A
Deserving Valuation of A.I.
8/22/26
Will the massive investment of the top 9 U.S. A.I.
companies pay off? What deserving valuation will Wall Street ultimately assign?
An 8/16/26 WSJ article writes:
“America’s blue-chip tech companies are placing these
huge bets ($3.60 trillion both on and off balance
sheet financing in 2026 and the next several years: 11.7% of 2025 U.S. GDP)
based on the assumptions about what the demand for AI computing – and
availability of AI hardware-will be in several years. Their hope is that they
will easily meet all their obligations with future revenue as consumers and
businesses adopt AI in every facet of American life. If these assumptions about
technology and demand prove wrong, these deals to clinch future capacity could
become a monstrous burden for the tech companies and their investors.” (through
added depreciation or chargeoffs)
The future of A.I. will severely impact both NASDAQ
and S&P 500 values, so we think it is a very good idea to look what
happened to market overvaluation in the past and how that is relevant to A.I.
in the future. We document the Internet and Sub-Prime mortgage market crashes,
to assess where the present stock market enthusiasms are the same, but also
where they are different.
We are interested in comparing the present valuation
of the S&P 500 with market peaks on:
S&P500-3/24/2000 – 1527 - CAPE 44.2
S&P500 – 10/9/2007 – 1565 – CAPE 27.2, with a
level around 17, the long-term average.
S&P500 – 8/18/2026 -7691 – CAPE 42.0
You didn’t want to be in the S&P 500 at the first
market peak, for it would take more than 7 years for the index to recover, but
dividends did help. The present level of stock market overvaluation relative to
the equilibrium Gordon model now almost reaches the level on 3/24/2000.
The present excessive market valuation had three main
causes:
1) Very
low long-term interest rates. (obviously now in the past)
2) The
invention of generative A.I. (now
requiring the A.I. be quickly wide-spread)
3) Increasing
corporate profit margins.
These three factors, the first two taken successively,
account for those on Wall Street that have never seen a bear market. In
addition, the investment horizons on the Street are very short.
But what about the future of A.I.? Ashutosh Padhi is
an A.I. consultant at McKinsey. He notes that A.I. is a general
purpose technology like electricity, computers, and the Internet.
Technology transformation takes years, but economic transformation, to embed
the technology, takes decades. The change, so far, is broad but shallow. One
billion consumers have adopted A.I. in three years, but only 6% of all surveyed
companies have found that A.I. had made a meaningful difference. A figure
around 5% has been found by U.S. government survey as well. The best use of
A.I., as we have chronicled, has been applications at the departmental level.
The eventual best use of the A.I. is not just a focus
on productivity and automation, but on growth. The emphasis should not be on
faster models, but on A.I.’s human effects. To put this another way: the
emphasis should not be, by necessity, partially FAIR data, but on good company
leadership that can exercise good judgement and can earn the trust from
customers and employees.
This suggests that Wall Street has been carried away
by A.I. and that a more sober analysis of A.I. would be:
From a
4/16/26 (b) analysis:
12/31/25 6/30/26 Calculation 8/18/26
S&P
500
6846 7499
7691
With 30%
A.I.
4264 4360 1.263
x the below 4408
A.I.
with 2% 3376
3452 +38 S&P points/quarter 3490 (A.I. with
Economic
Growth *
2% Economic
Growth)-3614
(previous drops)
Given my risk preferences and portfolio structure,
assuming A.I. with an economic growth around 2%, I would probably start
purchasing the S&P 500 around 3490-3614, leaving 10-20% of the portfolio in
cash which will probably earn a decent return. I choose to lower my portfolio
risk because the above does not include the effects of climate change and lower
crop yields, which will occur.
A graph
of the S&P 500 in 2000 and 2007 shows that past large market slides did not
result in an immediate bottom; slides occurred over an average of 2.0 years. We
assume a slide equal to the average of 53%. An A.I. slide could be faster and
more extreme. It then takes many years for the S&P 500 to assume a previous
peak.
* Our S&P 500 model is within 7% of the Gordon
Model, and we can use both interchangeably for the period 2015-2025 to judge a
market peak. But not beyond, because Gemini then assumes massive earnings
improvements due to A.I..
__
You don’t believe us? Read what Bill Dudley, former
president of the Federal Reserve Bank of New York has
to say. When considering investments, you should consult a
number of credible sources and then make up your own mind. That is the ultimate goal of our website, to create better investors.