Stripped to its essentials, finance is a race against time. What lies ahead of us is unknown, and the vast industry of banking and finance has developed to manage the risks that come with making commitments now that depend on an uncertain future. This is true of borrowers looking for credit to start up ventures, their creditors deciding to lend, and to investors who buy a share of a company’s future earnings when they purchase its stock but pay a price meant to reflect all that it has achieved.
It’s best to remember this when analyzing the extraordinary spending on the infrastructure for artificial intelligence, and the extreme reaction it has provoked in global markets amid increasing angst that we’re witnessing the bursting of another bubble to match the dot-com implosion in 2000. For example, the market value of Microsoft Corp., the world’s biggest software company and one of the leading “hyperscalers” attempting to dominate AI, has fallen by more than $1.1 trillion in three months. That’s roughly the same size as the gross domestic product of Switzerland. US software stocks as a whole have dropped 25% in that time, which qualifies as a crash or at least a bear market. Alphabet Inc., which controls Google, has somehow persuaded investors to back a bond that won’t pay out for a century.
These huge financial effects are driven by a material world in which the big four hyperscalers (who also include Amazon.com Inc. and Meta Platforms Inc.) between them report a $1.6 trillion backlog of orders placed for their cloud computing services that they haven’t yet been able to fulfill. This is not “build it and people will come.” The customers are already demanding the product.
There is a growing bottleneck in the supply of memory chips, which AI gobbles up at a scale never before seen. In what is now known as RAMageddon, this has driven the prices of some chips up by more than 500% over the last year — fantastic for the mostly Korean and Taiwanese companies that make them, and terrible for those who want to buy.
Electricity prices in the US rose 6.9% last year (the overall consumer price index rose less than 3%) as AI demand for power pushed up inflation. A backlash is swelling as communities often don’t want huge new data centers in their backyard. Citizens’ groups are mobilizing online, with support from across the political spectrum. Data centers now account for roughly half of all the electricity used in the state of Indiana. Opposition has grown into a critical political issue in the state.
Markets give a great window into the troubles that are plainly affecting an ambitious building program and will take time to correct. But their judgment on the core issue — whether huge spending now will pay off in extra revenues and profits in the future — remains nuanced.
Bubbles Are Inevitable
Market turbulence when a disruptive technology arrives is not surprising. It’s almost a necessity. History is littered with a succession of burst financial bubbles in new technologies that would generate big but uncertain returns long in the future. Canals, railroads, cars and, of course, the internet were all initiated amid speculative excess. Even though each was to prove truly transformatory, they all inflicted ruinous losses on many of their early backers when the gains didn’t come through in time.
It’s no surprise that AI is having a similar effect. As with the worldwide web 30 years ago, it’s not clear exactly what the technology will do in the future and who will make money from it, but anyone with an imagination can see that the potential is profound.
Judged by classic long-term metrics, the US stock market looks its most expensive ever, save only for the internet craze at the turn of this century. There are also clear signs of dangerously uniform bubble sentiment, particularly in the way that people are talking about the shortage of chips. The behavioral economist Peter Atwater examined Bloomberg Morning Briefing’s coverage of the issue and commented:
I’m not sure I have ever seen “insatiable,” “shortage,” “only expected to get worse,” “unprecedented,” “no end in sight” and “parabolic” prices in one news story! Given that, I expect it won’t be long before feast turns to famine for memory-chip makers. While smiles now span the globe, that won’t be the case once sentiment peaks and turns down.
Put slightly differently, this is all a matter of time. Higher chip prices will ensure that far more are made, and eventually bring down their price (and dent earnings and share values). This isn’t necessarily the critical bottleneck that will bring the AI revolution to an end, but both sentiment and market prices make clear that it’s a serious issue that has to be addressed.
When Bubbles Are Dangerous
Doing some damage to the French language, an American financier once expressed the critical question as follows: “Cherchez la leverage.” If people lose their own money, then that’s tough on them. Lose someone else’s money, and you can create the cascading losses that drove the Global Financial Crisis in 2008.
The good news so far is that the biggest hyperscalers have mostly played with their own money. They’re no longer using cash to reward investors by buying back stock, and their massive expenditures have eaten into free cash flow, leaving them without the safe cushion they once had. This isn’t like dot-com startups who financed themselves with capital from equity investors before ever turning a profit. Rather, the hyperscalers are taking their huge profits and deploying them to invest in what they hope will be continued dominance of the next transformative technology.
That is risky. Microsoft’s falling share price reflects this. It’s making investments that might not pay off, and that is a far riskier proposition for shareholders than using spare cash to buy back stock. But at least leveraged losses are not in sight.
Another way that burst bubbles, or equity bear markets, can create economic pain is through wealth effects. The more people have come to rely on savings in the stock market, the greater the suffering a bear market inflicts. This is far more important for the US than for anyone else since so many more Americans are invested in stocks, especially as retiring baby boomers draw down their 401(k)s. The vast majority can take comfort, for now, that they’re in funds tracking the main indexes, which are barely down from their all-time highs.
Further, Jennifer McKeown, chief economist at Capital Economics, points out that of the last 10 major US stock market downturns, none clearly led to a recession: “Stock market corrections rarely cause recessions on their own, and we doubt this episode will be an exception.” A serious threat to the economy, she says, would be “a complete loss of confidence in AI’s long-term economic value.” That could do very serious economic damage, but a lower stock market would merely be one of the effects.
Microsoft’s sudden fall was shocking, and events like this can easily prompt broader market reactions, but this has been a true rotation. Dhaval Joshi of BCA Research argued that there have been shifts within the technology sector (from software to hardware), and from tech to everyone else. The results have largely washed out, and are driven more by a reassessment of who will win most from AI:
The broader market has absorbed Microsoft’s huge drawdown. You might have expected that when a behemoth tech stock has lost over $1 trillion in market value, the Nasdaq, the S&P 500 and even the US-dominated world stock market would be in a full-on crash. Not this time though.
The money that companies feel obliged to spend to get ready for AI shows up perhaps most clearly in the phenomenal success that Nvidia Corp. has had in selling chips since ChatGPT launched in November 2022. Since then, its revenues have risen by 880%. Sales for the Magnificent Seven companies as a whole are up 94% in that time; revenues for the other 493 companies in the S&P 500 have risen only 16%.
Despite appearances, this isn’t a replay of the internet bubble. The risks — which are considerable — aren’t so much about a stock market boom and bust, but rather the possibility that the broader costs of all this infrastructure, in terms of rising prices, social displacement, and the growing backlash against AI could prove to be more than society can bear.
If such risk comes to fruition, then it will doubtless damage the stock market. But that will merely be evidence, or a symptom, of the immense difficulties involved in the multi-year effort to make AI work. The extreme market gyrations of the last few weeks should be taken as a warning about the difficulties of trying to fund any project so huge over such a long period of time, rather than as a trigger of disaster in themselves.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
