The competition for dominance in artificial intelligence was blown off on Monday after the release of Chinese chatbots was wiped out $ 1 TN from a major US technology index.
In addition to the emergence of Openai and ChatGPT tool competition, DeepSeek, the investor punished high -tech stocks around the world on Monday, shaking the religion of the US artificial intelligence boom, and providing the same performance in resources.
TECH HEAVY NASDAQ COMPOSITEX index has slightly exceeded 3 % in the afternoon transaction in New York, indicating a loss of about $ 10 from the 32.5 TN closing price last week.
NVIDIA, a major computer chip manufacturer that supplies power to AI models, has been overtaken by Apple as the most valuable listed company in the United States after the fall of 17 % and wiped nearly $ 600 million from the market value. 。 Google's parent company has lost $ 100 million and lost $ 7 billion in Microsoft.
NVIDIA's fall was the largest in the history of the US stock market.
The DeepSeek AI Assistant was also the top of the U.S. and the UK Apple App Store on the weekends on Openai's Chatgpt.
Deepseek claims that it has developed a model using chips, which is less rival, and argued that it has been raised and raised questions within the scope of billions of dollar AI spending by US companies that have been boosting the market in recent years. I am doing it.
According to research papers published in December, the company has developed a custom -made algorithm for building models using the low capacity H800 chips generated by NVIDIA.
NVIDIA's most advanced chips, H100, have been banned from exports to China by September 2022. Subsequently, NVIDIA developed a powerful H800 chip for the Chinese market, but was banned from exports to China last October.
Deepseek's success of the highest AI model that cannot access the most advanced US technology has raised concerns about the effectiveness of Washington's STYMIE high -tech accessor.
Mark and Leisen, a major US venture capitalist, compared the launch of DeepSeek's R1 model last Monday with an extremely important moment in the US space race, and X's AI's “SPUTNIK moment”. I posted it. A rival of the Cold War by introducing satellites into track.
According to DeepSeek, the R1 model exceeds Openai's O1-mini model with “various benchmarks”. Research by artificial analysis In terms of overall quality, it is placed on a model developed by Google, Meta, and humanity.
The company was established by Liang Wenfeng, an entrepreneur who operates HIGH-FLYER CAPITAL, a hedge fund that uses AI to identify stock prices patterns. LIANG began purchasing NVIDIA CHIPS in 2021, began developing AI models as a hobby, and has been banked by hedge funds. In 2023, he established DeepSeek, a base in Hangzhou, eastern China.
The company is purely focusing on research, not commercial products. Deep Sep and the basic code can be downloaded for free, but DeepSeek models are cheaper than O1 of OPENAI.
In an interview with a Chinese media, Liang said, “AI is affordable and can be accessed by anyone.” LIANG also stated that the gap between us and China was only one or two years.
DeepSeek development questions the need for a large amount of investment in AI infrastructure such as chips and the need to lead the role of leading the market for high -tech companies in the United States in AI. I'm threatening to be there.
After the newsletter promotion
Deepseek claims that R1 costs $ 5.6 million, but warns that experts may be underestimated compared to much higher estimated values of Western development models. Last year, DARIO Amodei, a co -founder of major AI corporate and humanity, set the current advanced model of the current training cost from $ 100 to 1 billion.
The US investment bank's Goldman Sachs analysts raised the alarm more than last year's AI spending. We will publish a note In June, there is a title, “GEN AI: Do you have too much expenditure, do you have too little profit?”
He asked if the $ 10 investment on AI in the next few years would be “rewarded” and expressed concerns about the benefits of expenditures contrasted by DeepSeek.
The Pan European Stoxx 600 fell on Monday, and the major European technical stocks have decreased. Dutch chip maker ASML slides 7 %, but German Siemens energy, which provides hardware on AI infrastructure, has decreased by nearly 20 %, and the French digital automation company Schneider Electric has decreased by 9.5 %.
Following the loss in Asia, Japanese chip company disco and advanced tests (NVIDIA suppliers) were reduced by 1.8 %, respectively, 8.6 %, respectively.
Richard Hunter, the director of the platform interactive investor, states: With major indexes for the past few years.
“It was suddenly a big problem to see if the investment of hundreds of billions of dollars in AI needed to be re -evaluated.”
Andrewdancans, the director of the basic AI of the Aranchugaling Research Institute in the UK, Deep Shek Development is a “really exciting access” to an advanced AI model by becoming an open source developer. I said. Following the pass, a meta with a mark suckerberg lama model continues.
“Academia and the private sector can play, explore it, and use it as a launch,” he said.
Duncan added: “It indicates that it can be surprising with relatively small models and resources, indicating that it can be innovated even if you do not have a large -scale Openai resource.”
