By Humphrey McQueen
Anniversaries are not harbingers of doom. Twenty-five years back, Walter Marks of Oakland Capital, which now manages $US200 billion, had warned his clients that the dot.com boom was a bubble about to burst. On March 23 that year, the market peaked. From there, the Millennium Bug was not in the race to the bottom.
This year, in a January 7 memo ‘On Bubble Watch,’ Marks held back from warning about a generative-AI (gAI) bubble because its boom still lacked one of the five criteria that had busted dot.com. No one was preaching that share prices knew no upper limit.
Here is the data
The ratio of share-price to earnings for the corporations on the Standard & Poor (S&P) Index hovers above thirty to one, compared with a long-term average of sixteen to one. Were you to buy a share portfolio today for $100 you could expect an annual return of $3.30 compared with almost twice that over the long haul. Meanwhile, a Treasury bond will pay 4.5 percent on the knocker.
Had you dared to buy into Advanced Micro Devices (AMD) at $3.00 in 2014, you would be laughing all the way to bank now that its shares sell at $30.00 with earnings to match.
The S&P Index has shot up 50 percent in twelve years, compounded at around 17 percent since the pits of 2009, with a brief drop in 2020 and despite being stagnant for nine months during 2022. The collapse of the Silicon Valley Bank and New York’s Signature Bank over the weekend of March 10-12, 2023, might as well never have happened.
Part of the reason why gAI keeps booming is the racket between the Wolf-Street backers of Initial Public Offerings and the founders of start-ups. Goldman Sachs, Morgan Stanley and their ilk issue parcels at par to mates who then unload a tranche onto ‘after-buyers,’ technically, ‘the Greater Fool.’
Today’s Magnificent Seven are Alphabet, Apple, Meta, Microsoft, NVIDIA, and Tesla. Six years back the talk was of the ‘Big Nine,’ which included Amazon, Baidu, IBM and Tencent.
The Seven account for a third of the market-value of the corporations listed by S&P. Before the 2000 crash, the top seven held only twenty-two percent.
Eighty-one percent of Wolf-Street’s recommendations today are to buy into the Magnificent Seven, or 86 percent without Tesla, which Bloomberg’s Jonathan Levin calls the ’most preposterously valued.’
Nvidia had magnificence thrust upon it by the gAI-generated demand for its hard-drive, developed for online games. Its share-price soared by 170% while the other six have averaged a miserly 61 percent.
These numbers confirm the first of Warren Buffett’s two golden rules for investing: the stock-market is a voting machine, not a weighing machine. Last year, his Berkshire-Hathaway sold off two-thirds of its Apple shares, and now sits on well over $300bn. in cash to ride out 2025.
The force of ‘the market’
Only three gunfighters are alive when the house-lights come up on Hollywood’s knock-off of Kurosawa’s Seven Samurai – and they were all fighting on the same side and not trying to knock each other off.
Nvidia’s customers worry that, with 95 percent of the sales of hard-drives for gAI, their supplier will use its monopoly to price-gouge. In response, AMD is spending $6bn a year on research to secure a slice of the generation after next.
Take a moment to distinguish the exposure of each of the Seven in terms of how it makes most of its revenue, whether on-line services like Alphabet, or physical commodities with Apple phones. A slump among the gAI unicorns would hit Nvidia more directly and immediately than the others.
High time to consult the Bank for International Settlements (BIS). Although light on numbers, what its September report on the gAI boom headlined as ‘Challenges’ by December became ‘Risks.’
In mid-2007, the BIS had forecast a 1930s-style Great Crash, an alarm which did not reach the front-pages for fourteen months while the ADMASS media conjured the confidence trick of ‘this time it’s different.’
When pondering whether the gAI boom is a bubble about to burst one cannot afford to forget the accounting malpractices of the 1990s. Given the doings at Price Waterhouse Cooper, they are not likely to be in need of much of an update.
Too busy confabulating and hallucinating about the latest start-up, few commentators acknowledge that generative Artificial ‘Intelligence’ (gA’I’) is not Artificial General (Gestalt?) Intelligence, and is never likely to be. When will the next OpenAI be ready to sell the latter to Gates? The answer is stuck where it was sixty-years ago: ‘soonish’.
