Google: The Innovator's Dilemma is real

Jan 27, 2024

Welcome to UncoverAlpha newsletter. The newsletter is primarily focused on deep dives and insights into great companies/subindustries in the technology sector.

This article is sponsored by Stream.

I use their platform regularly, and it is one of my main tools for analyzing a company also from a qualitative standpoint. The insights shared by ex-employees, big customers, and competitors about a company are very valuable in addressing the industry trends and path of where a company is going. You will see many of them in this article as well.

Stream by AlphaSense helps investment research teams access unique, high-quality expert insights faster and more cost-effectively than most traditional expert networks.

With Stream, you’ll have instant access to:

20,000+ on-demand transcripts of interviews between experienced buy-side analysts and company experts provide first-hand perspectives on the companies and sectors that matter to you.
The fastest-growing library in the industry, you're bound to find what you need, when you need it!
Insight from experts with 500+ hours of industry interview experience.

If you want to try the platform out, you can use the link below

Try Stream

If you are interested in sponsoring the next article, you can DM and reach me on Twitter @RihardJarc

Hi everyone,

In this article, I will explain my thoughts on Google. If you follow my newsletter for a while you know that I wrote a piece about it back in September 2022 when the competitve environment was much different. My concerns have recently grown stronger due to the emergence and strong adoption of LLMs. As you will see through this article, I believe Google is facing a real Innovator's Dilemma. I will continue to follow the company closely, but share my thought process here, as some of you might find it helpful.

So why is it a challenging time for Google right now? I will break down my thought process in this article into a few segments:

The need that Search solves.
Are LLMs a disruptive or evolutionary technology for Search?
The "unpenetrated moat" is showing cracks in my view.
Alternatives to Search are on the rise.
What can Google do?
Being the "king of AI" for Google might not look better than what the company is today in the mid-term.

The need that Search solves

When we talk about search, the basic need it fulfills for users is the need to get an accurate answer to their questions. While some might argue that some searches are unrelated to answering a question, I would argue that they all are. Even if somebody is trying to buy a bike, for example, and they write down "buy a bike" in a search, this query intends to find the best possible bike for their needs and desires within their budget.

Are LLMs a "disruptive" or "evolutionary" technology for Search?

A lot of my thinking around technological disruptions and the emergence of new technology and the understanding of that was influenced by the book "Innovator's Dilemma" by Professor Christensen. The book describes well the two types of technological advancements. One is evolutionary or sustaining technology, and the other is disruptive technology. Evolutionary technology is more common and refers to technological advancements that improve the underlying technology base. In contrast, disruptive technology basically brings everyone to ground zero with new value chains, customer needs, and solutions. In evolutionary technology, existing leaders and incumbents are the ones that generally benefit the most from these advancements because of their strong positions in the basic underlying technology that the advancement is addressing.

With disruptive technology, existing market leaders often fail and become losers as new entrants become the new leaders. The main reasons for existing leaders to fail at a new disruptive technology are the following:

Companies depend on customers and investors for their resources. To survive, they must provide customers and investors with profits. The problem with disruptive technology is that it doesn’t initially provide existing customers with value increases. Because of that, the company is not inclined to give resources and focus on developing technology that doesn't serve their customers.
Disruptive technology is often, at first, a small market and doesn't satisfy the growth needs of large companies.
The best employees of companies often do not want to be on teams developing this new technology, as the parent company often views it as a "side project." There is also a high chance that projects often fail with disruptive technology, and a lot of pivoting is necessary to find the right solution and market. In such cases, employees have a conflict of interest with their personal ambitions of visible success, promotion, etc. The best solution for this is for companies to create spin-offs that focus solely on the new disruptive technology and have customers who get value from this technology, despite this spin-off company being small with small results at first. Small "wins" in a big company are often seen with skepticism about whether we should even continue, while with small independent companies, small wins lead to new energy and enthusiasm.

But in general, it's great to remember that the reason why small companies often lead in new disruptive technology is that they are doing something that simply doesn't make sense for established leaders to do.

As a great quote from the book says:

"Disruptive technology rarely makes sense during the years when investing in them is most important."

The book also shows results from the past, where it is clear that the first-mover advantages are significant in disruptive technology. Of established firms that entered the new disruptive technology market, only 3 out of 51 firms (6%) that entered the new market reached a $100M revenue benchmark. In contrast, 37% of the firms that were the first movers in disruptive technology not only surpassed the $100M revenue level but also logged a cumulative total of 62 billion dollars in revenue between 1976 and 1994. After those markets had become established, those that followed into the markets later logged only $3.3 billion in total revenue. So, there is a strong case to be made that if technological advancement is disruptive for an industry, established leaders from the past cycle face a very tough road.

Now, let's try to identify if LLMs are a "disruptive technology" for the search market or not. There are two questions one needs to answer to see if a technology is disruptive:

Does it meet the performance demands in the market?
Is the trajectory of the technology's improvement higher than the growth rate of the performance demands?

To better illustrate what this means here are the performance demands in the automobile market example of EVs. The performance demand of the market, in this case, would be that a car can have a range of at least 120-150 miles before needing to refuel and that it can accelerate from 0-60 miles per hour in less than 10 seconds, with a top speed achieved of at least 80 miles per hour.

In the context of technological advancements, the growth trajectory's ability to meet or exceed performance demands determines its disruptive potential. In our case of search technology, the performance demands can be outlined as follows:

Providing answers to user queries as swiftly as Google Search does today.
Delivering answers with the same level of accuracy as Google Search currently offers.

To establish performance benchmarks, it's prudent to refer to Google Search metrics, given its dominant position with over 90% market share. Now, considering the application of Large Language Models (LLMs), especially advanced models like GPT-4, on the first point, which pertains to the speed of obtaining answers, it can be argued that LLMs already match or potentially surpass traditional Search.

However, when it comes to the accuracy of responses, LLMs still exhibit discrepancies for certain queries, occasionally failing to provide accurate answers. Nonetheless, examining the growth trajectory of technological improvement and comparing it to the growth rate of performance demands in the market, it's reasonable to assume that we are still in the early stages of LLM development. With each new iteration and fine-tuning, the rate of accuracy improvement in LLMs (reducing inaccuracies or hallucinations) appears to be on a much steeper upward curve than the performance demands of the market. There is also a strong case to be made that both the speed of getting an answer and the quality of the answer in the long term will be higher with LLMs than with traditional search.

Therefore, my conclusion would be YES, LLMs qualify as a disruptive technology to traditional Search.

Also, a topic that often comes to mind is the costs associated with training and running LLM models. At first, when GPT3.5 and GPT4 came out, the estimates were that an LLM answer was 100x more expensive than a Google search query. Those costs are coming down fast. This month, Sam Altman, the CEO of OpenAI, addressed this question while being a guest on the Bill Gates podcast. He said the costs of running their earliest version of GPT, GPT3.0, in 3 years came down by a factor of 40x. The costs of running GPT3.5, which has been running for just a little over a year, came down by a factor of 10x. To add to it, he said that the costs of running LLM models are on the steepest slope of cost reductions that he had ever seen, much steeper than Moore's Law.

It is also important to acknowledge that some technology might be disruptive for one segment but not for another. When I revisit my thought process regarding LLM’s distributive nature in the context of cloud computing, my perspective shifts. In cloud computing, LLMs seem more like an evolutionary technology than a disruptive one.

The "unpenetrated moat."

Over the years, Google has built its search business into one with the highest "moat." Currently, 57% of Google's revenues come from Search. The main reason for continuously sustaining that moat is well described in this interview with a Former Microsoft employee who worked on Bing, who answers the question of why Apple has not built out its own search engine:

»I think the short answer to that is if they knew that they could not build a search engine that was going to be better than Google, even if they put in massive investment, and the reason for that, again, gets back to the fact that the search engine that has the most people searching is going to have the highest signal of information about what people care about, and therefore, will have the best capability of delivering the best answer….

…Initially, it woudn't be as good , and then they'd have to be able to explain to their users why…«

source: Alphasense

The key in Search is having the most people searching via your product, giving you the highest signal of information about what people care about and, consequently, the best capability to deliver the best answers. This is also one of the main reasons Apple, even though they probably have that conversation internally often, is unwilling to compete and release their own search engine. Simply put, users won't use your new search engine if the results are not good and won't continuously use your product for six months or more until you gather enough data and scale for your product to deliver good results. So, the smart thing for Apple to do is take billions from their default Safari Search deal with Google instead of competing with them. This dynamic created an essential flywheel for Google, as in reality, nobody could compete in getting you the best answers like Google could, as nobody could gather enough data to make it work. But then we had the ChatGPT moment and, later on, a wave of other LLMs.

Now we can debate whether these LLMs, when trained, actually complied with copyright laws or if they just bluntly scraped data from the internet with a problematic legal basis. However, for this article, the important thing is that the genie is out of the bottle. Today, we have numerous LLM models for which most of them can answer questions for many use-cases as well as or even better than Google Search. To make matters worse, some of the models are open source, and closed-end models like GPT from OpenAI have now shown the first signs of giving access to an App Store-like model for other businesses. Today, the costs and effort to build a competitor to "search" using these open-source or closed-end LLMs are not nearly as high as before this happened. Google's unpenetrated Search moat is showing cracks. Google is the biggest animal in the ocean, and for the first time in a long time, it is dripping blood in the water, and the whole ocean can smell it.

Alternatives to Search are on the rise.

As we already discussed, the problem that LLMs have caused is that the basic need to answer a question has become accessible to almost everyone. And it's not just the super LLMs developed by OpenAI, Anthropic, or companies specialized in developing AI; it's also the smaller companies that will launch their chatbots powered by either open-source LLMs or LLMs that use GPT or some other closed-end model. The problem for Google is that now a company like Salesforce can train a model with their proprietary data and answer questions from their clients better than Google's Search can.

On top of that, it's very convenient to have an AI agent that answers questions at the touchpoint where the question arises. It's not just Salesforce; it's companies like Meta, inside WhatsApp, Instagram & Messenger. It's Microsoft inside Office or Dynamics, and at some point, it will also be Apple, with Siri finally becoming a useful tool. It's also a company like Airbnb for travel questions. I think in the next year or so, we will see almost every company with distribution and specific user data launch their own version to help their customers solve their problems and questions. If users can have their questions answered conveniently at the source of where the question has arisen (be it some platform, ERP system, CRM system, or any other place), the need for them to take a few extra steps in opening the browser and entering a search query in Google becomes less relevant.

It is true that a lot of the queries on Google Search are not monetized (according to the latest estimates, around 57% of queries are non-monetized by ads on Google). The reason is that many queries have just a straightforward answer, so it doesn't make sense for ads to be placed there. The problem for Google is still in the habit formation of people. Suppose people get their questions answered in a quality way on a different platform than Search. In that case, they are also more inclined to start asking other questions on that platform that are "monetizable" on Google.

To make matters worse, before the emergence of LLMs, Google already had problems with competitors going after some of their monetizable queries, like e-commerce queries. Amazon is one of the big reasons as more and more people start their online purchase path on Amazon instead of going to Google Search.

Then there is also the problem with the younger demographic who like to start their online purchase experience on social media platforms like TikTok or Instagram. Google acknowledged this. Recent data showed that many Gen-Z’s are starting to use TikTok and Instagram’s Discovery engine for search. At a technology conference in July 2022, Google senior vice president Prabhakar Raghavan acknowledged that:

“In our studies, something like almost 40 percent of young people, when they’re looking for a place for lunch, they don’t go to Google Maps or Search. They go to TikTok or Instagram,”

This visual that I made describes the significant change in solving the essential problem of answering a question. Under the »LLM chatbot & LLM powered search« bracket, I just named a small number of companies as many more will emerge and use their data with an accessible LLM:

The recent big push into social commerce by TikTok with its TikTok Shops is not helping Google either.

So right now, Search is being "attacked" by LLMs, social media, and e-commerce platforms. The worst part of everything is that we will only see even more alternatives in the coming years, not fewer, as some companies still need to release their chatbot products.

What can Google do?

For Google, I see only two possible paths when it comes to Search:

They stick with the existing traditional Search and only use LLMs to enhance traditional search.
Make your own pure LLM type of product, fully compete with traditional Search, and cannibalize it over the years.

The first option is something that Google has done in a way so far. It incorporated some LLM functionalities into traditional Search. The problem here is that history tells us that this strategy often fails. This strategy works well if the technology is evolutionary and not disruptive, as we have already analyzed in the first part of the article. The problem with this approach in disruptive technology is that, in reality, you are trying to "save" and make your legacy product relevant. By doing so, you don't amplify enough the new advantages and usefulness of the new technology. In some sense, you are trying to "save" your existing business. On the other hand, competitors will not be burdened with making "traditional search" relevant and have already created a pure LLM type of search experience (for example, Perplexity). In this case, the user interface is just more natural, and it can take full advantage of the new technology.

If choosing this path, Google, in my view, starts going towards the ways of the IBMs of the world who didn't die out but missed major technology moves without positioning themselves in them.

The second option is the one that makes the most long-term sense and is how established companies, based on history, survived and came out of technology disruptions as continuous leaders. In this scenario, the best thing for Google would be to make a spin-off company with separate employees, goals, and clients and let it compete with the parent company on Search. While this might seem counterintuitive, it's how many established companies in the past have carried their dominance into the next technology disruption.

There are several reasons why a spin-off is important:

• Companies are driven by the needs of clients and investors, and in Google Search's case, the clients (the advertisers) do not want Google Search to move to an LLM type of search as there is much less room for ads in that product. A separate unit would have new clients, and the goals of those clients and the spin-off company would be aligned.

• Small wins would be met with enthusiasm in the spin-off company. At the same time, in a big company, it would raise questions about whether it even makes sense for the unit to continue operating because the results are »too small« in the parent company's income statement.

• Getting the best talent. It is hard to attract talent, as nobody wants to work for a "side project" with a high chance of failure inside a big company, but a spin-off company being funded by a big company is perceived as positive.

The interesting thing is that Google already had a spin-off company called DeepMind. However, they recently decided to merge them with their Google Brain unit inside the parent company. While I agree that merging both AI units into one makes sense, I think the better solution would be to merge them into the separate unit and not into the parent company.

The real question for Google is, will they allow their new LLM product to compete with their existing Search business? It is a hard decision but a necessary one. In the short term, this will result in a hit to revenues and profits, but in the long run, it lifts the chances of being a successful leader in the LLM search experience.

My conclusion is that Google is facing the typical Innovator's Dilemma when it comes to its Search business.

Being the "king of AI" for Google might not look better than what the company is today.

Google has all the necessary ingredients also to be the leader in LLMs:

Access to compute and infrastructure
AI talent
Number 1 or 2 in the world when it comes to the amount of consumer data
Deep pockets
Distribution points (YouTube, Android, Pixel, Search, Browser, etc.)

The chances of them being the LLM leader in the next few years are very high. However, the problem for me as an investor is that the road to getting there might be bumpy, as they will have to cannibalize their current Search business, which brings in 57% of all revenues. On top of it, the monetization options on LLMs when it comes to ads are much smaller than on a surface like Search today. I also believe expecting them to have such a monopolistic position as they do in Search right now, with over 90% market share of Search, is unrealistic as the "moat" becomes smaller and distribution becomes a bigger factor.

Let's face it right now: Google is not the leader in LLMs.

And while the race is still early, Google has faced albeit more minor battles on the »AI playground« in the past, and not all of them were won:

All of these factors might mean that even though Google becomes the "king of AI," that might not look better than what the company is today, at least not in the mid-term. Now, the road to search disruption might take some time, but the process has begun; Google will face this dilemma head-on, and things might become rocky for investors.

Summary

The reason for me selling my Google position is that, despite having great assets like YouTube, Android, Maps, and GCP, Google still relies on Search for 57% of its revenue. Interestingly, when breaking down Search, Google directed nearly 12% of all search clicks back to Alphabet-owned properties (data from 2019). If search starts to lose relevance, this will also hurt other parts of their ecosystem.

I believe that LLMs are a disruptive technology for Search, which presents a few challenges for Google.

First, even if they become leaders in LLMs in the long term, they cannot maintain the same monopolistic position they have in Search.

Second, the LLM experience is not ideal for displaying ads. The primary goal is to provide fast and accurate answers without users going through multiple links. While I believe that ads will eventually be incorporated into these products, they will not reach the same level of ad saturation that Search currently offers, at least not in the short to medium term.

Lastly, the transition of Search from traditional to LLM type will be painful for Google. It will cause Google to cannibalize its existing business for the new one to grow.

Therefore, I believe Google will face significant challenges in the coming years. At this point, I prefer to stay on the sidelines and observe how things unfold. I want to gain more clarity, as these are significant changes to the conglomerate.

At the end of the day, when it comes to Google, two years ago, we wouldn't even be having this conversation as the »moat« was perceived as the biggest one in the world. That fact alone tells you a lot about today's search environment.

Until next time, take care.

As always, If you liked the article and found it informative, I would really appreciate sharing it.

Disclaimer:

I own Amazon (AMZN), Meta (META), Microsoft (MSFT) and Airbnb (ABNB) stock.

Nothing contained in this website and newsletter should be understood as investment or financial advice. All investment strategies and investments involve the risk of loss. Past performance does not guarantee future results. Everything written and expressed in this newsletter is only the writer's opinion and should not be considered investment advice. Before investing in anything, know your risk profile and if needed, consult a professional. Nothing on this site should ever be considered advice, research, or an invitation to buy or sell any securities.

Google: The Innovator's Dilemma is real

Discussion about this post