The detailed earnings reviews for all are coming soon. For Apple and Airbnb, revenue/profit is highly seasonal and Q/Q comps shouldn’t be focused on.

Demand vs. Expectations:

Profitability vs. Expectations:

Balance Sheet:

Guidance & Valuation:

Low-to-mid single-digit revenue guidance for next quarter. If we assume that means 3% Y/Y growth, it missed estimates by 5.6%. 46% GPM guidance missed 46.7% estimates. EBIT missed estimates by 11%.

As of right now, Apple trades for 29x forward EPS and likely closer to 30x following this report. EPS is expected to compound at a nearly 9% clip for the next two years.

Demand vs. Expectations:

Profitability vs. Expectations:

Balance Sheet:

Guidance & Valuation:

What an excellent quarter this was. I can’t wait to write the review. Spoiler alert — it will be positive.

Demand vs. Expectations:

Profit vs. Expectations:

Balance Sheet:

Guidance & Valuation:

Duolingo trades for 50x forward FCF and likely a few turns lower following this report. FCF is expected to compound at a 39% clip over the next two years. Those growth estimates will be revised higher.

“Demand signals for the commercial business have been consistent through April” per CFO Amy Hood. Their guide assumes that continues. Its range of outcome for the MPC segment is wider than it usually is to factor in tariff-related uncertainty and changing inventory behaviors.”

Cloud & AI Infrastructure – Runway

Microsoft continues to aggressively invest in more AI infrastructure. Like it said last quarter, a lot of that CapEx for this year is earmarked for long-lived assets like data centers, with plans to shift that to short-lived assets like chips and servers as we head into their new fiscal year. It also reiterated expectations for positive, yet slowing, CapEx growth next year. And in terms of signals to gauge the current GenAI infrastructure boom runway, we got one that’s very positive for players like Nvidia and Broadcom, with one that’s perhaps a bit negative. On the positive side, it now sees supply constraints lasting longer than June. This is based on stellar AI service demand (more later) and bodes very well for the hardware providers. On the other hand, it also said lead times for new GPUs are 20% lower Y/Y, which does point to them gaining easier, less bottlenecked access to this valuable asset.

There were a ton of rumors about Microsoft cancelling data center leases during the quarter. This was understandably the very first question in the Q&A. As Satya told us “they’re always making adjustments.” There are no fundamental changes to its footprint growth plans, as again, it’s still renting all the capacity it can build.

Cloud & AI Infrastructure – Performance

Microsoft continues to work hard on driving cost efficiencies across every piece of the AI stack… from chips to models to apps. Through this work, they’ve boosted performance by 30% and cost per token by 50%. They continue to be uniformly excited about training and inference cost deflation, as that deflation is expected to greatly spur demand and be a large net positive for the business. Like I’ve said many times… it’s the exact same process we saw play out in the cloud computing revolution. The industry continues to double model performance every 6 months.

“Cloud and AI are the essential inputs for every business to expand output, reduce costs, and accelerate growth. From AI infrastructure and platforms to apps, we are innovating across the stack to deliver for our customers.”

CEO Satya Nadella

Intelligent Cloud and Azure Migrations:

The pace of Azure migrations is accelerating, with ServiceNow, Coke and Abercrombie among companies to meaningfully expand deployments this quarter. AI services added 16 points to overall Azure growth (vs. 13 points Q/Q). That surely includes a lot of GPU re-selling, as part of the strength was thanks to more capacity than expected coming online. Still, it wasn’t the only source of the upbeat showing. Adoption of Microsoft Copilot (GenAI and Agentic AI assistant) across its suite of tools is also going very well. More on that throughout this piece. Encouragingly, AI services were not the only reason for MSFT’s success. Non-AI services performed very well and drove a material chunk of Azure’s outperformance. I realize this piece of Azure is not nearly as sexy, but it is still highly important. It’s almost funny that we call cloud computing the “legacy portion” of Azure now, considering more than 80% of compute is still on-premise and most of that will migrate in the coming decades. The runway for the “legacy” portion is still massive and still important to nurture. Last quarter, it didn’t do a great job of that. It shifted priorities too far away from non-AI products, and left revenue on the table. They vowed to rebalance their focus, and this great quarter is the byproduct of that. Really well done.

The Intersection of Data & AI:

Broken record alert: Data and AI services work better together, under one vendor. This leads to better interoperability for data transfer costs, better querying performance and, generally speaking, more productive work.

Microsoft’s PostgreSQL (standard query language) offering is doing very well. As a reminder, PostgreSQL is an open-sourced relational database that primarily handles structured data and deployment. Usage again accelerated, with 60% of the Fortune 500 already using the product.

Its Not Only SQL (NoSQL) product, called CosmosDB, is perfect for handling unstructured data, which is vital in the world of GenAI. It’s a great document-oriented addition to PostgreSQL’s more statically organized rows and columns of data. This has also enjoyed accelerated growth, as GenAI customers lean on MSFT for this highly complementary offering.

Microsoft Fabric is its overarching data platform, enabling seamless access to the data products already mentioned in this section. This is where data warehousing, Power Business Intelligence (aggregates and displays data from various sources to accelerate learning), and its overall data analytics businesses preside. It added 2,000 customers Q/Q following 3,000 net new additions last quarter, maintaining a rapid double-digit sequential growth rate and 80% Y/Y growth. Within Fabric, real-time intelligence is the shining star. It is already up to a 40% customer adoption rate less than two full quarters into launch. 20% of its total customers now use 3+ real-time intelligence products as well. Sticky.

Finally, its multi-cloud data lake is doing very well. I think this could become its most important data offering in the future. Nobody wants to be a single cloud shop anymore, and this reflects those involving tastes. It lets customers tap into whatever cloud provider they want to use within the overarching Azure environment. Data within this lake is up 500% Y/Y.

AI Platform & Tools:

Foundry is Microsoft’s cloud-native platform for experimenting with, creating and deploying complex AI agents. It already has 70,0000 customers a few quarters into launch. I realize all of MSFT’s products rapidly ramp to tens of thousands of customers, but we still shouldn’t overlook that. That’s just the luxury of owning arguably the stickiest enterprise bundle in the world. Cross-selling is easier. Token processing (indicator for usage and traffic) is up 5x Y/Y. The newer agentic offering under this service is already up 10,000 customers, and more “fine-tuning tools” to enable deeper agent and model customization should keep momentum strong.

Speaking of models, its Phi series of small language models (SLMs) is up to 38M downloads with more mini models and CPU-only models coming soon. CPU-only is interesting to me, considering GPUs are the chip type powering accelerated next-gen compute. Perhaps this is a way for Microsoft to help extract more value from those assets for its customers.

Quantum Computing – “the next frontier of cloud systems”

Microsoft is “making real progress on a path” to a scaled quantum computer. It recently introduced Majorana 1.

Developer Tools:

GitHub Copilot is adding more agentic, multi-step functions within code creation, bug uncovering and repairs with “Autofix.” It’s also previewing a software engineering (SWE) agent that can “asynchronously execute developer tasks.” Microsoft continues to push closer to automating the end-to-end source code-to-software package deployment process. It will take time to get all the way there, but these releases are important marks of progress. All in all, GitHub Copilot users are up 4x Y/Y, with Twilio, Cisco, Target and more using the service.

The “Future of Work:”

Copilot for Microsoft 365 keeps enjoying larger deal sizes every quarter. In addition to that, it also enjoyed a record number of customer seat expansions during the period. Overall, this now has hundreds of thousands of customers – with 200% Y/Y growth. To create a more intuitive and less siloed interface, it consolidated collaboration notebooks, search, and AI agents into one product. It also added deep reasoning agents and AI researchers to augment the complexity and richness of token outputs this product can provide. Just like everyone else, MSFT is pushing from GenAI (tell it what to do and how to do it) to agentic AI (tell it what to do and count on it to get there in the best way possible).

Copilot Studio:

As a review, Copilot Studio is Microsoft’s secure developer environment to create custom AI agents on top of existing apps and models. It offers no-code tools to make this highly accessible. 90% of the Fortune 500 have already used this, and (shockingly) MSFT is rolling out more agentic workflows to augment Copilot’s utility. Customers created 1 million customer agents during the period, representing 130% Q/Q growth.

More Personal Computing (MPC) Notes:

Microsoft (thanks to Activision M&A) is now the top publisher on Xbox and PlayStation. PC Game Pass enjoyed 45% Y/Y revenue growth, and a new cloud gaming LG TV integration provides more room for expansion. Generally speaking, cloud gaming strength was good, with MSFT crossing 150M hours played for the first time ever. It’s now adding Copilot for gaming, with custom game coaching.

We continue to see increased commercial traction as we approach the end of support for Windows 10.

CEO Satya Nadella

More Product Notes:

Satya on Macro Insulation:

“Software is the most malleable resource we have to fight any type of inflationary pressure or any type of growth pressure where you need to do more with less.”

CEO Satya Nadella

Microsoft, the enterprise software king, is quite insulated from tariff risks. It has no large import/export business outside of maybe some hardware components on the MPC side. Furthermore, its ad business is less important to growth than Meta, Amazon and Alphabet. Chinese sellers pulling back from Bing placements is not nearly as material as it is for those three.

Excellent quarter. There has been a lot of trash talk from the Marc Benioffs of the world about Microsoft Copilot being effectively worthless and chock-full of empty promises. Looking at these numbers makes it clear to me that this isn’t the case. Copilot is providing value. That is why customer growth and engagement per customer growth both remain rapid. This company is leading most of the world in terms of monetizing GenAI (outside of building the hardware) and this quarter just proves that once more. Rock-solid performance from a world-class company with a legendary leadership team. Congratulations to shareholders. Why don’t I own it? Because I’m a doofus. That’s why.

Q2 EBIT weakness was based on an expected step-up in stock-based compensation, recognizing a chunk of Kuiper-related expenses & some impact from tariffs. They’re planning for a “wider range of outcomes” than they normally would, and that’s reflected in the guidance.

FCF is expected to compound at a 39% clip for the next two years. EPS is expected to compound at a 16% clip for the next two years. Amazon trades for 31x EPS and 40x FCF. Estimates could modestly fall following the large Q1 beat and the Q2 profit miss.

Stores & Selections:

Amazon added several new brands to the marketplace, including Michael Kors and Oura Rings. I think more interestingly, however, it added a Saks “refined luxury assortment of fashion and beauty items from Dolce Gabbana, Jason Wu and more. I get excited when I hear this, as building a large luxury fashion business should be highly incremental to its overall marketplace revenue and the tastes that it can cater to.

On the other end of the spectrum, it’s fixated on controlling prices in this tariff-obsessed environment wherever possible. Amazon’s decision to localize outbound fulfillment over the last few years is overwhelmingly coming in handy right now. This has put cost to serve in a much better place and given them far more wiggle room to deal with current profit headwinds in a more seamless manner. This is also why it still makes sense for them to be selling low-priced everyday essentials, which is quite important considering growth for this category is doubling its stores business overall.

Tariffs & Stores:

Amazon has not seen demand weakness across any categories yet, but it has seen some pull forwards for certain goods. In terms of selling price, most sellers haven’t changed theirs yet. This is because many have loaded up on Amazon inventory ahead of expected tariffs to delay paying them for as long as possible. Amazon did the same thing with its own 1st-party inventory.

Despite nearly 50% of its top sellers being from China, it’s confident in relative insulation throughout the trade war. It still has over a million merchants outside of China and a massive array of goods, from a massive array of competitive substitutes. This should breed competition, the ability for some vendors to keep prices lower due to lower tariffs in some places, and continued effective price competition overall. Price tags may go up for Amazon; but it’s in better shape most others – despite the reliance on Chinese sellers. That is how they see things. They’ve always taken market share during tough times because of their economies of scale, fortress balance sheet, lower relative prices, world-class subscription and ubiquitous brand. They think the same thing will play out during the trade war and they will emerge stronger from it. To me, the worst-case is that higher tariffs are permanent, and profit growth is held back for 4 quarters while comps normalize and it gets creative on finding different sellers. But even if that begins to play out, we’re one headline away from the backdrop becoming increasingly favorable once more and profit upside coming.

Amazon’s guide provides room for a wider range of outcomes and does bake in some impact from expected tariffs. Again, because they’re so aggressively stocked-up (both 1st party and their sellers), they’re in pretty good shape to outlast a trade war through June. As it drags on beyond this, the hit from trade policy will ramp. Let’s hope for positive resolutions. We just got a Bloomberg headline 30 minutes ago on China considering USA trade proposals. A lot can happen in 90 seconds in this environment… let alone 90 days.

More on Fulfillment Cost & Efficiency:

As discussed in previous quarters, the next fulfillment efficiency project is inbound fulfillment regionalization. The redesigned workflows have been finalized and rolled out. This is already yielding incremental speed, fulfillment package combination and cost to serve gains. Jassy and his team are now focused on routing more inventory into their local fulfillment centers with this new approach. This is going to be a constant, iterative process over the coming quarters and years, with a very encouraging start. 

Advertising:

We didn’t hear much about this segment aside from it being pleased with Y/Y growth stability on a larger base. It continues to be extremely early across both top-of-funnel brand marketing opportunities and bottom-of-funnel performance marketing. It’s also extremely early for its demand-side platform, where it will look to grow its programmatic advertising presence. As that happens, impressions should get more and more relevant and pricing power should rise.

AWS:

AWS remains supply constrained and growth rates would have been higher if it were able to procure more capacity. The supply chain is still not in amazing shape, but it expects bottlenecks to ease throughout the year. Jassy strongly hinted at growth rates accelerating as that happens. This will also mean accelerated spending to place all of this capacity, which probably means the record 39.5% AWS EBIT margin will be weaker during the second half of the year.

“I think we could be doing more if we had more capacity, and I expect that the capacity to ease in the coming months… Before this generation of AI, we thought AWS had the chance to ultimately be a multi-hundred billion dollar revenue run rate business. We now think it could be even larger.”

CEO Andy Jassy

Chips:

Trainium 2 is starting to “land capacity in larger quantities with significant appeal and demand.” This isn’t because its GPUs are better than Nvidia’s. They aren’t. It’s because they’re 30%+ cheaper and many customers care more about cost than best-in-class performance.

Frontier Models:

Amazon continues to obsess over model choice within Bedrock. For all parts of the AI opportunity, it simply wants developers to have whatever tools they need or prefer. It doesn’t care if that’s Anthropic’s new model, DeepSeek R1 or its own Nova models. It mainly cares about workloads running on AWS. That’s how it will win.

For its own models, it just launched the Nova Premium model, which offers “leading price performance.” Slack, Coinbase, FanDuel and, Blue Origin are early users. Nova Sonic is its speech-to-speech model that unlocks audio-centric use cases with lower word error rates and higher win rates vs. comparable models. And for the world of Agentic AI, it debuted Amazon Nova Act. This is its actionable web browser agent that can conduct several-step tasks and break stated goals into “atomic commands like search and checkout. It hopes this product will raise accuracy rates for multi-step tasks from under 50% to over 90%.

More Notes:

This was an average quarter for a world-class company. They’re somewhat vulnerable to tariffs, and I think the Q2 guide does more than enough to account for that downside risk, while leaving room for upside surprise if things brighten a bit. If things don’t change before its Q2 call this summer, profit will likely be challenged again in the Q3 guidance. We’ll see how the trade war progresses, but I am exceedingly confident that Amazon will weather this storm better than anyone else can. The longer any potential economic turmoil from tariffs persists, the more market share this company will take from smaller players that are unfortunately less capable of overcoming these obstacles.  Either profit growth is challenged for four quarters before comps reset and normalize (with more market share), or we get a positive resolution and it’s business as usual. 

The AWS miss was tiny and a byproduct of capacity constraints… so is not at all concerning to me. Kuiper is progressing very nicely. Its ad business is rocking. This is a holding I really do not worry about much. I remain a confident shareholder and love having Amazon near the top of my portfolio.