Ben Jiang,Ann Cao,Wency ChenandZhou Xin

Published: 11:00am, 8 Feb 2025

In 2022, under the pseudonym “an ordinary little pig”, the founder of artificial intelligence (AI) start-up DeepSeek quietly donated 138 million yuan (US$19 million) to Chinese charities. The name, a nod to Wang Xiaobo’s dark-humour short story “A Maverick Pig”, reportedly masked the true identity of Liang Wenfeng, a man whose quantum trading operation was facing mounting regulatory pressure.

On January 28 that year, Liang’s hedge fund High-Flyer Quant issued a statement to the public denying that its quantum trading was responsible for a plunge in Chinese stocks.

“In recent days, there was market gossip that a coordinated sell-off by quant trading institutions had led to a market downfall,” High-flyer said. “We’ve discussed with managers of the major players in the industry and they all denied such a thing. Please don’t believe the gossip, thanks.”

Three years later, DeepSeek, a spin-off from High-Flyer, would make headlines for actually being responsible for roiling the US stock market. On January 27, a massive sell-off wiped out nearly US$1 trillion in tech stocks, including US$600 billion from Nvidia alone.

The catalyst was the release of DeepSeek’s R1 reasoning model, which came just weeks after it unveiled the more traditional V3 large language model (LLM). These models rivalled the performance of OpenAI’s offerings at a fraction of the cost, casting doubt on the assumptions underpinning the high valuations of US chip and AI companies.

Silicon Valley venture capitalist Marc Andreessen hailed the R1 model as AI’s “Sputnik moment”, referring to when the Soviet Union sent the first satellite into space, spurring the Space Race with the US. Deutsche Bank analyst Peter Milliken said “it is more China’s Sputnik moment”, showing the country’s overall strength in AI innovation and capabilities.

Even US President Donald Trump took notice, calling DeepSeek a “wake-up call” in the duel for AI dominance.

China’s press was exuberant. The official Xinhua news agency said in an editorial this week that DeepSeek “is expected to set off a new wave of AI innovation” that will accelerate the process of building up an autonomous and controllable industrial ecology.

Science and Technology Daily, a newspaper run by China’s science ministry, said DeepSeek’s breakthrough defied the West’s “compute hegemony”. “DeepSeek is like a strong light beaming through the foggy mist,” it said. Rao Yi, a renowned Chinese life scientist, called DeepSeek the greatest invention by China since the Opium War in the 1840s, and one tech executive wrote in a widely shared social media post that DeepSeek could change China’s national fate.

DeepSeek’s breakthrough comes as China has been pushed to the corner by the US in AI competition since OpenAI launched ChatGPT in late 2022. The US has tightened export controls on advanced chips to China, particularly Nvidia’s graphics processing units (GPU) that are seen as indispensable in training AI models, depriving China of the most powerful weapons in the battle. Meanwhile, Chinese AI companies are disadvantaged by a smaller capital market, making them unable to match the eye-watering expenditures of their US counterparts.

With a lead on AI advancements and a near monopoly on advanced chips, the US has become the most formidable global player in the industry. And where China is concerned, it has been flexing its muscles. Amid regulatory uncertainty, OpenAI has kept ChatGPT inaccessible in China. Last summer, it closed a loophole that allowed people to use its application programming interface (API) without a virtual private network, blocking the country’s internet protocol addresses. China was effectively being treated like Iran or North Korea.

DeepSeek-R1 reset the narrative by shattering notions of just how much money it takes to be a leader in AI.

Senior Chinese officials, including Guangdong communist party secretary Huang Kunming, have shared in the triumphalism of corporate executives and commentators, washing away a sense of inferiority and self-doubt since the launch of ChatGPT. China’s traditional advantages in the field, such as vast troves of data, suddenly looked irrelevant.

Does the arrival of China’s low-cost DeepSeek mean the end of Nvidia’s chip dominance?

Now it is OpenAI CEO Sam Altman doing the soul searching. He recently said his company was on the “wrong side of history” regarding its proprietary models, noting the advancements that open-source models had made. He said OpenAI needed a “different open-source strategy”, but added that it was not a high priority.

While OpenAI said it had evidence that DeepSeek used a process called distillation to train its models on the US firm’s data without proper authorisation, doubts about DeepSeek quickly dissipated when Altman said he had no plans to sue.

DeepSeek’s popularity overseas has also raised privacy concerns. Italy, Australia and South Korea, as well as the US Navy and Nasa, have banned the use of the Chinese firm’s AI chatbot on government devices over privacy and security concerns. Since DeepSeek’s models are open source, however, US tech giants such as Nvidia, Microsoft and Amazon.com have offered hosting solutions that do not send data to China.

“Never underestimate the ingenuity of Chinese scientists and engineers,” Fu Cong, China’s permanent representative to the United Nations, said at a press conference in New York on Tuesday. “From Huawei to TikTok, and now to DeepSeek – how many more does the US want to ban?”

Tilly Zhang, an analyst with financial services company Gavekal, noted that DeepSeek showed how Chinese companies were making remarkable strides in software innovation, mitigating the constraints imposed by US export controls on hardware. “The race for AI leadership is no longer just about who has access to the best chips, but about who puts them to best use,” Zhang said.

At home, DeepSeek’s influence has skyrocketed, turning what some once called an obscure hedge fund “side project” into a source of national pride and technological power. It has rapidly become the centre of gravity for China’s chaotic AI industry, with major players from GPU developers to cloud service providers paying tribute to the low-key firm.

The country’s chip developers and AI infrastructure service providers, including Moore Threads and Iluvatar Corex, rushed to adapt their hardware to run DeepSeek models in the hope that the start-up could lead the creation of a completely autonomous AI chain. For instance, Gitee AI, a Shenzhen-based one-stop service and platform for AI developers, said it was offering four DeepSeek-R1-based models through servers powered by GPUs from Shanghai-based chip designer MetaX. Gittee, China’s alternative to Microsoft’s GitHub, added that it has “realised the full use of domestically developed and made-in-China technologies from chips and platforms to computing power and models.”

Besides GPU developers, China’s major cloud service providers – including Tencent Holdings, Huawei Technologies, Baidu, and Alibaba Group Holding, which owns the South China Morning Post – worked overtime to support DeepSeek’s new models on their respective platforms, amid soaring demand from Chinese consumers and businesses.

Meanwhile, DeepSeek’s own app, released in early January, also saw rapid adoption. Chinese consumers rushed to download the chatbot during the Lunar New Year holiday, generating poems and articles that celebrated the season. The DeepSeek AI assistant, offered as a free app on Android, Apple’s iOS and the web, had an average of 22.2 million daily active users (DAUs) in January, surpassing Doubao’s 17 million DAUs in the same period to become the most popular AI app from China, according to the latest data compiled by Aicpb.com, a website that tracks the popularity of global AI services.

There are already early signs that the DeepSeek effect has started to trickle downstream to other sectors, empowering areas such as China’s smart manufacturing efforts. Hong Kong-listed UBTech Robotics, one of the leading makers of humanoid robots, has reportedly started to test DeepSeek on its robots. And in the healthcare market, when the AI medical platform ClouDr added DeepSeek-R1, the Chinese company’s Hong Kong-listed shares jumped.

Demand from developers has been so overwhelming that DeepSeek on Thursday temporarily stopped users from topping up their accounts for accessing the models through its API. In the onshore stock markets in Shanghai and Shenzhen, shares of companies even remotely related to DeepSeek surged amid an investor frenzy. In Hong Kong, the Hang Seng Tech Index, whose biggest members include Tencent and Xiaomi, approached a four-month high on Thursday, after rallying more than 10 per cent over the past two weeks, thanks to strong purchase following DeepSeek’s success.

Amid all the high-profile attention, DeepSeek has kept its head low. Its office in a commercial property in downtown Hangzhou became the Mecca for tech journalists and jobseekers, while Liang’s hometown gave the entrepreneur a hero’s welcome when he returned during the Lunar New Year holiday. Yet the company has avoided promoting itself, staying quiet on rumours and its own achievements.

In an early interview with local Chinese tech media outlet 36Kr, following the launch of DeepSeek-V2 in May 2024, Liang expressed disappointment that Chinese firms were reluctant to conduct their own frontier research. “China has to be a contributor [to global innovation] and not just always get a free ride,” he said. He pointed out that innovation is driven by curiosity and a desire to create, rather than just business needs.

In the widely shared profile about Liang and DeepSeek, the start-up is described as an “extreme case of technology idealism”, with a founder who has a clear vision for changing the world through innovation.

Far from being an “ordinary little pig”, all eyes are on Liang, as he is now firmly at the forefront of China’s AI revolution.

Ben Jiang

Ben is a Beijing-based technology reporter for the Post focusing on emerging start-ups. He has previously covered Chinese tech for publications including KrAsia and TechNode.

Ann Cao

Ann Cao is a Shanghai-based technology reporter for the Post, covering technology start-ups and policies in the city and eastern China. She graduated from the University of Hong Kong with a master's degree in journalism.

Wency Chen

FOLLOW

Wency tells stories that explore how technologies are reshaping society, with a focus on cross-border e-commerce, AI, the supply chain and others. Before joining SCMP, Wency contributed to KrASIA, Wired, Rest of World, World of Chinese, Tech in Asia, Vice China (BIE), Harper's Bazaar, etc. She attended Columbia Journalism School.

Zhou Xin

FOLLOW

Zhou Xin is Tech Editor of the Post, following stints as Political Economy Editor and Deputy China Editor. He has previously worked for Reuters and Bloomberg in Beijing.