top of page
Phil_McAvoy.jpeg

Phil McAvoy

Phil McAvoy is the founder of the Beyond Buy & Hold newsletter and a successful hedge fund manager (the Norwood Equity fund).  A dissatisfaction with the status quo and an unwillingness to accept that “Buy and Hold” is the best that the investment industry has to offer led to the creation of the proprietary strategy and the algorithms used in the Beyond Buy & Hold investing system. 

bbh-logo-large.jpg

GET OUR FREE
NEWSLETTER

Sign up and learn how to invest a better way.

Email *

By Submitting your email address, you are agreeing to our terms and conditions.

COMING SOON!

MARKET
SIGNALS

A NEW WEEKLY NEWSLETTER

COMING SOON!

YOU'LL RECEIVE:
 

  • Alerts Before Bear Markets Strike
     

  • Alerts Before Bull Markets are About to Run
     

  • Weekly Stock Market Risk Assessments
     

  • Training on How to Interpret and Respond to the Signals.

THE ONLY AI NUMBERS THAT MATTER

3 days ago
5 min read

Since I like to base investment decisions on data and not headlines, I have been looking for some indicators to evaluate the returns on the trillions of dollars being invested in AI infrastructure. 


I couldn’t find anything that was that useful, so I created my own analysis to determine if AI investments are producing returns that make those investments worthwhile.


There is some data on token pricing and agent usage in large companies, but for me those stats don’t help us answer the important questions related to return on investment.


First, some caveats.  I used some rough estimates to try to scope the problem.  A lot of the details needed for an exact analysis are not readily available.


But the large companies are publishing enough data to do a decent evaluation of the AI investments.  The AI capital investments are widely known.  This is how we know that in 2027, it is expected that the big tech companies will be investing about $1.1 trillion on AI infrastructure – up from $800 billion in 2026.


These spending levels are unprecedented and represent a huge slice of the total US economy.  These investments are at the center of the concerns about an AI bubble.


Since the big capital expenditures started to appear in 2024 and we know how much has been spent since then, we can calculate the profits needed to justify that spending.  There is a lag between the time when capital is spent and when it starts producing returns on all projects and AI investments will be no different.  I am using a two-year time lag for this analysis. In other words, I am expecting 2024 capital spending to start generating returns in 2026.


To measure AI profitability, I am using a combination of the financial results of Microsoft and Alphabet (Google).  They are two of the biggest AI spenders and they have similar profit margins and expected returns on invested capital.  Their financial statements also allow us to make decent estimates of the revenue they are generating from AI.  They also have similar annual capital expenditures on AI.   


The following data is for just one of those companies to keep things simple.  I used an average of the two companies and a healthy dose of rounding to arrive at the figures used below.  I am calling this average “The Big AI company”.  We can use this analysis to monitor the results at both Microsoft and Alphabet to see what kind of returns the AI capital expenditures are generating.  The AI returns are the only statistics that matter going forward.

 

AI CAPITAL EXPENDITURES


The annual capital spending on AI at our Big AI company for the last three years is as follows:



By the end of 2027, our Big AI company will have invested $600 billion on AI infrastructure. This is what AI capital spending looks like at both Microsoft and Alphabet.   


Using current operating profit ratios and expected returns on invested capital, I came up with annual sales increases needed to support this massive amount of spending.  I also assumed a two-year lag between capital spending and the sales generated from that spending. 


Each $50 billion spent on AI capital needs an additional $30 billion in annual revenue to provide a good return on that investment.  Looking ahead to 2029, here are the estimated incremental sales needed from these large AI capital expenditures.



It is interesting that both Microsoft and Alphabet will produce about $30 billion in AI revenues in 2026.  This puts them right on schedule to return the $50 they each spent in 2024 on AI.


Because the spending levels have increased so dramatically in 2025 and 2026, the required additional yearly sales must also increase significantly.  Our Big AI company needs to now add $90 billion in AI sales in 2027.  That number grows to $210 billion in 2028 and $360 billion in 2029.


What kind of sales growth rate does represent to for our Big AI company.  To calculate the sales growth rates, I had to make some assumptions about growth in the other lines of business in our big AI company.  Both Microsoft and Alphabet should see an average of $40 billion in sales growth each year from their other lines of business, so I added that amount to the AI sales to come up with a total increase in sales for our Big AI company.


Our Big AI company should post a 20% sales increase in 2026.  This represents the starting point.  Here are the sales increase percentages for 2027 through 2026 needed to generate the required return on AI capital investments.



Both Microsoft and Alphabet should come in close to the 20% increase in sales this year to stay on track with the growth needed to pay for the AI investments.  Historical growth rates have been closer to 15% per year. 


The growth requirement jumps to 31% in 2027 and 45% in 2028 and 50% in 2029. 


This means that Microsoft and Alphabet need to triple their current sales growth rate in the next two years.  As we move through 2027, the stock market should expect a 30% sales growth rate.  That does not seem impossible as AI revenues are growing rapidly at both companies.  We shall see.  The 45% sales growth rate in 2028 will be the real test. 


For context, Anthropic and OpenAI (stand-alone AI model companies) will post 2026 revenues of $57 billion and $35 billion, respectively.  Their projections for 2027 are $170 billion and $70 billion.  Our Big AI company’s projection of $90 billion in AI revenue in 2027 looks to be in line with the two AI model companies. The AI targets for 2028 and 2029 of $210 billion and $360 billion for our Big AI company will be big challenges.


I like to have some simple data points to determine whether these massive AI expenditures are delivering or not.  My numbers may be off, but they do provide a meaningful benchmark to measure AI success. 


The AI numbers are so large that big picture targets are more useful and informative.  I don’t know what to make of the token prices and AI agent usage data talked about in the media.   


You can ignore the noise and simply focus on the sales growth rate at Microsoft and Alphabet to see if AI lives up to the promise.



Stay Disciplined My Friends,


Phil

Disclaimers The Beyond Buy & Hold newsletter is published and provided for informational and entertainment purposes only. We are not advising, and will not advise you personally, concerning the nature, potential, value, or suitability of any particular security, portfolio of securities, transaction, investment strategy or other matter. Beyond Buy & Hold recommends you consult a licensed or registered professional before making any investment decision.


Investing in the financial products discussed in the Newsletter involves risk. Trading in such securities can result in immediate and substantial losses of the capital invested. Past performance is not necessarily indicative of future results. Actual results will vary widely given a variety of factors such as experience, skill, risk mitigation practices, and market dynamics.


 
 
 

Recent Posts

See All
IS THIS THE POPPING OF THE AI BUBBLE

Since the election, the downtrend in AI stocks has accelerated. Tech stocks and AI stocks, however, have been falling over the last six weeks. In the chart below, you can see that the semiconductor s

 
 
 
IS THE AI BUBBLE POPPING

Since the election, the downtrend in AI stocks has accelerated. Tech stocks and AI stocks, however, have been falling since over the last six weeks. In the chart below, you can see that the semicondu

 
 
 
UPDATED AI vs DOT COM BUBBLE

I last issued this analysis/comparison six months ago. This is a current update. Recent patterns in the stock market are eerily similar to what happened in the late 1990’s during the dot-com bubble.

 
 
 

Comments


bottom of page