The gathering of artificial intelligence has already reached historic proportions and is currently passing several famous, or rather infamous, milestones. The SOX Semiconductor Index’s peak price is 62% above its 200-day moving average, more than twice the spread on the Dow Jones Industrial Average leading up to Black Monday in 1987, and more than twice the spread leading up to Black Tuesday in 1929, Bank of America strategist Michael Hartnett wrote in a Thursday memo. This spread approaches the 55% margin the Nasdaq had before the dot-com crash of 2000, when the commercial internet was starting to take off and valuations of companies with no clear path to profitability reached hundreds of millions of dollars. This is even within the 73% spread of the French CAC All Trade Bulls Index before the Mississippi Bubble burst in 1720. In that episode, shares of the troubled French colony Mississippi Company were allowed to be used as legal tender, leading to a doubling of France’s money supply. “Why is everyone’s new base case collapsing, with exponential price movements, market concentration, collapsing volumes, stocks pushing up bond yields…here we go,” Hartnett mused on Thursday. AI stocks began a parabolic curve at the end of March. This is highly unusual in securities price charts. In particular, the shares of chipmakers such as Micron, Advanced Micro Devices, SK Hynix, Marvell, and Intel all exhibit this trend. Some economists are convinced that all investments in AI, which Wall Street banks believe will exceed $1 trillion next year, are a bubble. “We had to save over $1 trillion in cash to support investment… leading to what everyone would call a bubble,” economist Anne Pettifor, director of policy research at the Macroeconomic Organization, told CNBC. Comparing the AI ramp-up to other historic booms Despite the large capital expenditures, not all commentators are impressed with the scale of the AI ramp-up. Robin Wigglesworth of the Financial Times called it “a tiny gnat on an elephant’s butt” compared to the railway boom of the 1860s. The rail boom resulted in much higher bond issuance compared to total AI debt when adjusted for inflation and converted to GDP. “There’s about $5 billion, $6 billion worth of bonds outstanding, which doesn’t seem like a lot, but because the U.S. was a small economy, that’s equivalent to $10 trillion today in GDP terms,” the editors of the FT’s Alphaville blog said earlier this month on the Unhedged podcast, citing JPMorgan analysis. Still others don’t seem too worried and acknowledge the possibility of a bubble. “Railroads were the bubble that changed America. Electricity was the bubble and it changed America. The broadband rollout of the late 1990s was the bubble that changed America,” author Derek Thompson wrote last year in a column referenced in a memo by Oaktree Capital Management co-founder Howard Marks. “AI is unlikely to be the first innovative technology that isn’t overbuilt and doesn’t require short-term painful fixes.” Despite all the debt and the fact that companies are using conduit financing techniques to keep much of it off their balance sheets, real AI returns are actually being realized. Alphabet’s first-quarter cloud revenue rose 63% annually, the company reported last month. Amazon’s AWS Cloud division posted 28% year-over-year revenue growth in the first quarter, with AWS division revenue reaching $37.59 billion. Microsoft’s cloud revenue rose 40%, with the company’s division, which includes Azure, reporting third-quarter revenue of $34.68 billion. These numbers come as a relief to the overall stock market, where gains are increasingly concentrated in semiconductor and AI infrastructure stocks, suggesting that the foundations of the current stock boom may be weak. Piper Sandler’s note on Wednesday said the ratio of companies making profits to those losing money has declined even as broader market indexes such as the S&P 500 have soared since the end of March. “As SPX continues to hit new record highs, the rise and fall lines show a clear divergence, indicating more concentrated leadership, and more technology leadership,” Craig Johnson wrote in a letter to Piper.
