Amazon (AMZN) – Earnings Review – April 30, 2024
Demand
- Beat revenue estimates by 0.5% & beat guidance by 1.8%.
- AWS revenue beat estimates by 3.7%.
- Advertising revenue met estimates.
Revenue growth ex-leap year would have been 11.3% Y/Y instead of 12.5% Y/Y. AWS growth ex-leap year would have been 16%. Unexpected foreign exchange (FX) headwinds reduced revenue by $700 million, which is why Amazon did not exceed the high end of its outlook (exceeded the all-important midpoint).
Source: Brad Freeman – SEC Filings, Company Presentations, and Company Press Releases
Margins & Profitability
- Beat EBIT estimates by 38.5% & beat guidance by 53.0%.
- Beat $0.83 GAAP EPS guide by $0.15.
Source: Brad Freeman – SEC Filings, Company Presentations, and Company Press Releases
Balance Sheet
- $87B in cash & equivalents.
- $58B in debt.
- Basic shares +1.4%; diluted shares +3% Y/Y.
- Amazon plans to pay down about $25 billion in debt by the end of the year. That juices EPS via reducing interest expense.
Q2 Guidance & Valuation
- Revenue missed by 2.3%, which was partially related to incremental FX headwinds.
- EBIT missed by 4.2%.
- Europe was called weaker than the rest of the world.
Call & Press Release
GenAI:
In the last few Amazon earnings reviews, I’ve introduced the three layers of GenAI: hardware to support training and inference, models and consumer apps. To avoid redundancy for consistent readers, I’d point folks interested in reading that explanation to this article. Here, I’ll simply update Amazon’s progress across these layers. On the hardware front, it continues to work very closely with Nvidia; that won’t change. Still, Amazon is hard at work on building out its own custom silicon. Graviton is its basic computing chip, while (as we frequently discuss) Trainium and Inferentia are its training and inference GenAI semiconductors. Airbnb, Databricks and Meta’s Llama 3 are early customers here.
For models, Bedrock continues to resonate with developers for its broad lineup of partners like Anthropic, Mistral AI, Cohere and Meta. Tools like Model Evaluation (to pick the best one for you), question-answering guardrails and multi-step tasks are helping too. Leadership is especially optimistic about a new Bedrock tool called Custom Model Import. It’s a one-of-a-kind product that makes onboarding custom models more intuitive and rapid. This way, customers can use all of the compliance, security, hosting and other tools AWS provides before delivering a model or software package to runtime. This tool has become very popular to use alongside AWS SageMaker. SageMaker is its environment to build custom models on top of Bedrock. It’s similar to Microsoft’s Copilot Studio. Perplexity AI, trains models 40% faster with SageMaker vs. alternatives. Workday uses it to reduce inference latency by a full 80%. Overall, SageMaker Inferencing reduces foundational model costs by 50% and latency by 20%. These edges within inference matter as this area will eventually become the main part of the GenAI opportunity. After models are trained, they’re generally asked to constantly infer new insights (and only periodically re-train with new data).
Amazon Q was broadly released today. While Codewhisperer is its GenAI tool for general code writing, Q is purpose-built for AWS. Amazon thinks this is the best software assistant out there to generate, debug, and translate code. Q also offers a tool called “Agents,” which can perform a step-series of tasks like refactoring code and updating software. A developer conversationally asks it to perform a function, Q “analyzes existing code,” creates a plan and, upon approval, implements it. Pretty cool. Q Apps is another product to help developers build apps in the secure AWS environment — with more access to their data. Data is king for all cloud and customer applications… that may be even truer in the realm of GenAI. Models need to be trained; the best way to do that is with massive sums of relevant data. Enter Q Apps. Datadog, GitLab and the National Australia Bank are early users.
Amazon now has a few billion in annual GenAI revenue. This is a tiny part of its gigantic revenue base, but should be a high margin segment.
AWS:
We heard more of the same on AWS this quarter. Cost optimization trends have continued to fade and stabilize. It’s signing larger deals more quickly as migration appetite re-accelerates. Demand across all cloud apps for AWS was called very strong. And just like every other mega-cap, Amazon will invest into that strength. Being a global GenAI player requires massive infrastructure to ensure a lack of bottlenecks as use cases scale. Infrastructure is expensive. Considering this, Amazon said CapEx would materially grow in 2024 and called the $14 billion Q1 level a low point. This implies at least 15% Y/Y CapEx growth (likely faster). As a reminder, CapEx is expensed on the income statement as depreciation over the useful life of assets created. That is an EBIT and net income drag, which will greatly ramp up for Amazon in 2024. It added some needed detail here.
Leadership told investors that Amazon is at a point where it can balance growth and profits better than it was able to in the past. This ramping CapEx will not be Amazon throwing margin preservation out the window and accepting another long period of cash burn. This is it taking advantage of the far healthier position it finds its margins and balance sheet in to both walk and chew gum. There are margin tailwinds to offset this headwind, which we’ll dig into shortly. It didn’t say the tailwinds would make up for the added depreciation (like Alphabet did), but did assure us that it would keep a heavy focus on efficiency.
“We don’t spend capital without very clear signals that we can monetize it.” – CEO Andy Jassy
Customer wins and expansions included:
- Siemens is using Amazon Bedrock.
- Philips is using AWS Health Imaging and Bedrock.
- New Anthropic collaboration to help highly regulated sectors comply with and adapt to GenAI regulation and trends.
- PT Group is using Amazon Q to automate 12% of a software engineer’s job.
- Audi is using SageMaker to build a GenAI customer service app.
Cost & Speed to Serve:
The fulfillment localization project through 2023 was the big cost to serve lever for Amazon. There’s still more to be done there, but most of the margin accretive work is complete. When combining this with more 3rd party fulfillment services, the flex driver program and 3rd party seller momentum, we are left with the exploding North American EBIT margin you can see above. Amazon thinks it can do much more to reduce cost to serve, and this is where the CapEx offset will mainly come from. Ads, i nternational market scaling and 3rd party seller momentum will help too.
In 2024, cost to serve improvements will come from a few increasingly important places. First, it’s time for the company to fully leverage its same day fulfillment warehouses around the U.S. It is investing meaningfully in sharpening inventory algorithms to better place demand closer to the end consumer. This means fewer miles to fulfill, lower cost, and happier customers as they get things faster. It’s a big part of 60% of Prime Member orders in Amazon’s 60 largest U.S. markets arriving same or next day (75% in Tokyo, Toronto and London). Another piece of this idea is combining fulfillment boxes when shipping multiple goods to single destinations.
Amazon also continues to quietly work on proprietary robotics hardware and software to automate and streamline in-factory tasks. Beyond these items, it is changing its seller fee structure to motivate usage of local facilities. This should raise the proportion of fulfillment handled by these more efficient warehouses.
Store Selection:
Amazon is pushing hard to make best-in-class selection even better. It added several popular brands like Sonos, Oura, Parade and a new luxury re-selling collaboration with “Hardly Ever Worn” in Europe. Adding items is one way to broaden selection, which makes it easier for 3rd party merchants to add product listings is another. This quarter, Amazon debuted a GenAI tool to make that listing process absurdly easy. Now, merchants copy and paste a URL to the website, upload it, and Amazon builds you a beautiful product description page. It already has 100,000 sellers using GenAI tools like these… and the opening pitch in the top of the first GenAI inning has not even been thrown.
Grocery:
A positive byproduct of shrinking delivery times is gaining customer purchase intent for daily essentials and groceries. If I can get a toothbrush or an apple from Amazon in an hour, why drive to CVS or Kroger? That’s the thinking here, and Amazon’s increasingly timely delivery capabilities are facilitating great perishable goods growth. It was always a decent player in non-perishable foods; now fresh food is on the table. Its formula of Whole Foods and Amazon Fresh stores (which are performing well per the team), plus world-class fulfillment could make Amazon a large grocery player down the road.
- Debuted a newer $9.99 grocery subscription Prime benefit. This pays for itself after one use per month, which should fuel more momentum.
- Will launch “Whole Foods Market Daily Shop” in Manhattan this year as a smaller Whole Foods store concept.
Advertising:
Advertising growth remained quite strong at 24% Y/Y on an FX neutral basis. This is still being driven mainly by sponsored listings, which have a long way to go in terms of growing ad load. Separately, streaming ad performance from the nascent Prime Video ad-tier is off to a great start. The added granularity of targeting and measurement available in the CTV world is driving strong ad demand early on.
This quarter, AWS debuted the Amazon Publisher Cloud. This uses AWS Clean Rooms to ensure an ad buyer can access all 1st party data (and Amazon’s) in a safe and secure environment. This, paired with enhanced audience segmentation and campaign planning, should continue to bolster targeting and returns.
International:
- EBIT inflection reflects progress across mature and developing markets.
- Remains highly confident in all new markets eventually becoming profitable.
- New availability zones coming in Saudi Arabia, Mississippi and Mexico.
Other:
- Kuiper is almost ready for a commercial beta test.
- Added same-day prescription delivery in 2 cities with 10 more coming in 2024. Added more Eli Lilly drugs (including in obesity).
- Prime Video will get its first NFL Wild Card game next January.
- Launched Ring Battery Doorbell Pro.
- Zoox, its self-driving robo-taxi, is expected to debut later this year. It just got a California Public Utilities Commission Driverless Autonomous Vehicle Pilot permit. Quite the mouthful. It will soon operate in Las Vegas too.
Take
The quarter was excellent. All of the AWS commentary was positive while pretty much everything else was too. Ramping FX headwinds powered the next quarter guidance miss, which is one of the least concerning reasons out there. Additionally, the aggressive CapEx guide was all but inevitable.
Amazon is the king of e-commerce and the largest member of a three-headed cloud monster. Its leading position in two highly compelling structural growth markets shouldn’t be taken lightly. I expect far more profitable compounding in the years ahead; I expect Amazon to remain near the top of my portfolio for a long time. More of the same… solid print…. next.
