The world’s 500 richest people, led by Nvidia Corp. co-founder Jensen Huang, lost a combined US$108 billion on Monday as a tech-led selloff tied to Chinese artificial intelligence developer DeepSeek sent major indices plunging.

Billionaires whose fortunes are linked to AI were the biggest losers: Huang saw his fortune fall US$20.1 billion, a 20 per cent drop, while Oracle Corp. co-founder Larry Ellison’s US$22.6 billion loss was larger in absolute terms, but represented just 12 per cent of his fortune, according to the Bloomberg Billionaires Index. Dell Inc.’s Michael Dell lost US$13 billion, and Binance Holdings Ltd. co-founder Changpeng “CZ” Zhao shaved US$12.1 billion.

Tech-sector titans as a group saw US$94 billion of wealth evaporate — roughly 85 per cent of the Bloomberg index’s total decline. The Nasdaq Composite Index fell 3.1 per cent, and the S&P 500 dropped 1.5 per cent.

Hangzhou-based DeepSeek has been developing AI models since 2023, but the company first came onto the radar of many Western investors this weekend as its free DeepSeek R1 chatbot app topped download charts worldwide. So many new users piled in that DeepSeek struggled to keep the app online, suffering outages and forcing it to restrict signups to users with Chinese phone numbers.

DeepSeek’s dark-horse entry into the AI race, which it says cost just US$5.6 million to develop, is a challenge to Silicon Valley’s narrative that massive capital spending is essential to developing the strongest models. That delivered a serious blow to billionaires whose fortunes are tied to the Western AI supply chain that’s been the equities market’s biggest driver over the past two years.

Similar playbook

Soaring valuations for so-called AI hyperscalers — including Meta Platforms Inc., Alphabet Inc. and Microsoft Corp. — have generated billions in wealth for their owners since OpenAI unveiled ChatGPT in November 2022. These companies have for the most part operated on a similar playbook: Spend huge sums to develop and run AI systems by hoarding top-of-the-line semiconductors and the energy supplies needed to run them.

Meta chief executive Mark Zuckerberg announced Friday that the company planned to spend US$60 billion to US$65 billion on projects related to AI this year, well above Wall Street estimates. Capital spending across all Big Tech firms is on pace to reach US$200 billion in 2025, according to a Bloomberg Intelligence report.

Despite limited revenue to show for all their investment so far, markets have rewarded U.S. tech stocks with record-high valuations, which have in turn generated historic wealth gains for their owners. Nvidia has stood out as the AI boom’s biggest single winner so far, with Huang’s net worth increasing almost eight-fold to US$121 billion since the start of 2023 through Friday. Zuckerberg’s fortune soared 385 per cent to US$229 billion over the same period and Amazon.com Inc.’s Jeff Bezos gained 133 per cent to US$254 billion.

While Huang and Ellison suffered losses, other major tech billionaires’ fortunes escaped unscathed. Zuckerberg’s net worth ended the day up, gaining US$4.3 billion as Meta rebounded from an early-session decline. Bezos’ wealth climbed by about US$632 million.

Capital spending

The fact that DeepSeek was able to develop a free model that potentially rivals or beats competitors including ChatGPT and Anthropic’s Claude at a fraction of the development cost led investors to question the logic behind Silicon Valley’s dependence on capital spending.

A key reason why DeepSeek didn’t rely on big investment and top-of-the-line chips to develop its model is that Chinese firms have had limited access to the powerful GPUs, or graphics processing units, most Western companies rely on ever since the U.S. government instituted strict export controls on the most advanced chips.

In an interview with CNBC last week, Alexandr Wang, CEO of training data provider Scale AI, said that despite the export controls, DeepSeek and other Chinese developers likely have more GPUs than Silicon Valley is aware of.

“The Chinese labs have more H100s than people think,” Wang said, referring to Nvidia’s top-of-the-line AI chip. “My understanding is that DeepSeek has about 50,000 H100s, which they can’t talk about, obviously, because it’s against the export controls that the U.S. has in place.”

—With assistance from Devon Pendleton and Tom Maloney.

Bloomberg.com