generally favorable to volatility.

The good news is ASML is always ready to bounce on the next cycle. With a 15% Revenue to R&D expense ratio and almost a 50% expense ratio against net income, ASML is well aware of the investments necessary to keep them at the cutting edge of the manufacturing business. A consistent 15% R&D spend means well more than $3B in R&D over the next few years. 

Judging Defensibility

So far we know that ASML’s business model heavily relies on the ability to stay ahead in photolithography. The industry also operates in boom/bust cycles which causes long and silent winters. For context, take a look at the Philidelphia Semiconductor Index (PHLX). $100 bought into the fund between 2001 and 2002 would be worth almost the same 18 years later.

The story of ASML is a story of innovation. The industry innovates beyond imagination when ASML takes something theoretical and turns it into reality (like with EUV lithography). A prime example is the internet bubble and what came after. ASML’s advancements allowed value creation in the form of better computers and mobile phones.

The dot-com bubble isn’t the only example of this. Recent advancements in chip technology coupled with a greater need for data processing enabled the so-called “AI boom”. Nvidia pushed itself to lead the market, TSMC made a name for itself, and ASML, the backbone of the operation, added over $125 billion in value to the company just over the past 4 years.

So when it comes to defensibility, here is the thesis. As long as ASML can innovate in a way that fuels the next wave of innovation (Internet, Cloud, AI, and the next big thing), there are no questions about their moat.

That’s all for today folks! If you enjoyed this piece covering ASML’s domination in the semiconductor manufacturing industry, hit the like button or drop a comment with your thoughts.  Simply contact us by replying to this email if you have any insider takes or leave a comment on our Substack page!

If you set aside supercars and high-end luxury, as long as you last and produce at a fair price point, you may eventually hit economies of scale. In 2023, Toyota produced 11.2 million and thus enjoyed margins of 11% on its cars. Although, the per-vehicle margin story is expected to be abysmal. Elon notes that revenue from replacement parts and servicing expands bottom-line margins.

“Prototypes Are Easy, Product Is Hard”

Think about entering this market as a startup. Let’s say you have your prototype. To successfully build your first car, you need to spend hundreds of millions on R&D, building a manufacturing plant, hiring workers, procuring materials to make your car, and ensuring its quality. Then you have your first car, but nobody knows about you. Alongside production, you need to build relationships with dealerships and marketing your brand.

It’s hard to call Rivian company a startup. The company has been in the industry for over 15 years and raised 13 rounds since 2009 for over $10.5 billion from BlackRock, Amazon, Volkswagen, Cox Automotive, T. Rowe, Ford, and others. Despite the constant cash infusion, the company still struggles its footing. Issues with scaling production and demand for its vehicles haunt the automaker in 2024.

Subscribe now

A Product Focused Company

RJ Scaringe loved two things as a kid: cars and nature. Tinkering with car restorations and spending time rowing down the Indiana River during his childhood, RJ had an intimate relationship with both hobbies. In due time, he realized the competing forces at play: these gas-guzzlers that brought him so much joy were destroying the environment he loved.

This dilemma followed him to MIT, where he wrote his thesis on low-emission alternatives to traditional ICE engines, but low-emission would soon become no-emission. RJ and a team of scrappy engineers founded Rivian in 2009, more than 12 years before they delivered their first vehicle.

Rivian was founded as more than just a car company. RJ focused relentlessly on building a culture of adventure and catering to people who enjoyed the Great Outdoors. Rivian built on this mission by mirroring the likes of Figma, Notion, or even Apple - passionately building great consumer products.

The R1 speaks for itself on this matter. Rivian spent months observing people use cars at parking lots of beaches, mountains, trails, and even strip malls to understand how to design and add valuable features that fit the mission of adventure. They looked out for DIY hacks people had, like unconventional ways to fashion a bike rack or things that people relied on like flashlights.

Each feature on their flagship truck was built with a customer need in mind. Take the Gear Tunnel. This empty gap of space between the truck’s bed and passenger compartment could have been closed off, but after observing surfers and golfers weirdly storing their boards and clubs, Rivian decided to create a comfortable space to store both. Getting dark on your hike or drive? Rivian R1 series cars have a built-in torch for this specific reason. The feature design was so nuanced that Rivian even added a rubber strip at the end of the tailgate to set items against it after noticing they would slip off on other trucks.

Betting The House

“From the very beginning, we said that we need to own the electronics in the vehicle—that allows us to own the network architecture, software stack, and hardware within the vehicle as it evolves…it’s really important to control the full stack.”

RJ knew that third-party manufacturers, equipment, or software providers were eating the margins of legacy automakers. To be successful in the long run, Rivian was going to need to own the entire stack. Obviously, this isn’t new. Tesla long owned its entire production and distribution line. But at least Tesla had a first-mover advantage and even some revenue and a small profit in its early years with the Tesla Roadster.

Rivian attempting to do the same was essentially betting the house because if anything didn’t work their way, the company would likely be closed for good (ie Fisker and Lordstown Motors). Plus, their first production car would be delivered over 12 years from their first fundraise.

Owning the entire stack would require billions (see total funding from introduction) and over 15,000 employees to make it happen. The struggle has been quite evident. As recent as the first few months of this year, Rivian lost around $39,000 per car. Zooming out paints a better picture.

In 2021, Rivian sold 580 cars at around $95,000 per car, losing over $790,000 per car. Then in 2022, the company tightened up margins by selling significantly more cars (~20,000) and losing $239,000 per car. Finally, in 2023, the company sold ~50,000 cars and again trimmed margins by saving around $100,000 per car, losing around $129,000 per car.

Finding Clarity

Today, Rivian’s stock ($RIVN) is down almost 91% from a high of $130 after a post-IPO pop. Competitors are either ramping down or cutting plans due to high per-unit losses. Rivian’s own ambitions to scale to 1 million in production by 2030 have taken a hit.

In March 2024, the roadmap became crystal clear. Rivian announced plans to deliver its next line of cars, the R2 series. R2 is set to take up part of the 215,000 production capacity in Rivian’s Normal, IL plan and sell for almost half the price of the R1. If the company can reach a profit with the R1 series as they claim, they can use that along with substantial funding from Volkwagen to stay afloat and roll out the R2 line.

A revamped R1 line and a lighter price tag on the R2 should drive demand for the business, and material cost reductions on R1 Gen 2 and R2 should expand per-unit margins for the automaker. Bottom-line improvements would cycle back into the business and allow production to scale and headway for R3 and future car lines. Easy said, but impossible to do.

That’s all for today, folks! If you enjoyed this piece covering Rivian’s rise to the number two EV producer in the US, hit the like button or drop a comment with your thoughts. Simply reply to this email if you have any insider takes or leave a comment on our Substack page!

Motorola was a pioneer in the cell phone world with the launch of the DynaTAC 8000X. This bulky phone cost a whopping $4,000 and took almost 10 hours to charge, offering just 30 minutes of talk time. Fast-forward nearly two decades, and mobile device usage had exploded, driven by features like full keyboards, internet access via WAP, and the transition to 3G networks that made connectivity faster and more affordable.

By the early 2000s, the adoption of flip phones peaked, solidifying a mobile-first world. However, messaging technology struggled to keep pace with hardware advancements. SMS, launched in the 90s, was clunky and expensive - costing up to 20 cents per message in 2008. As a result, SMS usage was low, with the average user sending only 35 texts a month at the turn of the century.

SMS had its limitations: messages were capped at 160 characters, leading to abbreviations like "LOL," "TTYL," and "BTW." Issues like unreliable delivery and out-of-order message threads were common frustrations.

Building WhatsApp

Hidden within this mobile boom was Yan Koum, a young engineer who had recently left a 9-year stint at Yahoo. After taking some time off to travel, Koum found himself messing with the newest development in the mobile phone space: the 1st generation iPhone. XCode 3.1 brought SDKs for iPhone, allowing developers to build and publish applications on the iOS AppStore.

So, Koum decided to download XCode and play around with what it had to offer. The possibilities were crazy - the iPhone had a full TCP/IP capability, full internet and networking capabilities, and access to selected Apple APIs that allowed developers to tap into parts of the end user’s experience.

These networking capabilities are what propelled the launch of WhatsApp v1.0. When Koum was away taking time off to travel, he struggled to communicate his status to people in his contact book. PC-native instant messaging platforms still prevailed, meaning it was almost impossible to communicate with people if you were away from a computer.

Enter WhatsApp, a status app to help users communicate whether they are busy, away from their computers, traveling, or more. Despite signs of organic growth, this first version ended up being a failure.

Finding The Value Prop

The focus in business is generally fortifying revenue streams and figuring out new ways to monetize the product. Cost is simply a by-product or even something justified to drive future growth. But what if it couldn’t be that way?

The business of messaging platforms is tough because you’ve got to maintain a highly performant platform with little promise of money or monetization. People already pay for their phone plan - what’s your platform providing that iMessage or SMS doesn’t already do? Some platforms argue for privacy, some cater to professional or academic use, and others are industry-specific. Bottom line, catering to a niche seems to be an acceptable justification for charging a fee.

WhatsApp wasn’t substantially unique or different. Instant messaging platforms, SMS, and even copycats popped up along the way to try to take their own piece of the pie. Furthermore, WhatsApp’s adherence to user privacy and ad-free principles was strong. Even when Koum sold the company to Facebook for $19.3 billion, he made the Zuck swear to never mess with user privacy or run ads. His strong adherence to these values eventually even led to him leaving the company in 2018.

The app’s revenue model in 2012 was simple but flimsy: 1 dollar to use the app and 1 dollar every year to keep your subscription. However, it sufficed! At the time Koum sold the company, the platform was already pulling in $700 million a year, far exceeding expectations. Growth exceeded rival platforms, as the user count far exceeded 450 million well into Year 4.

Regardless, a flimsy long-term revenue model meant viewing the business the other way around: optimizing cost rather than focusing on revenue drivers. Fun fact, WhatsApp maintained a mind-boggling $7 million average revenue per employee due to a lean team of only 35 engineers. This meant simplifying the stack and focusing on what was needed - a minimalistic approach to problem-solving.

“Do one thing, and do it well”

Koum has frequently been quoted for his emphasis on focus, and this was especially true for WhatsApp. Because of WhatsApp’s long-term funding restrictions, the engineering effort had to be laser-focused towards a single pillar. Given the core of WhatsApp is message-based communication, this had to be messaging. An emphasis on minimalism even carried down to their tech stack. Choices of Erlang, FreeBSD, and XMPP helped simplify what their engineers were working with while also being able to handle throughput and scale.

Erlang specifically is a great choice for a messaging platform. Erlang was suited to juggle communications between a huge number of users and was very malleable with deploying new code on the fly. Additionally, the language was built to handle concurrency which was especially useful in real-time messaging platforms.

FreeBSD was the OS of choice for WhatsApp’s servers due to its cost-efficiency and open-source stability. Koum and Acton became familiar with it during their time at Yahoo and observed it in use at other tech companies such as Dell, Netflix, Juniper Networks, and more. XMPP was a natural choice as the messaging protocol standard and due to its ability to offer almost near-real-time exchange of data.

WhatsApp Today

Since Facebook acquired WhatsApp in 2014, the platform has grown to over 2 billion active users, driven by major markets like India, Brazil, and Indonesia. India alone has over 500 million users.

Source: Rasayel Blog

While WhatsApp's original simplicity remains, it has relaxed a bit, thanks to financial support from Meta and revenue from WhatsApp Business. The company now employs nearly 3,000 people and brought in over $1.27 billion in revenue in 2023.

That’s all for today, folks! If you enjoyed this piece on WhatsApp’s humble growth story, hit the like button or leave a comment with your thoughts. If you have any insider takes, simply reply to this email or leave a comment on our Substack page!

Artists have always been lone wolves, working for themselves or compartmented away at some large company.

Thus, design software was created with this in mind. Photoshop, Sketch, and other design tools all existed in a vacuum, hidden away in the individual desktops of each user.

Decentralization defined design software until almost 2018. If you were a designer before Figma, your job may have looked like this. Your product manager sends you a bunch of requirements, you mock up a design and email it to the team for feedback. Repeat this process until everyone on the team is pleased with your work. Seems normal right? But that’s pain point #1. This iterative process led to a ton of version control since you had to keep updating your Photoshop or Sketch file. This only worsened if you were working with a team of designers since you had to keep track of who did what and merge it all together.

It was impossible to design an entire platform in harmony. Version control should have all been consolidated under a single document, allowing multiple users to edit, comment, and view throughout the creative process. That’s exactly what Figma’s multiplayer solved.

However, there were concerns with collaborative editing. Since everyone on the team could hop on the file, designers were worried that their managers would be activity-tracking or micromanaging employees. Several users also pointed out their preference to work alone, feeling annoyed or anxious with others watching.

Bottom line, Multiplayer was a groundbreaking feature that had to be added. Let’s get into how Figma accomplished it in a year of work.

Building Multiplayer

Figma already had a reputation for having a great feel. The challenge here was to maintain the same level of performance with the added complication of handling multiple edits on a single document in real time.

The technology already existed. Google Docs and the rest of the suite supported real-time editing so to say it was impossible was an exaggeration. Figma ended up doing things differently, but by no means were they trying to reinvent the wheel.

As Evan recalls, Figma used a client/server architecture with several “Multiplayer” servers dedicated to resolving and syncing changes between multiple users.

On the client side, each user had a local copy, where updates would be synced over a WebSocket connection to the cluster of aforementioned servers. This sort of implementation allowed robust handling of various edge cases. Offline editing, undo handling, and conflict resolution would be much easier with both a local copy and a master copy on the server.

Here’s where Figma started freeballing. Up until that point, the popular algorithm for handling collaborative sessions was an Operational Transformation (OT). OTs made it possible to handle index changes caused by simultaneous edits. 

For example, imagine 2 people were trying to change the string “abc”. If one was trying to add an “x” to the beginning and the other was trying to delete the “c”, the result would be a “xab”. These functions typically operated by specifying the index position the user wanted to change; Insert[0, “x”] or Delete [2, “c”]. But of course, if one operation is executed, that changes the index position of the second operation. OTs handled that index change to make sure consecutive operations were handled appropriately.

Turns out the most effective use-case for OTs was indeed…text editing! Figma had to search for a solution that was flexible and had robust data type support. Thus, they chose to use something called CRDTs (conflict-free replicated data type). CRDTs guaranteed convergence at some point in time, even though individual files may look different up until that point. 

Many sorts of CRDT approaches exist. For Figma’s case, one known as the “last-writer wins register” suited them the best. Here, edits for one object can be made by any number of different people. The final value would simply be determined by who made the last change.

Obviously, Multiplayer involved much more complication like scaling servers when needed, syncing properties, syncing trees, and handling editing of ordered sequences. But with CRDTs and a client/server WebSocket architecture, they were off to a good start.

More Product-Led Growth

The concept of Product-Led Growth even existing seems so bizarre to me. Why are we as customers using products that other people have to convince us are good? All companies should be product-led growth companies and the ones that aren’t can only succeed for a limited time. Think of the world’s greatest technological inventions up until this point. No one had to be convinced to use the first web browser, buy the iPhone, query on ChatGPT, and others. They spoke directly to our needs as humans.

And with budget restraints and recessionary fears, sales-led companies are finally catching on. Recently, Dell announced they were laying off 10% of the workforce, with cuts specifically to their sales team. Klarna, a BNPL company, laid off its customer support roles after it realized the power of AI assistants. It is the survival of the fittest, and in today’s environment that means companies with actual product innovation, not just those that can just throw money at their problems.

That’s all for today folks! If you enjoyed this piece covering Figma’s unprecedented approach to collaborative design software, hit the like button or drop a comment with your thoughts. Simply contact us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Regardless, its impact was so existential that the evolution of social media would look much different today without it. Photoshop wasn’t just some app. It was a verb - you didn’t “edit” an image, you photoshopped it. All those celebrities like Beyonce got exposed for editing their videos just a tad too much. Much lesser-known stories exist too. Zilla van den Born went viral back in 2014 for faking her trip through Asia, recreating photos of East Asia from her house and around her hometown of Amsterdam.

The term’s ubiquity conveyed the impact of the product, but also the company, Adobe. Ever since its early days, Adobe has dominated the photo, video, and design space growing to $238 billion in market cap while they were at it. Photoshop alone has over 40% with Adobe overall having comfortably more than 80% market share amongst enterprise use graphics software.

If you were to introduce a new product that stood a chance against Adobe, it had to be objectively better. That’s exactly what Figma did.

Subscribe now

WebGL

Launched in 2011, WebGL established a new way for browsers to take advantage of the graphics capabilities of any device. Remember our post on NVIDIA? In their early days, OpenGL was the standard way for anyone to interact with a GPU. OpenGL provided an API that people could use to harness the power of their GPU.

Furthermore, OpenGL offered multiple language bindings, one of which was used to create WebGL in 2011. With WebGL, you could render high-performance 2D and 3D graphics through a browser.

This was a crazy breakthrough. Before WebGL, there was no way for cloud-based web apps to take advantage of the opportunities offered by a device’s GPU. That’s why if any software ever needed to work with a GPU, it would only be available as a desktop app. Consider Photoshop. Since it was a desktop app, it worked directly with the computer’s hardware to do what it needed.

Naturally, this meant that creativity wasn’t the only limiting factor for why Figma launched when it did. The tech was barely there. WebGL was a primitive version of browser-based graphics rendering. From 2018 and beyond, lower-level languages like WebAssembly and future versions of WebGL made browser rendering more robust.

Launching Figma

If people cannot create, then that’s a moral wrong. Our tools should make it simple enough so that everyone can be creative.

That was the gist of Bret Victor’s talk over 12 years ago, the very talk that inspired Dylan Field and Evan Wallace to quit their jobs and find a solution to their own issues with design tools. Dylan and Evan had previously worked in creative industries at Flipboard and Pixar, using design tools that they quickly got frustrated with. The software wasn’t intuitive, lacked collaborative features, and missed a sense of fluidity.

You’d think that VCs were fishing for a lean startup that launched asap, but Figma was funded in 2013, more than 3 years after they were first funded. Even though they started in 2012, it took more than 4 years before Figma launched to the public, as they worked to create a platform that had the right feel.

That may sound super ambiguous, but consider the engineering problems for a minute. Figma was trying to create a browser-based vector editing software using a brand-new rendering API. Eventually, that platform had to support simultaneous editing, which meant considering load and how to handle versioning.

But back to the feel - ambiguous - but if you see it, you’ll know it. To get that feel, Figma had to ensure that their platform maximized FPS but also supported lower end and older devices. It needed to beautiful everywhere it existed.

Product Led Growth

Long before the days of Figma, Notion, Slack, Dropbox, and the other numerous SaaS companies of the modern age, there was Oracle. Greg Isenberg’s post properly sums up their business model.

No one likes the landlord, but you gotta give them credit, they make a boatload of money. Today’s enterprise software companies choose to focus on letting the product sell itself. Figma’s equivalent of Oracle’s enterprise sales team is Figma’s engineering team. And by god, those engineers had a selling to do. Before they could get companies hooked on their product, they had to drastically change the fundamental pillar of every designer.

Tune in for next week’s post to discover how Figma changed the landscape of graphic design by focusing on product-led growth. We’ll even break down what product-led growth means. However, that’s all for today! If you enjoyed this piece covering Figma’s early days and the development of WebGL, hit the like button or drop a comment with your thoughts. Simply contact us by replying to this email if you have any insider takes or leave a comment on our Substack page!

In November 1999, Priceline launched a crazy idea; a name-your-own price program for gasoline. As a customer, you could load in your credit card information, and input how much you were willing to pay, and if the seller accepts your price - you got the deal! WebHouse Club, the program’s name, also offered the same system for groceries.

Offering consumer essentials over the internet wasn’t a new idea. Other classic dot-com busts launched with the sole idea of providing groceries. Names like Webvan and PeaPod grew to billions in valuation in the same industry.

Just looking at the top-line numbers, WebHouse was a massive hit. Within a year, Priceline claims it grew WebHouse to over 2 million customers and 13,000 participating retailers. Buzz also spread through the investing community, with investments from Paul Allen’s Vulcan Ventures, Wit Capital, and Goldman Sachs. 

WebHouse wasn’t alone. Around the same time, Priceline also launched a more scrappy version of eBay - a marketplace for unwanted items called YardSale. People could sell off anything they owned and others could bid for it.

Travel Economics

The story doesn’t end as well as it started. In late 2000, Priceline divested both YardSale and WebHouse after being unable to “raise more capital for a business that was not yet profitable”. The company took a $189 million hit on outstanding obligations. How was it that a name-your-price tool worked so well with Hotels and Flights but not so well with other markets? 

Simply put - economics. Consider the mechanics of a flight or a hotel room. The travel company that hosts the seat on the airline or the room in the hotel can do their best trying to judge demand, choose the right locations, and even offer the right deal to get you to book them, but at the end of the day, a large amount of seats and rooms go empty. 

According to the St. Louis FRED, anywhere between 70-85% of a given flight is full, meaning at most 1 empty seat for 3 seats. The numbers are even more grim for hotels. After accounting for pandemic recovery, the yearly peak for hotel occupancy is 70% only during the summer, with yearly lows at around 50%. 

If 1 of every 3 seats or half of every hotel is empty, the company hosting the seat/room loses out on the opportunity to capture any revenue at all. Even if that seat was sold for a 50% discount, they would be able to create some cash to help operating costs. Airlines long knew that but still refused to sell below a certain point which they called the “price line” (hence the name of the company).

That is, they didn’t want one area of the internet to show the full price and another to show a discounted price - cause people would migrate to the discounted price. Priceline’s team figured - why not let the customer guess the price? The company wouldn’t give away its secret “price line” and it would enjoy re-couping some operating costs. The flight is gonna take off at the same cost to Delta either way, so it makes logical sense to fill that seat at any non-zero price (adjusting for per-person costs).

The thing is, gasoline doesn’t work the same way. If you and I bid on gasoline at a certain price, the gasoline provider doesn’t lose out if he doesn’t take your bid. He has a variety of options.

He could easily just offer the next week, or wait until someone comes to his pump and fills up the tank. We’ve previously talked about price sensitivity - a concept in economics that measures the change in demand after a unit change in price. Gasoline and Groceries are prime examples of price insensitivity - people are willing to buy at the market price because it's a necessary good, they need to fill their car up and need food either way.

Notching Billions

Priceline’s success was evident, but the numbers revealed a different picture. In 2003, the company notched $836 million in revenue, significantly lower than other American competitors like Expedia at $2.41 billion. The name-your-price tool catapulted them, but they knew they would need to add other features. Starting in 2004, they expanded past auction pricing, eyeing to become a one-stop shop for travel.

However, no feature they could add could make a greater impact than their 2005 acquisition of Booking.com. Priceline paid just around $135 million for the company but increased their profit by more than 10x that in the first 10 years. Between 2003 and 2012, Priceline’s profit catapulted from $10 million to $1.1 billion.

By acquiring Booking.com, Priceline had made a strong move towards distinction. American competitors were focused on the local market - but missed the global opportunity. Booking, a Europe-based company, had strategic reach into a European market that was laid out significantly differently than the US.

For example, Booking.com primarily operated off the agency model, something which Europeans were much more used to. Rather than a client paying through Booking.com’s website, they would pay the hotel directly, with Booking.com taking a cut after.

Other factors of difference made an impact too. Booking.com enjoyed a European landscape with much less consolidation, meaning smaller hotel joints benefitted from marketing themselves on the platform. And of course, there’s the classic trope about Europeans - they just simply vacationed more. Amazing in terms of customer retention. Fast forward to 2023, Booking.com notched $21 billion while Expedia barely cracked $13 billion. 

That’s all for this one folks! If you enjoyed this piece covering Priceline’s growth story, hit the like button or drop a comment with your take. Simply contact us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Since then, investments in AI and computing have propelled the market to new heights. Over the past 2 years, the S&P 500 rallied over 2,000 pts - from around 3,500 to past 5,500.

This market rollercoaster has brought a few names to the forefront. As the world tried to make sense of the storm, names like NVIDIA, AMD, TSMC, and ASML quickly gained notoriety. After all, anyone with a pulse over the past year has heard about the astronomic rise of NVIDIA. And then there’s the story of the little engine that could - and very much is - SMCI.

Super Micro Computer Incorporated - an enterprise server company - has been on a tear since the end of 2022. Before recent declines, the stock rose over 1,200% over the past 18 months.

That’s until recently - SMCI took a major hit last week. Hit with a major short report from Hindenburg Research for allegations of improper revenue accounting, the company tanked over 25% during the trading session.

Something must be intrinsically driving the company up. How did this no-name company doing $11 billion in revenue rise to over $60 billion in market cap? Let’s dive into SMCI’s competitive advantage in the enterprise computing industry.

The Server Market

Supermicro is a global technology leader committed to delivering first-to-market innovation for Enterprise, Cloud, AI, Metaverse, and 5G Telco/Edge IT Infrastructure. We are a Rack-Scale Total IT Solutions provider that designs and builds an environmentally friendly and energy-saving portfolio of servers, storage systems, switches, and software, along with global support services.

This sounds like a lot of buzzwords. Basically, SMCI provides everything related to servers for companies that want to build out their data centers. Server racks, storage, networking, and all the small components that are involved in data centers - a “Rack-Scale Total IT Solution”. 

With all the hype around the stock, you’d assume they must control the market. But SMCI is literally the underdog on the block. Competitors in the infrastructure industry include Dell with EMC, HPE, Foxconn, and others. SMCI only has a 5% market share, compared to Dell’s 51% or HPE’s 14% market control.

Regardless, SMCI is a lean machine. Think about their largest US competitors. Dell, HP, and Lenovo weren’t originally enterprise server suppliers! Each of them started out as a consumer technology company and rode that wave through the dot-com bubble. Now, those divisions exist as mandatory deadweight. They do nothing other than bloat and pollute the margins of a lethargic and slow-moving company.

SMCI shines here. Even after laying off 10% of its staff recently, Dell has over 100,000 employees. SMCI can develop, build, and ship their maximum production with only 5,100 employees. In financial terms, SMCI’s operating costs are only around 3-5% of their overall revenue. Dell’s is much more than 10% - a key figure when gauging the outflows of the company’s revenue.

Modularity

Current CEO Peter Liang co-founded the company with his wife back in 1993 - a time when titans walked around in Silicon Valley. To establish any presence, SMCI had to double down on its competitive advantage - size and speed. “Design the best product” Liang believes. If our product isn’t the best, how do we compete with others?

Innovate or die - A common phrase you hear in Silicon Valley, but even more apparent as a small operation. SMCI targeted PCs out of the gate but quickly observed the saturation in that market with Apple, Dell, HP, and others controlling the market.

“Thirty years ago, when I designed our product based on building-block architecture, I did not know it would be as powerful as it is today”

In hindsight, getting ousted from the personal computer market was a boon for SMCI. Liang saw past the consumers, straight into the real golden goose - enterprise. Over the next 30 years, SMCI would develop on two parallels - modularity and green computing. Unlike their competitors, SMCI offers what they call a “Rack Scale Plug-And-Play” solution - a hot-swappable server rack that can support changes when there are improvements in the industry.

Isn’t it all coming together now? SMCI rallying 300% just YTD makes so much more sense knowing that the company’s core competency is literally to build server racks that are excellent in accommodating improvements in the computing industry.

For training massive foundational AI models, Supermicro is prepared to be the first to market to release NVIDIA HGX B200 8-GPU and HGX B100 8-GPU systems.

Before NVIDIA even unveiled its newest chip in March - the Grace Hopper - SMCI was prepared to unveil their liquid-cooled solutions to support the newest NVIDIA GPUs. Being so close to the largest GPU manufacturers helps SMCI work in lockstep with the leaders - accelerating time to market and innovation for the next-generation server.

Being a tiny player isn’t even really hurting them. When addressing their new liquid-cooling server racks, Liang mentioned how they can produce over 1,000 new liquid-cooled racks monthly - something for which they already have a backlog. Overall, however, the company pumps out over 5,000 full-scale racks every month.

SMCI’s core competencies - being lean and modular - are at the crux of what drove the stock up multitudes during the AI wave. If Liang sticks to these strengths, the argument could even be made that SMCI is undervalued. Let me know your thoughts in the comments below!

That’s all for this one folks! If you enjoyed this piece covering SMCI’s recent rise to fame, hit the like button or drop a comment with your take. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Last week Starbucks announced a groundbreaking change to their current leadership, ending the short tenure of CEO Laxman Narasimhan. 

This week, the company made headlines again with the compensation package of Brian Niccol, the incoming CEO. Here’s what the SEC fillings indicate: 

During Mr. Niccol’s employment with the Company, he will receive an initial base salary of $1,600,000 per year and…be eligible to receive annual equity awards with a target value of $23,000,000…Mr. Niccol will receive a cash signing bonus of $10,000,000, in consideration for cash incentive opportunities that are being forfeited from his current employer [and]...will also receive a replacement equity grant with a targeted grant date fair value of $75,000,000.

Bonkers…that adds up to more than $100 million in possible compensation in the first year alone. This is the type of Blind post that needs to show up on my feed. In addition to the $100 million in first-year comp incentives, Niccol can work remotely and take a PJ to the headquarters (~1,600 miles away) 3 times a week to comply with their RTO policy.

What went so wrong with Laxman that Starbucks had to cough up $100 million to fix their problems? Is the fix simply to swap out the CEO? Let's break it down.

Venti Woes

Even though his tenure was short, there were some memorable moments from both the former and incoming CEOs. You guys remember Niccol’s viral “head nod” interview. And obviously, there’s the famous CNBC squawk box when Laxman gets grilled by Jim Cramer (of all people).

I’ll save you the time of watching it. Narasimhan uses the words “action plan” more than 5 times in the 8 minutes Cramer grills him but only once drops the specifics - weekly drops and bundle deals on the app. Nothing original. Instead, Narasimhan blames the company’s troubles on a shrinking wallet and a drop in occasional customer visits.

I beg to disagree. If anything, customer demand is currently booming. Coffee consumers are mostly split into two categories: those who make quick drive-thru visits and those who grab a drink and sit down.

Check out any local coffee in Austin. Go after 9 am, and you struggle to find a spot. In a city with over 2 million people, Austin has over 14 coffee shops for every 100,000 in population. 16% more than the average city. I bet you could find the same for those in the Bay Area, New York, Seattle, and the other major hubs of the nation.

And customers prefer local coffee shops. In a survey by Premise, customers indicated that they are more likely to spend an hour sitting at a local coffee shop than at Starbucks due to ambiance and preferences for quirky, colorful, and different. 

Such a preference alone is daunting for Starbucks. Rewinding 40 years back to their roots, Schultz initially envisioned Starbucks as the American version of Italian coffee houses. A symphony of people and interactions. 

“In each shop I visited I began to see the same people and interactions, and it dawned on me that what these coffee bars had created, aside from the romance and theater of coffee, was a morning ritual and a sense of community,”

That is all out of the door. Over Narasimhan's term, prices increased by more than 30%. While wallets may have tightened up over the pandemic, price increases outpaced any stringent budgets. Just think about the prices 10 years ago!

And then you have the drive-thru crowd. Let’s put the anecdotes and fluffy stories aside and look at the facts. 

According to Placer.ai, among four of the major coffee chains — Starbucks, Dunkin’, Dutch Bros and Biggby Coffee — only Starbucks saw negative traffic measured in visits per location.

Isn’t the drive-thru industry booming? The growth of drive-thru-only chains like Dutch Bros over the past few years further cemented the rise of the new coffee generation. With double the sales growth of Starbucks - 24.2% versus 12.5% for Starbucks - and a leaner business model that focuses on output and value than a large store footprint, Dutch Bros is driving the future of the drive-thru-only category.

Dutch Bros isn’t alone. Scooter’s, 7 Brew, and The Human Bean are among other pure drive-thru restaurants with massive growth recently.  

Swap Solved?

Does Brian Niccol's swooping in change Starbuck’s trajectory? Maybe. The ex-Chipotle CEO took a large role in the company’s success. Just 2 months ago, the company announced a 50-to-1 stock split, splitting the stock price at record highs at $3,000 after Niccol’s 6 years of 550% growth.

Niccol comes in with a specialty in understanding digital channels and strengths in optimizing in-store operations. Last year, Chipotle saw an 11% increase in comparable sales while competitors faced declines. It’s safe to say that Niccol understands the restaurant space, but where’s he spending time optimizing Starbucks?

Feel free to drop your thoughts - but I think Starbucks’ mobile app (while acting as a huge cash infusion for them) killed their product. Focusing on the mobile app added to issues. By hammering down promotions and incentivizing mobile app use, people shifted to Starbucks as a touch-and-go option, diverting from the company’s original vision. All Niccol has to do is focus on driving prices down through operational improvements or whatever else he can think of. Once prices are back down, Starbucks’ 16,000 store locations and ubiquity will do the rest.

Fortunately, there’s a lot to be excited about. Even though it’s hitting 100-degree weather here in Texas, Starbucks dropped its fall menu this past Thursday. Pumpkin spice is back, along with my personal favorite - the Iced Pumpkin Cream Chai!

Oh and about the subtext for today’s newsletter - Chipotle’s HQ was originally Denver, CO till Brian Niccol swooped in and moved it to his town of Newport Beach. Will he do the same with Starbucks? Time can only tell.  

That’s all for this one folks! If you enjoyed this piece breaking down the downfall of Starbucks and if they can get back in the game, hit the like button or drop a comment with your take. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

SEEKING HUSBAND FOR MAIDEN, COMPANIONSHIP FOR WIDOWER, WEALTHY YOUNG MAN LOOKING FOR TRUE LOVE.

Lonely Hearts Ads - The dating app before there were dating apps. Since the 1600s, eligible individuals seeking companionship used to place matrimonial ads in their local newspapers. You’ve probably heard the rhetoric, but here are a few examples.

Eventually, newspaper ads became dated when the telegram revolutionized real-time communication. Phone dating took form, and people could leave their numbers on a dedicated line for others to find. Paid services offered the opportunity to hear messages from people looking for someone. Usually, you could take their numbers (which they left on these lines) and call them to initiate a conversation.

Then, the early days of the internet evolved phone dating into chat rooms offered by the early internet giants like AOL. Classifieds previously in newspapers evolved into classifieds on Craigslist. Remember Craigslist Personals? People could post that they’re looking for companionship and just like you’d reach out to someone selling their used PlayStation, you could connect with a future partner.

Personals shut down in 2018, signaling an end to centuries of dating without a third party involved. 

Matchmaking and dating services existed in vacuums and local markets, but scalability was never a factor until Silicon Valley came into town. Gary Kremen, a Stanford Business School graduate, noticed the piles of cash local newspapers made from personal ads and sought the opportunity to do the same on the internet.

In 1995, Gary launched Match.com, an online dating site where people could exchange personal messages and photographs through email. The growth was insane. Within months, 100,000 users signed up for the service. By 2010, the platform grew to over 1.5 million subscribers and spread to over 30 countries. End of story, right?

Subscribe now

Churn

With users comes churn. Ever heard of BeReal? The social media company that had grown to 53 million users in 9 months over 2022 has seen a 61% drop in daily active users. 15 million DAUs quickly became 6 million. Churn was a massive issue, with rates as high as 20.7%.

If you’re running a subscription service like Uber One, Netflix, or hell - even a Utility company, your customers generally sign up, use your product for a while, and typically never leave if they are satisfied. Take Netflix, which arguably has the best churn rate in the business at around 2% monthly.

Tinder, Hinge, Bumble, and practically every other dating app deal with an added issue. 

People trickle in and fill the platform, swiping every which way and making matches with each other. Eventually, some find love, and others drop out and give up. Whether you find success or not, you end up churning out of the platform!

What do churn rates look like for dating apps? Generally, 20% monthly! Annualized, that is around 90% of your entire user base churning out each year.

If you have a 70% annual churn rate, you have to have a strategy to replace almost your entire customer base each year, plus a bunch of percentage points to drive topline growth. 

- Andrew Chen, GP at Andressen Horowitz

Imagine you’re a founder of a dating app you have to coax an investment from Andrew at a16z. How do you explain near-hundred percent rates of churn to your investors? What is your sustainable competitive advantage versus a competitor?

The good thing is that churn can be positive. Positive churn is when a customer enters the app, has a great experience, finds a match, and ultimately leaves the platform because it worked out. Yes they churn, but while they use the platform they are super engaged, spend a boatload of time and money, and let the app consume all their attention.

When your experience is great, your partner’s is too. That’s two people who are both spreading the word about the platform.

Consolidation

Positive churn sounds like a great rainbow wonderland explanation for your investors. That coupled with statistics that 1 in 5 American adults have used online dating services paints a good picture about top-line growth and word-of-mouth success.

However, the reality is that your app’s experience is essentially the same as the next one out there. Platforms are already hyper-optimized. Hyper-gamification with swipe surges and user retention tactics through push notifications and external messaging.

I open them just because a push notification has informed me that I have new recommended matches, or because “swipe surge is now in session”; soon the smooth-flowing interface is inviting me to peruse and “like” profiles.

These apps do have differences in the markets they cater to. Grindr, Farmers Only, Muzz, Dil Mil and others choose smaller pockets to emphasize. Past that, these apps are commodities. Once a market becomes commoditized, customer acquisition costs skyrocket and lifetime values get put in jeopardy. 

Apps also lack stickiness. More than 30% of people between 18 and 44 claim to use multiple dating apps at a time. 

All of this brings us to consolidation. Remember the hyper-growth story about Match.com and Gary? Today, the company controls almost the entire dating ecosystem. Tinder, Hinge, Plenty of Fish, and other smaller apps all fall under the holding company’s reach. 50% of relationships that start on an online dating site are owned by Match Holdings.

So when that investor asks you how your app is going to succeed - here’s your pitch. We’re going to grow our user base to a critical mass, hitting organic growth points and great CACs. Once we have a critical mass, we’re hunting for an acquisition. Sell our platform, the users we have, and the data we gathered for a 10x on your investment.

That’s all for this one folks! If you enjoyed this piece on the dating app ecosystem, hit the like button or drop a comment with your take. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

We had one hundred employees and a mere three hundred thousand subscribers and were off to a rocky start. That year alone, our losses would total $57 million. Eager to make a deal, we’d worked for months just to get Antioco to respond to our calls.

The media giant rejected the two co-founders after they asked for a mere $50 million. But, we all know how that story ended. Hastings reminisces about the situation, citing culture as the real reason why Netflix lasted and Blockbuster whittled down to one store.

A culture that valued people over process, emphasized innovation over efficiency, and had very few controls. Our culture, which focused on achieving top performance with talent density and leading employees with context, not control.

300,000 subscribers in 2 years ballooned up to 260 million by the pandemic. Competition from the likes of Amazon Prime, HBO Max, Disney Plus, and Hulu grew strong. Consumer appetite for spending looked more bleak than ever. In 2022, the average rate of inflation was 8%.

Challenges crept strong once again. Let’s first zoom out and understand what Hastings means when he boasts about Netflix’s culture.

Culture Fueled Success

In 2009, Netflix published a slide presentation on the company’s culture. A 125-page slideshow posted on the company’s website laid out the company's values, many of which defied classic corporate America's norms.

With millions of views and over a thousand modifying the thesis, this cultural doctrine made it clear that what may have worked for companies in the past won’t work moving forward. Hastings believed in that very thesis:

As a society, we’ve had hundreds of years to work on managing industrial firms, so a lot of accepted HR practices are centered in that experience…Industrial firms thrive on reducing variation (manufacturing errors); creative firms thrive on increasing variation (innovation).

Increasing variation meant creating an environment that encouraged thought. Unlimited PTO, hiring only “A” players, and giving employees full control over expenses are a few of the practices that Netflix followed. The infamous saying around the company was “adequate performance gets a generous severance package”.

While Hastings spearheaded the cultural revolution at the company, but he wasn’t alone. Patty McCord, Chief Talent Officer till around 2013, was pivotal in changing the tides. Patty recalls several tenets that Netflix upheld to encourage innovation, one of them being to encourage employees to understand how the business that they work for actually functions.

For example, Netflix heavily emphasized subscriber growth during its early years. Every employee was laser-focused on increasing the subscriber count. Oftentimes, this meant missing the bigger picture: growth comes with costs.

We were spending huge amounts buying DVDs, setting up distribution centers, and ordering original programming, all before we’d collected a cent from our new subscribers.

Having employees see their roles through the lens of Netflix was crucial to maintaining quarterly success. All in all, modeling responsible behavior did indeed create responsible employees.

Forging A Future

Responsibility comes in handy in times of crisis. In 2022, Netflix hit yet another snag. For the first time in its 20+ year history, the company reported a loss of 200,000 subscribers with projections for another 2 million by mid-2022. 

A 37% drop in the stock followed the news. Management was stuck with a problem that needed solving. Fast. Blaming performance on macroeconomic pressures worked short-term but unless the real reason was addressed, the public would become wary soon.

Hidden among this pile of excuses by the company was one factor key to their plan over the next coming years. The same year they reported their subscriber loss, Netflix revealed the estimated households sharing legitimate passwords. 100 million households worldwide with 30 million just in the United States.

Boom. Therein laid their next plan. Netflix announced its password-sharing crackdown policies in May 2023, sealing an end to their 26-year-long laissez-faire approach to this practice, which they technically always prohibited in their terms and conditions.

Success came fast. Within just half a year of announcing the plan, Netflix reported growth of 15% and a new 9.3 million subscribers just in the first quarter of 2024. They also realized that they are no longer a fledgling startup. Starting in 2025, we’re no longer going to receive subscriber numbers from the company. This makes sense. After any consumer product gains a critical mass of users, reporting slowing subscriber growth (or even decline) only sheds a negative light on a company.

Following the likes of Facebook and Twitter, ceasing subscriber count reports means the public will be more likely to fixate on other metrics such as revenue growth, profitability metrics, and business prospects.

And those aren’t just buzzwords. Since that disastrous 2022 earnings call, Netflix introduced an ad-supported tier and expanded live sports and video game deals. Combined with increasing existing investment in original content and expanding content licensing from other 

All of this screams the increasing variation that Hastings wants. Pouring money into live programming, Netflix Originals, licensed content, advertising, and even video games means being ready for something to fail. But Netflix will surely be ready to change course when something strikes.

That’s all for this one folks! Apologies for the break there, but we are hopefully back to regularly scheduled programming. If you enjoyed this piece on how Netflix used culture as a lever of innovation, hit the like button or drop a comment with your take. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

In the late 1990s, the US carried out its first operation in Iraq. In a series of two campaigns, Operation Desert Shield and Operation Desert Storm, the US ran large bombardment campaigns in Iraq, resulting in the establishment of many US military bases in the Gulf region. 

Fast forward 15 years, and the general sentiment around the wars in the Middle East was that it was more than just stopping Saddam Hussein. Even mentions of oil spark memes and jokes about the US government coming to invade or claim whatever it may be.

Oil as a resource of need isn’t much of a problem anymore. We’ve become reliant on our natural gas reserves and the oil supply of our North American neighbors. However, a new oil has emerged. Today’s corporate giants have made it their mission to protect their data collection and usage rights. Check out last week’s post for a gut check on the amount of data that Google holds today. 

But it’s not just Google. Practically every company you interact with engages in some form of data collection, whether that means storing information that you exclusively provide or actively mining for your information every time you interact with their platform or other connected services. Beyond that, companies aren’t only fighting regulators but also other competitors to protect the integrity and their right to continue collecting data on their users.

All of this has led to the rise of another industry. Data warehousing and management. Of course, you have your legacy players - the Oracles and IBMs of the world. But then there is Palantir, Informatica, Databricks, and Snowflake - the subject of today’s piece.

Setting The Scene

Snowflake IPO’d on September 16, 2020, as the largest software IPO on the American markets. The stock soared to more than double the IPO price within minutes, closing at $240 on the day of the IPO. In the frothy days of the pandemic, the company’s valuation peaked at $130 billion with price targets much above that.

For all the hoopla, it's worth exploring what the company does. Imagine you’re a large company that hasn’t fully migrated away from your data centers. Maybe you also have a few services on AWS, Google Cloud, and Azure and want some way of connecting all of them.

That’s where Snowflake comes in. Their infrastructure allows firms to consolidate multiple data sources to easily process and visualize data. Snowflake also separates storage and compute, meaning customers have more control over how they want to optimize their costs.

The clear advantages taper from there. Today, the company is struggling to justify its value proposition in the face of competitors and pessimistic market conditions. To add fuel to the fire, Snowflake swapped CEOs earlier this year, moving away from the legendary Frank Slootman towards the more AI-focused Sridhar Ramaswamy, a Google executive. Initial reactions were stark, with the stock dropping more than 20% right after the announcement.

Fortunately, Ramaswamy’s few words over the first two earnings calls have revealed his priorities about the direction of the company.

I’ve been focused on 3 key priorities in my first quarter as CEO: listening to and learning from our customers, driving execution and alignment within our go-to-market teams, and fueling our innovation and product delivery

I apologize for the load of business talk here. Simply put, Sridhar is emphasizing exactly what Snowflake is focusing on now. Sales. Spending all their time and money on customer acquisition. Generally, this is a sign that a company is starting to look a lot more like its competitors, meaning dollars toward sales are more valuable than dollars toward research and development. Let’s test that theory out by seeing how the data warehousing market looks.

Subscribe now

The Data Warehousing Market

Outsourcing data to third parties was still nascent in the late 2000s. Numbers were growing, as companies handed off services like photo storage and computing-intensive operations. But alas, the sentiment at the time could best be described as pessimistic.

It could either be a service-outage-type catastrophe or a security-based catastrophe. In either case, it will be big enough. It will be the kind of disaster that makes you say, 'That's why I didn't get involved with the cloud.

Concerns with service capabilities and obvious mistrust of the security provided were the largest friction points for cloud migration. Even in 2015, only 30% of corporate data was stored on the cloud.

Many see Snowflake as the company that pioneered data warehousing on the cloud. The reality is that the market existed before Snowflake. When they launched in July of 2012, Amazon’s Redshift and Google’s BigQuery were already live in some initial form. However, being tied to their respective incumbents limited their capability to offer robust compatibility with other cloud providers. Why would AWS want to allow you to be compatible with Google or Microsoft Cloud?

Source

More than 10 years of runway allowed the players to catch up. As of 2020, Redshift had the largest amount of adopted companies, with Snowflake at third behind BigQuery. Nuances in use cases give each provider a compelling argument to be used over another.

And then there are the disruptors. Palantir, Databricks, Informatica, and other companies have also entered the industry, each with a unique edge. Palantir’s already locked down government contracts, Databricks is making a name for itself in the AI space, and Informatica eases ETL workloads of firms using other data warehousing solutions.

The clear takeaway here is the lack of differentiation moving forward. Within 5 years, each warehousing solution will eventually converge to be the same product with different tech components and pricing. The commoditization of data warehousing could mean one of two things. Either existing large-scale cloud providers win (Amazon, Google, Microsoft) - eating up the competition and creating a feature within a larger suite of products. Or, a true data warehousing company emerges as a victor.

That’s all for this one folks! If you enjoyed this more informational piece on Snowflake and want to know more about data warehousing and cloud computing, hit the like button or drop a comment with your take. If you enjoyed the piece, please feel free to share it with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

But Brin and Page hated the idea. They thought banner ads were ugly and distracting. Worse, banner ads took time to load, and Google’s founders possessed an almost religious devotion to efficiency and speed. 

And then came Google AdWords. Imagine a world where you had control over the ads you saw. When AdWords launched on October 23rd, 2000, that's exactly how it was! 

Users would sign up on a completely self-serve basis and set how much they were willing to pay per thousand impressions (or times the ad showed); whoever was bidding the most would appear higher up on the page.

In 2000 when they launched, AdWords was using a pay-per-impression model, meaning that marketers would be charged regardless of whether someone engaged with their ad or not. Change came soon. In 2002, Google introduced cost-per-click advertising options to compete with the highly successful competitor, Overture.

Google didn’t create this business model. Overture (GoTo.com) launched in 1998 with the radical idea of search being fueled by advertising. No organic results, just the culmination of the highest bidders of a certain keyword. Talk about a dystopia.

Subscribe now

Creating Value

You know that one quote, “Nothing is ever free. Everything is paid for by somebody”? 

Last I checked, I never had to pay to use Google, Facebook, Instagram, or even YouTube. Going even further, I probably spend a 4th of my day online - using several Google and Meta services while I’m at it. For completely free. Or is it? What’s the cost of a free search? The obvious answer is - duh, it's free because they make money off you watching ads. But it's worth digging a bit deeper.

October 23rd, 2000 is significant for another reason. It was the day that our privacy became commoditized.

While diving into research for last week's post, I stumbled upon this feature that Google offers that allows you to download all the data they have on you. Google Takeout. Thanks to the California Consumer Privacy Act (CCPA), companies are legally required to hand over data they store about you. Neatly tucked inside their Data & Privacy page, Google allows me to download all my data. Figured I might as well give it a go!

There’s no harm in trying right? Now of course, since I have a Google account, naturally they store data about my emails, photos, documents I have on Google Drive, and other data I elect to store on Google services.

But the buck doesn’t stop there. All in all, Google had 16 gigabytes of data on me. You know how you’re always running out of storage because of photos? Well, 16 gigs of data is approximately 6,000 large photos. And it’s not like Google is storing your data in JPEGs or PNGs. Just plain text files. That's right, Google has 16 gigabytes of mostly JSON data on me.

The sheer intricacy of all of this is what blows my mind. Google splits up its data into each service it offers. Google Ads, Gmail, Google Flights, Google Play Store, Google Drive, etcetera.

16 gigabytes of data split into 6 folders. What you see above is just one of the sub-folders in the FIRST folder. I’m barely scratching the surface, but so far I’ve seen activity from every Ad I’ve interacted with, every search I’ve made, every IP that I have logged in from, every location Google has on me, every flight that I’ve searched and every time I’ve clicked through to the booking website, all data from Google Analytics.

Ad Saturation

Alright well, my privacy has fully been compromised. But more so, mine and our privacy on the internet has been commoditized. Our identities and data are being used to optimize the advertiser experience and increase the effectiveness of Google AdWords.

Just think about this. All that data I have is likely used for targeting me on a very granular level. Apart from the basic stuff I provide when I create a Google account - Age and Gender - Google knows my household income, my parental status, the device I use, my location, whether I own a home or not, and my education status.

And this is all binary data, called “detailed demographic data”. All my activity is also subject to placing me in “affinity buckets”. Categorizing me in segments such as Banking & Finance, Beauty & Wellness, Food & Dining, Home & Garden, and more.

But Ads have gotten saturated. Search results are filled with sponsored content and unrefined results. Take a search for “best dog food” for example. Every listing that appears in the main view is sponsored. If I’m looking for the best dog foods, I am much more interested in objective content on why one dog food brand is better than the other, whether it be nutrition facts, consumer reputation, or something else. 

Fortunately, all hope isn’t lost. As a community of Google users, we’ve resulted in a common heuristic. To get real and raw opinions from people, we resort to appending “Reddit” to anything we search. Just by changing my search query to “best dog food reddit”, I eliminate every sponsored result that shows up.

This isn’t just some observation I made. People are catching on! In this post, Dmitri Brereton breaks down exactly why this is. Google search is dying because of its optimization towards ads and subpar content. Google isn't a search engine anymore, but rather it's the result of hyper-optimizers taking advantage of its PageRank algorithm. Results that show up are just those of SEO junkies and sponsored listings.

Conclusion

I realize that the last post implied this would be a second DuckDuckGo piece and I have not mentioned the company even once.

Bear with me. The underlying theme in the search market is that Google’s search dominance is being challenged. DuckDuckGo is a direct competitor, but people also search on Yelp, Reddit, and other sites dedicated to industries like travel, jobs, and more. 

As consumers, we still prefer having one place to make an all-encompassing search. But we hate the ads and aren’t fans of the data collection. And that’s exactly what DuckDuckGo’s mission statement is. Google Tracks You. We don’t. That’s what it all boils down to. 

Sure, today DuckDuckGo only has 0.54% of the search market, but it's hard to look past people’s preferences. 61% of people believe that any effort they make to protect their privacy will not make a difference. Only 20% of people believe that those who hold access to personal information will treat it responsibly.

Yahoo’s data breach was by far the largest so far. But even this week, Evolve Bank reported a massive cyberattack that leaked personal data to the dark web. If you aren’t familiar with them, Evolve is the bank behind several Fintech companies including Affirm, Stripe, Mercury, and other smaller startups.

Change is for sure imminent. As data breaches become more common and consumers question the integrity of the products they use, a pro-privacy movement could be coming. DuckDuckGo may not be the one to lead the coming change, but it is the search engine that is so pro-privacy that even the government can’t get you. Why? Because it isn’t stored anywhere!

They can get to your Google searches, but if you use DuckDuckGo it’s completely encrypted between you and us. We don’t store anything. So there’s no data to get. The government can’t subpoena us for records because we don’t have records.

That’s all for this one folks! If you enjoyed this piece on the dominance of Google and the rise of internet advertising, hit the like button or drop a comment with your take. If you enjoyed reading, please feel free to share it with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

If you’ve watched our Yahoo video, it's pretty obvious that there were some pretty large misses. Nevertheless in the autumn of 2016, Yahoo finally had an offer of $4.8 billion on the table from Verizon. They’d be shipped off to be some vestigial part of a broadband telecommunications company, left to just operate the platforms they had at the time. 

By this time Yahoo had already made the public aware of the massive data breach that had occurred. It was revealed in a press release in 2014 that the names, email addresses, dates of birth, and phone numbers of 500 million Yahoo users had been compromised. A couple of months later, the estimate was revised to 1 billion. But it took over 2 years for the real number to come out. Every single Yahoo account was affected. All 3 billion accounts on Yahoo were compromised. Quite possibly the largest data breach of the internet age.

Yahoo’s hack exposed information critical for the user’s use of the platform (debatably), but there have been some other infamous data breaches. Equifax, Cambridge Analytica, Marriot Hotels, and even Dunkin Donuts have all faced massive data breaches.

All of this is part of a larger money making opportunity. The Attention Economy. At its core, the attention economy is a theory that claims that our focus as humans is limited and therefore a resource that can be captured and monetized. Platforms we use capture our data, data brokers sell them, and companies use them straight back at us. But what if it didn't have to be that way?

The Search Engine Market

The search market was already mature by 2008. When Tim Berners-Lee coined the term the “World Wide Web” back in the early 90s, the internet was just an amalgamation of text documents that were linked together. However, its use case soon became clear.

In a matter of just a few years the use of search proliferated across private communities like universities, governments, and private corporations. Finally in 1993 search was made available more publicly with the advent of web browsers like NCSA Mosaic and later Netscape.

Only in the late 90s did search actually become reliable. PageRank - Google’s search algorithm - was built off observations that Larry and Sergey made about academic papers. Seeing that a large part of the credibility of a certain research paper was based on the number of citations it received, it only made sense that a webpage was more reputable if other high quality pages were linking to it.

Numerous competitors sprang into action with different takes on search engine algorithms. For example, AskJeeves built its niche off being the “natural language” search engine that used actual human beings to curate search results based on queries from users. LookSmart, an early search engine that positioned themselves as a provider for other engines, relied heavily on the human curation of web listings.

Search engines were starting to become a dying breed by the early 2000s. Consolidation happened quickly, leaving the smaller upstarts with little oxygen to survive. With users flocking to their favorite and the dot-com bubble killing any oxygen remaining for smaller engines, the clear leaders emerged: Google, Yahoo, and MSN.

In the late 2000s, Google slowly grew its market share to close to 70% because of increases in search traffic and by eating up Yahoo’s and Microsoft’s share. Given its ubiquity, Google was just way too sticky in 2008 for any startup to try to compete.

Subscribe now

Creating Value

Around the same time that Google was hitting its inflection point (currently at 92% market share), Gabriel Weinberg was tinkering with some projects of his own. Fresh off a $10 million acquisition of a social network site he built, he decided he’d work on something he actually found interesting. Search.

Rather than breaking his head trying to make sense of the market opportunity, Gabriel just focused on the product - removing spam, increasing instant answers, and prioritizing high traffic websites like Wikipedia. A privacy feature was also added just to give it some flair. Dontrack.us (the OG DuckDuckGo) racked up 16 million in searches by 2010. 

DuckDuckGo’s team was obsessed with growth. All of this sounds basic today, but focusing on simple things such as search engine optimization, content marketing, microsites, and embedding the DuckDuckGo browsers everywhere they could were pivotal reasons that search query volume grew. There was also the infamous billboard ad. That’s when it went viral. 

This billboard went up in early January 2011 in San Francisco. People instantly loved it. It was the classic David vs Goliath story, painting Google as the large evil corporation that spies on everyone and painting DuckDuckGo as the underdog. The company for the people. 16 million in 2010 became 99 million in 2011, 496 million in 2012, and 1 billion by 2013. While their 36 billion today is only 0.54% of the market, it's more than enough to fulfill the needs of a team of 250 employees.

The secret to challenging Google was to start out by not competing with them. Even though their search volume is a tiny fraction of the market, DuckDuckGo has been able to stick around for 16 years. So, is Google’s dominance dwindling? If you’re confused by the intro to this piece, it’ll all make sense next week.

That’s all for this one folks! If you enjoyed this piece on the rise of DuckDuckGo, stay tuned for next week’s piece to hear more. If you enjoyed reading, please feel free to share with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Thumbnail creds

At $1.9 trillion, Amazon quadrupled Walmart’s valuation in what is essentially half the time. Given the company’s size today, it’s quite unimaginable to hear the crazy startup stories that Bezos and company went through. For example, back when Amazon was just an online bookstore, the company was often strapped for cash. Whenever someone placed an order, Amazon had to source that order from book distributors who required that retailers order 10 books at a time. But they found a loophole - Amazon realized that while they were required to order 10 books, they didn’t need to receive that many. To work around the situation, Amazon would fill up the order with a bunch of books that were out of stock, meaning they never got charged!

Then there are the classic stories about the team working minimum 60 hours a week, sleeping in their cars, using Bezos’s garage, and choosing between names like Cadabra and Relentless before landing on Amazon.

During its journey, Amazon’s retail business has always been a substantial revenue driver for the business. Last year the company recorded revenue of $574 billion, with retail revenue making up the vast majority. “Online Stores” sales made up $232 billion, “Third-Party Seller Services” accounted for $140 billion, “Advertising” revenue was at $47 billion, and “Subscription” sales amounted to $40 billion.

Growing to the second largest company by revenue in 30 years is no easy feat, but in hindsight, I’m pretty sure that the company was no stranger to the “up and to the right” trajectory it has seen. You see, Amazon built its business in a very intentional way - enabling themselves to grow with little added capital.

Subscribe now

Growth Capital at No Cost

If you wanna grow your company faster than revenue allows you to do so, you have a few options. Either you can issue shares through venture capital or public offerings or you can finance yourself through debt. Pretty basic stuff. Well, retail-platforms are structured in a way that opens the possibility for another channel of capital.

For the most part, e-commerce and retail joints like Amazon, Walmart, or Costco have built their business by providing a place for consumers to buy goods from third-party sellers. While they do have their own brands, third-party sales dominate. In 2024, 60% of sales on Amazon’s platform came from third-party sellers!

So as a platform, you typically buy some goods and sell them to consumers. Once someone buys a product, it’s your job to deliver it and pay back the third-party involved. Here’s the catch. Compared to other retailers, Amazon structures these payments vastly differently. Amazon pays their sellers 38 days after a customer purchases something. This is huge. This means that the company has a $1 billion daily cash pile completely interest-free! 

In accounting, there’s this term called working capital. Working capital is the amount of cash or assets available for a company to use for their day-to-day operations. If companies need to increase their cash pile, it is because they want to have more working capital in the business. The more cash you have, the more working capital you have (in most cases).

But if your Amazon and your inventory is sold months before you pay your vendors, your vendors literally finance your business. No need for selling shares or raising debt, every sales cycle you are sitting on a boatload of cash that you know you’ll be able to cover by the next sale cycle. But how did Amazon end up in this situation?

Third-Party Sellers

Unlike Walmart or Costco, Amazon does not rely on a select few suppliers. Costco has the lowest amount of products at around 4,000 unique SKUs and Walmart’s stores climb up to 140,000 SKUs. But because Amazon is primarily a web platform, they’ve created a low barrier to entry for suppliers that want to sell goods on the platform. Consequently, it's not surprising to hear that they have 600 million products on the platform.

It’s more than just having an e-commerce platform. In comparison to Amazon, Walmart spends enormous time and resources selecting and negotiating with their vendors. The company has long been known for having the cheapest retail goods. Since the Sam Walton days, Walmart has achieved this by putting suppliers through the wringer. Walmart is notorious for aggressive negotiations with suppliers and regularly has vendors travel all the way to the HQ in Bentonville to experience a negotiation of a lifetime.

Obviously, this takes time and resources. Walmart’s supply managers spend considerable time and effort to choose which products to stock. Once chosen, each supplier has to go through contract negotiations. All of this adds overhead to the business. And that is where Amazon is different.

Amazon is infamous for being a deathmatch for sellers. For years, Amazon has been known to be a gulag with sellers fighting each other for rankings on Amazon. Sellers have long been known to sabotage their competitors with manufactured five-star reviews, counterfeits, defacements, or even fake fires. 

You’ve probably seen all those dropshipping videos on YouTube that claim to make you a millionaire in a month - well they are all courses on Amazon’s Fulfillment by Amazon (FBA) program. Amazon’s FBA program allows anyone to sell goods on Amazon’s platform while not taking any responsibility for the logistical nightmare of storing and shipping goods to customers.

Remember the revenue breakdown from the introduction? A quintessential seller will likely find themselves contributing to each and every one of those line items. First, you sign up for Amazon’s FBA, typically a 20-35% fee off the cost of every product. Then of course you still have a transaction fee that directly comes from each product - about 8-15%. But once you list your product, selling it is no easy feat. Products compete against each other through reviews, popularity, and even paid advertising. Estimates put advertising at around 15%. By the end of it, Amazon is taking nearly 50% of each sale!

Engulfing the size of Walmart in half the time isn’t a happy accident. If they were simply an e-commerce platform, Amazon would likely be much smaller than it is today. Look at eBay - much less revenue but also only worth $26 billion. Creating a huge power dynamic between the platform and its third-party sellers means Amazon can create its own reality. They can push back payments to create more working capital ($1 billion daily!). They can bid up top spots on search for free cash, and they can even turn around and charge customers for faster shipping (Prime). Over time, this means the more they grow their platform, the better they can set themselves up for the future.

That’s all for this one folks! If you’re interested in more content about how companies take advantage of accounting magic, let me know by liking this post or dropping a comment. If you enjoyed reading, please feel free to share with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Thumbnail provided by Vecteezy

With $5.5 million of their own, they decided it was time to build the next big thing. The three spent the next few months in a conference room in Seattle pouring through problems of their own, evaluating the market for them, and trying to come up with solutions.

One personal need was creating a seamless way to manage all the photos and media they amassed over their time building companies and relationships. Could there be a need for a product that enables people to edit and share photos and other files over the internet? Pursuing the idea for a short while, they figured the revenue opportunity wasn’t there, theorizing that file sharing would be a race to the bottom over time with larger players offering those services for free. Well, Dropbox was founded just over a year later and has created $7 billion of value simply off that very idea.

During this time, Rich, Lloyd, and Spencer were also in need of a home. Coming off some pretty successful stints and flush with cash, they were trying to characterize the best home buying experience for themselves. Aggregating county level economic data, google maps geographic data, and all MLS data they could get their hands on they were able to create an approximate representation of how much homes in their locality were worth.

This massive research project brought on a stark realization. There was a massive amount of information asymmetry in the real estate market. Quality home data available through the Multiple Listing Service (MLS) was available with a huge barrier - you had to be a licensed real estate agent.

Realtor.com and listings on any realtor’s website showed the current homes available, but that was about it. Absolutely no information about the value of one's home existed, meaning a huge knowledge gap for the average person who was in the market for a house. So that’s what they chose to zero-in on.

Subscribe now

Finding A Market

Here’s an interesting segue. Back in 2005, Realtor.com controlled the online share of real estate attention. But, Realtor.com was not really a scrappy startup. Rather, it was the online presence of the nation’s largest real estate cult - the National Association of Realtors (NAR). In an effort to appeal to the internet age, the NAR launched Realtor.com as a way for people to see publicly available MLS Data.

But Realtor.com made a grave mistake. You see, they offered a service where third-parties could license MLS data for a fee - something that Zillow jumped on immediately. Starting with the Seattle area, the Zillow team aggregated MLS data from homes all across the country. With a mission to create the estimates for every home in the nation, Zillow ate through all the data that Realtor.com had, combining it with neighborhood income, school zones, and wealth statistics. 

After a few months, they were able to generate estimates of home prices for around 40 million homes - coined “Zestimates”. And with this, they went live in February 2006. Got this picture from Wayback Machine.

Over 1 million visits on the first day just from people searching their addresses! Even today, this beats out Instagram, Snapchat, and even ChatGPT! Over the course of that month, they were able to scale to 4 million - but then the troubles started. Zillow finally regained the same level of traffic after spending 2 years adding to the core product.

An Emotional Chord

How did Zillow get so viral? There wasn’t a crazy innovation like AI with ChatGPT. There were no massive network effects like YouTube or Instagram had. And it wasn’t like they had a crazy marketing stunt like Dollar Shave Club pulled. You saw how their website looked - the landing page was basically just some text and a couple input boxes.

What if you had a $0 marketing budget? Bill Gurley, partner at Benchmark Capital, posed this very question to Rich when Zillow was raising their Series A in late 2005. Rich, a massive proponent of bringing “Power to the People”, was obsessed with information asymmetry in industries where customers and agents existed. Just take a look at his accolades - Rich founded three separate billion dollar companies all with the aim of bringing transparency to previously closed off industries.

Trulia built the one-stop shop for seamless home searching, but never added a proper light to house valuations. Trulia became the “Google of Homes”. Zillow on the other hand was focused on the larger opportunity. Using the vast amount of data readily available to solve the knowledge gap in valuations.

There’s one other harder way to create a spark - tapping into a person’s emotions. Zillow’s go to market plan was to use controversy to fuel popularity. By creating a tool that made the general value of everyone's home public - you’re provoking a ton of voyeuristic searches. Maybe you want to look up the value of your ex-girlfriends home, or maybe the boss you really dislike, or even all of your relatives!

Today, Zillow is the 29th most visited site. You might use it when you go on vacation in a new place. Or maybe you pull out the app when you see a cool looking house while driving around town. At the end of the day, Zillow still drives the attention of the average person - the core target market for realtors. When it's time to buy, you can bet the first stop is Zillow.

That’s all for this one folks! If you’re interested in more content about viral growth stories, let me know by liking this post or dropping a comment. If you enjoyed reading, please feel free to share with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Palantir’s ride to the top is nothing short of phenomenal - but let’s take a closer look at some of the partnerships they had during this growth. The startup started out with the public sector but even took some private sector contracts during this time. You’ll start to notice a pretty obvious trend here.

Early on, Palantir was the X-Factor during the War On Terror, working with the Pentagon and CIA in Iraq and other areas in the Middle East. When the conflict in the Middle East started, the US Military used various drones and balloons to gather tons of surveillance footage. This raw data is then washed through Palantir to help avoid IEDs and identify terrorist hotspots.

Non-conflict uses through the government also existed. Palantir has worked with the Department of Health and Human Services (DHHS) to offer Gotham as a tool to detect Medicare Fraud. By combing through Medicare claims data, Palantir was able to visually map hotspots where doctors were overprescribing and allow the DHHS to crack down on prescription abuse. These efforts were made to tackle a $100 billion fraud market.

JPMorgan infamously used Palantir to quite heavily spy on their employees. Gotham ingested tons of emails, browser histories, GPS locations, printer activity, and transcripts of phone conversations to flag employees for potential abuse of corporate assets and point out which employees were getting disgruntled. 

I think you’re starting to get the gist of it - Palantir has enabled numerous agencies to ingest and analyze amounts of raw data that were previously unimaginable. Sometimes even if it meant toeing the line of ethics or blatantly disregarding it.

All of this sounds awfully familiar right? Isn’t that what Google and Facebook have been doing for years? Contrary to what all this might be indicating, Palantir firmly believes that it does not collect and monetize off data at scale.

Subscribe now

The Problem with Gotham

We build digital infrastructure for data-driven operations and decision-making. Palantir’s platforms tie these together by bringing the right data to the people who need it, allowing them to take data-driven decisions, conduct sophisticated analytics, and refine operations through feedback. We license this software to organisations, who receive secure and unique instances of our platforms in which to conduct their own work on their own data.

If you believe what they say, Palantir does not collect data, but rather simply enables those who do to better manipulate the data in ways more useful to their business. 

Palantir’s first product, Gotham, was built around the rather archaic requirements of the government. When pivoting to other private sector use cases, this meant limited flexibility in pricing and structure around the product model.

Here’s how Gotham worked. If you wanted to contract Palantir, you would first need to pay for a server which comes with 1 year of maintenance. Then you’d need to gauge how many people in your organization might need access to Gotham. Let’s say it’s just 10 - maybe you need a 1 core system. That’s around $150k for 1 core before maintenance. For every year after the first, maintenance is $30k. For just 3 years, you’re already paying $200k before software updates, training costs, power, security, IT support staff, etc. All of these can easily double the cost.

There’s one more cost to mention. Operating Gotham requires onsite engineers, called “Forward Deployed Engineers”. These employees are essentially just onsite contractors that work fully on the specific business use case for a particular customer. As part of their contract, the customer is required to pay the $150/hr salary commanded by the FSR.

All of this was just way too expensive for the prospective Palantir customer. Large agencies with obvious use cases could deploy these expenses but enticing new customers was nearly impossible. Even the Street didn’t really love Palantir. The company was back to IPO levels just two years after going public. With Gotham as the start of the show, Palantir was not expected to make any money unless they were able to get massive government and private contracts several times, every year, forever.

To turn heads and create profit, Palantir needed to find a way to offer the same product for a substantial operating margin. Palantir had to create a SaaS product.

Competitive Advantages with Foundry

Welcome to Foundry. No more onsite engineers, core server costs, maintenance and software update costs, nothing. At its core, Foundry is a data ingestion and integration platform that was designed with cloud infrastructure in mind. As a data engineer, you are able to use a web platform to link multiple data sources to better use unstructured data.

Ok, if you put a magnifying glass to Foundry, you might be reminded of competitors like Snowflake, BigQuery, Databricks, Azure, and other data aggregators. These companies allow you to ingest multi-platform data into one data lake and get a holistic view what you want. But let me advocate for Palantir here. Palantir does not want to be seen as a replacement for Snowflake or Amazon or Microsoft. Similar to how they built Gotham, Foundry also has a “sky compute” engine that makes it easier for business level employees (non-technical) to manipulate data sources similar to Tableau.

Palantir has been in the AI space long before it was seen as cool. Similar to NVIDIA handling the hardware infrastructure that’s powering AI, Palantir has developed the software infrastructure for the past 20 years at this point. In 2019, Palantir won a bidding war for a US Army contract to create a dashboard that would pull together data from human resources. You know who they were competing against - Microsoft, Deloitte, Accenture, and Ernst & Young. Isn’t that crazy? Just 5 years ago, they were bidding against companies that you would never consider software business - EY is literally an accounting firm!

It’s impossible to project where the data aggregator industry is headed in the next. In 2021, the total addressable market was already $119 billion with the potential of doubling in the next few years because of developments in AI. Palantir, Databricks, and Snowflake - all leaders in the industry - all sit at a $45 billion valuation. Even without killing the competition, Palantir can thrive simply by growing its business. And if they can figure out how to further pivot away from Gotham, they can once again pave the way as the leading company for data mobilization

That’s all for this one folks! If you’re interested in more content around the data cloud, let me know by liking this post or dropping a comment. If you enjoyed reading, please feel free to share with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Thumbnail from Medium

While easy for small businesses, breaking into the defense industry at a large scale was almost impossible. Lockheed Martin, General Dynamics, and Northrop Grumman have been around since World War 2. Sure, there are a few other players - but for the most part - the past has only seen consolidation, not addition. Here’s how the defense industry has basically been since the turn of the century. 

King, David. (2019). Validating Stock Market Reactions to Acquisition Announcements. 1. 1. 

These companies are legacy manufacturers. Even in the general space of govtech, there really has not been much movement in terms of contracts or public-private partnerships at scale (“-tech” is just a suffix nowadays). Honestly, it’s hard to blame anyone for shying away from winning government contracts. For one, the government is so fractured, and each layer has its own political headache.

After the federal government, we’ve got 50 state` governments, 3,000 counties—which play different roles in different parts of the country—and 20,000 municipalities.

The list continues. Bureaucracy, security clearance, and confusing registration guidelines have plagued the industry over the last few years (we’ll get into this soon). Ironically, one company completely out of radar was able to break into the government sector not only once, but twice. PayPal. We all know that Elon Musk went on to create SpaceX, which launches everything from rockets to satellites for the government. And, Peter Thiel went on to found Palantir - the topic of today’s story.

The Makings Of Palantir

In the early 2000s, the integration of finance in the internet was still quite nascent. PayPal was still just establishing itself as the pioneer of payments on the internet. With any emerging technology, the internet brought both early adopters and bad actors.

PayPal especially faced this firsthand since it was the gateway for internet payments. In 2000, PayPal was losing more than $10 million each month due to credit card fraud. The team had to create a robust algorithm to overcome the fraudsters - a tough challenge in yesterday’s world. After some iteration, Max Levchin (CEO) and team created a dual-pronged approach where computer algorithms would flag suspicious transactions and humans would either approve or deny them. PayPal finally posted their first quarterly profit in Q1 2002, with almost all of the credit going to this algorithm.

Thiel immediately recognized the opportunity here. PayPal didn’t just figure out how to turn cash-flow positive, they created a system that better empowered humans to detect red flags - for practically any use case! To add to the party, the FBI already identified a use case that was heavily topical at the time. Identifying terrorist networks.

And surely, on May 6th, 2003, Thiel and his law school buddy Alex Karp (current CEO) founded Palantir. Honestly, I didn’t know what “a palantir” was but it's apparently a reference to an indestructible crystal ball in Lord of The Rings - sorry LOTR fans. Thiel’s VC Founders Fund as well as the CIA’s venture investing arm (In-Q-Tel) invested a total of $32 million to get the company up and running towards their first product.

It was all quite patriotic. Palantir spun up the algorithm soon after 9/11, set up ties with the government through the FBI and CIA, and created the product that is known today as Gotham. Within no time, the government and Palantir were synonymous. CIA, FBI, NHS, HHS, and ICE all used Palantir systems to analyze data streams real time for detection of red flags and fraud. Get this - Palantir even developed the platform to help locate Osama bin Laden!

Shaping The Future Of The Government

There’s this concept in the government space referred to as the Valley of Death. Essentially, its the period of time in between when a vendor approaches the government with a prototype and when that prototype becomes a contract. Typically, many companies run out of funding and go belly up just waiting to hear back from the government. Well, Palantir was able to cross this chasm. Today, Palantir generates 43% of its revenue from US government contracts. Solely the US Government. That’s more than a billion dollars of revenue coming from the government for a software as a service business. Within 20 years, Palantir changed the landscape of how startups interact with public agencies in the United States.

Again, “govtech” was not a thing in the early 2000s. The primes - Lockheed, Northrup, Raytheon, and a few others monopolized all contracts. In order for Palantir to win, they had to successfully break past every layer of friction in the government. Forge through useless bureaucrats and comply with the stringent requirements that vary between different state and federal agencies. They had to play politics.

And that's exactly what they did. Palantir used their early ties with the CIA to learn how the game is played. A few years later, they loaded up on the nation’s best lobbying firms and brought in top politicians to lobby for funding for intelligence analysis tools and defense appropriations bills.

Palantir was completely aware that in addition to playing politics, they had to deliver like the big guys. Successful lobbying efforts and perfect execution meant more contracts with the Army, Marine Corps, Defense Intelligence Agency, DoJ, FBI, Homeland Security and anything else related to US defense.

Structurally, the only way for Palantir to break in, deliver value, and build a business was through the creation of vertically integrated products that could be delivered as complete, opinionated, end-to-end solutions, just like the primes.

The earlier graph outlined consolidation in the defense industry over the past 50+ years. This one shows how Palantir shifted the tide for new entrants. Palantir’s lobbying efforts paved the way for Anduril, Saronic, HawkEye, Epirus and many more. From the looks of it, they are just getting started. Also, their massive government contracts used to mobilize the government tech stack also unlocked a key point of value enabling them to move into the next generation of their company. Data. Tune in for next week’s post to learn more.

That’s all for this one folks! If you’re interested in more content around how tech companies are defense, let me know by liking this post or dropping a comment. If you enjoyed reading, please feel free to share with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!

Since then, Airbnb has experienced hockey-stick growth peaking at $134 billion in February of 2021. Today, there are over 7.7 million active listings on the platform spanning 100,000 cities worldwide. Fun fact, more than 6 guests check into a listing every single second.

However, the company has faced a slew of headwinds over the past couple years. The stock is down 13% since the end of March with most of that movement coming from their recent earnings call.

Remember all those stimulus checks from the pandemic? This influx of cash into people’s pockets combined with nothing to waste it on created the perfect storm for prospective hosts. Midway through 2023, the inventory of available homes for sale sat at 570,000 while Airbnbs touched almost 1 million. Airbnb even reported that between Q2 of 2021 to Q2 of 2022, the number of new hosts grew by more than 50%.

Common logic would have us believe that all this supply must have meant low prices for renters. But, this spending spree created a positive feedback loop of high inflation and increasing home prices. High inflation meant looming interest rate hikes (first graph) and massive home buying activity meant the average sales price of a home almost doubled between 2012 and 2022 (notice where it started picking up).

Hosts bought in high and got left to dry. And to make the best of it - most hosts are turning to profit maximizing methods of charging high cleaning fees, surge pricing in peak seasons, and various extra fees. It's become more of a full on business rather than simply renting out a spare  bedroom. To add fuel to the first, top cities for new listings saw huge declines in rental revenue.

That’s just half of the issue. It did not take long for HOAs and municipalities to see the number of short-term rentals and decide enough was enough. New York City enacted “Local Law 18”, requiring hosts to register with the city and be present during a customer's stay. San Francisco has similar rules, enforcing hosts to live in the unit for 275 days a year and enforcing them to obtain a short-term rental license. Santa Monica, a suburb of LA, is no different in requiring hosts to live on the property during a renter's stay. This has wiped out 80% of the Airbnbs in the suburb.

Prices are high, hosts can barely make mortgage payments, and cities like LA, New York City, San Francisco, Denver, and Portland are all trying to ban short-term rentals. Let’s dive into how Airbnb is tackling this.

Subscribe now

What Is Airbnb Doing?

Brian Chesky had a couple chances to address the future of Airbnb over the past couple of weeks during their 2024 Summer Release and the Q1 Earnings Call. And boy did he tell us quite a bit.

Well, total nights and experiences booked on the platform have increased since the same time last year to 132 million, up 9.5%. Airbnb reported revenue for the first quarter of 2024 at $2.1 billion and a profit of $264 million. That’s the most profitable Q1 ever with a record high Q1 take rate of 9.3% because of Easter weekend.

Airbnb also revealed Icons during their 2024 summer release. Icons is an experience that's “more than just a stay” - tied to something like a brand, movie, cultural movement, sports, or music. Chesky drew on a campaign Airbnb did 10 years ago where they listed an IKEA sample room on the platform which generated tons of buzz. Examples of these experiences include: the last operational Blockbuster in Oregon, the house from Pixar’s Up, and an interactive stay with Jahnvi Kapoor or Khaby Lame.

The level of detail and investment in tapping into the nostalgia with these stays are quite simply insane. And if you take a look at the prices, everything is under $100. Why go through all this pain? Chesky says it is to keep Airbnb’s brand relevant and change the way that people think about Airbnb. Cheap, quality, exclusive stays go a long way in reverting the laundry list of issues circling the brand of Airbnb.

From the looks of it, Airbnb is trying to create their own experiences for people to enjoy - adding more excitement to the platform and giving customers a potential hint at where they are headed to next - entertainment. The added flair in these Icon stays is that its not just aesthetically similar, but there’s an interactive element. For example, the house from the movie Up actually floats up and Xavier’s Academy has a built in adventure. Going even further, if you are joining Kevin Hart on a show or Feid on tour, it’s not even about the stays anymore.

I'd say it's working too. In one week, Icons has generated over 8,100 pieces of global media coverage and driven 371 million social media impressions (more press and attention than their IPO!). It’s similar to how people forget about Disney’s controversies when they walk into Magic Kingdom.

The sheer volume of social media impressions did wonders from mobile downloads. Just between Q1 2023 and Q1 2024, mobile downloads increased 60% YoY. Airbnb touted this multiple times during the call but the reality is if anyone wants to see more about a certain Icons stay, Airbnb directs them to the app. I really wonder how many of those downloads came from the release of Icons.

In the past, Airbnb was able to enjoy a take rate between users and hosts on the platform, but it seems as though they are trying to diversify revenue streams by offering an end-to-end experience that they take the revenue on.

Setting The Standard

As a platform, Airbnb’s powers cannot reach into local regulation or the real estate investing of hosts. The main issue that Airbnb is trying to solve is in the space of optimization. To get users who look for Airbnbs to actually book them. The type of vacationer that book an Airbnb more or less obvious. Groups of people that stay for a longer period of time. In fact, 50% of nights booked are of long stays of a week or more. Cleaning fees and total charges just don’t make it economical for the transient vacationer.

Well, prices are trending down. The platform has implemented features like “total price before taxes” to show the sum of all small fees before taxes when searching for a stay. Users can compare full prices of stays leading to much more upfront price transparency. So far, its been somewhat helping. Not only are many hosts reducing cleaning fees - Airbnb is setting the standard for industry.

Right now, Airbnb’s highest penetration markets are US, Canada, Australia, France, and UK. The next step is to move into markets like Mexico, Brazil, and Latin America. Longer down the line - China and India. Expanding into untapped markets means taking advantage of huge price discrepancies and unlocking huge markets (Asia itself is more than 4.5 billion). Asia is an untapped market because of many young people who are not predisposed to hotels and are on social media.

Airbnb also removed thousands of low quality listing from the platform during the quarter leading to 15% net growth in listings. This effort to maintain quality as well as offer features like “compare listing” for hosts and guest oriented tools like group lists and group chats with hosts further involve multiple people during a certain stay.

That’s all for this one folks! If you’re interested in more content around how tech companies are transforming real estate, let me know by liking this post or dropping a comment. If you enjoyed reading, please feel free to share with your friends. Reach out to us by replying to this email if you have any insider takes or leave a comment on our Substack page!