The Chinese model of artificial intelligence, DeepSeek, is on the verge of changing the belief that the development of AI will require massive investments, vast computing power housed in energy-consuming data centers, and that this race will be won by America. It is now a battle between superpowers, in which China has just taken the lead.
The latest version of the Chinese artificial intelligence model developed by the Chinese tech startup DeepSeek, which appeared on the Apple and Google Play app stores a week ago, has demonstrated capabilities seemingly equal to its more well-known and much more expensive rivals, led by ChatGPT, owned by the American company OpenAI.
In a research paper published last year, DeepSeek showed that the model was developed using a "limited capacity" of Nvidia chips (the most advanced technology was banned in China under export controls from 2022 - ed.), and the development process cost only $5.6 million.
Its success seems to pose a fundamental challenge to the established idea that the development of AI will require massive investments, vast computing power housed in energy-consuming data centers, and that this race will be won by America, as stated in an analysis published by Sky News.
Last week, Donald Trump announced an investment project in AI of up to hundreds of billions of dollars. OpenAI, Google, Meta, Microsoft, and the ubiquitous Elon Musk are all in this race, desperate to be the first to find the Holy Grail of artificial general intelligence - a theoretical concept that describes the ability of a machine to learn and understand any intellectual task that a human can perform.
Now this race is taking place not only among the tech giants of California with macho budgets but also among the superpowers of the planet.
The implications have been clear on the US stock markets. The launch on January 10 of DeepSeek's AI assistant, powered by the DeepSeek-V3 model, as well as the launch on January 20 of its R1 model, shocked Silicon Valley and led to the collapse of technology company stocks, with the presumed low development and usage costs of the Chinese startup prompting investors to question the huge spending plans of the main AI firms in the United States.
The Nasdaq stock index, based on the technology sector, dropped by 3%, a loss of $1 trillion, and Nvidia lost nearly 17% of its stock value, reducing its market value by $600 billion and losing its status as the largest company in the world. Google's parent company lost $125 billion in market value, and Microsoft, which holds a stake in OpenAI, lost $8.7 billion, as detailed by Sky News.
DeepSeek, which has developed two models, V3 and R1, is now the most popular free app on the Apple App Store in the US and the UK.
In addition to questions about the cost and capacity of American models, all these financial losses also demonstrate investors' desperation to bet on the winner in the race for arguably the most important "general-purpose technology" since the discovery of electricity. It seems that AI will change the world, but no one can say for sure how, when, or in what way.
As Morgan Brown, Vice President of Product and Growth in Artificial Intelligence at Dropbox, said, it is currently "insanely expensive" to train top AI models. "DeepSeek just showed up and said: What if we did this for $5 million?. And not only did they say it, but they actually did it. Their models match or surpass GPT-4 in many tasks. The AI world is (…) shaken," he wrote on X.
There are many aspects about DeepSeek that we still don't know, for example, how reliable this development figure is.
Moreover, while established models in the United States have "hallucinations," inventing facts, DeepSeek seems to have selective memory. Ask it about Tiananmen Square or other censored issues and events in China, and you will see that it cannot help you, as stated in the cited analysis.
Some have expressed reservations about the Chinese company and the manipulation of user data. American billionaire Bill Ackman, a hedge fund manager, has labeled DeepSeek as "a Trojan horse" and said that, in comparison, TikTok, which was temporarily banned in the US earlier this month due to national security concerns, "is just a toy."
This denotes broader concerns about the role of Chinese technology, which have prompted US authorities to call for the banning of TikTok and the British government to remove Huawei technology from the UK's communications network.
Addressing House Republicans on Monday, Donald Trump called the development of the Chinese AI model "a wake-up call for our industries that we need to focus on competition to win."
And DeepSeek is just the beginning of this game that China is taking to the next level. The Chinese technology company Alibaba launched a new version of its artificial intelligence model, Qwen 2.5, on Wednesday, which it claims surpasses the DeepSeek-V3 model.
"Qwen 2.5-Max outperforms almost across the board, GPT-4o, DeepSeek-V3, and Llama-3.1-405B," Alibaba's cloud unit said in an announcement posted on its official WeChat account, referring to the most advanced open-source AI models from OpenAI and Meta.
The predecessor of the DeepSeek V3 model, DeepSeek-V2, triggered a price war among AI models in China after its launch in May of last year. And DeepSeek's success has sparked China's "tech frenzy," leading to a battle among its national competitors to update their own artificial intelligence models.
T.D.