There's a popular meme on Twitter that goes like this:
"I'm starting to be a real believer in AI. Will be a megatrend like no one is really imagining right now. I'm sure lots of ramps and drops along the way though"
Like any good meme, there is indeed an ounce of truth to it. Maybe not an ounce; I actually think there is likely to be a lot of truth to it. However, as former Bernstein Internet Analyst Corry Wang (who is incidentally currently at Google AI Strategy team) put it : "There is a misguided obsession in many tech circles around predicting the future of technology. In contrast, I would posit that predicting the future is actually pretty easy - the hard part is making any money on it."
Indeed, a couple of years ago, it was somewhat easy to foresee that the narrative around the future of search would likely continue to sour. It did, but the details have been somewhat more difficult to be right about. For example, since the rise of GenAI and chat bots, one common concern has been a potential double whammy for incumbents such as Google i.e. GenAI queries are harder to monetize and at the same time, the cost to serve such queries would be materially higher than traditional search queries. I doubt anyone was expecting Google’s operating margin would go higher from 30.6% in 2021 to 32.1% in 2024. I myself thought Google Services operating margin in 2021 might be "peak margin" for Google as they would have to navigate much higher cost to serve queries and lower monetization phenomenon (admittedly, part of this margin expansion was due to extending depreciation schedule, but even if you adjust that, 2021 and 2024 Google Services operating margin would be somewhat similar). What actually happened was Google reduced cost to serve these GenAI queries by 90% in 18 months through " hardware, engineering and technical breakthroughs" (3Q'24 call). Moreover, Google has stumbled onto new query markets e.g. in 4Q'24 call, they mentioned how lens is used for 20 billion visual search queries, majority of which are incremental. How about monetization of these GenAI queries? It's less reassuring so far, but in 1Q'24 call Google management mentioned "based on our testing so far, I'm comfortable and confident that we'll be able to manage the monetization transition here well as well."
This isn't a Deep (or even shallow) Dive on Google, so I'm not going to focus much on numbers or try to address/discuss all the Google related concerns (read this piece if you're more interested in numbers; I may also do an update later in the year).
Look, when I sold Google in December 2023, I did mention I expect search to go through a significant transformation in the next 10-15 years. It's always dangerous to change your mind on something in two years that you expect to play out over much longer period. I still have plenty of sympathies for search related concerns. And even from personal perspective, while it is true that Google did lose my query share mostly to OpenAI, Google hardly lost any of my monetizable query share. However, if you want to disrupt Google, you always were going to disrupt the non-monetizable queries first and then work your way to monetizable queries later. So, I am certainly not claiming Google to be out of the woods yet. My appreciation of the risk is certainly why I made Google only a 3% position so far with an average cost of $150. I will pay close attention to price and broader developments in AI before sizing it up more.
However, I do want to mention a couple of things to give you a sense what prompted to change my mind.
Let me be very frank. When I sold Google in 2023, I hardly knew much about semiconductors and I knew even less about TPUs. As I gained a bit more understanding over the last 15 months, I started appreciating there are certainly scenarios in which Google's deep infrastructure advantage and breadth of existing relationship with users can mask almost all of their weaknesses. I have been following big tech since 2018-19 fairly closely and it never became easier to predict how the narratives will shift just in 2-3 years down the line. Just imagine trying to predict the narrative around Meta today back in 2022 (or "Apple 2021 narrative" back in 2018). I certainly do not rule out the possibility that search can chug along just fine, and thanks to Google's almost end to end control over their infrastructure (unfortunately, they still do have dependency on TSMC for fabricating TPUs), Google's narrative can change materially especially in the post-regulatory clarity era.
I am not super confident in predicting how the big debates will be settled in AI, but it does seem there is a distinct scenario in which Google can potentially be materially ahead of everyone else in 5 years. Outlining the exact path is hard to do, but I have come to the view that while the search related debates consume all the airtime, most investors are potentially missing or underappreciating a scenario in which Google may just topple everyone else by leveraging their infrastructure advantage.
But isn't Google pretty bad at productizing their AI advancements? They have already deployed quite impressive models and capabilities, but if they're bad at productizing it, will this infrastructure advantage matter at all?
While I was pondering about the point about infrastructure advantage for almost a year now, the recent Sharp Tech podcast episode really drove the point home and also reminded me why Google may be able to succeed despite its weakness in productizing model capabilities. A listener of the pod sent the following musing for Ben and Andrew:
"When exactly was Google good at building new products? The answer is never. Google has always sucked at building new products. Consider the epic failures of Google Buzz, Wave and Plus or G chat, and Google Hangouts.
Google was only ever successful at innovating and building one of the following three things:
•Google search
•A critical infrastructure required and custom tailored for Google search.
•Products where search-like characteristics, huge scale, and a data flywheel turned out to be critical for success.
Google pioneered amazing concepts such as map reduce, batch processing, zero trust security, containers, software defined networking to name a few. However, it sucked when it came to wrapping them as general purpose infrastructure products. This goes back to the famous 2011 Stevey's platform rant.
Stevey's platform rant predicted so well why Google Cloud Provider was destined to lag behind AWS. This is also why despite employing a phenomenal collection of talent with security researchers and engineers, Google could never have built a product like Wiz for external use. All of this is to say, I don't think there's anything new going on. Maybe this is just Google being Google, the awkward nerd that gains his advantage by staying up in the data center all night figuring out how to stack 20 times more servers compared to Inktomi who occupied the cage next door. Of course, OpenAI has the better polished product. That was never Google's advantage. And if that's what what it takes to win an AI, then well, Google is just not going to win. Not now and not ever in its history.
The only hope is if it turns out AI models aren't a commodity. Search engines were considered a commodity back in the Inktomi, Alta Vista days very much like LLMs today. Google's only hope is if just like in the early 2000s, it turns out the consensus is wrong and there is a sustainable long-term advantage to be gained by better infrastructure engineering."
But is this a good time to buy Google when majority of the revenue is largely dependent on advertising revenue given the recently introduced tariffs may even cause a global recession?
Tariffs are certainly a risk, but let me offer some brief thoughts on them. Let me provide a historical analogy that I think can have some resemblance to today. Imagine we are in late 2001 to early 2002. Tech bubble just crashed and people just experienced 9/11. I bet 9/11 consumed almost all of our attention (and rightly so), and perhaps more and more people started laughing at people who thought internet would revolutionize everything. In my mind, these tariffs (if they remain unchanged) are like 9/11 i.e. incredibly impactful for the world and will certainly affect us in ways we may not be realizing today, just as 9/11 did. But from purely long-term market perspective, 9/11 wasn't the main story that mattered; what mattered was the internet.
Today's "internet" is "Generative AI". In a decade (likely lot sooner), it is much more likely than not that Gen AI will be dominant driver of the market, and while tariffs may be impactful in the meantime, they are mostly going to be adapted and absorbed by different stakeholders in the value chain.
Google, Meta, Amazon have all been down 30% from their highs. I have started deploying my capital to all of them today (3% Google at $150, and adding more to my existing Amazon and Meta positions: 1% Amazon at $168, 1% Meta at $500). I am keeping an eye on Microsoft as well, but haven't pulled the trigger yet. I expect myself to buy more if stocks keep going down. There is a very good probability that people may "forget" about AI amidst the tariff tantrum, but I am of the opinion that might lead to a compelling opportunity to steer my portfolio to the future of the world even more.
Great update, thank you MBI