Is There A Bubble In AI Stocks?

I've seen and heard a lot of things that have made me cynical about the investment industry, but nothing irritates me when I hear "valuation" used as a plea for retail investors to buy expensive assets (one of the best I've heard is from a former colleague who stated that a company was expensive on a price-to-earnings (P/E) basis but cheap on an earnings-yield (E/P) basis…they are the same thing!).

On that basis, my credulity was recently tested when I heard from a Morgan Stanley analyst
EM
, the big investment bank, declares that technology stocks are "attractively valued" based on the "price/innovation" ratio. Price/innovation is the relationship between a company's price and its profits plus what it spends on research and development; As such, it is designed to make companies with low profits and large, risky investment budgets look attractive.

As a general rule, when the financial community starts inventing new valuation measures, it's time to worry. In the early 2000s, analysts introduced valuation ratios such as "price/clicks" for Internet stocks, attempting to lend a veneer of respectability to bubble-level valuations. Of course, everything ended badly.

Introducing the idea of ​​a "price/innovation" relationship into the market narrative should raise fears of a stock market bubble, especially in stocks powered by artificial intelligence (AI). Bubbles are difficult to define, but, following a famous legal opinion on pornography, "you know it when you see it."

The ingredients of a bubble are typically based on an innovation (railroads, the Internet) sprinkled with cheap money, and together they produce significant outperformance of a select group of assets resulting in historically extreme valuation levels.

Most likely, the core of a bubble exists in a group of large technology stocks, increasingly known as the Magnificent Seven, headed by chip design company Nvidia. To use a very simple valuation metric: the price/earnings ratio, the European stock market trades at a price/earnings multiple of 1.3 times, the American market is 2.5 times (big tech stocks drive this multiple up rise), Microsoft
MSFT
The sales ratio is 13 times and Nvidia's is 33 times (for comparison, UBS is trading at a ratio of 1.7 times and Siemens at 1.3).

A similar valuation bubble exists in private markets, where venture investors pay very high multiples for AI-focused startups. The underlying reason is that the potential sales growth of these big tech companies is so promising that they want a higher valuation multiple.

Last week, many of them reported earnings, and while most of these tech companies posted healthy profits, the revenue growth is not impressive, suggesting great optimism regarding the future impact of AI.

In that context, if AI is going to be the organizing logic of the next bubble (in the last fifty years we have seen asset bubbles in gold, Japan, Asia, the Internet, housing, China... to name the main ones) , will be We Need the Boost, or the 'little hit of whiskey' (as New York Fed President Benjamin Strong put it in 1927...guess what happened next).

In general, since the global financial crisis monetary conditions have been very loose, causing asset inflation rather than consumer inflation. That trend has changed recently when a burst of inflation led to a sharp rise in rates. Despite this, credit markets have performed very well (another bubble?). With inflation falling, expectations are growing that central banks could cut interest rates, which could add more fuel to the AI ​​bubble hypothesis.

In the days before quantitative easing there was a decent debate about whether central banks aided and abetted the creation of asset bubbles. At the time, the orthodoxy was that bubbles were difficult to identify and even if they could be identified, central banks found it difficult to deflate them. The difference today is that few, if any, central bankers worry about this risk and instead talk about the "wealth effect" of asset prices.

The reasons to be more vigilant here are that asset bubbles generally destroy wealth, invariably transfer it from poorer investors to richer ones (the rich buy early and the poorer investors buy late), they distort investment in all economies and when they collapse, their consequences can be costly (as demonstrated by Japan's lost decades). Bubbles often leave useful infrastructure behind: railroads in the late 19th century^th century and the Internet/telecom infrastructure of the 2000s, but at a high price.

In my opinion, we are not yet in a proper AI bubble. So far it is in a small number of companies that, very unusually, represent a large proportion of the stock market. BCA Research calculations show that the ten largest US companies represent 75% of the stock market, something that happened only in 2000 and 1929.

For the AI ​​bubble to grow into a mania (Manias, Panics and Crashes by Charles Kindleberger remains the best analysis of bubbles), companies in sectors that will be positively affected by AI (such as healthcare and life sciences, financial data-focused companies) must be captured by the AI ​​narrative and see their stock prices rise accordingly. In the same sense, the AI ​​mania will have to spread to countries in Europe, Japan and perhaps China.

Keep an eye on other indicators. When taxi drivers start talking about error correction in quantum computing, we will surely be in an AI bubble.