Alphabet reiterated its $75B CapEx guidance for the year. It continued to guide to accelerating depreciation expenses due to all of this CapEx growth. EPS is expected to grow by 16% this year and by 9% next year. EBIT is expected to compound at a 13% clip for the next two years.

Full-Stack AI:

The main theme of Sundar’s prepared remarks was again the firm’s full-stack AI approach. It’s this vertical integration that it views as a key differentiator vs. the field. I think they’re right. Nobody else builds and designs their own custom data centers and semiconductors, while also featuring a world-class research team, the top-ranked foundational model in Gemini 2.5 Pro and 7 products with distribution to over 2 billion people. Nobody enjoys all of that and the coinciding multi-modal data byproduct to more powerfully season models. In turn, that continues to spin the product flywheel.

The three layers of the opportunity are infrastructure, research/models, and apps/products. We’ll take these in order, as there were updates on each. Starting with infrastructure. Alphabet teased its 7th generation Tensor Processing Unit (TPU) called Ironwood. It boasts 10x compute power boosts and nearly 100% efficiency gains vs. the 6th generation. This builds on its value proposition of offering cloud workloads that run on a wide array of GPUs, including Nvidia’s. Like Meta and Amazon, Alphabet building its own chipsets in no way means it will stop working with Nvidia and offering its world-class Blackwell GPUs. The Search Giant was among the first to grant access to these and that should be the same for Nvidia’s next-gen Vera-Rubin platform too.

Fantastic AI infrastructure not only unlocks bountiful workload and new product growth opportunities for Google Cloud, but ensures bottleneck-free research and foundational model work. Its talent has all of the resources they need to work. The reception to Gemini 2.5 Pro has been “extremely positive,” with several 3rd-party research firms currently ranking it as the best foundational model on the planet – including securing top-ranked chatbot status “by a significant margin.” Gemini API traffic is up 200% year-to-date, showing how eager developers are to use this for their source code automation and other tasks. AI Studio active users are also up 200% year-to-date.  AI Studio is its environment for low-stakes testing and tinkering with its roster of 1st and 3rd-party models.

Finally, for distribution of all this AI work to its apps. Google has 15 products with over 500 million users now actively using Gemini. Android and Pixel can access its multi-modal capabilities via camera, voice or screenshot and the Google Assistant product will soon get a large Gemini-inspired upgrade (first on mobile and then across other devices). A lot of this product work so far can be seen in its search and advertising businesses. We’ll get more into the specifics of each of those areas right now.

Search:

AI continues to expand the types and depth of queries that consumers can tap into, which is supporting continued strong engagement growth and double-digit Y/Y revenue expansion. In the most recent news of the week issue, we talked about a viral chart showing ChatGPT growing users well in excess of Gemini. While that’s not ideal, using that as evidence to be bearish is misguided. The vast majority of AI search activity continues to happen within AI overviews, which specifically has 1.5 billion monthly active users (MAUs). Overview usage continues to nicely grow as it adds more markets, languages and query types, while customer satisfaction scores are in excess of legacy search. And finally, AI Overview monetization continues to also trend positively, further eroding the previously perceived risk of these ad placements not being lucrative.

Its new AI Mode “expands what AI overviews can do” with more agentic, multi-modal capabilities (thanks to the family of Gemini 2 models). AI model queries thus far are yielding “really positive feedback” from users.

From a multi-modal querying perspective, circle-to-search is available for 250M devices as of this quarter and growth reached 40% Y/Y. Lens search (use your camera to search) is up to 5 billion queries per month. That’s tiny compared to its core search business (1.2% of total annual queries), but is scaling very nicely. And encouragingly, Lens continues to be highly incremental to overall search volume.

Advertising:

Infusion of Gemini into its ad placements, targeting, reporting & campaign building is going well. The batch of advertising upgrades since 2024 has already netted a 26% boost to Y/Y conversion rates. Investors (including myself) often worry about GenAI investments yielding the kind of value creation required to effectively monetize. This is explicit evidence of that value creation happening exactly as needed. Whether it’s new audience asset recommendations to nudge best practices or easier content generation for campaigns, things are quickly improving. These products all boast tight integrations with Demand Gen (finds larger relevant audiences for brands) and Performance Max (PMax; its end-to-end automated campaign-builder across all Google impressions). 

Chinese Seller Ad Demand:

Removal of tariff exemptions for cheap Chinese goods will have an impact on 2025 advertising revenue growth. That impact was called “slight,” which I took very positively. As we previously worked through, Chinese ad demand was around 10% of revenue for Google in 2024. It seems as though a lot of these placements are getting filled by other merchants rather than remaining empty. Pricing is probably being negatively impacted, but Alphabet is showing they can recover much of that headwind. Fantastic news. 

More on YouTube:

Shorts views rose 20% Y/Y, as it continued to close the monetization gap between this content form and the rest of YouTube. Progress is encouragingly fastest in the USA, where the revenue per user ceiling is highest.

More on Cloud:

Alphabet talked about a few newer products for the cloud business. “Agent Designer” is a user-friendly, low-code template for building AI agents and agentic workflows, while its new Agent Development Kit (ADK) offers open-source tools for building AI agents. Finally, Google Agent Space is a managed product for helping companies identify needed data and instruct AI agents to use that data for a specific task.

Cloud remains capacity constrained, which is why the company will continue to lean heavily into CapEx to build its footprint. Just like previous quarters, most of the CapEx this Q was for short-lived assets like servers that are directly connected to near-term revenue opportunities. That will be the case throughout 2025; because of the uncertain timing of capacity coming online, growth rates will be a bit volatile from quarter to quarter.

Waymo:

Waymo grew paid rides by 5x Y/Y to 250,000 per week. While people criticize it for its geofenced, LIDAR-based approach, I actually think it’s preferred. It’s hard for a vision-only approach to see if conditions aren’t perfect, and LIDAR is why Waymo already has a rapidly growing business across many cities. And on Waymo being limited to major urban areas, that’s absolutely right. But? Those areas represent most of the demand… so that’s not a big negative in my mind. And to make it less of a negative, Waymo is now mapping highways and airport routes as we speak. Furthermore, the argument that Waymo can’t scale due to sensor costs is just misguided in my mind. Yes, this version of the hardware is too expensive for global scale. But? This is version 1. Version 2 will soon come out with 50% cost efficiency gains and things will keep getting cheaper thereafter… especially with Nvidia’s world-class synthetic data generation technology that vastly lowers the cost of collecting physical data and sharpening models.

Waymo continues to experiment with several business models and that will remain the case. It called early results and user satisfaction “very pleasing,” which was good to hear. And in Miami, where it’s partnered with other players like Moove for fleet management, Uber has a large equity stake and board seats.

Network:

The network business, which includes its advertising marketplaces, continues to be deprioritized and shrink. It declined Y/Y again, which I think is a good thing. This is the source of the current FTC monopoly lawsuit for the mega-cap’s ad tech business, as the network is where Alphabet double dips in terms of servicing the brands and the publishers, while perhaps unfairly routing some impressions to its own inventory. There’s going to be a ton of noise this year for this business, but it’s already becoming an irrelevant piece of the growth engine. That should mean more durable and less risky growth, and lower traffic acquisition costs.

Solid quarter for Alphabet. All major revenue buckets were ahead and profit beats were even larger. But perhaps more importantly, the headwind from Chinese ad demand seems to be much smaller than feared as we move into Q2 and tariffs take hold.

The company continues to lead in several pieces of GenAI innovation and continues to get very little credit for it. I view this as the highest-quality out-of-favor business model in markets and view a 14x EBIT multiple as a no-brainer. I think its continued execution will take care of sentiment and profit compounding will win out like it always does. Some see Alphabet owning most of search instead of all of it in the future as a bear case and a red flag. I just continue to think AI is vastly expanding the search pie, and owning most of a bigger piece works just fine for me. Oh… and it helps that this firm is a leader in streaming, autonomous driving, cloud computing, model innovation and quantum computing.