bodyType

customContents

entityId

entityUuid

articlePageQuery

filter

article

articleBaseAdvertisingInfoArticle

articleListArticle

articleSeoArticle

hooksArticleAcknowledgementGateOverlayArticle

hooksTrackPageViewArticle

hooksContentInterestArticle

articleLinkingDataContent

hooksAmpRedirectArticle

articleSchemaContent

liveArticle

advertZone

sentiment

contentLock

topics

types

sections

paywallTypes

createdDate

publishedDate

urlAlias

moreOnThisArticles

sponsorType

commentCount

authorLocations

authors

corrections

displaySlideShow

multimediaEmbed

printHeadline

seriesContainer

articleSpeeches

socialHeadline

headline

images

flag

firstTopic

glossary

flattenTypeIds

image

relatedLinks

relatedNewsletters

__typename

sponsor

pdfFiles

seriesReverseOrder

series

disableOffPlatformContent

published

updatedDate

summary

subHeadline

body

keywords

readingTime

customTimezone

liveContents

Josephine Ma

After DeepSeek became a worldwide talking point and sent shock waves through the US stock market, many observers have been trying to find out how a relatively small company could become a game-changer for the AI industry.
The Hangzhou-based start-up was founded only two years ago by a hedge fund High Flyer and it was little known before the launch of its large language model (LLM) last month.
It not only challenges the dominance of similar US products such as OpenAI’s ChatGPT, but also conventional wisdom about the cost of producing these products.
In recent weeks, the company has launched two powerful AI models – DeepSeek-V3 and DeepSeek-R1 – which were built at a fraction of the cost and computing power used by US tech giants for their products.
The announcement sent shares of semiconductor giant Nvidia and other AI-related stocks tumbling.
Another surprise was that a Chinese company had been able to make such a breakthrough despite US sanctions.
The US has banned the shipment of advanced Nvidia AI chips A100 and H100 since 2022, prompting the firm to develop modified versions for the Chinese market. The restrictions were further tightened in 2023, prompting Nvidia to release another set of less powerful variants for the Chinese market.
DeepSeek said it had trained its R1 model using just 2,000 or so of Nvidia’s less-advanced H800 chips and at a cost of US$5.6 million.
The company’s founder Liang Wenfeng has previously said that the firm had also stockpiled 10,000 graphics processing units from Nvidia before further export restrictions came into force.
While DeepSeek’s success is impressive, it highlights two important lessons for China.
Firstly, although it managed to use only a fraction of the graphics processing units (GPUs) and computing resources to train its chatbot that US firms needed, it still had to rely on advanced AI chips from the US.


While DeepSeek and other Chinese AI giants have reportedly been stockpiling such chips, some analysts believe supplies may run short as they try to upgrade their products – especially if the US further tightens export controls.
A second lesson is that the breakthrough was made possible in a company with a corporate culture and governance style that is totally different from the Chinese norm.
Media reports say that DeepSeek’s team consists almost exclusively of Chinese nationals from the country’s top universities, most of them with little to no previous experience. This may encourage them to try new ideas rather than copying what other tech giants have done before.
Liang has said in one of his rare interviews that he prizes curiosity and passion more than experience when he hires people. “A lot of people thirst for the opportunity to do research – and that desire far exceeds their need for money,” he told 3KR, a Chinese tech website, in 2023.
The company reportedly does not use key performance indicators (KPIs) and has a very flat structure. Instead of having a corporate hierarchy, a common approach to business and governance in China, it adopts the collaborative approach which is more common in academia and allows people to apply their specialist skills to different projects.
“Our conclusion is, to have innovation we should minimise intervention and management to give everyone the space for trial and error. Innovation usually occurs naturally instead of being a result of intentional arrangement or teaching,” Liang said during the same interview.
DeepSeek’s team is also relatively small with about 200 people, compared with the tech giants that hire thousands of people, and the company reportedly only hires those who excel in universities or in competitions.
Liang himself also reportedly spent much time on coding.


All of these factors make innovation possible as a pool of young talent with no baggage are allowed to collaborate and explore new ideas instead of being constrained by the bureaucracy, hierarchy, harsh working culture and politics of conventional Chinese companies.
And most important of all, the success of DeepSeek highlights the value of the private sector because it is willing to explore new management styles to boost innovation.
While China may not be able to produce advanced AI chips soon, at least it should do more to protect and encourage the private sector by giving them a stable investment environment.


Has China’s stock market gone from ‘uninvestable’ to ‘irresistible’?

ByteDance, other major Chinese tech firms dominate local consumer AI market: report

Japan announces US$6 billion support for home-grown AI development

Protection of vulnerable workers must be part of China’s AI strategy

DeepSeek

DeepSeek’s success isn’t just a wake-up call for the US, there are lessons for China too

Chinese AI start-up DeepSeek was founded two years ago by a hedge fund and was little known until last month. Photo: Reuters

Tech war: US firms embrace DeepSeek’s AI model despite scrutiny

US tech firms are rushing to adopt DeepSeek as part of their AI offerings. Photo: Reuters

Nvidia CEO Jensen Huang. Photo: AFP

What DeepSeek’s success means for Nvidia and costly GPU-driven AI growth

Nvidia’s stock has taken a beating amid concerns that DeepSeek’s AI breakthroughs could reduce reliance on the US semiconductor giant’s hardware. Photo: Reuters

News

China

Science

Opinion

China Opinion

The AI firm’s challenge to US dominance was achieved through a corporate culture that is very different from the hierarchical norm in China


The AI firm’s challenge to US dominance was achieved through a corporate culture that is very different from the hierarchical norm in China.


My Take

Artificial intelligence

China technology

Technology

China’s private sector

Expresses the views, ideas and conclusions of the author. (Eg. Enough of Crazy Rich Asians, think of crazy poor Hongkongers.)


Column

.css-1m2rp1p{color:#34b8b5;}My Take | DeepSeek’s success isn’t just a wake-up call for the US, there are lessons for China too

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}The AI firm’s challenge to US dominance was achieved through a corporate culture that is very different from the hierarchical norm in China

My Take | DeepSeek’s success isn’t just a wake-up call for the US, there are lessons for China too

The AI firm’s challenge to US dominance was achieved through a corporate culture that is very different from the hierarchical norm in China