When ChatGPT burst onto the global scene, a pertinent question arose: did it pose a threat to China, the United States' main technological competitor?
Two years on, a Chinese AI innovation model has instead asked the question: can American progress be stifled by China.
For a while, Beijing seemed to struggle with its response to ChatGPT, which is not accessible in China.
Baidu's chatbot, Ernie, was met with skepticism by users, who made fun of it. The technology companies Tencent and ByteDance then launched their own chatbots, but they were viewed as merely copying the format of ChatGPT and not improving upon it.
The administration expressed confidence that it held a leading position and wanted to maintain its status, prompting the Biden administration to strengthen restrictions on exporting advanced chips and technology to China.
That is the reason DeepSeek's launch has astounded Silicon Valley and the world. The company claims its strong model is significantly more affordable than the billions of dollars US companies have spent on artificial intelligence.
So how did a relatively unknown company – whose founder is being celebrated on Chinese social media as an "AI hero" – achieve this feat?
The challenge
When the United States restricted the world's top chip manufacturers, including Nvidia from selling sophisticated technology to China, it was a significant setback.
Those chips play a crucial role in creating advanced AI systems capable of carrying out a spectrum of tasks that typically require human intelligence, encompassing everything from providing straightforward answers to resolving intricate mathematical problems.
DeepSeek's founder Liang Wenfeng identified the chip ban as the company's "main challenge" in interviews with local media.
Prior to the ban, DeepSeek amassed a "substantial stockpile" of Nvidia A100 chips - estimated to range from 10,000 to 50,000 - according to MIT Technology Review.
Leading AI models in the West use approximately 16,000 specialized chips. But DeepSeek claims it trained its AI model using 2,000 specialized chips, and numerous lower-grade chips - which makes its product more cost-effective.
Certain individuals, including Elon Musk, a well-known US tech entrepreneur, have expressed skepticism about this claim, suggesting that the company may not be able to freely disclose the exact number of advanced chips it has used due to prevailing regulations.
According to authorities, Washington's ban has created both hurdles and possibilities for the Chinese artificial intelligence sector.
It has "forced Chinese companies like DeepSeek to innovate," so they can do more with less, says Marina Zhang, an associate professor at the University of Technology Sydney.
Achieving technological self-sufficiency is a strategic objective that has led China to innovate and adapt in the face of limitations, propelling the country's progress in multiple spheres.
The world's number two economy has made significant investments in leading-edge technology - including the batteries that power electric vehicles, solar panels, and artificial intelligence.
Turning China into a global technological powerhouse has been President Xi Jinping's goal for a long time, so the restrictions imposed by Washington were also a challenge that Beijing took up.
The release of DeepSeek's new model on 20 January, when Donald Trump was inaugurated as US president, was intentional, according to Gregory C. Allen, an AI expert at the Center for Strategic and International Studies.
"The timing and the manner in which this is being communicated is precisely what the Chinese government wishes everyone to believe - that export controls are ineffective and the US is no longer the global leader in AI technology," says Mr. Allen, former director of strategy and policy at the US Department of Defense's Joint Artificial Intelligence Center.
In recent years, the Chinese government has fostered AI talent, offering scholarships and research grants, and promoting partnerships between universities and industry.
According to Ms Zhang, the National Engineering Laboratory for Deep Learning and other state-backed initiatives have aided the training of thousands of AI specialists.
And China had a plentiful pool of skilled engineers to draw upon.
The talent
For example, take DeepSeek's team - Chinese media reports that it consists of less than 140 people, majority of whom are talented individuals from elite Chinese universities, referred to by the internet as "home-grown talent".
Western observers have overlooked the emergence of a new generation of entrepreneurs who prioritize foundational research and long-term technological progress over rapid financial gains, according to Ms. Zhang.
China's top universities are creating a "rapidly growing pool of AI talent" where even senior managers are often under the age of 35.
"They've grown up alongside China's rapid technological development, which has imbued them with a strong motivation for self-sufficiency in innovation," she adds.
Liang Wenfeng, the founder of Deepseek, is a prime example of this. At 40 years of age, he studied AI at Zhejiang University, a prestigious institution. As described in an article on 36Kr, those familiar with him describe him as "more like a geek than a boss".
Chinese media refer to him as a "technical idealist," attributing this label to his commitment to maintaining DeepSeek as an open-source platform. In reality, experts agree that a thriving open-source culture has enabled young start-ups to collaborate and progress more rapidly.
Unlike larger Chinese technology companies, DeepSeek prioritized research, which has enabled a more experimental approach, according to professionals and individuals who worked at the company.
"We may not have the top 50 talents in this field already here in China, but we can definitely train people to reach that level," Mr. Liang stated in an interview with 36Kr.
However, experts ponder how far DeepSeek can extend. Ms. Zhang notes that "further US limitations may impede access to American user data, possibly affecting global perspectives for Chinese models akin to DeepSeek".
Some argue the US still holds a significant advantage, as Mr Allen put it, "their vast quantities of computing resources" are substantial - and it is unclear how DeepSeek will continue to advance its model using cutting-edge chips.
But for now, DeepSeek is basking in its newfound attention, given that most people in China only learned about it over the weekend.
The new AI heroes
His sudden rise to fame has turned Mr Liang into a sensation on China's social media, where he is being welcomed as one of the "three AI heroes" from southern China's Guangdong province, bordering Hong Kong.
The other two are Zhilin Yang, a leading specialist at Tsinghua University, and Kaiming He, an educator at MIT in the United States.
DeepSeek has pleased the Chinese internet before the Lunar New Year, the country's largest holiday. It's good news for a struggling economy and a tech industry that is preparing for further tariffs and the potential sale of TikTok's business in the US.
A Weibo comment with many likes says 'only if you are true will you make it through time'.
"This is the best New Year's gift. I wish our motherland is prosperous and strong," another reads.
A "flurry of shock and interest, especially within the open-source community," is how Wei Sun, a principal AI analyst at Counterpoint Research, described the reaction in China.
Fiona Zhou, a tech worker in the southern city of Shenzhen, says that her social media feed was suddenly flooded with DeepSeek-related posts yesterday.
They call it 'the glory of made-in-China', and say it stunned Silicon Valley, so I downloaded it to see how good it is.
She asked it for "four pillars of her destiny", or ba-zi - like a personalized horoscope based on the date and time of birth.
To her dismay, Deepseek was mistaken. Though she received a comprehensive explanation of its "thought process," it did not contain the "four pillars" of her actual ba-zi.
She intends to give it another try at work, as she believes it will be more productive before tasks of that nature.