As AI startups continue to proliferate and the sector rallies the stock market, family office investors are increasingly wondering whether artificial intelligence is a bubble that’s about to burst.
This year, AI has been the one hot area in tech, climbing to $17.9 billion in startup investments in the third quarter while overall investments fell 31% during that time.
But some high-profile setbacks have also occurred.
Startup Olive AI Inc., which made AI-powered software to automate hospital administrative tasks, raised $848 million in venture capital before shutting down this month. And Jasper, a generative-AI darling that raised $125 million in a Series A round last year and helped fuel the AI funding fever, just cut its internal valuation by 20%.
Some prominent investors are cashing out.
Stanley Druckenmiller, whose family office profited from its shares in chipmaker Nvidia Corp., just sold about 75,000 shares for $37.2 million. Likewise, Soros Fund Management exited Nvidia in the third quarter, selling stock for $4.9 million. Some family offices, such as that of hedge fund billionaire David Tepper, are still confident, buying more Nvidia stock in the quarter.
The AI hype cycle’s rosy financial predictions and the role of a few big players like Nvidia in boosting the Nasdaq are exactly what you’d expect to see in a bubble, said Allyn Robins, a senior consultant at Brainbox, a think tank specializing in law, technology and policy. “There is a strong financial incentive for companies of every size to hype AI, to make investments in it and to claim that they're 'AI leaders,' " Robins said. "Because at the peak of a bubble, doing so can deliver quick boosts to their stock price.”
Other potential signs of a bubble cited by analysts who look at AI startups: high capital requirements to train their models, which can limit growth and innovation; hesitancy in adopting generative AI by companies; and legal, ethical and intellectual property concerns surrounding the data that AI systems are trained on.
Paul Hsu, the founder and CEO of Decasonic, a blockchain and AI venture fund, said many of his family office clients are asking questions about the AI hype cycle and whether it compares to the dot-com bubble of the late 1990s.
“There will be businesses that will be decimated and new upstarts that will help build the future,” said Hsu, whose firm makes up to 10 deals a year. “AI is moving so fast now, it’s a question from an investor standpoint: How fast can you iterate or pivot to these exponential changes? Those that can’t change fast enough will die a fast death.”
The comparison with the dot-com bubble is a natural one but unwarranted in the case of AI, said Beerud Sheth, the CEO and founder of Gupshup, a conversational messaging platform for commerce. “The Internet 1.0 bubble was caused by a number of factors, including overinvestment in speculative companies and unrealistic expectations about the growth of the Internet,” Sheth said. “However, the AI market is more mature, and there is more understanding of the challenges and opportunities associated with AI.”
Strategists such as Peter Oppenheimer at Goldman Sachs agree, noting that the valuations of the stocks leading the market are not as stretched as in previous periods, such as the internet bubble that collapsed in 2000. Today's companies have unusually strong balance sheets and returns on investment, he said. “We believe we are still in the relatively early stages of a new technology cycle that is likely to lead to further outperformance,” Oppenheimer wrote in a recent research report.
Maybe overhyped, but not a true bubble
It’s possible to be overhyped without being a true bubble, said Jake Miller, the co-founder of the private markets platform Opto Investments. “How can you avoid crossing over that line?”
Miller highlights a few red flags for family offices to look for when assessing AI startups, including some that involve more gut instincts and art than science:
- If you see that most of the activity is insiders selling to outsiders — “why are the people who know the most trying to get out?”
- If most of the purchases are on leverage, then any small downsizing can knock them out. “How do I have faith that you have the collateral to pay this back?”
- Do the founders need the world to change fundamentally for the value to be created? “I like to know how my investments can make money in the world as it exists,” Miller said.
- Does it solve real-world problems?
“When you put those things together on AI, it doesn’t check a lot of those boxes," he said.
"Has it replaced human beings? No. But is it useful? Absolutely,” said Miller, pointing to biotech, cybersecurity and infrastructure as areas that will be transformed.
In a way, family offices and other long-term pools of capital are the ones to invest in the space since there is risk involved, but they have more patience.
Decasonic's Hsu advises some of his clients, such as Jason Pritzker, to sort out the hype and look for opportunities that are truly innovative.
“There may be strong signals or faint signals," Hsu said. "With general-consensus faint signals, the returns aren’t as attractive. But nonconsensus faint signals have opportunity” and the potential to be transformative, with a larger vision to shape the industry and transform the tech architecture.
“Does this forward valuation match the potential in the vision you’re sharing?” he said.
Hsu said his firm sees potential in AI’s intersection with blockchain, such as solving for bias, and with mixed reality and the metaverse, such as deploying generative AI to build characters for virtual reality or augmented reality and other immersive experiences.