This article originally appeared on Dean Baker’s Patreon. It is reprinted here with permission.
A bit less than 20 years ago, a nationwide housing bubble collapsed, giving us the Great Recession. Millions of homeowners had their houses foreclosed. We had high unemployment for the better part of a decade. And the subsequent falloff in construction created the basis for another extraordinary run-up in house prices during the pandemic. In other words, it was pretty bad news.
The current bubble in AI is laying the groundwork for another bad story. As was the case both before and after the collapse of the housing bubble, there is a tremendous premium in intellectual circles on making the problem more complicated than it is.
My latest poster child for this point is a column in the New York Times by Richard Bookstaber, a hedge fund manager who had predicted the financial crisis that followed the collapse of the housing bubble. His column notes the AI bubble, but then argues that the big problem is that we are also facing risks from the private credit market, as well as geopolitical risks, like the fact that China could cut off the supply of chips from Taiwan and also the price shock associated with the cutoff of the oil flow through the straits of Hormuz.
The collapse of the stock prices of the companies that are big factors in AI will then have huge spillover effects, devastating people’s 401(k)s, as well as whacking pension funds. This will lead to a huge fall in consumption, which would likely lead to a recession.
The warnings are well-taken, but the story is actually not complicated. Bookstaber tells us at the start of his piece:
“Yet they [the potential problems he notes] are different entry points into the same underlying structure — a complex and tightly coupled system where the specific source of stress matters less than how quickly that stress can spread.”
As was the case with the financial structure supporting the growth of the housing bubble in the first decade of this century, there are some complex issues. But the housing bubble itself was simple. House prices had grown hugely out of line with the fundamentals of the housing market. Nationwide, real house prices had grown by 70 percent between 1996 and 2006. This followed a century in which house prices on average had just kept pace with the overall rate of inflation.
The run-up in house prices took place despite a relatively high vacancy rate. There also was no corresponding growth in rents, which had largely kept pace with inflation.
The rise in house prices led to an unprecedented boom in residential construction, which peaked at 6.7% of GDP in the fourth quarter of 2005. After prices peaked and started to fall, construction plummeted, bottoming out at 2.4% of GDP in the third quarter of 2010.
This was the story of the Great Recession, not the financial crisis. We have no easy mechanism, apart from massive government stimulus, to replace the 4.3 percentage points of lost demand that resulted from the ending of the construction boom. This would be equivalent to $1.3 trillion in annual demand in today’s economy. In addition, the loss of trillions of dollars in housing wealth by homeowners led to a further reduction in annual demand of 1-2 percentage points of GDP, an additional $320-$640 billion in today’s economy.
The financial crisis provided good entertainment, as we watched leading politicians from both parties insist that we couldn’t let the Wall Street bankers be ruined by the free market and their own incompetence, but this was a sidebar. The collapsed bubble was the story of the Great Recession: full stop.
To be clear, the flood of fraudulent loans that the industry greedily issued and securitized allowed the bubble to grow much larger than would otherwise have been the case, but the key issue was house prices. If they had not grown so out of line with fundamentals a wave of defaults, which would have been far smaller, would have had a limited impact on the economy.
It is the same story now with the AI bubble. The problem we have is a grossly inflated stock market driven by the AI bubble. The various problems identified by Bookstaber would not be a big deal if this was not the case.
A freeze-up in private credit would not matter much to the economy if it was not the fuel source for the AI bubble. Furthermore, if Ai was not in a bubble, the loss of one specific source of credit would not have huge impact. Other lenders would be happy to make loans to the sector. But because it is a bubble, there are no alternative sources to fill the gap, just as the fuel for the housing bubble’s expansion disappeared after the subprime mortgage market froze up.
In addition to Bookstaber’s risks to the AI bubble, let me add my current favorite, Chinese AI. Chinese AI companies have been rapidly expanding market share, focusing on easy use and low cost. According to some accounts, they had already captured 30 percent of the world market by December. Given the rapid growth of Chinese AI (it likely would have been less than 10% a year earlier), their share would almost certainly be considerably higher today.
As the U.S. frontrunners focus on massive computing power, the Chinese AI leaders are developing low-cost practical applications. I can’t claim any great expertise on the specifics of AI, but on the surface, the Chinese route would seem to be the better long-term or even near-term path. If China’s AI leaders manage to capture a large share of the market and drive down the prices charged by U.S. competitors, the massive profits stock investors are banking on will never be there.
In this context, it’s probably worth mentioning that Trump’s war in Iran is not going to make potential AI users around the world more inclined to turn to the American AI industry. No one is going to want to be dependent on important systems from a country where the president can shut off access any time he gets angry or has his feelings hurt.
At the end of the day, the exact reason the AI bubble will burst is impossible to predict, but the key point is that the existence of a huge bubble driving the economy is a real problem, not the specific cause of its bursting. Our elites like to make things complicated so that they can appear like great intellects when they unravel the mystery, but that’s just a myth.
The web of financing that supported the housing bubble was quite complicated, but the housing bubble itself was very simple. It’s the same story with the AI bubble.
Republish our articles for free, online or in print, under a Creative Commons license.
