David Godes remembers his first year as a professor at Harvard Business School, when teenage graduates started dropping off like flies. It was 2000, the dawn of the contemporary Internet, and potential Harvard MBA graduates decided they would be better off starting and joining nascent Internet companies.
Little did they know that it was all a bubble of historic proportions.
“It was crazy,” recalls Godes, whose class of more than 100 students quickly shrunk to about 80 students that year. Even faculty have left academia to get caught up in the Internet craze. It was a “FOMO thing,” he said. “You know, I have to be a part of this. All my friends from college are part of startups.”
Investors ultimately he threw too much money into risky startups like Pets.com – pushing its shares well above levels justified by its business. Eventually it all came crashing down, and the bursting of the bubble caused trillions of dollars in market capitalization to be lost before the recession of the early 2000s.
Godes said today’s craze for generative AI is different. He currently teaches at Johns Hopkins, and his students aren’t going to Silicon Valley any time soon. He said they show a hearty skepticism towards emerging technology. That’s just one reason he sees the excitement around artificial intelligence as completely different from the early internet era.
Over the past year, generative AI has attracted multi-billion-dollar investor interest. In particular, companies producing AI hardware and software chip giant Nvidia, saw their shares skyrocket. This has led to skeptics warn against another technology bubble which will inevitably break. One economist said as much earlier this year AI craze has made companies even ‘more overvalued’ than in the delayed 1990s.
But for those on the side of the debate, the feeling of anxiety is short-sighted.
“When we first had the dotcom bubble… it was hype. This is not hype.” JPMorgan Chase CEO Jamie Dimon told CNBC in February. “That’s true.”
Hype about artificial intelligence and buzz on websites
“Generative AI is the most disruptive technology since the Internet,” said Gil Luria, an analyst at DA Davidson.
But Godes said there is a different level of skepticism about artificial intelligence than there was in the dot-com era. Much of the political and cultural discussion around AI is doom and gloom: State sponsored groups using it to interfere in elections, chatbots sending disturbing messagesartificial intelligence creating music that imitates real artists – and of course ongoing debate above or AI will take nations Job offers. (It’s complicated.)
“It’s more fear than delight,” Godes said.
People were skeptical about the Internet, too. But now, with the evolution of the Internet as a cultural reference point, concerns about the shortcomings of artificial intelligence are more defined. Rules all over the world, academic institutionsand even AI software companies they investigate potential risks with a level of scrutiny that didn’t exist in the 1990s.
The buzz on Internet portals was so great that in the spring of 1999 one in 12 Americans surveyed said they were in the process of setting up a company. The bubble has started developed in the mid-1990s and exploded in 2000. There was a massive influx of cash into Internet-related technology companies as global interest in personal computers and the World Wide Web exploded. Everything happened as it happened in the US is experiencing the longest period of economic expansion since after World War II era.
Internet companies including Priceline, Pets.com and eToys have gone public, attracting investors who have driven their market value to novel heights – all while ignoring the shaky fundamentals of their businesses. Banks had a lot of cash when the Fed printed money in 1999, and they were pushing it onto the same websites. This fall, the capitalization of 199 Internet companies monitored by Morgan Stanley was the same valued at a total of USD 450 billion — even though their actual companies lost a total of $6.2 billion. Pets.com went bankrupt less than a year after going public.
“In the delayed 1990s, there were websites that just didn’t make sense,” Godes said. “There was nothing complicated about this technology.”
AI startups are different, he said, because the technology is quite complicated.
“It is more arduous for an MBA student without technical training to prepare a business plan, jump in and get started [an AI] business,” he said.
“AI bubble” or “AI hardware bubble”?
Gil Luria of U.S. Attorney Davidson doesn’t even agree with using the word “bubble” to refer to artificial intelligence. Assets can inflate and enter a bubble, he said, while the underlying technology goes through cycles. Like any novel technology, artificial intelligence may be in the “hype” phase. But that doesn’t mean all AI companies are overvalued, Luria said.
Luria said there is a significant difference between the rally in AI hardware stocks and the rally in AI software maker stocks. Although the share prices of several companies, especially Microsofthave gained great benefits from artificial intelligence, it is thanks to AI software actually increased its profits — unlike the websites of the Internet boom. Today’s AI software stocks continue to “trade at reasonable prices within historical limits [price] multiples,” Luria said. (In other words, even though Huge Tech stock prices are much higher than they used to be, theirs price to earnings, sales and free cash flow ratio are not radically different.) These companies’ software will continue to boost sales for years to come.
However, “equipment is a one-time sale,” he said, so “the bigger disappointment may be in hardware inventory.” Luria compared AI chip giant Nvidia to Cisco Systems, the company whose products helped build the early infrastructure of the Internet and whose explosion defined the dot-com era. Luria said Nvidia chips are to artificial intelligence what Cisco networking equipment was to the early Internet.
“By 1999 and 2000, we had enough tools. We had enough equipment, fiber optics and routers to support the growth of the Internet in the coming years,” Luria said. “And we believe that’s where we are right now. … By the end of this year, Microsoft, Amazon, Google and the like will have had enough [AI chips]”
Cisco shares fell 80% in 2001–2002 as revenues fell below expectations as demand for networking equipment fell to record levels. As with Cisco, Luria said, demand for AI hardware will not continue at a breakneck pace.
“If investors are hoping for a continuation of the current pace of growth in hardware that supports AI development,” he said, “they may be disappointed.”
He pointed Nvidia’s largest customers creating their own AI chips. Just this month Google AND Meta, two of Nvidia’s five biggest buyers, have released the latest versions of their own custom AI chips. While Meta’s doesn’t support its AI apps yet, Google’s AI Gemini chatbot runs on its novel chip. With Nvidia’s five largest customers generating two-thirds of its revenue, Huge Tech is moving its AI hardware internally could seriously harm Nvidia’s results– Luria said.
Even bubble supporters believe that artificial intelligence is not an internet portal
Even among experts who see an artificial intelligence bubble forming, many say it won’t end as badly as the dot-com burst. Richard Windsor, founder of research firm Radio Free Mobile, said people are using “complicated and untested methods to justify very high valuations [AI] companies” as it was in the dot-com era.
But, as he said, “the bursting of the Internet bubble [was] worse than the bursting of the AI bubble.” This is partly because even in its “immature” form today, artificial intelligence is capable of generating much more revenue than the Internet did in the 1990s and early 2000s. The Internet in the 1990s was “very ponderous,” he said, “and it took a long time to realize” its full potential. Windsor, meanwhile, said he sees AI’s full potential as ultimately circumscribed. Even if the AI bubble bursts, “what the Internet will become will be greater than what AI will become in its current form,” Windsor said.
Windsor said one reason he sees AI models as ultimately circumscribed is that machines can’t distinguish causation from correlation.
“Because of this, they will never really get to the point where they are superintelligent because they can’t reason,” Windsor said.
Windsor said he doesn’t know when the AI bubble will burst, but there are signs to look for, including price erosion or when a product’s price drops over time due to customer demand and competition. Windsor said he’s already seeing signs of price erosion starting to intensify. These signs include OpenAI allowing users to exploit products without an account, which Windsor said looks like the company is trying to attract more users, and a search engine Embarrassment AI starts selling ads despite previously stating that search should be “free from the influence of ad-based models” – which Windsor sees as a signal that its monetization is not going well. He also pointed to research showing that enormous companies are cautious about implementing generative AI, largely due to security concerns.
“There is a general expectation in the market right now that AI will be available soon,” Windsor said. “I respectfully disagree with that statement.”
Luria predicts Nvidia stock will return to Earth in 12 to 18 months.
“We may not see the peak of the hype until next year,” he said. “But when we do, a lot of people will be very disappointed.”
