Since the internet boom in the mid-1990s, San Francisco has been economically isolated from the rest of America. The Bay Area is on a unique trajectory, as evidenced by everything from home rents to per capita growth rates that are much higher than the U.S. national average.
I couldn’t help but think about this when I flew into San Francisco a few weeks ago. Today’s excitement is driven by artificial intelligence, not consumer web pages. But the atmosphere is the same as it was 30 years ago. There is one story and everyone is buying it.
Every billboard we passed from the airport to the city had something to do with AI. Huge white blocks of brand new housing and office space lined the main roads. Outside the window, I could see young men (Silicon Valley’s workforce is still mostly male, as it was then) typing away at computers while playing sports or video games on oversized flat-screen TVs.
So far it’s the 1990s. So I sometimes wonder what’s different this time. The topic has been much debated in recent months, as it becomes clear how much of the US growth outlook depends on AI.
While some of the answers are easy, and most agree that even if we see a short-term correction, the long-term impact of AI will be far more widespread and profound than the rise of the consumer internet, I would like to point out three comparisons between the 1990s and today that are more worth considering for investors.
First, AI is very capital intensive. The dot-com bubble wasn’t like that. At that time, anyone could start a website with almost no investment. Today, investments in AI represent all of the growth in capital spending for American companies.
Contrarians might point out that from the mid-1990s to the end of the dot-com boom, more than 80 million miles of fiber-optic cable were installed, and much of that investment took years to pay back. U.S. carriers spent $444 billion on capital expenditures between 1996 and 2001.
Still, compare that to the $342 billion that top AI data center and computing infrastructure investors like Microsoft, Alphabet, Amazon, and Meta will spend in the US this year alone. At current electricity consumption rates needed to accelerate AI development, estimated investments will reach nearly $7 trillion by 2030.
This brings us to another important point on the capital investment side. That means energy is a huge constraint on the development of AI today. That wasn’t the case in the 1990s.
If the current energy economics of AI hold, the investment story will become even more costly and complex. That’s because it depends on the trajectory of grid upgrades, how more nuclear power comes online, how quickly renewable energy replaces fossil fuels, and how much oil prices rise amid these changes.
Another difference between the two eras is that today’s AI investments are funded with equity rather than debt. At first glance, that seems like a very good thing. While the $5 trillion in telecom investments during the dot-com bubble were mostly financed with debt, today’s big AI players are huge and profitable companies.
So the Magnificent Seven (and probably four in terms of AI) will not fail, so we probably won’t see a repeat of the failures of WorldCom and Global Crossing. The richest corporations in the history of global capitalism have cash to burn.
But this leads to the most important comparison between the 1990s and today.
Many market participants argue that the fact that the AI hype appears to be propping up the overall U.S. stock market shouldn’t be too worrying because the profits of the big tech companies driving the boom are so large.
But if you look under the hood, there’s more to it than that. Thorsten Slok, Apollo’s chief economist, is known for his ability to pinpoint the metrics that really matter, and a few days ago he released a remarkable chart. It showed that since President Donald Trump’s tariff “release day” in April, companies in the Russell 2000 with negative profits have outperformed those with positive profits.
This small-cap index includes many companies in sectors with strong AI investments, such as software, biotech, and healthcare. However, it also includes a number of companies in sectors that are similarly betting on an AI future, including finance, energy, materials, and communications.
This is an unpleasant nod to the history of the dot-com boom, where loss-making companies overtook profitable companies for a long period of time.
What should we take away from all this? A few things.
First, “AI will be a revolution in many ways, but we need to pay close attention to the influence of the 1990s,” Throck says.
Moreover, given that more Americans now own more stocks than ever before, even if the AI bubble bursts in the short term, the economic impact could be greater than it was during the dot-com era.
Second, the underlying dynamics of this bubble are much more complex than they were 20 years ago. Will productivity gains offset AI-related job losses? Will energy remain a challenge? Will Chinese innovators disrupt the market? Or will U.S. growth continue to outpace the world, as it did in the 1990s?
Possibilities beyond Silicon Valley depend on the answer.
rana.foroohar@ft.com
