To Buy or Not to Buy AI?

To Buy or Not to Buy AI?

Important Disclosure: None of the references to specific publicly traded companies in the text below should be construed as a recommendation to buy, sell, or hold.  The references are purely intended to convey my understanding of where these companies fit into the AI marketspace.

By John H. Robinson, Financial Planner (February 2025)

As I noted in a separate post, the most common question I have been fielding lately is “How should I invest in the Trump era?”  “Should we be buying AI?” is the second most common question. 

With respect to leaning into AI stocks with your investment dollars, my first response is something along the lines of, “You know this isn’t exactly a sleeper sector, right?” As a thought exercise, I wonder how many times per day, I hear, see, or read a reference to AI.  I have not actually devoted a day of my life to counting, but I would not be at all surprised if the figure is in the hundreds. We are now at the hype-cycle stage where your barber or hairstylist is telling you about a customer who has a cousin who works for a company that is going to be the next OpenAI or Nvidia. 

I also cannot help but wonder how many people who ask that question truly understand what they are asking to invest in. For the record, I am not pretending to know either, and I submit that spotting investment opportunities in AI requires technical expertise and experience that neither your barber nor you nor I possess.  One kernel of knowledge I have picked up from listening to the Acquired podcast is that stock picks that turn out to be the next big thing are not easy to spot in the early stages even if one has requisite technical skills to be “in-the-know.” Still, for those readers who are eager to head down the rabbit hole, here is an AI primer that took me all of five seconds to find (using AI-powered search) to include in this article to give the appearance that I am merely being modest in downplaying my AI acumen.

The Basic Components and Branches of AI

Flexing My Thimble Full of Knowledge

The following text is pretty close to the full extent of my AI knowledge.

AI refers to the general ability of computers to emulate and/or materially extend human thought capabilities.  Machine learning, deep learning, neural networks, and natural language processing are all considered pathways or tools for creating this ability.

None of the tools mentioned above are new.  In fact, if you listen to the Acquired Podcast on Renaissance Technologies, you will learn how machine learning was applied in the 1970s and 80s to enormous amounts of market trading data to detect trends and patterns that would be undetectable by even the most brilliant human mind.  By applying their proprietary large language learning model to this data set, Renaissance Technologies created the most successful, most profitable algorithmic trading program the world has ever seen.

The reason why AI is suddenly such big news is that innovations in chip technology have dramatically increased computing power.  Nvidia created the graphics processing unit (GPU) as an electronic circuit for processing graphics and video for the gaming industry.  Nvidia GPUs are built using highly specialized silicon wafers produced by Taiwan Semiconductor.  The chips have billions of transistors etched upon them to control the flow of electricity. The advantage that GPUs have over CPUs (central processing units) - both of which are essential elements of a computer - is the ability to perform many calculations at one time and to breakdown computational tasks into smaller components that can be completed in parallel. In simple English, Nvidia’s GPUs dramatically increase processing power and speed.

Interestingly, the market for this data processing power did not exist much beyond gaming applications. In fact, Nvidia created the chips with the idea that they might one day help speed up scientific research by speeding up analytical tasks that might take scientists years to as little as a few hours or weeks.  Although the scientific community did indeed warm to the idea, it was more than a decade before a true market for the chips developed and the size and scope of the applications for these chips became clearer.  Beginning around 2023, the market size in terms of potential extraordinarily useful real-world applications for AI exploded.

If all of this seems like I may have at least some keen professional insight into AI investment opportunities beyond what you are already getting from cocktail parties and/or your hair stylist, the following links to Acquired Podcasts represent the source of pretty much everything I know about AI: 

Nvidia Part 1: The GPU Company (1993-2006)

Nvidia Part 2: The Machine Learning Company (2006-2022)

Nvidia Part 3: Dawn of the AI Era (2022-2023)

The Complete History & Strategy of TSMC 

How ARM Became The World’s Default Chip Architecture (with ARM CEO Rene Haas)

The Software Behind Silicon

Complexity Theory and Investing in Semicondutors

If you listen to these podcasts, you will realize, just as I did, how complex this space is and how brilliant many of the people who are leading this evolution truly are.   Absent a strong computer science background, I believe financial planners (including me), are poorly qualified to be dishing out stock-picking advice in this sector.

How FPH Clients Have Exposure to AI

First, I am proud to say that many Financial Planning Hawaii clients were early adopters and have had AI in their portfolios for years… in their index funds and ETFs. To the extent that the largest companies in the S&P 500 Index and Total U.S. Market Index represent the largest players in the publicly traded AI space, you have participated in the current boom.  (I am imagining readers rolling their eyes as they read this and thinking aloud, “Yeah, yeah, yeah…but where do you keep the good stuff.”)

In terms of the individual players, there are, of course, the suppliers of the AI technology. The most prominent of these is Taiwan Semiconductor (TSMC) as the maker of the highly specialized chips that are required for Nvidia’s data center GPUs.  There are also the scores of companies that supply the technology TSMC needs to fabricate the chips.  These include ARM, ASML, KLA, Applied Materials and probably dozens more that I do not know. 

Then, of  course, there is Nvidia which, at this time, is the dominant GPU maker for the massive data centers that are being built to power AI.  Its H100 chip that powers its GPUs is the most advanced chip ever built with more than 80 billion transistors contained on a single TSMC 4 nanometer wafer. As the first company to recognize the potential value of GPU technology, Nvidia is  estimated to be years ahead of its competitors and seems to be only constrained by the limits on the supply of chips that TSMC is able to produce.  It should also be noted that Broadcom, Qualcom, and Apple are also among the largest customers of TSMC.

Beyond these “picks & shovels” companies, are the companies that are among the first (after Ren-Tech) to exploit the value of machine learning LLLMs to great profitability.  These include social media giants Google and Meta.  Microsoft, Apple, and Amazon have also heavily invested in and are reliant on AI for current and future revenue generation.

In terms of publicly traded companies that are applying AI to drive revolutionary change in other industries, Palantir is a high profile example.  As per its Wikipedia page, Palantir specializes in software platforms for big data analytics.  With its primary emphasis  on intelligence and defense systems for counter-terrorism, it is disrupting the staid defense industry space.  There are scores of other companies that are applying AI technologies to create or disrupt in other industries.  Sorting them out or predicting which new entrants will disrupt or expand the business models in other industries is every investor’s dream, but is a challenge that is above my pay grade.

FOMO and the Buy Decision

I have absolutely no doubt that the fear of missing out (FOMO) is a huge driver of the AI client inquiries I am receiving.  You hear and read about how much money has been made from investing in stocks that are leading the AI boom, and you want a piece of the action too.  I get it. 

I am in no position to tell you whether all or any of the companies mentioned above will outperform the broader stock market, but one important bit of guidance I want to convey in this missive is that I believe most of the easy/lucky money on the biggest names in the space has already been made. All of the companies I have mentioned in this post have market capitalizations of $100 billion or more and eight of them (Broadcom, TSMC, Nvidia, Apple, Microsoft, Google, Facebook, and Amazon) have market caps of more than a trillion dollars. 

People who invested their money in these companies may have seen 1000%+ returns (10X+) on their investment cost.  Is it possible for a company  with a $1-2  trillion market cap to double again?  Maybe. The estimates for the size of the total addressable market for AI is estimated to be many trillions, but for a company to have a market cap in the $2-4 trillion range probably requires it to have hundreds of billions of dollars of earnings.  That is a tall, tall task.  To get to 10 trillion or more in market cap is almost an unimaginably big number (See addendum below), but to provide some context, as of January 1, 2025, the Federal Reserve estimated that the total amount of all U.S. currency in circulation to be $2.37 trillion. In other words, it would literally take more than all of the U.S. dollars in the world to buy out the company!

If you are hoping for the same 10X+ returns on your money, you will likely need to invest in the companies that will be the next Nvidia. Perhaps a better comparison might be to find the next Palantir.  In other words, find a company that is applying AI software to completely reshape a different industry space.  Again, this is far easier to do in hindsight than in foresight. 

Buying these companies early also means that you are investing in companies whose future success is far from guaranteed.  Most of the early stage companies in venture capital funds fail.  Even if you are buying individual early stage public companies, failure remains a significant too. For this reason, some investors may wish to consider AI sector ETFs as a way to get more diversified exposure.  For an introduction to some of the ETFs in this sector, here is a link to VettaFi’s Artificial Intelligence ETF List.

In closing, my general opinion of the AI-Sector is that it is over-hyped.  The term “AI” has become both a marketing buzzword and a meme.  As I floated the idea of  AI ETFs in the preceding paragraph I was reminded of the smoldering pile of Lithium Battery ETFs, Genomics ETFs, China Internet ETFs, and Cannabis ETFs from some of the recent manias.  With respect to AI, I feel like I have seen this movie before.

John “J.R.” Robinson is the owner/founder of Financial Planning Hawaii and Fee-Only Planning Hawaii and is a co-founder of personal finance software maker Nest Egg Guru.

ADDENDUM