“The AI Investment Boom Is No Different from the Past” ... WSJ Warns of “Overinvestment”

[Financial News] Astronomical sums are pouring into the artificial intelligence (AI) industry, but a warning has emerged that the history of overinvestment could repeat itself, as in the railroad and internet industries of the past, with large-scale investments failing to deliver the returns investors expected.
James Mackintosh, a Wall Street Journal (WSJ) columnist, wrote in a column published on the 18th local time, titled “Lessons from the AI Stock Market’s Turmoil: This Time Is No Different,” that there is still no reason to believe the current AI investment boom can avoid the problems seen in past waves of massive capital spending.
Citing the phrase “This time it’s different,” which legendary investor John Templeton called the most dangerous phrase in investing, he stressed that the enormous profits AI-related companies are earning now do not guarantee future returns on investment.
Mackintosh cited the railroad investment bubble in the United Kingdom in the 1840s and the construction of fiber-optic networks during the dot-com bubble around the turn of the 2000s as representative examples.


These investments brought enormous benefits to society as a whole, but they disappointed investors because it took a long time to realize returns or because intensifying competition kept profits below expectations.
The warning is that a similar situation could arise in the AI industry.
In a recent Brookings Institution report, Stijn Van Nieuwerburgh, a professor at Columbia Business School, estimated that total U.S. investment related to AI would reach $9 trillion (₩12.04 quadrillion) by 2032.
That would mean devoting an average of 3.2% of annual U.S. gross domestic product (GDP) during that period to AI investment, potentially making it the largest capital investment project in U.S. history.
Mackintosh noted that even if vast sums are spent building data centers, a substantial portion of the investment may not be recouped if AI technology fails to generate sufficient economic value.
He also said it was uncertain whether the high profits currently earned by AI-related companies would continue in the future.
The rapid expansion of data center construction has sharply increased revenue and profits for companies that supply semiconductors, cooling equipment and generators, but demand for these products could fall once the construction boom ends.


Investors, too, appear to believe that current high profits are unlikely to last, and AI-related companies’ share prices have risen relatively less than their earnings, according to the analysis.
Nvidia, the U.S. company with the largest market capitalization, saw its share price rise 15%, but its price-to-earnings ratio (P/E), based on forecast earnings for the next 12 months, fell from 23 times to 15 times.
Micron, a memory chipmaker, has also seen its share price triple, yet its forward P/E remains below five times.
Concerns have also grown recently that data center investment could weaken as higher-than-expected U.S. inflation pushes up financing costs.
Amid these concerns, shares of data center operator CoreWeave plunged 24% last week, wiping out all of their gains so far this year, while Micron fell 11%. South Korea’s Samsung Electronics and SK hynix also dropped 17.8% and 24.3%, respectively, recording their worst weekly declines since the 2008 financial crisis.
Mackintosh noted that even if AI investment proceeds as planned, the vast sums invested could be wasted if it fails to generate sufficient profits; conversely, if investment slows, the performance of related companies, including semiconductor makers, could suffer.

[email protected] Lee Seok-woo, International Affairs Specialist Reporter