The Bank for International Settlements (BIS) cautions that the current "exuberance" surrounding AI investments might culminate in a significant and extended investment bust. While the BIS report, published in January, avoids extreme alarm or over-optimism, it highlights that the present AI spending boom, though substantial, is not as large relative to the economy as some past bubbles. For instance, it's roughly 1% of US economic output, comparable to the mid-2010s shale boom, but only half the size of the late-1990s tech frenzy and less than a fifth of property and mining booms in Japan and Australia. Consequently, the BIS assesses the immediate financial stability risk as moderate, rather than acute.
However, the BIS points out that the sustainability of current AI valuations hinges on these companies eventually generating the enormous profits their stock prices anticipate. A key risk identified is the significant amount of AI financing, estimated by one Financial Times count to be over $120 billion, much of it debt, being held off-balance-sheet. This debt is often channeled through special-purpose vehicles (SPVs) or shell companies that finance data centers and chips, renting them back to the parent companies. Examples include Meta, which raised about $30 billion this way for a Louisiana site and another $30 billion in bonds, and Oracle, which utilized similar methods for $38 billion across two projects. The BIS warns that debt does not become safer simply because it is harder to locate.
While lenders are currently willing to fund these projects on ordinary terms, suggesting some financial health, the concern remains about the gap between current spending and future revenue. Sequoia, a respected Silicon Valley investor, estimates the industry needs to find approximately $600 billion in new annual revenue to justify its current spending, a gap that is presently widening. The BIS's warning aligns with other financial experts who see red flags, such as J.P. Morgan, which notes extreme market concentration with the top ten US stocks accounting for about 40% of the S&P 500's market capitalization compared to 17% in 2015. NYU finance professor Aswath Damodaran explicitly warns that an AI crash, fueled by massive, debt-funded capital expenditure, could be more severe than the dot-com bust.
