Inside China, it was called the tipping point for the global technological rivalry with the United States and the “darkest hour” in Silicon Valley, evoking Winston Churchill. It was possibly a breakthrough that could change the country’s destiny.
The news that the Chinese start-up DeepSeek can build artificial intelligence models that are as good as OpenAI’s, and at a fraction of the cost, tanked the stock market on Monday and sent Silicon Valley into a panic.
The claim about DeepSeek’s success was viewed in China as a shot in the arm for a discouraged tech industry and a public that’s suffering through a stagnating economy. On social media posts and state news outlets, DeepSeek was nothing less than a testament to the country’s ability to innovate, especially when facing efforts by the United States to limit China’s access to the most advanced technologies.
“A nation like China, which is equipped with substantial technological resources, cannot truly be suppressed,” wrote Hu Xijin, a retired editor in chief of the Communist Party tabloid Global Times. “U.S. sanctions in one area will only spur more comprehensive and resilient progress in China, potentially leading to breakthroughs that outpace the U.S.”
The American semiconductor policies toward China “may ultimately backfire on the U.S.,” he wrote.
On Monday evening, four out of the 10 most popular topics on the social media platform Weibo were related to DeepSeek.
“DeepSeek, keep up the momentum!” a Weibo user in Beijing wrote.
“The nation must protect the founder of DeepSeek at all costs! Seriously!” wrote another user in Shanghai who usually posts about entertainment news.
Even a hashtag about the DeepSeek chief executive, Liang Wenfeng, visiting his hometown in southern Guangdong Province for the Lunar New Year, which falls on Wednesday, was a hot topic on Weibo. It had more than 50 million views.
Much of the outpouring of attention emphasized the U.S.-China tech rivalry.
The assumption that the United States would lead the next wave of the technological revolution was now open to challenge, Li Chengdong, an e-commerce investor, wrote on his WeChat timeline.
Fancaiju, a business blog on WeChat, had a post saying that DeepSeek had burst the U.S. stock bubble in one fell swoop — a more significant strike than when George Soros bet against the British pound in 1992.
The DeepSeek breakthrough had turned the $100 billion A.I. initiative known as Stargate that President Trump announced last week into “Interstellar Graveyard,” said a post on Fancaiju.
The commentariat took immense pride that DeepSeek was stocked with talented Chinese technologists educated in China.
DeepSeek dispelled the myth of the dominance of American A.I. talent and demystified companies like OpenAI, said Tom Zhang, a human resources expert who has worked at several big tech companies in Silicon Valley. “A group of ‘homegrown Ph.D. graduates’ from Tsinghua and Peking University outshines their counterparts from Stanford and MIT,” he wrote on his WeChat timeline.
Entrepreneurs and investors said that DeepSeek demonstrated that China’s A.I. sector had an edge in innovation. They also said they believed that the U.S. government’s export restrictions on specialized chips from the Silicon Valley tech giant Nvidia forced Chinese companies to be more efficient.
That was not what Washington had intended, people in China said, but that was what happened.
DeepSeek trained its A.I. chatbot with 2,000 specialized Nvidia chips, compared with as many as the 16,000 chips used by leading American counterparts. It’s also not the only Chinese company to prove the efficiency of its engineering: 01.ai, a startup founded by Kaifu Lee, a Beijing investor and entrepreneur, trained its A.I. models with computing power that cost about $3 million, the company said, compared with the $80 to $100 million OpenAI has tapped.
“In my book AI Superpowers, I predicted that US will lead breakthroughs, but China will be better and faster in engineering,” Mr. Lee, who studied artificial intelligence at Carnegie Mellon in the 1980s, wrote on X on Sunday. “With the recent DeepSeek releases, I feel vindicated.”
Among the most popular articles on the Chinese internet were two interviews of Mr. Liang, the reclusive chief executive, with a tech blog.
In the interviews, Mr. Liang, who founded a quantitative stock trading firm called High-Flyer after graduating with a master’s degree in artificial intelligence, came across as a geeky billionaire full of idealism and optimism. He started DeepSeek as a side project in 2023 because he wanted to explore the limits of A.I., he said.
Mr. Liang said he believed that innovation was, first and foremost, a matter of belief.
“Why is Silicon Valley so innovative? It starts with daring to try,” he said. Mr. Liang noted that when OpenAI’s ChatGPT came out, China was suffering from a lack of confidence to pursue such innovation. “From investors to major tech companies, many felt the gap was too wide,” he said.
As China’s economy develops, he said, China should gradually become a contributor to tech innovation, rather than a follower.
He said he believed that China’s economic slowdown wasn’t necessarily a bad thing because it could force company founders to be driven less by financial success.
“When many people realize that making quick money in the past was likely due to the luck of the times,” Mr. Liang said, “they will become more willing to focus on genuine innovation.”
Mr. Liang told the blog that he had hired mostly young graduates or even graduate students with little work experience. Every team member who worked on an A.I. model that was released last spring graduated from a Chinese university, he said.
“The top 50 talents might not currently be in China, but perhaps we can cultivate such talent ourselves,” he said, a quote that has been reposted many times. People on social media portrayed DeepSeek employees as geniuses, posting their names and citing their educational background and academic papers.
Mr. Liang, born in 1985, didn’t attract much public attention until last week when he joined a group of businesspeople and academics for a meeting with Li Qiang, China’s premier.
The meeting was a sign that Mr. Liang had risen to the top, but it could also put him in an awkward position, tech executives said. The relationship between Chinese entrepreneurs and the government has been tricky after Beijing’s crackdown on the tech sector in recent years. The government wants the companies to help make China a tech power less reliant on the United States. But it’s also wary of the companies’ influence. The crackdown crushed the promise of the Chinese internet.
A Hong Kong investor told me that he would not invest in Chinese internet stocks as long as Xi Jinping, the Chinese leader, was in office — even with the DeepSeek breakthrough. Mr. Xi is known for his dislike of the internet sector.
Business executives hope the government can refrain from interfering with DeepSeek.
“As is often the case with this government,” Zhang Fuyu, an entrepreneur, wrote on WeChat, “any promising company gets incorporated into a national strategy — receiving funding and resources but being subjected to state directives — or is directly placed under control, then DeepSeek’s future prospects may be entirely stifled.”
Under government regulations, A.I. models that serve consumers are subject to censorship rules. Many DeepSeek users posted photos and videos of its chatbot censoring topics including Xi Jinping, the war in Ukraine, the Cultural Revolution and the Tiananmen Square massacre.
“If DeepSeek is truly as remarkable as claimed, so impactful that it shakes the U.S. stock market, yet it remains confined to being just an AI model with Chinese socialist characteristics,” a journalist using the handle Xiaoming wrote on her Threads account, “then that would be truly tragic.”
