This is a MW article which came to me for free (and i found interesting). Please delete my “discussion” if you see a copyright concern:

12:26 PM ET 8/1/24 | MarketWatch

By Gary Smith

Stocks valued on potential profits from AI are trading on thin ice

Confirmation bias is the all-too-human tendency to seek and remember evidence that confirms our beliefs. In investing, someone who likes a stock (especially if they own the stock) treasures good news and dismisses bad news.

Confirmation bias was rampant during the late 1990s dot-com bubble when internet enthusiasts proclaimed the arrival of a new economy that made traditional stock valuation metrics obsolete. Forget profits and dividends – we were told to value internet stocks based on how many people visited their websites, the number who stayed for at least three minutes, and the number of files that websites downloaded from servers. Some said we should look at a company’s burn rate, how fast it spent the money it had raised. (The faster, the better.)

In an investor survey near the peak of the dot-com bubble in 2000, the median prediction of the annual return on stocks over the next 10 years was 15%. This wasn’t just naive amateurs with this conviction. Supposedly sophisticated hedge funds were buying dot-com stocks just as feverishly as retail investors. They didn’t see the bubble because they didn’t want to see it. The actual annualized return over the next 10 years was negative 0.5%.

I remembered these delusional times when a reader forwarded me an article published in a peer-reviewed Elsevier journal entitled “Artificial Intelligence in Finance: Valuations and Opportunities.” The author examined “the financial opportunities arising from the new Artificial Intelligence (AI)” and introduced “a valuation model for AI stocks and ETFs, incorporating both AI fundamental and sentiment analyses.”

The author’s new valuation model is an equation stating that a stock’s future price is equal to its current price multiplied by the ratio of the current price/earnings ratio to the future price/earnings ratio:

This equation doesn’t make sense mathematically or logically. The author even gives an example using Nvidia (NVDA) stock that inadvertently demonstrates that it is nonsense:

“The current price of NVDA as of 6/6/2023 is $400 and the current PECurrent = 200 while the forward 1 year and the future PEFuture = 150; then the future price is 400*(200/150)=$533.3. Based on these figures, NVDA is expected to further rise and grow; and the upside is great for the future 1 year target. Hypothetically, if the PEFuture = 250, then the future price will be 400*(200/250) = $320. That means the current price for NVDA is way overbought and expensive.”

Huh? The author is arguing that when a company’s P/E goes down in the future, its price will go up, and vice versa. Thus, by the author’s reasoning, an astronomical P/E for NVDA (or any stock) means that “the upside is great” because the future P/E will almost certainly be lower – which, according to his valuation model, means that the stock price will be higher. So don’t be frightened off by high P/Es; embrace them.

Read: Nvidia’s stock surges as Microsoft says the magic words

The author’s second model is a linear equation that predicts a company’s revenue growth rate based on two factors: the S&P 500’s SPX revenue growth rate and the company’s research and development (R&D) spending. According to this model, every extra dollar in R&D spending increases a company’s revenue growth rate by 0.00075 percentage points. This implies that $1,000 of R&D will increase revenue growth by 0.75 percentage points and $1 million of R&D will increase revenue growth by 750 percentage points. This makes no more sense than did dot-com burn rates and the numbers are utterly implausible – but hype evidently doesn’t need to make sense or be plausible.

The author ‘s third model (“sentiment”) correlates Nvidia, Microsoft (MSFT), and Alphabet’s (GOOGL) stock prices with Google trends data for six technology terms: artificial intelligence, neural network, large language model, machine learning, generative AI, and deep learning. Among the findings: “for our analyses we study the period between 6/2/2023 and 6/8/2023. We find that the search peaks in the midnight and early am.”

The author also finds relatively small, positive and negative correlations between the three stocks’ returns and the various search terms. For example, during the period studied, all three stocks’ returns were positively correlated with the number of searches for “artificial intelligence” and negatively correlated with the number of searches for “neural network.” Duly noted.

We are evidently advised to buy AI stocks because they have high P/Es, have substantial R&D expenditures, and are correlated with certain search words. The real takeaway here is that this kind of nonsense is a good barometer of the reality that the AI bubble is alive and unwell.

Gary Smith, Fletcher Jones Professor of Economics at Pomona College, is the author of dozens of research articles and 17 books, most recently, The Power of Modern Value Investing: Beyond Indexing, Algos, and Alpha, co-authored with Margaret Smith (Palgrave Macmillan, 2023).

More: For Nvidia, AMD, Apple and other tech titans, earnings reports are all about AI

Also read: Fisker, Invitae and other fallen ‘unicorns’ are casualties of Silicon Valley’s broken venture capital system

-Gary Smith

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

> Dow Jones Newswires

August 01, 2024 12:26 ET (16:26 GMT)
Copyright (c) 2024 Dow Jones & Company, Inc.

This is a discussion topic or guest posting submitted by a Stock Gumshoe reader. The content has not been edited or reviewed by Stock Gumshoe, and any opinions expressed are those of the author alone.