January 28, 2025
A little-known Chinese business called DeepSeek has shocked the world of technology by releasing an artificial intelligence (AI) model that can compete with Google and OpenAI's innovations. The developer of DeepSeek-R1 claims that its model was created with fewer and less sophisticated computer processors than those used by American IT companies.
Less than $6 million was spent on computer power to train the model, according to a research article published last week by the model's development team. This is a small portion of the multibillion-dollar AI budgets enjoyed by US tech giants like OpenAI and Google, who created ChatGPT and Gemini, respectively. Assumptions regarding US superiority in AI have been called into question by the unexpected rise of a small Chinese business that can compete with Silicon Valley's leading firms. This has also sparked concerns that the exorbitant market values of companies like Nvidia and Meta may not be grounded in reality.
The company that produces the semiconductors that drive generative AI, Nvidia, lost about $600 billion in market capitalization on Monday 27th January,2025 as its shares fell 17%..
Information about DeepSeek –
Hangzhou-based DeepSeek was established in late 2023 by Liang Wenfeng, a seasoned businessman who also manages the hedge fund High-Flyer. Liang has a long history of fusing emerging technologies with investment, despite its lack of recognition outside of China. According to Chinese news agency Sina Finance, he and a fellow Zhejiang University alumnus co-founded Hangzhou Jacobi investing Management in 2013, an investing company that used AI to execute trading methods. In 2015 and 2016, Liang founded two other computer-directed investment-focused businesses: Ningbo Huanfang Quantitative Investment Management Partnership and Hangzhou Huanfang Technology Co.
Liang rejected the notion that AI should be viewed as unaffordable or that it was too late for companies to engage in the technology in a 2023 interview with the Chinese media source Waves. “Reproduction on its own is reasonably priced; based on open-source code and public papers, it requires little training or even fine-tuning. But research requires a lot of comparisons, trials, and more expertise and computing power," Liang stated, as reported in a translation of his remarks by the ChinaTalk Substack.
Why has DeepSeek become so popular in the tech industry? –
In other words, the company's success has sparked existential concerns about the US government's and Silicon Valley's approaches to AI. It has been commonly believed that US IT companies have a significant advantage in AI, in part due to their size, which enables them to attract top talent from around the globe and make significant investments in the construction of data centers and the acquisition of vast amounts of expensive high-end chips.
The US government's approach of limiting shipments of high-end chips in an attempt to curb China's AI advancements is called into question by DeepSeek's entry into the market.
According to DeepSeek's study paper, Chinese companies may still get chips in sufficient quantities, or the most sophisticated processors are not required to produce high-performing AI models.
Although most tech observers concur that DeepSeek-R1 performs on par with ChatGPT, or possibly better for some tasks, the sector is developing quickly.