China’s start -up DeepSeek has been concerned about competitiveness in the US lead in the AI and the US technology company has been a free market trading.
Sharing of chip designers nvidiaThe huge beneficiaries of AI over -advertising fell 11.8% to 07:05 am before the market opened. Dutch -based chip company ASML and ASM International 8.9%and 13.6%in European trade, and Japanese chip -related stocks in Asia were very low.
Deepseek argued that it was developed in just two months at a cost of less than $ 6 million, starting with a free open source large language model at the end of December. Last week, the company announced a better reasoning model than Openai’s latest work in many third -party exams.
Development raised questions about the amount invested by a large company in the AI model and data center.
SRINI PAJJURI, a semiconductor analyst of Raymond James, said on Monday, “Deepseek could not access more computing as much as our hyperscalers and develop a competitive model.”
Pajjuri said in a positive sense that DeepSeek said, “It can cause more urgent among the hyper -scalers in the United States.” Amazon and Microsoft -If you access to the graphics processing device (GPU), it uses the advantage to differentiate it from cheaper options. The GPU is a key part of the infrastructure needed to train a huge AI model. NVIDIA is the market leader of the GPU.
CITI’s analysts said that the large model of DeepSeek said, “Inquiry inquiries.
Around computing costs. “
They say, “In the most advanced AI model, US companies can potentially be challenged.” “So we don’t expect leading AI companies to get away from more advanced GPUs.”
Donald Trump’s $ 500 billion Stargate AI project announcement was “nodded for the necessity of advanced chips.”
On the other hand, Bernstein’s analysts expressed doubt about whether the DeepSeek tool was actually built less than $ 6 million, and this figure “does not include all other costs related to prior studies and experiments on architecture, algorithms, or data.” I insisted.
They emphasized that “the model looks fantastic, but we don’t think they are miracles,” and said that the panic of the “death of death” of the AI infrastructure that we know is “exaggerated.”
-CNBC’s Lee Ying Shan and Michael Bloom contributed to this story.
This braking news article is updating.
