Meta and Microsoft commanded most of my attention tonight and those are the two full reviews in this article. Still, I know there’s a lot of interest in these three companies, so I wanted to include very brief snapshots of each report tonight. While my reviews function as complete, detailed dives into reports, these snapshots offer 30,000 ft. views.

For Lemonade, I will include more detailed coverage, quantitative data and my views on the report in Saturday’s article. I don’t follow the other two names closely, but I also plan to cover Robinhood’s call in more detail Saturday if time permits. HOOD has implications for other names that I do focus on.

Lemonade expects to be EBITDA positive in 2026. Wildly speculative investment. High risk/high potential reward.

EPS is expected to double this year and fall by 3% next year.

Recent Content to Read:

Amazon & Apple coming tomorrow. Dozens more reports coming this season.

Q4 revenue guidance beat expectations by 0.6%. For the year, it slightly lowered its annual OpEx guidance, while slightly raising its CapEx guidance by a modest 1%. It continues to expect CapEx and associated depreciation growth in 2025 vs. 2024. It’s still working on a finalized 2025 CapEx budget. As a reminder, it has great flexibility to allocate these investments to different areas if need be. The dollars will not go to waste.

Meta EPS is expected to grow by 44% this year and by 15% next year.

“So when we think about the Q4 outlook and when we think about going into next year, we generally expect growth to continue to benefit from the healthy global advertising demand that we've seen.”

CFO Susan Li

Meta splits its AI work into near-term and longer term financial opportunities. The near-term work is called Core AI and is really where GenAI is already boosting key performance indicators (KPIs), which should lead to more financial success. All in all, its Meta AI assistant has 500 million monthly active users (MAUs) and is well on its way to becoming the most used free GenAI assistant in the world. Nice to have half of the planet on your apps.

For evidence of immediate Core AI impact, feed and video recommendations over the last year have increased Facebook and Instagram engagement by 8% and 6%, respectively. The former unification of its video recommendation algorithm on Facebook has also boosted Facebook video player engagement by another 10% since launching last quarter. It’s now moving onto “phase two” of its Facebook video user experience refresh, which entails changing the video tab to a full screen watching experience.

Generally speaking, GenAI means Meta can re-imagine how much data processing is economically rational to conduct for a given app or task. Accelerated compute makes extensive processing cheaper and faster, which is what makes cutting edge GenAI apps realistic. Otherwise, general compute would lead to exploding costs and poor performance. With GenAI, The ability to infuse far more data into every content matching event makes that content more relevant. This work has mainly started with Facebook, but is now expanding to more apps. 

As an important aside, GenAI-inspired engagement optimizations are a near term opportunity and a long-term opportunity. Today, it’s about tweaking and perfecting existing experiences. In the future, that attention will shift to creating entirely new ways for customers to engage… like through Meta AI Studio (more later) making it wildly simple for creators to build interactive AI avatars for fans to banter with.

All of this engagement work means more ad impressions, which means more advertising revenue. And to support this added demand, it has rapidly rolled out content creation tools like background generation, text expansion and image generation. The names explain what each product does. Image generation specifically is raising conversion rates by 7% for early adopters. Additionally, its Advantage + campaign building and automation tool is driving a 22% return on ad spend (ROAS) boost for its U.S. advertisers. Meta has 1 million advertisers using these GenAI tools as of right now. Real impact. Today.

“We're seeing strong retention with advertisers using our generative AI powered image expansion, background generation and text generation tools, and they're already driving improved performance for advertisers even at this early stage.”

CFO Susan Li

There’s much more work to do on optimizing ad load to mine more value from existing traffic, as well as perfecting ad timing to ensure ads are as effective and lucrative as possible. And if it executes here, there’s a long way to go for selling more ads. Facebook, for example, saw engagement growth greatly surpass ad impression growth while monetization lags product updates as always. That’s actually encouraging, as Meta has proven time & time again that dollars will follow eyeballs.

The aforementioned ability to process exponentially more data for content matching also unlocked a new targeting tool for Meta. Now, its targeting algorithms can provide a timeline or history of customer engagement to inform impression purchases. To this point, it was only able to generate the list of actions, without any time stamps. This upgrade is raising ad conversion rates by 2%-4% since debuting recently.

WhatsApp continues to enjoy its best success in the USA, with the app crossing 2 billion calls per day globally. Considering Apple’s near-monopoly here, this is highly notable. The USA is where revenue per user potential is the highest, which makes this even more compelling. Monetization is also sharply improving here as Meta delivered 48% Y/Y growth in business message to power its “other revenue” segment. This opportunity remains highly untapped and Meta is now showing real signs of capturing it. Another gigantic app to extract financial value from… it already has the traffic… time to get paid.

Speaking of more apps to grow the ad revenue opportunity, Threads continues to rapidly expand. Anecdotally speaking, I’ve noticed my own engagement on the app has ticked materially higher. It now has 275 million MAUs vs. 30 million about a year ago. It’s also up to 1 million new sign-ups per day; Zuck is convinced it can get to 1 billion. The launch is clearly going very well, as it successfully nudges Instagram users to try out the new product. Again… it’s nice to have half of the planet on your apps. Engagement and retention trends are also “strong,” but this still will not be a material revenue driver in 2025.

Instagram growth was robust across all markets and Facebook’s emerging trend of accelerating young adult usage in the USA continued. Marketplace is becoming extremely popular for the whippersnappers.

Meta’s open source approach to GenAI model building is showing clear signs of working. Broken record alert: it is hard to drive differentiation vs. a sea of competing models. How do you do that? Through more data and lower cost. Meta has all of the needed data.

What about cost? The open sourced approach (paired with world-class frontier Llama models) is a highly compelling combination for independent developers. It strongly incentivizes them to do their work in Meta’s ecosystem, as that gives them the best chance of finding users and making money. Some of this work will inevitably be on Llama efficiency and performance optimizations, which Meta can borrow for itself. Meaning? This approach invites elite developers from around the globe to basically do Meta’s work for it. And as Zuck puts it, as this traction grows, the work will help improve all Meta products… near-term and long-term. Zuck talks constantly about the world wanting to move to open source for better transparency and interoperability. Meta is simply skating where the puck is going and creating a rare model edge in the process. Access is what drives usage is what drives monetization. It’s the same playbook as all of its apps and its Quest business too.

Meta remains firmly in spend mode and sees underinvesting as the true GenAI risk. It needs to lean into infrastructure growth to ensure it has the capacity to support Meta AI, Llama and all of its other work. That costs money. And while investor concern on the call centered on 2025 CapEx following a (very small) 2024 raise, I’m not concerned.

Zuck has clearly shown us for decades how he can walk and chew gum at the same time. He has earned the right to lean into investments here because the core business is so healthy. And? 37% Y/Y EPS growth while talking about being in an investment phase is absurdly positive. I wonder what profit mode looks like.

The commercial query potential from Meta AI will eventually be massive when we get through this phase of building consumer adoption to attract enterprise attention. That should mean significant financial gain down the road. How far? Maybe closer than you’d think. Its Business AI products (automate customer service, sales leads etc.) and its Meta AI studio (environment for easy custom chatbot creation) are one year behind Meta AI in terms of maturation, per Zuck. As a reminder, Meta AI has 500 million MAUs.

AI work and aggressive spend are paving the way for Meta to lead the next computing form factor for another potential value unlock. Just look at how quickly the Ray Ban glasses are flying off of the shelves. That directly tells us how popular the clunky Quest devices can be when they’re miniaturized to the point of looking like shades. The glasses need to be as powerful as Quest 3 and as slick as they look now. That will take time, but it’s coming. Leading technological waves is always expensive and Meta’s founder mentality is bold enough to incur the losses, take the incessant public heckling and focus on the long term. I say go for it.

I remain convinced that these GenAI investments will come with strong returns. Not just within Meta AI and Llama… within Reality Labs too. And for now, FOA is more than capable of putting the financial burden on its back.

Leadership punted two questions about the rumored AI search engine product from Meta AI. It used the opportunity to say nice things about Alphabet.

This was a strong showing. Expectations were sky-high heading into the report, and Meta delivered. I have absolutely nothing negative to pick at and have both of my thumbs pointed firmly up. Great quarter from an elite company and an elite team.

This was a relatively light earnings call compared to most Microsoft reports. We didn’t learn much. Here’s what we did learn:

Microsoft changed revenue bucket categorizations during the quarter and didn’t offer much reconciliation help. For this reason, the table below is a bit lighter than it has been in the past.

“Just like in the past, we will allocate capital to build-out cloud based on the demand signal we were seeing and then we would then project the demand, and that's what we would build for.”

CEO Satya Nadella

The outlook assumes stable business trends.

Microsoft EPS is expected to grow by 11% this year and by 16% next year.

The theme of this call was unsurprisingly Microsoft Copilot and how Microsoft is poised to compete and win in GenAI. Here, we will set the table. As Copilot is being infused into pretty much every Microsoft product, it will be laced throughout this entire section as well. As Satya Nadella puts it, CoPilot is the “AI user interface” and an efficiency force multiplier for good, expeditious work. It is the driving force behind Microsoft reaching a $10 billion AI revenue run rate next quarter and lifting Azure’s growth rate by a full 12 points during the quarter. Success here is highly important, as competition such as Salesforce has been very vocal about thinking the product is low quality and a waste of money. These results offer a compelling counter, while Copilot now works to add agentic AI workflows to unlock goal-oriented tasks with more steps and more model autonomy.

Microsoft also continues to work on its own processors through a partnership with Arm. Its newest Cobalt 100 processor offers 50% better price performance vs. the previous version. Snowflake, Databricks and Elastic are using it.

Azure remains supply constrained. Growth would have been even faster without this bottleneck.

Azure continued to take more market share per leadership. Its $10 million and $100 million contracts drove the significant commercial bookings outperformance for the overall company. Azure Arc also maintained rapid Y/Y customer growth of 80%+. This is its platform for enabling integrated multi-cloud access to a client’s apps and data. Most large companies want to be multi-cloud shops to balance usage across vendors based on cost and performance needs. This makes Microsoft a better partner for enabling that, which is especially important in the GenAI age of access to data being the difference between leading or following in an industry. 

Azure AI is its toolkit for developers to create and operationally deploy high performance GenAI apps and workloads. This, as always, is done in a fully managed way thanks to Azure’s footprint. The product offers templates to create natural language processing, anomaly detection and more apps. Usage here doubled year-to-date as LG and many more were added as customers. GE Aerospace is now using this to build a digital assistant for its 52,000 employees as well.

It continues to work hard on building models and services internally, while also leaning heavily on its OpenAI partnership to help Azure blaze GenAI innovation trails. Microsoft brings the capacity, traffic and balance sheet; OpenAI brings the researchers and models.

Microsoft Fabric is the oil to Azure AI’s engine. This 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. And with Copilot studio customers can customize their own GenAI assistant for more domain-and-company-specific tasks. 

All Fabric now calls 70% of the Fortune 500 its customers.

Github Copilot saw 55% Q/Q customer growth. Copilot Workspace is its developer tool to help companies embrace agentic AI and “go from spec to plan to code with natural language. This is helping customers resolve coding vulnerabilities 3x faster.

LinkedIn growth in Brazil and India was over 10% Y/Y as the app accelerated overall. Its video investments are bearing fruit for advertisers by driving 6x Q/Q viewership in immersive content. Total video viewership on the app is up 36% Y/Y. More time spent… more advertising… more revenue. LinkedIn also added a new GenAI tool to offer granular career development advice and another hiring assistant to find qualified leads. This is yielding a 44% hiring lead acceptance rate boost.

More consistent execution from this iconic software company. The Azure guide was a bit light, but not because of demand. There’s every reason to believe that resolving supply scarcity will yield faster growth in the quarters ahead. It remains one of one in terms of ability to monetize GenAI software early on and is doing a great job balancing needed investments with margin preservation. There was a lot of negative sentiment surrounding Copilot’s value heading into this call. Salesforce CEO Marc Benioff has been quite vocal and this should work wonders in quieting that noise — at least for now.