On Intelligent Cloud and Azure specifically, the AI portion of Azure passed a $13 billion revenue run rate vs. more than $10 billion guided to last quarter. AI revenue rose 175% Y/Y and contributed 13 points to Azure growth. That portion of the intelligent cloud was better than expected, as it overcame some expected capacity constraints with strong execution. It’s killing it. Conversely, the segment-level miss for Intelligent Cloud overall was related to two things. First is a larger-than-expected FX headwind, which is outside of their control and not concerning. Even when excluding that, however, the segment still missed expectations slightly. That’s due to disappointing execution within the non-AI portion of Azure. This was especially true for channel partner selling, where MSFT tweaked their sales motion last summer to emphasize AI. It thinks that it deprioritized non-AI selling too much in the go-to-market and has left some demand on the table. It understandably wants to focus on AI services, but thinks it needs to shift the balance back a bit towards the middle. Candidly, I think this bodes very well for Google Cloud, Oracle Cloud Infrastructure (OCI) and AWS when they report earnings next month. If a key competitor is misstepping… they’re probably benefitting. Notably, during the Q&A, Amy Hood hinted at risk to its previous Azure growth acceleration due to this specific issue. As a reminder, they had expected this to come during the second half of the year, which means starting next quarter. That isn’t meaningfully happening based on the guidance laid out below. She didn’t explicitly say the reaccel isn’t coming in Q4… but did pour a bit of cold water on that expectation.

Next Quarter:

On the Azure guidance, AI service momentum remains very strong and capacity restrained. They’re selling everything they can provide, and that’s slowing down growth just a bit for now along with FX. It expects to have the capacity it needs in 6 months. For non-AI Azure services specifically, it will take time to fix the aforementioned problems that arose this quarter, and that’s likely another modest source of the miss.

For the full year, it continues to expect double-digit revenue and EBIT growth and now sees EBIT margin rising a bit Y/Y. This compares to initial 2025 guidance calling for a 1-point margin decline. For next year, it thinks CapEx growth rates will begin to slow. Furthermore, some CapEx will begin to shift back towards short-lived assets like GPUs, which are “more correlated to revenue growth.” This year, the majority has been allocated to long-lived assets (like building data center buildings). 

EPS is expected to compound at a 13% clip for the next two years. FCF is expected to increase by 7% this year (CapEx), and then compound at a 25% clip over the next two years. I think CapEx commentary could support estimate stability for FCF, while I think the EPS beat this quarter paired with the profit guide miss next quarter will keep EPS estimates mostly stable.

CapEx Thoughts Following DeepSeek News:

Like Meta and Tesla, Microsoft doesn’t seem deterred by DeekSeek innovation for now. According to Satya Nadella, AI scaling laws (more compute with more parameters means better models) “continue to compound across both pre-training and inference.” And? Efficiency gains are nothing new. They’ve routinely delivered large price performance gains each time they add new hardware or optimize existing models with software. They’re not afraid of this… they encourage it and simply view DeepSeek as pushing forward the trend of consistent model improvement. Per Nadella, “as AI becomes more efficient, they will see exponentially more demand.” As I spoke through in last night’s DeepSeek piece, cheaper models could potentially weigh on GPU renting demand and/or GenAI workload pricing power. At the same time, Microsoft thinks that pressure will coincide with an explosion in data processing, app traffic and overall workload demand. All of those things will greatly help Azure offset this potential hit. Furthermore, it will also mean it costs far less for it to upgrade Copilot and its other AI-powered tools, to drive affordability and accessibility. And finally, OpenAI can use all of this work for its own advancement too. That doesn’t mean they’ll stay ahead of competitors, but it does mean this news won’t push them too far behind.

“I think DeepSeek has had some real innovations. There are some things that even OpenAI found in it. And those [innovations] will be broadly used… When token prices fall and computing prices fall, that means people can consume more, and there'll be more apps written… this type of optimization means AI will be much more ubiquitous. For a hyperscaler like us, this is all good news as far as I'm concerned. ”

CEO Satya Nadella

There are puts and takes, and leadership thinks the net effect is positive. Going forward, it will remain fixated on scaling its global infrastructure with the “right balance of training and inference.” To me, this was Nadella telling us the same thing that Zuckerberg did last night: Compute demand will not shrink from DeepSeek or other companies making models more efficient; finite budgets will simply shift to other parts of the GenAI model journey. The overall amount of compute and Azure infrastructure will be governed by near and long term demand signals (as always) and it doesn’t see those demand signals starkly changing in the wake of this new model competitor. Still, it did begin to introduce some language hinting at the CapEx cycle potentially peaking, with the aforementioned commentary on flat Q/Q growth for the rest of 2025 and slowing CapEx growth in 2026. We are starting to see preliminary signs from others pointing to this CapEx cycle slowing too. Tesla talked about 0% Y/Y CapEx growth in 2025; Meta seemed open to pulling back if that became the correct decision.

“We remain very happy with the partnership with OpenAI.” –

CEO Satya Nadella

In summary, Microsoft continues to see proprietary advantages in having OpenAI’s models; it also sees a clear path to durable model differentiation in general. It doesn’t see these products racing to commoditization like I do, but that could be partially due to defending a massive investment in OpEnAI. It thinks cheaper models benefit OpenAI and itself by delivering an explosion in demand across other parts of Azure and GenAI. It does not think cheaper models mean OpenAI is in trouble.

Copilot & AI – Table Setting:

Just like with previous quarters, a main theme of this call was Copilot. Because Copilot is being added to basically everything Microsoft offers, it will also be laced throughout this section. Copilot is a GenAI chatbot application that leadership calls the “user interface for AI.” It can summarize text, help write code, answer queries and so much more. 

Azure – The AI Infrastructure Layer

Pace of cloud migrations remained quite strong for Azure, as nine-figure contracts helped power part of the commercial bookings strength (in addition to new OpenAI commitments). In a single deal this quarter, UBS migrated 2 full petabytes of data to Azure, to give an idea of how large some of these migrations truly are. With more than 80% of digital infrastructure still hosted on-premise, this migratory revolution shouldn’t be slowing down anytime soon. That’s why Azure, AWS and GCP are such incredible businesses: They have infrastructure moats that nobody can touch… within a gigantic sector… controlled by a few vendors… that should provide reliable, multi-decade structural growth. Nice spot to be in.

Azure AI Foundry is Microsoft’s cloud-native platform for experimenting with, creating and deploying complex AI agents. This debuted 60 days before the quarter ended, and already has 200,000 monthly active customers. That’s the beauty of debuting new products to a giant, entrenched, sticky customer base that is reliant on your enterprise software bundle… It's very easy to cross-sell at scale. Part of the utility of Azure AI Foundry is the broad selection of models offered to customers. These include exclusive access to OpenAI’s latest work, its “Phi” family of small language models (SLMs) and several open-source options, including DeepSeek’s R1 model. Willingness to offer that product so quickly gives more evidence of Microsoft seeing them as supporting the overall opportunity, rather than a structural risk. The DeepSeek model on Azure AI Foundry comes with “automated red teaming (attack simulation), content safety and security scanning.” 

Microsoft Fabric – The AI Data Layer:

Microsoft Fabric is the fuel to Azure AI’s fire. As a reminder, Fabric is Microsoft’s data platform, enabling seamless, rapid access to querying and data processing. The interoperability associated with using one vendor for cloud and data storage means lower data transfer costs, lower query latency and often lower costs. This gives it the ability to affordably train GenAI models within Azure AI and drive inference to ensure GenAI apps like Copilot are as valuable as they can be. 

Per Microsoft, “Fabric is breaking out” with now 19,000 customers vs. 16,000 Q/Q. This has quickly become the fastest-growing data analytics product Microsoft has ever offered… and that makes sense considering data is the key ingredient for differentiated model and app development. If a company has better data and identical model algorithms vs. another… they’ll have better products. Just like if one similar athlete has better coaching than the other. Microsoft thinks it has the largest source of organization knowledge data on the planet.

Furthermore, leadership explicitly talked about the AI explosion fostering an emergence of “new data patterns” and storage, database service and app growth. For example, Azure OpenAI apps running on Azure’s database and app services doubled Y/Y. This directly emphasizes the notion that having unmatched datasets, distribution and analytics tools is what will provide sustainable differentiation. There may be modest cost and scale advantages to enjoy down the road within models, but this is really where companies will durably separate themselves from the pack. Because Microsoft has all of these things, models becoming cheaper to train via DeepSeek isn’t an existential threat to this business. GenAI favors incumbents.

Enterprise AI Deployment Journey:

Per Satya, companies are now pushing beyond experimentation and beta testing, as they gear up for “enterprise-wide deployments,” and Microsoft is ready to support them with a wide range of products across various layers of the AI stack.

Copilot Studio – The Developer Environment:

For review, Copilot Studio is Microsoft’s secure developer environment to create custom AI agents on top of existing apps and models. The product now has 16,000 customers, which represents 60% Q/Q growth, while agents created with it have doubled Q/Q. So? Far more customers are using the product more frequently than they did 3 months ago. This quarter, it added its own agents to help “facilitate meetings, manage projects and resolve HR queries” while software leaders including ServiceNow, SAP and Adobe are leaning on it heavily to create their own agents.

Microsoft 365:

Microsoft’s productivity suite is seeing “accelerated customer adoption across all deal sizes” for Microsoft 365 Copilot. What’s even more encouraging is that once customers opt into Copilot, usage is consistently expanding. Since launching this product 18 months ago, the first quarterly customer cohort has raised their consumption by 1000%. 365 Copilot daily active users continue to double Q/Q and engagement rates are setting new highs, with usage intensity up 60% Q/Q. Companies like Salesforce love to rip on Microsoft for its Copilot products being all promise and zero delivery. Whether this fantastic, outperforming momentum is due to elite bundle stickiness or because these products are simply valuable (probably a combination), these results show that Copilot is more than capable of sustainably growing after a historically strong start.

More Products:

In Personal Computing, 15% of all premium laptops sold during the holidays were Copilot+ PCs. It thinks this will rise to over 50% in the coming years. In the near future, companies will be able to use DeepSeek’s R1 model on their Copilot+ PCs. Elsewhere in Personal Computing, gaming growth continues to be challenged by an intentional shift to prioritizing higher-margin products, such as Game Pass. Game Pass subscribers rose 30% Y/Y. Black Ops 6 enjoyed the best launch in the franchise’s history.

LinkedIn is enjoying “record engagement” across every single product it sells – from marketing to sales to premium subscriptions. Within subscriptions, the segment crossed a $2 billion revenue run rate while 40% of all subscribers are using AI tools to perfect profiles. Short-form video is growing at 2x the pace of all other content formats on that platform and it’s now adding AI agents to automate lead generation. 

In Search Advertising and News, Microsoft Edge rose to 30% U.S. market share for Windows-powered devices and took market share for its 15th straight quarter. Its work to improve ad targeting and measurement is leading to higher returns for advertisers and so more pricing power for Microsoft.

f. Take

This wasn’t Microsoft’s strongest showing. The go-to-market misstep on non-AI Azure revenue is surprising and so is the Azure miss paired with the weaker-than-expected guidance. FX is one cause of this, which isn’t within their control. The Azure execution item is within their control and needs to be expeditiously fixed. That’s very likely to happen, as this team is world-class and so is their bench of talent. Aside from this, AI product momentum is strong, Copilot engagement trends are fantastic and Azure’s core business is growing rapidly on a massive revenue base. Cloud computing tailwinds are wildly strong and durable, while its productivity business is thriving. This should be a blip on the radar when zooming out. 

Apple guided to low-to-mid single-digit revenue growth, which missed 5% Y/Y growth estimates. At the same time, FX headwinds are stronger than expected for it, just like for Microsoft and Meta. Without this incremental hit, it would have guided to roughly 4% Y/Y growth for the quarter, which still modestly misses estimates. Its 47% GPM guidance slightly beat estimates by 10 bps. I thought that was a strong result, considering incremental FX headwinds also are impacting gross profit dollars. Assuming revenue guidance means 3% Y/Y, its $15.2 billion OpEx guide implies about $28.5 billion in EBIT, which represents a 4.4% beat.

Apple trades for 32x forward EPS and 29x forward FCF. GAAP EPS is expected to compound at a 15% clip for the next two years (adjusted EPS at 11%). FCF is expected to compound at a 9% clip for the next two years. Profit estimates should modestly rise following this report.

Apple Intelligence:

For review, Apple Intelligence is the company’s GenAI-powered “intelligence system” offering the power of world-class models from OpenAI and its own proprietary tools. The rollout of Apple Intelligence took place in October, with the second iteration of iOS 18 released in December. For now, this comes with email summary and writing tools, emoji generation and help with organizing schedules. It also comes with visual intelligence, which can contextualize a person’s surroundings with the phone camera. Pretty cool. Additionally, it features a more conversational Siri, although there’s a lot more work to do there as leadership will tell you. Apple was bluntly asked why adoption here has been rather underwhelming in the eyes of analysts and why it hasn’t yet delivered any kind of meaningful iPhone growth acceleration. 

Cook explained this with a few simple items. First, the second iteration of iOS 18 came out with just two weeks left in this past quarter, with it only available in a few markets. Next, it has only been rolled out in a few countries, and the list does not include China. Considering that’s the most competitive smartphone market on the planet, and is probably hurting their positioning there a tad. We saw some research come out mid-quarter on Apple losing its top market share spot in that country, and this could be contributing. Lastly, Apple Intelligence is only available for iPhone 15 Pro and the new iPhone 16 models. It’s still rolling out in many places, with recent launches in New Zealand, South Africa, the UK etc. And to give confidence in these rollouts being successful, iPhone markets with Apple Intelligence available are outperforming markets without it. The focus will remain on introducing this to the rest of the globe and continuously updating the software to inspire more upgrades. For signs that this might be beginning to work, Apple set a new record for iPhone upgraders during the quarter. At the same time, another weak quarter for total iPhone revenue casts doubt over how powerful of a tailwind this will be.

With the Apple Intelligence product suite and utility still developing, the company is leaning on its privacy-first approach to GenAI to drive innovation. As reviewed in my coverage of the 2024 Worldwide Developer Conference (WWDC), Apple is fixated on maximizing data processing while minimizing data sharing or leakage. It isn’t impermissibly collecting and using personal data to season GenAI models. And when more processing power is required for a query than a device can provide, Apple’s private cloud (run by Apple Silicon M4 chips) will field traffic.

More on iPhone:

Apple set new revenue records in “dozens of markets” across North America, LatAm, Southeast Asia and Europe during the quarter. That’s tough to juxtapose with things like roughly 0% growth for the product over the last two years, the aforementioned upgrade record note and its install base reaching a new high. This is what led many to question how compelling of a recipe Apple Intelligence and iPhone 16 is for consumer upgrades. Apple remains confident, but it has more to prove beyond abstractly saying markets with Apple Intelligence are performing a bit better early on than those without it. 

This strange iPhone performance is also related to -11% Y/Y growth in Greater China. Interestingly, half of that decline was due to a needed inventory channel reset throughout the quarter, while the rest of the decline was due to weak macro in the region and heated competition. As Cook reminded us on the call, a national-level stimulus package from the Chinese government was issued a little over a week ago, which should help every single player in that region. It now thinks Chinese revenue is in a good spot, partially thanks to a better-than-expected end to the quarter, and Apple doesn’t see inventory bloat in any other markets.

451 Research continues to report iPhone’s customer satisfaction score at a sky-high 96% and Kantar Research reported that iPhone remains the top seller in Urban China (not Greater China), the USA, India, France, Japan and more markets. Again… a strange mix of extremely positive items and some more underwhelming items.

Other Hardware:

iPad enjoyed 15% Y/Y growth thanks to strong product iPad Mini and Air launches and demand. Incredibly, more than 50% of all iPad’s purchased this quarter were from brand new customers. Where are they finding these people? So impressive. Customer satisfaction rates sit at a lofty 96% per 451 Research.

For Mac, 16% growth was also thanks to strong product launches like the M4 chip-powered Macbook Pro. Apple also delivered more than 10% Y/Y growth in upgrades and also brand new Mac customers. Apple Silicon continues to be a core point of differentiation for the product. In the eyes of Tim Cook, this gives Apple the “best AI PC out there for running workloads.” Customer satisfaction rates sit at 94% per 451 Research.

For Apple Watch, we got some heartwarming stories about the product saving lives of folks with low heart rates, sleep apnea and atrial fibrillation. While these notes are anecdotal, they do point to the unique utility that the Apple Watch provides. Growth for the segment was challenged by comping over the successful Watch Ultra 2 launch Y/Y, but the install base still rose to a new high and 50%+ of customers were brand new to the product. I personally love mine and 94% of customers overall love theirs too – per 451 Research. 

Services:

Apple’s services runway continues to be supported by an increasing active hardware install base. The more products in circulation, the more services will be consumed. This quarter, Apple crossed 2.35 billion active devices vs. 2.20 billion 9 months ago. Apple TV+ continues to resonate with viewers and Apple keeps layering on services like a new Airtag integration with airlines to track suitcases with their help. Engagement levels remain strong as transacting and paid accounts again set new highs, with paid account growth north of 10%. They were asked about the margin ceiling for this segment a few times, but didn’t share much more detail about where they think things can go.

Enterprise Adoption:

Global Store Expansion:

Final Notes:

Apple has virtually nothing to lose from DeepSeek and other model vendors driving model building efficiency. It has everything to gain in terms of using these now cheaper input costs to build better apps and products. It never got into the game of building the best model or investing in the company trying to do that. It has partnered and instead leaned on its fortress ecosystem and distribution moats. 

They had nothing to say when asked about tariffs. Good decision. Nothing to gain from talking about this.

Apple is one of the best companies that our world has ever known, if not the best. Their humongous cash pile is the byproduct of that and creating products beloved by me, you and a large chunk of the planet. We often joke about the “blue text bubble” moat, but that’s actually somewhat real. Not having an iPhone can often make people (especially North Americans) feel like outsiders in group chats. All of this is why Apple can sell expensive hardware at higher margins than its competition can.

But? Where’s the growth? A 3% 2-year revenue CAGR doesn’t include any pandemic pull-forward to create tougher comps. I don’t know where Apple will find this growth aside from layering on another service here and there and selling a few more products during upgrade cycles. I think this company can continue to thrive in the coming years. But I don’t think it’s currently positioned to enjoy decades of future dominance like it has up until this point. I see very little optionality and little room for expanding profits beyond buybacks. I see very little desire from the current team to do anything but rely on the previous, historic success of Steve Jobs. In my mind, they need to use the balance sheet to expand to new product categories and entirely new businesses going forward. And that’s an extremely hard decision to make. Shareholders will not uniformly support fewer buybacks and lower near-term profitability for this cash printing machine. They’re in a tough spot, in my opinion.

When will Apple create anything new? Not sure. When will that begin to matter? Also not sure. But I think there are better places to invest at this time. Owning some Apple will surely not kill you… but I don’t think it will generate much alpha either.