What exactly is DeepSeek and what is the reason it is changing AI?

DeepSeek is a Chinese startup announcement of the latest AI models that, according to the company, are as good or even better than the top models that are available in the United States at a fraction of the price could upset the order of the world of technology.

The company has gained attention from global AI circles following the publication of an article in the last month stating that the training of DeepSeek-V3 needed just $6.9 million of computing power using Nvidia H800 processors.

The DeepSeek AI Assistant, powered by DeepSeek-V3, has outperformed its rival ChatGPT and has become the highest-rated free app on Apple’s App Store across the United States.

The issue has prompted questions regarding the rationale behind U.S. tech companies’ decision to commit billions of dollars for AI investments and shareholdings of a number of major tech companies such as Nvidia were slashed.

Here are some details about the company that is shaking things up in the AI sector across the globe.

WHY IS DEEPSEEK CAUSING A STIR?

The release of the OpenAI’s ChatGPT in the latter half of 2022 resulted in a flurry of Chinese tech companies which rushed to design their own chatbots based on artificial intelligence.

After the announcement of the initial Chinese ChatGPT equivalent, made by the search engine giant Baidu (9888.HK) There was a general displeasure in China at the lack of AI capability in between U.S. and Chinese firms.

The cost-efficiency and quality of DeepSeek’s models has turned the narrative upside down. Two designs that were lavished with praise from Silicon Valley executives and U.S. tech industry engineers as well, DeepSeek-V3 as well as DeepSeek-R1 are in line with OpenAI and Meta’s top models as the Chinese startup said.

They’re also less expensive to utilize. The DeepSeek R1, which was launched last week, is between 20 to 50 times more affordable to run than the OpenAI the o1 model, based on the application as per an article on the DeepSeek official WeChat account.

However, some have expressed doubts about DeepSeek’s success story.

Scale AI Chief Executive Officer Alexandr Wang said during an interview with CNBC on Thursday, but without providing any evidence, that DeepSeek owns the 50,000 Nvidia H100 chips. He said he said would not be disclosed as it would be in violation of Washington’s export restrictions which ban these high-end AI chip from ever being offered to Chinese firms. DeepSeek has not yet responded on a demand for clarification regarding the allegations.

Bernstein analysts reported on Monday in a note of research that DeepSeek’s total cost of training for the V3 model are not known however they were significantly greater than $5.58 million that the company claimed was allocated to computing power. Analysts also pointed out that the cost of training for the likewise praised R1 model were not revealed.

WHO IS BEHIND DEEPSEEK?

DeepSeek is a Hangzhou-based company who’s controlling shareholder is Liang Wenfeng, co-founder of the hedge fund that is quantitative High-Flyer that is which is based on Chinese corporate information.

Liang’s fund made an announcement in March 2023 via their official WeChat account that it was “starting again”, going beyond trading and focusing its resources in the formation of an “new and independent research group, to explore the essence of AGI” (Artificial General Intelligence). DeepSeek was founded later that year.

ChatGPT creators OpenAI identify AGI as systems with autonomous capabilities that outperform humans in the majority of economically profitable tasks.

It’s unclear the amount High-Flyer has poured into DeepSeek. High-Flyer is headquartered inside the same space as DeepSeek and holds patents on clusters of chips used to create AI models, as per Chinese company records.

High-Flyer’s AI department announced in its official WeChat account in July 2022 that it owned and manages an A100 cluster comprising 10,000 chips.

HOW DOES BEIJING VIEW DEEPSEEK?

DeepSeek’s growth has been noted in China’s most prestigious circles of power. On the 20th of January on the day that DeepSeek-R1 was made available to the general public the the company’s founder Liang was a guest at a private symposium for business professionals and experts, which was held by Chinese premier Li Qiang, according to the state-owned information agency Xinhua.

Liang’s presence at the event could be a signal that the DeepSeek’s achievements could be crucial to Beijing’s goals of breaking Washington’s export control and becoming self-sufficient in strategically important industries such as AI.

Similar to the one last year, it included Baidu Chairman Robin Li.