Young Chinese AI startup DeepSeek said Monday that its highly competitive, and potentially surprisingly cost-effective, model is raising questions about the hundreds of billions of dollars that America’s biggest companies are spending on artificial intelligence. This stirred up a huge crash in US technology stocks.
DeepSeek’s arrival is shaking investors’ faith in the AI narrative that has powered the U.S. bull market for the past two years. That cast doubt on the hype surrounding Nvidia’s chips and sent ripples through the market for power company stock prices, which should have been boosted by demand for AI data centers.
Here’s how the market crash on Wall Street caused by DeepSeek unfolded.
new inference model
DeepSeek was founded in May 2023 by Liang Wenfeng, whose company was partially funded by an AI-powered hedge fund. In late December, AI developers launched a free, open source, large-scale language model that they say took just two months to develop and cost less than $6 million to build.
On January 20, Hangzhou, China-based DeepSeek released its inference model R1, which beats Open AI’s latest o1 in many third-party tests.
DeepSeek differentiates itself from its competitors with its inference capabilities. That is, before providing a final answer, the model first generates a “chain of thought” to improve the accuracy of the response.
highest performance model
The buzz surrounding DeepSeek’s R1 comes after ScaleAI CEO Alexandr Wang touted its competitiveness over the best products of the US mega-cap tech giant, which was considered by most to be leading the AI wars. , appears to have gained momentum.
“What we found is that the leading AI lab in China, their model, Deep Seek, actually performs the best or is almost as good as the best models in the United States.” said Wang, whose company provides data to help train AI tools, on CNBC from last week’s World Economic Forum in Davos, Switzerland.
Wang pointed out that DeepSeek actually has more Nvidia H100 chips than expected, about 50,000 of them. These chips are the processor of choice for U.S. AI companies such as OpenAI, and the U.S. bans the sale of advanced AI chips to China.
NVIDIA stock fell 3% on Friday as buzz around DeepSeek started to pick up steam.
No.1 app
But it was over the weekend that the hype surrounding Deep Seek reached a fever pitch on social media.
Marc Andreessen, co-founder and general partner at venture capital firm Andreessen Horowitz, said the R1 model was “one of the most amazing and impressive advances I’ve ever seen,” and said that DeepSeek praised X. Andreessen’s portfolio includes Airbnb and dozens of AI companies.
Chamath Palihapitiya, a tech investor at , we have essentially broken through one of the holy grails of AI.”
Chinese AI app DeepSeek in Apple’s US App Store on iPhone 12. In the ranking of free apps, DeepSeek was even ahead of OpenAI’s ChatGPT.
Picture Alliance | Getty Images
At this point, DeepSeek’s mobile app shot to the top of Apple’s app store download charts in the US over the weekend, making it a clear threat to more expensive models such as OpenAI’s ChatGPT.
U.S. futures fell sharply on Sunday night, and investors woke up Monday morning to a sea of red.
Stock chart iconStock chart icon
Nasdaq Composite, 1 day
Shares of Nvidia, the AI darling and chipmaker, fell more than 12% on Monday, marking its worst day since March 2020.
Stock chart iconStock chart icon
Nvidia, 1 day
Other chipmakers and power companies were also hit hard. The tech-heavy Nasdaq Composite Index fell as much as 3.6% on Monday, dragged down by mega-cap stocks.
