Thesis: Google (GOOGL, GOOG)
An inevitable AI winner
This post is a legacy article converted to an evergreen Research article. Mentions of “Weekly Brief” or “Friday Filter” or specific time mentions (“this week”), are from the old format of the original post, now discontinued. This is not a stock report, but an overview. All my full stock reports are paywalled.
See here for an overview of this Substack’s content types:
In the meantime, onto the Google thesis.
Google I/O ‘25 strengthened my Google thesis
Try not to brag. Try not to brag.
Well, I hate to say I told you so… just kidding (kind of).
Hopefully you don’t start thinking I’m some arrogant, short-term-minded investor. That would been crazy, eh?
In all seriousness, this week was an amazing week for Google. The stock rose 5% at its peak, and sentiment has changed dramatically.
The cause? Google I/O 2025.
What was learned:
Google as a whole is not being heavily disrupted by chatbots and AI search engines like Perplexity,
Google Search is not disintegrating, and
Google is not a victim of The Innovators Dilemma1.
Before I get into the announcements at I/O and how they made me more bullish on Google, I’d like to give you a rough breakdown of my Google AI thesis and how important conviction has been in holding Google up to this point. The fog has cleared slightly, but many, many tailwinds remain ahead.
Once upon a Google thesis
Back in October 2024, I wrote the following about Google’s AI strategy:
“Most of Google’s AI spending is focused on building out its data centre capabilities to improve its cloud computing offerings (similar to what the other cloud giants, AWS and Azure, are doing). This is the main focus for Google at the moment.
Google Cloud is the fastest-growing revenue segment of Google by far, and the cloud computing market growth is on an exponential uptrend along with it.
Secondarily, with the remainder of the spend, Google is investing heavily in developing AI models and LLMs. Google’s eventual plan in this area is to incorporate these models into Pixel hardware, Google Workspace, and the many other products, apps, and services Google offers, to enhance capabilities and improve the value proposition. Unlike OpenAI, which is the current undisputed leader in the AI model and LLM space, Google has much more user data to train these models and has been developing AI for over a decade.
OpenAI is falling short, or will eventually fall short, of data to train its models. What this means in the end for OpenAI, is high licensing costs to get hands on more data. Google does not have this problem and, quite frankly, probably never will.
Artifical intelligence, if put in a single word, would be data—and Google has a lot of it. Remember those many popular products and services that Google owns and the 2 trillion(+) annual Search queries Google processes every year? With those alone, Google has an almost endless source of data in its current position.”
From October 2024 to April 2025, there was lots of talk and very little action from Google on this strategy. “We’re spending money, we’re spending money. We’re a leader in AI,” said management in earnings calls. Well, duh.
I didn’t doubt for a second that Google wasn’t spending money—they were spending tens of billions every quarter. But seeing no real push-forward announcements to back up my thesis? It’s easy to say: holding Google made me sweat. Because a month after I wrote that blurb above, Perplexity launched its advertising platform, and ChatGPT announced their Search feature.
By this point, the stock had barely moved. All of this was on top of ongoing litigation with the U.S. Department of Justice, where it faced potential breakup threats and would eventually be ruled an illegal monopoly for its ads business, and with Search generally.
Google’s stock price would essentially end up stagnating during this time.
“Should I sell? Should sell Google and buy Amazon, MercadoLibre or Shopify instead of holding? Opportunity cost!”
These were all thoughts that circled around my head. Honest.
I’m not going to lie by saying I knew everything was going to be fine for certain, because I didn’t. I simply had a thesis (as shown above), and I was waiting for my thesis to progress. Remember: long-term holding.
Google wasn’t dying. Its financials weren’t degrading. (Still aren’t.)
Google was still generating positive returns on capital during the time I was holding—even with the wide list of concerns facing its business(es)—so why would I be selling? This was all sentiment, and sentiment isn’t a thesis.
But speaking of sentiment, the cherry on top for negativity came at the beginning of May 2025, when Apple swooped in and told the world Search queries were down for the first time ever on Safari, and that they would be looking for new AI browser options. Google released a statement the next day (below), but before this, the stock dropped 10%, and I added to my position.
Here was Google’s public statement:
“We continue to see overall query growth in Search. That includes an increase in total queries coming from Apple’s devices and platforms. More generally, as we enhance Search with new features, people are seeing that Google Search is more useful for more of their queries — and they’re accessing it for new things and in new ways, whether from browsers or the Google app, using their voice or Google Lens. We’re excited to continue this innovation and look forward to sharing more at Google 1/0.”
Here was my thesis on the Safari concern (why I added) that same week:
“Overblown concern. Even if this is an option for Safari, most consumers will just choose Google anyway assuming Google continues with transforming Search into an LLM-AI engine.
[…] I want to come back to the risk section of Google’s recent annual report (which I skimmed through during the conversation in the server today). Like all companies, Google is legally required to report its business risks.
And last year, at the top of its list was competition “from large companies and startups,” with a big, fat, huge, bolded text disclaimer above saying “these risks are only apparent if we do not innovate and acknowledge our current changing market position, and we are doing both.” This is hugely paraphrased, but that’s the gist. If you listened to the call two weeks ago [Q1 2025], and you saw the report, and you pieced everything together from Q to Q, you’ll know that management is staying healthy. They’re acknowledging these pest problems and they’re loading up on pesticides. That big fat disclaimer was put there for a reason. Google is experienced in dealing with changing market conditions. Adaptability is high for the company. Much like Amazon. I can only imagine Google can win, if they try.
They’re fighting the lawsuits, they’re innovating and changing Search (for the better) to destroy competition, and they’re doing all of this while investing heavily in growing a profitable distinct cloud business that’s set to benefit greatly from secular trends in the distant future. Plus, many small and medium sized businesses depend on Google Ads. There is no viable alternative to this if they were to cease to exist today. Not by LLMs, not by Facebook. And I think because of that alone it makes a good argument for further Google search dominance (until this changes).
[…] I didn’t find out about this statement until the morning after, but once I read it, I thought a few things:
If Apple—the company that receives $20 billion per year to keep Google as its default search engine in Safari—is saying Safari search traffic is down for the first time ever during a court hearing on Google’s antitrust/monopoly case… that’s a huge help to their case. It has to be strategic testimony, right? Apple wants to keep its $20 billion.
If Safari search traffic is down according to Apple, but total Google Search queries are rising even on Apple devices and platforms, all that means is people are moving away from Safari, not Google—from Safari to Chrome, or to the Google app.
Both of these scenarios are bullish signs for Google, even during such a challenging time in its history. And so, Google Search isn’t dead. This Safari concern is overblown. (What isn’t overblown is Search’s AI competition. However, I do believe AI is something they can adapt to over time, and we’re seeing this happen now [with AI Overviews].”
Concern after concern, even though I was never fully sure of what was going to happen, I stuck to my thesis. The thesis that Google had vast amounts of data at its disposal to feed AI models, vast amounts of compute and money to spend on training them, and billions of people for distribution when they’re finished.
It’s a simple thesis to make too. Even now, while many competitors need shovels, Google already had a pile of gold.
All Google needed to do was execute. And they did.
The company continued to innovate with Search via AI Overviews, and has continued to see user growth of Gemini—one of the best AI models on the market now at 400 million monthly active users (Google CEO Sundar Pichai mentioned it was the best on the Q1 2025 earnings call).
This, all without mentioning Google’s fast-growing Cloud business, piggybacking off the growth of AI innovation and the secular trend of cloud adoption, where it’s facing huge demand.
And so, throughout doubt—but as a consequence of conviction—I held. I even went as far as adding to my position amid this doubt.
This is the value of having a thesis. Research matters so much. It takes time, yes, but without it, you’re doomed. This thesis was rock solid all the way from October, when so much changed since then. That’s research. I mean, this is why I publish quality research on this newsletter: (1) I actually quite enjoy it, and (2) why would I not want to provide you with a rock-solid thesis to reap 12–15% CAGR growth? That’s incredibly valuable.
But how’s the thesis playing out? Positive or negative leaning?
I’ll let you decide…
What was announced at Google I/O?
Like I said at the start, I’m not a news aggregator. I’d more than recommend watching Google I/O 2025 on your own time—or at the very least, watching Google’s 10-minute summary of the event—to judge how well this thesis has turned out. This post won’t be a summarization for you.
Words can only do so much when it comes to describing technology like this. You need to see it with your eyes.
As for my thoughts? Here are a few highlights from the event that strengthened my thesis as described above:
Gemini 2.5 Pro is now ranked #1 across all LMArena2 categories.
Part of the original thesis that Google’s “vast compute power and data access” would produce a leading model with proper execution.
Google announced screen sharing and camera functionality for free in the Gemini app (features ChatGPT has paywalled), and enabled its new image-generation model, Imagen 4, in the app as well.
This Redditor agrees:
Google Search AI Mode
The most influential announcement from this event that put aside “Innovator’s Dilemma” arguments:
Now, Google’s most powerful AI models will be integrated directly into core Search as an option called “AI Mode;” a Gemini-powered, ChatGPT-like update built into Search that answers complex questions with detailed results like any LLM would. Graphs and charts will be generated in the answers for financial and sports questions, coming mid 2025.
(AI Mode is an option consumers will need to switch to if they want it. Normal Search queries will be shown AI Overviews. Consumers can continue to type in longer and more complex queries into normal Search with AI Overviews. I can only assume Google will integrate the experience over time.)
This mode also integrates Google Shopping and Google Pay, which adds seamlessness to the checkout experience. You can also now browse AI Mode with generated mosaics of the items you wish to buy, ask it prompts until it helps find what you’re looking for, and try on clothes virtually to see how they look on you—all within Search. (This strengthens the thesis on Google’s distribution and integration.)
Veo 3 (Google’s newest video generation model) now has native audio generation.
One of the more fascinating announcements that strengthened the thesis on Google’s incredible amount of data.
Google now has a model that can turn your text into a video—with sound. No output containing fifteen fingers or unnatural lip movement either; it’s scarily realistic. Here and here are a few examples of this model.
It’s only because of YouTube that Google could train this video model as effectively. Yeah yeah, sure, OpenAI and the other competition can unethically download millions of YouTube videos and train off them too, but data integration matters. Massive compute power matters. And Google wins in both of these areas.
This isn’t even the tip of the iceberg. I haven’t even begun to mention the new Google Glasses, Android XR, or millions of other things like Project Astra that have possibly killed thousands of startups.
Again, I would highly recommend you watch the event. I have that, and the 10-minute summary of it, linked above.
Overall, my Google thesis is intact. And after this event, it has strengthened.
Things can change, of course—we haven’t seen OpenAI’s future announcements, for example—so I’m not going to pat myself on the back just yet. I wouldn’t recommend you do so either. This is a rapidly evolving space, and we continue to have the DOJ probes, monopoly rulings, breakup concerns, European privacy law attacks, etc. The overall sentiment is probably still negative.
So am I going to be complacent herein with my thesis because it’s been intact so far? No. But is it gratifying to have a thesis play out correctly nonetheless, after the hard work of research?
Absolutely.
Google is currently my third-largest portfolio position.
—-
Thank you for reading. Happy investing.
Great book. Would recommend.
“LMArena is a competitive benchmarking platform where large language models (LLMs)—like ChatGPT, Claude, Gemini, etc.—compete head-to-head in a kind of “arena.” Users act as the judges by submitting prompts, comparing anonymized model responses, and voting on which one is better. Over time, this builds a leaderboard.” (LLM-generated definition) [LMArena link]
The DOJ decision resolved has made a big impact on sentiment. Even with the remedy of sharing data, who can compete? Does sharing the search index make Bing better? No. The user habit is the moat. Great work 👍
Well done Jacob. Google seems to be on the right track