I know… Amazon is hardly an insider tip.
But take a look at the stock chart, and you might be surprised to see that it’s gone nowhere over the last four years.

The business, however, has changed dramatically.

Behind the world-famous storefront, Amazon has built one of the most complex and far-reaching ecosystems in tech — spanning cloud, logistics, media, and advertising.

Amazon spent hundreds of billions in CapEx, doubled the size of AWS and its ad business, and turned both into two of the most profitable segments in tech. Oh, and they started launching satellites into space.

After building out physical infrastructure in the pandemic-era, Amazon has now entered its next phase: AI and cloud investments.

The question now is whether that strategy will be enough to put the stock back among the market’s top performers and into our Intrinsic Value Portfolio.

Let’s dive in!

— Daniel

*Special Announcement*

We’ve announced our Intrinsic Value Community in the past few posts and the response has exceeded all expectations.

The Intrinsic Value Community is a place for sophisticated long-term investors to exchange ideas, challenge each other’s thinking, and build meaningful connections with like-minded individuals.

We give each other feedback on our latest research, post our best ideas, and bring in guest speakers, from industry experts to investment professionals, who share actionable insights and explain how they invest and think.

We’re currently reviewing applications and will begin closing the process soon. If you’re interested in joining, now’s the time to apply!

Amazon: Is It Prime Time for Investors?

Amazon — Not Just a Retailer

When most people hear "Amazon," they still think of boxes on doorsteps — a giant digital storefront where you can buy books, tech, clothes, or household items with a single click and near-instant delivery.

But that’s not the business that makes Amazon tens of billions of dollars in profits. Over the past decade, Amazon has transformed from a retailer into a full-stack platform with deep infrastructure across cloud computing, advertising, logistics, and subscription services. And while these segments might look like side businesses from the revenue breakdown, they are the engine of Amazon’s growth and profitability.

AWS, Amazon’s cloud division, now generates the majority of the company’s operating profit. Advertising, which didn’t even warrant its own revenue line a few years ago, is now generating $55 billion, $20 billion more than YouTube. Prime subscriptions not only bring in billions in recurring revenue but also act as the glue that keeps customers embedded in the ecosystem.

Today, Amazon is less about selling things and more about powering the infrastructure behind digital commerce. Whether it’s a merchant using Buy with Prime on their own site, a startup building on AWS, or a brand paying for sponsored placement.

Amazon continuously expands margins, grows network effects, and has turned into a business that looks more like a mix between Microsoft and Google — not Walmart.

The Amazon Flywheel

When we think about Amazon’s success, what comes to mind is its scale and logistics network. But that’s only a consequence of what Bezos really focused on: customer experience. Warren Buffett once said: “There are certain people you do not want to try to beat at their own game, and certainly Jeff Bezos would be No.1.”

When I was a kid, ordering something online meant days, sometimes even weeks, of waiting. With that much waiting time, you don’t make impulse purchases.

And don’t get me started on return policies. If the product didn’t arrive as promised, the money was pretty much gone. Amazon changed all of that. Delivery within 24 hours, returns without giving any reason, and of course, lower prices than anywhere else.

Amazon is the definition of high convenience. But how did they get there? A large part of that answer is Amazon Prime.

Prime is Amazon’s subscription service that started offering fast and free delivery. Today, it’s so much more than just that. Prime addresses every customer out there.

Music, movies and shows, sports, and of course, the good old free and fast delivery. All of that is part of the modern Amazon Prime subscription.

It’s Amazon’s behavioral lock-in, the core of a flywheel that drives virtually every part of the business. Over 230 million people are Prime members. They shop more often, spend twice as much, and with 90% retention rates, Prime seems to be uncancellable for its users.

Prime increases order frequency, lowers customer acquisition costs, and makes the marketplace more attractive to third-party sellers, who in turn spend more on advertising to win visibility.

That demand gives Amazon leverage to invest even more in logistics and infrastructure, investments that reinforce the customer experience, creating a self-reinforcing loop — the Amazon flywheel.

The magic of this model is that each segment reinforces Amazon’s ecosystem. The advertising revenue, along with Prime subscriptions and AWS profits, funds continued investment in logistics, infrastructure, and customer experience. Faster delivery and better service make Prime more valuable, which keeps customers loyal and further strengthens the entire system.

It’s a loop that never really ends, and it’s what makes Amazon so hard to compete with. Other platforms might beat it on price for some products or on a single vertical. But no one has the full-stack integration of subscription, logistics, marketplace, cloud, and advertising at this scale.

AWS — Amazon’s Profit Machine

While Prime focuses on customers like you and me, Amazon’s most profitable business is not connected to its marketplace at all. At least not beyond being subsidized by it in the beginning.

Today, it sounds surreal, but not too long ago, Amazon Web Services, or AWS, was part of the “Other” segment in Amazon’s reporting. Today, AWS makes up more than half of the company’s total operating income, despite only being responsible for a little less than 20% of total revenue.

That’s because AWS runs at 35% to 40% margins, while the retail business only runs at margins in the low single-digits. The international business has even been loss-making in three of the past five years.

But perhaps I should explain what exactly AWS even does before I dig deeper. AWS is the backbone of the modern internet. It provides cloud infrastructure to businesses and organizations of all sizes, from startups to governments, allowing them to store data, run applications, and scale on demand without owning a single server. Companies like Netflix, Airbnb, NASA, and even Amazon itself rely on AWS to power their digital operations.

Think of AWS as the utility provider for the digital world. Just like you wouldn’t build your own power grid to run a factory, most companies don’t build their own cloud infrastructure. They rent it — and AWS is still the biggest landlord in town.

And although it doesn’t directly benefit or participate in the flywheel I’ve outlined above, AWS has a strong moat built around switching costs. Once a company migrates to AWS, leaving isn’t easy. Over time, businesses adopt more of Amazon’s storage types, data tools, machine learning services, and backup systems, each with its own configuration, retrieval setup, and costs. That makes switching expensive and operationally risky. It’s what industry experts refer to as vendor lock-in, and investors call high switching costs.

Even though Microsoft Azure and Google Cloud have gained market share over the past decade, AWS continues to grow in absolute terms and has maintained a stable share of a rapidly expanding market. Azure and Google Cloud didn’t steal AWS’s customers, they onboarded new ones.

Azure onboarded a lot of customers thanks to its strong relationships with big enterprise clients who use its Office products and Windows Server. Azure was simply bundled into a lot of those broader corporate deals.

Google Cloud is the youngest of the three but also quickly catching up in terms of market share and margins. Ultimately, I expect the cloud market to be dominated by five to six players, with the three current leaders continuing to lead the space.

AI Strategy — Building the Rails, Not the Buzz

Besides AWS’s core, it’s also becoming the foundation for Amazon’s next big bet: artificial intelligence.

But Amazon is taking a different route than its big tech competitors. Amazon is not trying to compete with ChatGPT or build the next large language model (LLM).

At the core of Amazon’s approach is a three-layer AI stack: infrastructure, models, and applications. Together, these layers form a platform that enables customers to build, run, and scale AI systems entirely within the AWS ecosystem.

1. The Infrastructure Layer: Owning the Compute Stack

At the bottom of this stack is raw computing power. Training large language models like GPT or Claude requires massive amounts of processing, and that has historically meant relying on Nvidia GPUs. But Amazon is now aggressively investing to reduce that dependency.

It’s building out data center capacity and designing its own custom chips. Trainium is used for training deep large language models, and Inferentia for running models at scale.

These chips are already being used in AWS services. Owning the hardware stack gives Amazon pricing control and long-term cost advantages, crucial in a world where compute is the most expensive input in AI.

But for now, instead of saving money, Amazon is spending tens of billions of dollars on AI infrastructure — including $100 billion in CapEx this year alone. That’s more than many national infrastructure budgets.

2. The Model Layer: Titan, Nova, and the Bedrock Bet

On top of this infrastructure, Amazon offers a mix of proprietary and third-party AI models. Its Titan family of models supports tasks like text generation and embeddings. The Nova model is Amazon’s newest multimodal foundation model, trained to understand and generate both text and images.

But here’s where Amazon differentiates itself: instead of pushing its own models exclusively, it acts as a neutral platform via Amazon Bedrock.

Bedrock is a managed service that allows enterprise clients to access and fine-tune multiple foundation models from various providers. Those models include Claude by Anthropic, in which Amazon has a $4 billion investment itself; LLaMA by Meta, Command by Cohere, and, of course, Amazon’s own Titan and Nova models.

Clients can customize these models using their own proprietary data, all while keeping that data securely inside AWS. That makes Bedrock extremely attractive for enterprises that want to use powerful LLMs without exposing sensitive data or managing complex infrastructure.

Instead of competing with its suppliers, Amazon’s model-agnostic approach positions it as somewhat of a middleman. It reminds me a bit of Adobe, which doesn’t try to compete against AI tools, but instead positions itself as the service where institutional clients can safely and legally use these tools for their work.

3. The Application Layer: Embedding AI Across the Ecosystem

The final layer of Amazon’s AI strategy is where all the infrastructure and models start to show up in the “real world” — in the tools and products that Amazon delivers to customers and uses internally to optimize its operations.

One of them is Rufus, Amazon’s AI-powered shopping assistant. I don’t know about you, but my experiences with digital shopping assistants have always been more than frustrating. It’s about time that AI changed that.

According to Andy Jassy, Amazon’s CEO, the long-term vision with Rufus is to be able to offer everything that nowadays only a human in a physical store can deliver. Such as finding the right product based on a specific event like Valentine’s Day, or comparing the size of clothes by looking at your order history and the reviews of other shoppers with similar order histories.

Physical Stores — The Limits of Amazon’s Model

Amazon is known as the company that made physical retail obsolete, at least to a certain extent. Nevertheless, it decided to venture into brick-and-mortar stores itself.

In 2017, Amazon acquired the premium grocery chain Whole Foods in a $14 billion deal. This gave Amazon an instant footprint of over 500 high-end stores across the U.S. and Canada.

Weekly Update: The Intrinsic Value Portfolio

Notes

• Ulta Beauty reported a solid set of earnings on Thursday, and the stock jumped 15% in response — finally pushing it into positive territory for the year. Given that the position had grown larger than initially intended, mostly due to it being one of the first added to the portfolio, we used the strong market reaction to reduce it from 7% to 5%.

  In hindsight, the original valuation estimate and position sizing were a bit too optimistic. While we still want to own Ulta longer term, it’s not our highest-conviction bet at current prices. For the sake of consistency with our views on other holdings, it made sense to take advantage of a day when Mr. Market was likely pulling forward future returns and overreacting to earnings that were good — but not spectacular.

  The position is now in line with our allocations to Adobe and Airbnb, reflecting a level of conviction that’s more consistent across all three companies.

  • As for the earnings, Ulta beat revenue estimates of $2.8 billion by nearly 2%, reporting $2.85 billion. EPS came in at $6.70, significantly above analysts’ expectations of $5.80.

  • Full-year guidance was also raised, but only slightly, from $11.5 billion to $11.7 billion. While that’s a positive update, a 15% jump in the stock price adds roughly $2.5 billion in market cap, which feels like an overreaction to what was a solid but not transformational quarter. That’s why we viewed this surge as an opportunity to trim the position, though the primary driver was to remain consistent with our overall sizing discipline.

  • The results confirm that Ulta continues to see strong demand despite concerns about a weakening consumer and tariffs. Competitors have also posted solid numbers, suggesting the beauty sector as a whole remains resilient.

Quote of the Day

"Take a simple idea and take it seriously.”

— Charlie Munger

What Else We’re Into

📺 WATCH: First Principles of Valuation with Tim Koller

🎧 LISTEN: Professor Galloway and Aswath Damodaran discuss Markets

📖 READ: The Collection of Shareholder Letters by Jeff Bezos

You can also read our archive of past Intrinsic Value breakdowns, in case you’ve missed any, here — we’ve covered companies ranging from Alphabet to Airbnb, AutoZone, Nintendo, John Deere, Coupang, and more!

Your Thoughts

See you next time!

Enjoy reading this newsletter? Forward it to a friend.

Was this newsletter forwarded to you? Sign up here.

Use the promo code STOCKS15 at checkout for 15% off our popular course “How To Get Started With Stocks.”

Follow us on Twitter.

Read our full archive of Intrinsic Value Breakdowns here

Keep an eye on your inbox for our newsletters on Sundays. If you have any feedback for us, simply respond to this email or message [email protected].