The word ‘’invest’’ originates from the Latin words ‘’to clothe’’ evolving through time to become the Italian word ‘’investire’’, which meant to ‘’commit money to earn a financial return’’.
At the heart of this definition, is the word ‘commit’ which implies a length of time; it calls for attachment that yields returns.
To illustrate this to an extreme, when the annual fruits from an investment (be they dividends or retained earnings) exceed the investor’s base or entry cost, the investment becomes an infinite free option on future growth. The ultimate investment never needs to be sold; it does not rely on the existence of a buyer.
But the investing world of 2025 shows limited interest in the definitions above. Media outlets and dinner parties are instead awash with discussions on price predictions for gold, crypto and AI stocks.
All these investments are dependent on the existence of buyers. Assuming rational, these buyers must believe that they have greater insight into the future prices of these assets than the sellers.
Such buyers can also not be motivated by the fruit (profits or cash flow) from these assets as they do not generate any – particularly for gold and crypto, but by the idea of felling a large tree – the capital return materialized on selling out.
We are not arguing that each of these assets is worthless, but that they are extremely risky. Gold has gained value over thousands of years through evolutionary instincts, symbolism, and psychological attachment. It has a scarcity factor. Its only real usefulness in portfolio management is that its price behaviour is at times opposite to other conventional asset classes.
Crypto is considered by many to be modern gold. Like gold, it ignites investor backlash when one explores its virtues and is impossible to value using conventional valuation techniques. It relies on pure belief.
This sentiment is reminiscent of technology stocks in 1999 (e.g., JSE-listed Persetel/Comparex and DiData); financial stocks in 2008; building and construction companies in SA in 2016 and let us not forget Steinhoff. Unlike gold, crypto’s price characteristics have not been assessed over time.
This leads me into AI, undoubtedly a life altering industry and an easier place to demonstrate irrational exuberance with math.
The mega technology stocks or hyper scalers are expected to deploy $3trn of capital into the AI ecosystem, of which $1.7trn is for tech hardware and chips, over the next four years. Their combined market capitalization has risen by $12trn since ChatGPT was released in 2022. The market is pricing them at four times the replacement cost of these assets, this is called Tobin’s Q ratio.
In the technology industry, companies aim to generate a 15% to 25% free cash flow margin. This margin is simply the ratio of free cash flow to net sales.
Considering $1.7trn of hardware investment by 2028, to meet a 20% free cash flow margin, the hyper scalers need to generate $500bn of free cash flow just to justify their cost of capital, which is assumed to be 10%.
Knowing the cash flow required in 2028 and that 20% is a normalized free cash flow margin, we can calculate that $2.5trn of sales is needed in 2028 to justify all this investment.
If we then look at the clients of these hyper scalers and assume that they want to make a 20% profit margin on the $2.5trn of revenue that they give the hyper scalers, then the industry needs to generate $3.1trn of revenue to justify the $1.7trn of investments.
Based on various reports from large global investment banks that we follow, the technology industry is generating AI revenue of between $2.5bn to $43bn in 2025 and the most bullish forecasts that we can find to 2030, lifts this number to $780bn. This is vastly different from the required $3.1trn calculated above.
In conclusion, moving from an industry analysis to an individual company, to add credence to our concerns of distortions in markets, look at the valuation of Palantir. Palantir trades at a forward P/E multiple of 268 times and ninety-seven times its current revenues.
If we place Palantir on a P/E that is more in line with a good technology business of 25 times, then the company needs to grow its earnings per share at 159% annualized for the next 3 years, 77% annualized for 5 years and 21% annualized for 15 years. To contextualize this against research undertaken by McKinsey on sustainable growth rates of high growth businesses, only 21% of companies have sustained growth rates exceeding 15% over 10 years.
I conclude with a great quote from Akira Kurosawa: “In a mad world, only the mad are sane.”
NOTE:
This article was written and first published on the 3rd of November and Bitcoin and the share prices of numerous AI companies mentioned have materially changed since then.